null
nil
-
0
0
0
0
false
Vibe Coder
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
7 then
taskLabel.Text = "Analyzing the problem…"
end
if time_elapsed > 15 then
taskLabel.Text = "Writing code…"
end
if time_elapsed > 25 then
taskLabel.Text = "Still working — hang tight…"
end
end
end)
if not okThink then
pd.model_thinking_frame = nil
end
end
local current_model_object = model_data
if not current_model_object then
pd.current_model = "auto"
current_model_object = AISendHandler.get_model_data(pd.current_model)
end
if not current_model_object then
ComposerControls.update(pd)
DiffReviewUI.update(pd)
return
end
if is_auto then
-- Auto uses the free backend; no BYOK key needed
pd.widget.App.BottomThread.ApiKeyInput.Visible = false
pd.widget.App.BottomThread.ApiKeyHiddenLabel.Visible = false
pd.widget.App.BottomThread.ShowApiKeyButton.Visible = false
pd.widget.App.BottomThread.ApiKeyInput.PlaceholderText = "(Auto — no key needed)"
else
pd.widget.App.BottomThread.ShowApiKeyButton.Visible = true
module.update_api_key_visibility(pd)
if current_model_object.can_use_gg then
pd.widget.App.BottomThread.ApiKeyInput.PlaceholderText = "(optional) API key"
else
pd.widget.App.BottomThread.ApiKeyInput.PlaceholderText = "API key (required)"
end
end
if current_model_object.not_recommended then
pd.widget.App.WarningFrame.WarningContent.Text = "We don't recommend using " .. current_model_object.name .. " for long scripts."
pd.widget.App.WarningFrame.Visible = true
else
pd.widget.App.WarningFrame.Visible = false
end
ComposerControls.update(pd)
DiffReviewUI.update(pd)
-- Busy affordance on Send
local send = pd.widget.App.BottomThread:FindFirstChild("SendButton")
if send then
send.Active = pd.model_bussy ~= true
if send:IsA("ImageButton") or send:IsA("GuiButton") then
send.AutoButtonColor = pd.model_bussy ~= true
end
pcall(function()
send.ImageTransparency = if pd.model_bussy then 0.45 else 0
end)
end
end
return module
]]>
{3B271064-AA46-4DFA-8FB5-448472F8FFB0}
0
false
WidgetWindow
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{2B2A504F-4FB9-4325-9CF1-41F8B5593494}
0
false
WidgetModelSelectFrameOpener
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{B3E5806A-799D-4A7A-82B3-0D86B1FB15C6}
0
false
GgHandler
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{CE577970-C424-4A58-B52A-D460D6FFD4A9}
0
false
ModelRegistry
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
= FREE_MESSAGES_LIMIT then
pd.widget.BuyProFrame.Visible = true
pd.widget.BuyProFrame.Header.Text = "You have reached your FREE daily messages limit."
return
end
-- Mark busy BEFORE any yielding UI work to prevent double-send races.
pd.model_bussy = true
refresh_composer(pd)
local model_data_early = module.get_model_data(pd.current_model)
local is_auto_early = model_data_early and model_data_early.is_auto
-- Optional focus change (Cursor-style). Don't wipe Auto memory on deselect.
if pd.last_message_selected_script ~= pd.selected_script then
if pd.selected_script ~= nil and pd.last_message_selected_script ~= nil then
spawn_system_bubble(pd, "--- focused script changed ---")
if not is_auto_early then
pd.history = {}
end
elseif pd.selected_script ~= nil and pd.last_message_selected_script == nil then
spawn_system_bubble(pd, "Focused: " .. tostring(pd.selected_script.Name))
end
end
pd.last_message_selected_script = pd.selected_script
pd.widget.App.BottomThread.ChatInputTextBox.Text = ""
-- Persist API key on send (not every UI tick)
pcall(function()
pd.plugin:SetSetting(
"vibe-coder-secret-" .. (pd.current_model:gsub("[^%w]", "")),
user_api_key
)
end)
pcall(function()
local version = CheckUpdate.is_update_required()
if version ~= "no" then
spawn_system_bubble(pd, version .. " is out! Please go to https://rbxvibecoder.com and update your Vibe Coder.")
end
end)
spawn_user_bubble(pd, user_prompt)
pcall(function()
AtMentions.capture(pd, user_prompt)
end)
local mentionSummary = ""
pcall(function()
mentionSummary = AtMentions.summary(pd) or ""
end)
if mentionSummary ~= "" then
spawn_system_bubble(pd, mentionSummary)
end
local model_data = module.get_model_data(pd.current_model)
if model_data and model_data.is_auto then
pd.model_thinking_frame = spawn_model_thinking_frame(pd)
pd.model_thinking_started = tick()
pd.on_agent_step = function(kind, text, extra)
pcall(function()
if kind == "tool" then
local detail = ""
if typeof(extra) == "table" then
detail = tostring(extra.detail or extra.path or "")
end
ChatUI.spawnTool(pd, text, detail)
elseif kind == "diff" then
local path = text
if typeof(extra) == "table" and extra.path then
path = extra.path
end
path = tostring(path):gsub("^Queued change:%s*", "")
DiffReviewUI.notifyQueued(pd, path)
end
end)
end
local okRun, pack = pcall(function()
local text, err = AgentLoop.run(pd, user_prompt)
return { text = text, err = err }
end)
if not okRun then
spawn_model_bubble(pd, "Agent crashed: " .. tostring(pack))
else
local pending = 0
for _, e in ipairs(pd.pending_edits or {}) do
if not e.applied then
pending += 1
end
end
if pack.err then
local errText = tostring(pack.err)
if errText == "Stopped" then
errText = "Stopped."
end
spawn_model_bubble(pd, errText)
elseif pack.text and pack.text ~= "" then
spawn_model_bubble(pd, pack.text)
bump_usage(pd)
elseif pending > 0 then
-- Tool-only turn that queued edits still counts as a successful run.
bump_usage(pd)
end
if not pack.err and pending > 0 then
pcall(function()
ChatUI.spawnCardsForPending(pd)
end)
spawn_system_bubble(pd, string.format("%d pending edit(s) — Accept/Reject per file, or use Diff Review.", pending))
end
end
cleanup_thinking(pd)
return
end
local system_prompt = SystemInstructions.get_system_instructions(pd)
local okByok, byokErr = pcall(function()
return handle_ai_interaction(pd, user_prompt, system_prompt, user_api_key)
end)
if not okByok then
spawn_model_bubble(pd, "Request failed: " .. tostring(byokErr))
end
cleanup_thinking(pd)
end
return module
]]>
{161C3D43-D28B-4B6D-93E0-249568D69AAA}
0
false
AISendHandler
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
")
end
local function build_hierarchy_string_recursive(
current_obj,
indent_level,
feature_object_to_mark,
current_lines,
printed_objects_count_ref,
max_depth
)
if printed_objects_count_ref.count >= MAX_OBJECTS_TO_PRINT then
return
end
if not current_obj or current_obj == game or indent_level > max_depth then
return
end
local indent = string.rep(" ", indent_level)
local marker = current_obj == feature_object_to_mark and " (selected)" or ""
local entry = indent .. "- " .. current_obj.Name .. " (" .. current_obj.ClassName .. ")" .. marker
table.insert(current_lines, entry)
printed_objects_count_ref.count = printed_objects_count_ref.count + 1
local children = current_obj:GetChildren()
table.sort(children, function(a, b)
return a.Name < b.Name
end)
for _, child in ipairs(children) do
if printed_objects_count_ref.count >= MAX_OBJECTS_TO_PRINT then
break
end
build_hierarchy_string_recursive(
child,
indent_level + 1,
feature_object_to_mark,
current_lines,
printed_objects_count_ref,
max_depth
)
end
end
function module.get_game_hierarchy_string(feature_object)
if not feature_object or not feature_object.Parent then
return "Unknown location"
end
local context = {}
local printed_objects_count = { count = 0 }
local selected_object_path = getPathUpTo(feature_object)
table.insert(context, "Selected object path: " .. selected_object_path)
table.insert(context, "")
local hierarchy_snapshot_root = feature_object
while hierarchy_snapshot_root.Parent and hierarchy_snapshot_root.Parent ~= game do
hierarchy_snapshot_root = hierarchy_snapshot_root.Parent
end
table.insert(
context,
"Game Hierarchy Snapshot (max "
.. MAX_OBJECTS_TO_PRINT
.. " objects, max depth "
.. MAX_DEPTH_FOR_CONTEXT
.. "):"
)
table.insert(context, "----------------------------------------------------")
build_hierarchy_string_recursive(
hierarchy_snapshot_root,
0,
feature_object,
context,
printed_objects_count,
MAX_DEPTH_FOR_CONTEXT
)
if printed_objects_count.count >= MAX_OBJECTS_TO_PRINT then
table.insert(context, "\n... (Hierarchy truncated due to object limit)")
end
local marked = false
for _, line in ipairs(context) do
if string.find(line, "(selected)", 1, true) then
marked = true
break
end
end
if not marked then
table.insert(
context,
"\nHierarchy note: selected object may be deeper than snapshot depth — see path above."
)
end
return table.concat(context, "\n")
end
return module
]]>
{423CDCBD-4250-4532-9331-3A3B9B169D6C}
0
false
PrintGameHierarhy
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
false
0
{4E27C2EA-0816-468F-8E92-B68F91982143}
0
false
Plugin
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{AF020E13-BBDB-4D31-8392-ED767FAFC236}
0
false
SystemInstructions
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{E3965072-934B-4334-AEC3-C74F72B82CEA}
0
false
FeedbackWidgetWindow
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{CCF0E175-A015-4436-90E4-3EA083AC7C87}
0
false
GamepassHandler
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
l[i] then
return true
end
if r[i] < l[i] then
return false
end
end
return false
end
local function fetchVersion(url)
local success, response = pcall(function()
return HttpService:GetAsync(url)
end)
if not success then
return nil
end
local okJson, decoded = pcall(function()
return HttpService:JSONDecode(response)
end)
if okJson and typeof(decoded) == "table" then
return decoded.display or decoded.version
end
return nil
end
function module.is_update_required()
local localVersionString = script.Parent.VersionTitle.Value
local remoteVersionString = fetchVersion(VERSION_API_URL)
or fetchVersion(BACKEND_VERSION_URL)
or fetchVersion(LOCAL_VERSION_URL)
if not remoteVersionString then
return "no"
end
if isNewer(remoteVersionString, localVersionString) then
return tostring(remoteVersionString)
end
return "no"
end
return module
]]>
{E7D61F42-6DFD-4F2D-9E6E-E6C0630F983A}
0
false
CheckUpdate
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
0
0
0
false
Prefabs
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
1
0
0.345098048
0.427451015
0.490196109
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.0624782071
0
0
false
[null]
0
0
0.850000024
0
0.0624782741
0
0
true
1
true
[null]
0
0
0
0
false
0
false
UserBubble
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
User Content
1
1
1
0
false
14
0
0
0
1
0
0
true
0
0
false
0.5
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.5
0
0
false
[null]
0
0
0.899999976
0
0.75
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ContentLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
1
0
0.345098048
0.427451015
0.490196109
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.0624782071
0
0
false
[null]
0
0
0.850000024
0
0.25
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ModelBubble
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Model Response
1
1
1
0
false
14
0
0
0
1
0
0
true
0
0
false
0.5
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.5
0
0
false
[null]
0
0
0.899999976
0
0.949999988
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ContentLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
1
0
0.345098048
0.427451015
0.490196109
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.221606731
0
0
false
[null]
0
0
0.850000083
0
0.141555712
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ModelThinkingFrame
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0
0
0
0.156862751
0.156862751
0.156862751
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
false
[null]
0
0
0.649999976
0
1
0
0
true
1
true
[null]
0
0
0
0
false
0
false
Bubble
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Model is thinking for
1
1
1
0
true
14
0
0
0
1
0.25
0
true
0
0
false
0.5
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.635996699
0
0.244562343
0
0
false
[null]
0
0
0.639432073
0
0.239124462
0
0
true
1
true
[null]
0
0
0
0
false
0
false
TaskNameLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
3 seconds
1
1
1
0
true
14
0
0
0
1
0.25
0
true
0
0
false
0.5
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.635996699
0
0.619992435
0
0
false
[null]
0
0
0.639432073
0
0.51173532
0
0
true
1
true
[null]
0
0
0
0
false
0
false
TimeLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://126455106069276
1
1
1
0
0
0
0
0.25
0
3
0
0
0
0
1
1
0
1
0
false
0
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
1.83940596e-08
0
0.5
0
0
false
[null]
0
0
0.316280693
0
0.5
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ImageLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
1
0
0.345098048
0.427451015
0.490196109
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.221606731
0
0
false
[null]
0
0
0.850000083
0
0.141555712
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ModelEditedFrame
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0
0
0
0.156862751
0.156862751
0.156862751
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
false
[null]
0
0
0.649999976
0
1
0
0
true
1
true
[null]
0
0
0
0
false
0
false
Bubble
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Model edited your scripts.
1
1
1
0
true
14
0
0
0
1
0.25
0
true
0
0
false
0.5
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.635996699
0
0.244562343
0
0
false
[null]
0
0
0.639432073
0
0.239124462
0
0
true
1
true
[null]
0
0
0
0
false
0
false
TaskNameLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://80289912970625
1
1
1
0
0
0
0
0.25
0
3
0
0
0
0
1
1
0
1
0
false
0
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
1.83940596e-08
0
0.5
0
0
false
[null]
0
0
0.316280693
0
0.5
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ImageLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Revert Changes
1
1
1
0
false
14
0
0
0
1
0
0
false
2
1
true
[null]
false
[null]
false
0
true
0
0
0
0.435294151
0.239215702
0.239215702
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.316280752
0
0.465525955
0
0
true
[null]
0
1
0.649999976
0
0.349999994
0
0
true
1
true
[null]
0
0
0
0
false
0
false
RevertButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Accept AI Changes
1
1
1
0
false
14
0
0
0
1
0
0
false
2
1
true
[null]
false
[null]
false
0
true
0
0
0
0.321568638
0.435294151
0.235294133
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.316280752
0
0.465525955
0
0
true
[null]
0
1
0.649999976
0
0.349999994
0
0
false
1
true
[null]
0
0
0
0
false
0
false
AcceptButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
true
0
1
2
true
true
1
true
[null]
0
0
0
0
false
0
false
FeedbackGUI
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0
0
0
0.117647059
0.117647059
0.117647059
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
false
[null]
0
0
1
0
1
0
0
true
0
true
[null]
0
0
0
0
false
0
false
App
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
false
rbxasset://fonts/families/FredokaOne.json
400
1
-1
true
0.529411793
0.529411793
0.529411793
Tell me bout your Vibe Coder expirience!
false
true
1
1
1
0
true
false
20
0
0
0
1
0.150000006
0
true
0
0
true
0.5
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.49955985
0
0.0386468396
0
0
true
[null]
0
1
0.826584518
0
0.60202527
0
0
true
1
true
[null]
0
0
0
0
false
0
false
FeedbackInput
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Send
1
1
1
0
true
14
0
0
0
1
0
0
true
2
1
true
[null]
false
[null]
false
0
true
0
0
0
0.258823544
0.258823544
0.258823544
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.0862676054
0
0.832203388
0
0
true
[null]
0
1
0.826584518
0
0.141525418
0
0
true
1
true
[null]
0
0
0
0
false
0
false
SendButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
If you want me to contact you back, please leave your email in the feedback.
1
1
1
0
true
20
0
0
0
1
0.25
0
true
0
0
false
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.0860000029
0
0.677053869
0
0
false
[null]
0
0
0.824852169
1
0.117847033
1
0
true
1
true
[null]
0
0
0
0
false
0
false
TextLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
true
0
1
2
true
true
1
true
[null]
0
0
0
0
false
0
false
WidgetGUI
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0
0
0
0.117647059
0.117647059
0.117647059
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
false
[null]
0
0
1
0
1
0
0
false
10
true
[null]
0
0
0
0
false
0
false
NoSelectionFrame
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.5
0
0
false
[null]
0
0
0.5
0
0.150000006
0
0
true
11
true
[null]
0
0
0
0
false
0
false
Frame
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
0
0
0
0
false
1
1
2
1
0
false
UIListLayout
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://79115352535620
1
1
1
0
0
0
0
0.25
0
3
0
0
0
0
1
1
0
1
0
false
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
false
[null]
0
0
1
0
0.569999993
0
0
true
11
true
[null]
0
0
0
0
false
0
false
ImageLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Select an script in Explorer first.
1
1
1
0
true
14
0
0
0
1
0.25
0
true
2
1
false
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
false
[null]
0
0
1
0
0.5
0
0
true
11
true
[null]
0
0
0
0
false
0
false
TextLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/GothamSSm.json
400
rbxasset://fonts/Montserrat-Regular.ttf
1
-1
false
Select a script to edit, or continue with Auto — the agent can work across your place.
0.70588237
0.70588237
0.745098054
0
false
13
0
0
0
1
0
0
true
2
1
false
0
0
0
0.639215708
0.635294139
0.647058845
1
0.105882354
0.164705887
0.20784314
0
1
false
false
true
0
[null]
[null]
[null]
[null]
0.0500000007
0
0.620000005
0
0
false
[null]
0
0
0.899999976
0
0
40
0
true
1
true
[null]
0
0
0
0
false
0
false
AutoHint
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/GothamSSm.json
500
rbxasset://fonts/Montserrat-Medium.ttf
1
-1
false
Continue with Auto
1
1
1
0
false
14
0
0
0
1
0
0
false
2
1
true
[null]
false
[null]
false
0
true
0
0
0
0.215686277
0.352941185
0.627451003
0
0.105882354
0.164705887
0.20784314
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
-100
0.720000029
0
0
true
[null]
0
1
0
200
0
36
0
true
1
true
[null]
0
0
0
0
false
0
false
ContinueAutoButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0
0
0
0.117647059
0.117647059
0.117647059
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
false
[null]
0
0
1
0
1
0
0
true
0
true
[null]
0
0
0
0
false
0
false
App
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
1
0
0.156862751
0.156862751
0.156862751
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.98999995
0
0
false
[null]
0
0
0.949999988
0
0.167168677
0
0
true
1
true
[null]
0
0
0
0
false
0
false
BottomThread
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
1
0
0.482352942
0.541176498
0.78039217
0.899999976
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.256017506
0
0.220721155
0
0
false
[null]
0
0
0.480832011
0
0.154647633
0
0
false
1
true
[null]
0
0
0
0
false
0
false
SelectedScriptFrame
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Script
1
1
1
0
true
18
0
0
0
1
0.25
0
true
2
1
false
0.5
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.520498514
0
0.499998957
0
0
false
[null]
0
0
0.685664654
0
0.899999976
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ScriptTitle
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://113661567766482
1
1
1
0
0
0
0
0.25
0
3
0
0
0
0
1
1
0
1
0
false
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
3
0
0
0
false
[null]
0
0
0.200000003
0
1
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ScriptIcon
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
false
rbxasset://fonts/families/FredokaOne.json
400
1
-1
true
0.529411793
0.529411793
0.529411793
Chat with Vibe Coder… (@ to mention, Auto to edit the place)
false
true
1
1
1
0
true
false
14
0
0
0
1
0.150000006
0
true
0
0
true
0.5
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.497061342
0
0.299999893
0
0
true
[null]
0
1
0.944122851
0
0.41400972
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ChatInputTextBox
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://123181546754641
1
1
1
0
0
0
0
0.25
0
3
0
0
0
0
1
1
0
1
0
true
[null]
false
[null]
false
0
true
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.892009795
0
0.733529866
0
0
true
[null]
0
1
0.0900033265
0
0.160360187
0
0
true
1
true
[null]
0
0
0
0
false
0
false
SendButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/SourceSansPro.json
400
rbxasset://fonts/SourceSansPro-Regular.ttf
1
-1
false
0
0
0
0
false
14
0
0
0
1
1
0
false
2
1
true
[null]
false
[null]
false
0
true
0
0
0
0.137254909
0.137254909
0.137254909
0.899999976
0
0
0
0
0
false
false
true
1
[null]
[null]
[null]
[null]
0.0299999993
0
0.706502259
0
0
true
[null]
0
1
0.271118104
0
0.216216221
0
0
true
2
true
[null]
0
0
0
0
false
0
false
CurrentModelButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://79274818810097
1
1
1
0
0
0
0
0.25
0
3
0
0
0
0
1
1
0
1
0
false
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.800000131
0
0.125
0
0
false
[null]
0
0
0.179667592
0
0.741667807
0
0
true
1
true
[null]
0
0
0
0
false
0
false
DropdownIcon
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
no model selected.
1
1
1
0
true
18
0
0
0
1
0.25
0
true
0
1
false
0.5
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.414509743
0
0.499999851
0
0
false
[null]
0
0
0.779018283
0
0.899999917
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ModelTitle
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
false
rbxasset://fonts/families/FredokaOne.json
400
1
-1
true
0.529411793
0.529411793
0.529411793
API key
false
true
1
1
1
0
true
false
14
0
0
0
1
0.150000006
0
true
0
1
true
0.5
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.678268731
0
0.733529329
0
0
true
[null]
0
1
0.287856996
0
0.151351243
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ApiKeyInput
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://73515595170574
1
1
1
0
0
0
0
0.25
0
3
0
0
0
0
1
1
0
1
0
true
[null]
false
[null]
false
0
true
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.819884002
0
0.733529747
0
0
true
[null]
0
1
0.0900033265
0
0.160360187
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ShowApiKeyButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
API KEY HIDDEN
1
1
1
0
true
14
0
0
0
1
0.25
0
true
2
1
false
0.5
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.677999973
0
0.734000027
0
0
false
[null]
0
0
0.287999988
0
0.150999993
0
0
false
1
true
[null]
0
0
0
0
false
0
false
ApiKeyHiddenLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://130372484515289
1
1
1
0
0
0
0
0.25
0
3
0
0
0
0
1
1
0
1
0
true
[null]
false
[null]
false
0
true
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.928200066
0
0.0738055855
0
0
true
[null]
0
1
0.0531407259
0
0.146914348
0
0
true
1
true
[null]
0
0
0
0
false
0
false
OpenMenuButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/GothamSSm.json
500
rbxasset://fonts/Montserrat-Medium.ttf
1
-1
false
Stop
1
0.862745106
0.862745106
0
false
12
0
0
0
1
0
0
false
2
1
true
[null]
false
[null]
false
0
true
1
0
0
0.470588237
0.156862751
0.156862751
0
0.105882354
0.164705887
0.20784314
0
0
false
false
true
0
[null]
[null]
[null]
[null]
1
-58
0
6
0
true
[null]
0
1
0
50
0
22
0
false
5
true
[null]
0
0
0
0
false
0
false
StopButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
6
0
6
0
6
0
6
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
1
0
0.0980392173
0.0980392173
0.0980392173
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.274051219
0
0.940936327
0
0
false
[null]
0
0
0.440239549
0
0.220514685
0
0
false
2
true
[null]
0
0
0
0
false
0
false
ModelSelectFrame
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Select AI Model
1
1
1
0
false
14
0
0
0
1
0.75
0
false
0
1
false
0.5
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.363621145
0
0.014999954
0
0
false
[null]
0
0
0.677242219
0
0.129761562
0
0
true
2
true
[null]
0
0
0
0
false
0
false
Header
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
2
rbxasset://textures/ui/Scroll/scroll-bottom.png
0
0
0
0
0
200
0
0
rbxasset://textures/ui/Scroll/scroll-middle.png
0
0
0
0
12
2
true
rbxasset://textures/ui/Scroll/scroll-top.png
0
0
true
0.5
0
0
1
1
1
1
0
0
0
0
0
true
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.185738772
0
0
true
[null]
0
0
0.949999928
0
0.814261019
0
0
true
2
true
[null]
0
0
0
0
true
0
false
ModelsList
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
1
0
0
8
0
false
1
0
2
1
0
false
UIListLayout
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
ModelName
1
1
1
0
true
14
0
0
0
1
0.25
0
true
0
1
true
[null]
false
[null]
false
0
true
0.5
0
0
0.0980392173
0.0980392173
0.0980392173
0.899999976
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.0243607201
0
0.0309071168
0
0
true
[null]
0
1
0.75
0
0
20
0
false
2
true
[null]
0
0
0
0
false
0
false
ModelSelectButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
0
0
0.678431392
0.494117647
0.956862748
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.842000008
0
0.0149999997
0
0
false
[null]
0
0
0.211999997
0
0.129999995
0
0
true
2
true
[null]
0
0
0
0
false
0
false
ProLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
PRO
1
1
1
0
true
14
0
0
0
1
0
0
true
2
1
false
0.5
0.5
0
0.678431392
0.494117647
0.956862748
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.5
0
0
false
[null]
0
0
1
0
0.75
0
0
true
2
true
[null]
0
0
0
0
false
0
false
Text
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
2
rbxasset://textures/ui/Scroll/scroll-bottom.png
0
0
0
0
1
0
0
0
rbxasset://textures/ui/Scroll/scroll-middle.png
0
0
0
1
12
2
true
rbxasset://textures/ui/Scroll/scroll-top.png
0
0
true
0.5
0
0
1
1
1
1
0
0
0
0
0
true
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.0546247326
0
0
true
[null]
0
0
0.949999928
0
0.756085217
0
0
true
1
true
[null]
0
0
0
0
true
0
false
ChatHistory
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
0
0
8
0
false
1
0
2
1
0
false
UIListLayout
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
1
0
0.0980392173
0.0980392173
0.0980392173
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.753058791
0
0.81245333
0
0
false
[null]
0
0
0.440239578
0
0.155998558
0
0
false
2
true
[null]
0
0
0
0
false
0
false
MenuFrame
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
0
0
0
0
false
1
0
2
1
0
false
UIListLayout
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
1
1
1
0
true
14
0
0
0
1
0
0
true
2
1
true
[null]
false
[null]
false
0
true
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
true
[null]
0
1
0.899999976
0
0.330000013
0
0
true
3
true
[null]
0
0
0
0
false
0
false
ClearButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Clear chat
1
1
1
1
true
14
0
0
0
1
0.25
0
true
0
1
false
0
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.184383377
0
0.499999672
0
0
false
[null]
0
0
0.815616727
0
0.75
0
0
true
3
true
[null]
0
0
0
0
false
0
false
TextLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://72456202240333
1
1
1
0
0
0
0
0.25
0
3
0
0
0
0
1
1
0
1
0
false
0
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0.5
0
0
false
[null]
0
0
0.200000003
0
0.5
0
0
true
3
true
[null]
0
0
0
0
false
0
false
ImageLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
1
1
1
0
true
14
0
0
0
1
0
0
true
2
1
true
[null]
false
[null]
false
0
true
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
true
[null]
0
1
0.899999976
0
0.330000013
0
0
true
3
true
[null]
0
0
0
0
false
0
false
FeedbackButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Send Feedback
1
1
1
1
true
14
0
0
0
1
0.25
0
true
0
1
false
0
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.184383377
0
0.499999672
0
0
false
[null]
0
0
0.815616727
0
0.75
0
0
true
3
true
[null]
0
0
0
0
false
0
false
TextLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://90206302915644
1
1
1
0
0
0
0
0.25
0
3
0
0
0
0
1
1
0
1
0
false
0
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0.5
0
0
false
[null]
0
0
0.200000003
0
0.5
0
0
true
3
true
[null]
0
0
0
0
false
0
false
ImageLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
1
1
1
0
true
14
0
0
0
1
0
0
true
2
1
true
[null]
false
[null]
false
0
true
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
true
[null]
0
1
0.899999976
0
0.330000013
0
0
true
3
true
[null]
0
0
0
0
false
0
false
JoinSupportServerButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Join Community Server
1
1
1
1
true
14
0
0
0
1
0.25
0
true
0
1
false
0
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.184383377
0
0.499999672
0
0
false
[null]
0
0
0.815616727
0
0.75
0
0
true
3
true
[null]
0
0
0
0
false
0
false
TextLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://97572524929990
1
1
1
0
0
0
0
0.25
0
3
0
0
0
0
1
1
0
1
0
false
0
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0.5
0
0
false
[null]
0
0
0.200000003
0
0.5
0
0
true
3
true
[null]
0
0
0
0
false
0
false
ImageLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
1
0
0.403921574
0.223529413
0.223529413
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.80445838
0
0
false
[null]
0
0
0.949999988
0
0.0321747102
0
0
false
1
true
[null]
0
0
0
0
false
0
false
WarningFrame
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Warning
1
0.956862807
0.956862807
0
true
14
0
0
0
1
0
0
true
2
1
false
0.5
0.5
0
1
0.913725555
0.913725555
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.5
0
0
false
[null]
0
0
0.75
0
0.800000012
0
0
true
1
true
[null]
0
0
0
0
false
0
false
WarningContent
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://106040231377035
1
1
1
0
0
0
0
0
0
3
0
0
0
0
1
1
0
1
0
false
0
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0.5
0
0
false
[null]
0
0
0.100000001
0
0.75
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ImageLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0
0
0
0.125490203
0.141176477
0.164705887
0
0.105882354
0.164705887
0.20784314
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
8
0
8
0
false
[null]
0
0
1
-16
0
72
0
false
10
true
[null]
0
0
0
0
false
0
false
DiffReviewFrame
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/GothamSSm.json
500
rbxasset://fonts/Montserrat-Medium.ttf
1
-1
false
Pending edits: 0
0.90196079
0.90196079
0.941176474
0
false
13
0
0
0
1
0
0
false
0
1
false
0
0
0
0.639215708
0.635294139
0.647058845
1
0.105882354
0.164705887
0.20784314
0
1
false
false
true
0
[null]
[null]
[null]
[null]
0
8
0
6
0
false
[null]
0
0
1
-16
0
22
0
true
1
true
[null]
0
0
0
0
false
0
false
Title
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/GothamSSm.json
500
rbxasset://fonts/Montserrat-Medium.ttf
1
-1
false
Accept all
1
1
1
0
false
12
0
0
0
1
0
0
false
2
1
true
[null]
false
[null]
false
0
true
0
0
0
0.156862751
0.470588237
0.274509817
0
0.105882354
0.164705887
0.20784314
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
8
0
36
0
true
[null]
0
1
0
90
0
26
0
true
1
true
[null]
0
0
0
0
false
0
false
AcceptAllButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
6
0
6
0
6
0
6
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/GothamSSm.json
500
rbxasset://fonts/Montserrat-Medium.ttf
1
-1
false
Reject all
1
1
1
0
false
12
0
0
0
1
0
0
false
2
1
true
[null]
false
[null]
false
0
true
0
0
0
0.470588237
0.196078435
0.196078435
0
0.105882354
0.164705887
0.20784314
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
106
0
36
0
true
[null]
0
1
0
90
0
26
0
true
1
true
[null]
0
0
0
0
false
0
false
RejectAllButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
6
0
6
0
6
0
6
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Vibe Coder v1.0 — just chat. Selection optional.
1
1
1
0
true
14
0
0
0
1
0.25
0
true
2
1
false
0.5
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.00903614424
0
0
false
[null]
0
0
0.899999976
0
0.0355421677
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ReleaseInfo
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0
0
0
0.117647059
0.117647059
0.117647059
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
false
[null]
0
0
1
0
1
0
0
true
8
true
[null]
0
0
0
0
false
0
false
BuyProFrame
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://115853486982676
1
1
1
0
0
0
0
0
0
3
0
0
0
0
1
1
0
1
0
false
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.222636372
0
0.0271084346
0
0
false
[null]
0
0
0.596456707
0
0.191987902
0
0
true
8
true
[null]
0
0
0
0
false
0
false
ImageLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
You have discovered a Vibe Coder PRO feature.
1
1
1
0
true
14
0
0
0
1
0
0
true
2
1
false
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.0811600089
0
0.218373492
0
0
false
[null]
0
0
0.834194362
0
0.176204816
0
0
true
8
true
[null]
0
0
0
0
false
0
false
Header
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
1
1
1
0
true
14
0
0
0
1
0.25
0
true
0
1
false
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.0850794911
0
0.409638554
0
0
false
[null]
0
0
0.868840098
0
0.213855416
0
0
true
8
true
[null]
0
0
0
0
false
0
false
Paragraph
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
In order to unlock Vibe Coder PRO, buy T-Shirt that is available at the following link:
1
1
1
0
true
14
0
0
0
1
0.25
0
true
0
1
false
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.0868525654
0
0.701807201
0
0
false
[null]
0
0
0.868840098
0
0.0542168692
0
0
true
8
true
[null]
0
0
0
0
false
0
false
Paragraph
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
false
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
0.699999988
0.699999988
0.699999988
NO LINK
false
true
https://www.roblox.com/catalog/79884753121491/Vibe-Coder-PRO
0.678431392
0.494117677
0.956862807
0
false
true
14
0
0
0
1
0
0
true
2
1
true
0
0
0
0.156862751
0.156862751
0.156862751
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.0811600238
0
0.766566277
0
0
true
[null]
0
1
0.834194362
0
0.0798192769
0
0
true
8
true
[null]
0
0
0
0
false
0
false
GamepassLinkInput
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
8
0
8
0
8
0
8
0
false
UICorner
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
Copy this link and paste it in your browser.
1
1
1
0
true
14
0
0
0
1
0.75
0
true
0
1
false
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.0868525133
0
0.858433723
0
0
false
[null]
0
0
0.828041196
0
0.0180722885
0
0
true
8
true
[null]
0
0
0
0
false
0
false
Paragraph
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
IMPORTANT: Restart Roblox Studio after purchase.
1
0.674509823
0.674509823
0
true
14
1
0
0
1
0.25
0
true
0
1
false
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.0868524835
0
0.885542154
0
0
false
[null]
0
0
0.82850194
0
0.0223202705
0
0
true
8
true
[null]
0
0
0
0
false
0
false
Paragraph
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
I am not interested.
1
1
1
0
true
14
0
0
0
1
0.75
0
true
1
1
true
[null]
false
[null]
false
0
true
0
0
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.549645364
0
0.959337354
0
0
true
[null]
0
1
0.365248233
0
0.0256024096
0
0
true
8
true
[null]
0
0
0
0
false
0
false
CloseButton
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
false
0.5
1
0
0.345098048
0.427451015
0.490196109
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.0624782071
0
0
false
[null]
0
0
0.850000024
0
0.25
0
0
true
1
true
[null]
0
0
0
0
false
0
false
SystemBubble
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxasset://fonts/families/FredokaOne.json
400
1
-1
false
System
1
1
1
0
false
14
0
0
0
1
0.5
0
true
2
0
false
0.5
0.5
0
1
1
1
1
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0.5
0
0.5
0
0
false
[null]
0
0
0.899999976
0
0.949999988
0
0
true
1
true
[null]
0
0
0
0
false
0
false
ContentLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
0
0
0
false
Libs
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
256 or n - (n - 1) ~= 1 or m - (m - 1) ~= 1 or n == m then
return k, false -- floating point datatype
elseif n == prev_n then
return k, true -- integer datatype
end
end
end
-- Make sure Lua has "double" numbers
local x = 2/3
local Lua_has_double = x * 5 > 3 and x * 4 < 3 and get_precision(1.0) >= 53
assert(Lua_has_double, "at least 53-bit floating point numbers are required")
-- Q:
-- SHA2 was designed for FPU-less machines.
-- So, why floating point numbers are needed for this module?
-- A:
-- 53-bit "double" numbers are useful to calculate "magic numbers" used in SHA.
-- I prefer to write 50 LOC "magic numbers calculator" instead of storing more than 200 constants explicitly in this source file.
local int_prec, Lua_has_integers = get_precision(1)
local Lua_has_int64 = Lua_has_integers and int_prec == 64
local Lua_has_int32 = Lua_has_integers and int_prec == 32
assert(Lua_has_int64 or Lua_has_int32 or not Lua_has_integers, "Lua integers must be either 32-bit or 64-bit")
-- Q:
-- Does it mean that almost all non-standard configurations are not supported?
-- A:
-- Yes. Sorry, too many problems to support all possible Lua numbers configurations.
-- Lua 5.1/5.2 with "int32" will not work.
-- Lua 5.1/5.2 with "int64" will not work.
-- Lua 5.1/5.2 with "int128" will not work.
-- Lua 5.1/5.2 with "float" will not work.
-- Lua 5.1/5.2 with "double" is OK. (default config for Lua 5.1, Lua 5.2, LuaJIT)
-- Lua 5.3/5.4 with "int32" + "float" will not work.
-- Lua 5.3/5.4 with "int64" + "float" will not work.
-- Lua 5.3/5.4 with "int128" + "float" will not work.
-- Lua 5.3/5.4 with "int32" + "double" is OK. (config used by Fengari)
-- Lua 5.3/5.4 with "int64" + "double" is OK. (default config for Lua 5.3, Lua 5.4)
-- Lua 5.3/5.4 with "int128" + "double" will not work.
-- Using floating point numbers better than "double" instead of "double" is OK (non-IEEE-754 floating point implementation are allowed).
-- Using "int128" instead of "int64" is not OK: "int128" would require different branch of implementation for optimized SHA512.
-- Check for LuaJIT and 32-bit bitwise libraries
local is_LuaJIT = ({false, [1] = true})[1] and _VERSION ~= "Luau" and (type(jit) ~= "table" or jit.version_num >= 20000) -- LuaJIT 1.x.x and Luau are treated as vanilla Lua 5.1/5.2
local is_LuaJIT_21 -- LuaJIT 2.1+
local LuaJIT_arch
local ffi -- LuaJIT FFI library (as a table)
local b -- 32-bit bitwise library (as a table)
local library_name
if is_LuaJIT then
-- Assuming "bit" library is always available on LuaJIT
b = require"bit"
library_name = "bit"
-- "ffi" is intentionally disabled on some systems for safety reason
local LuaJIT_has_FFI, result = pcall(require, "ffi")
if LuaJIT_has_FFI then
ffi = result
end
is_LuaJIT_21 = not not loadstring"b=0b0"
LuaJIT_arch = type(jit) == "table" and jit.arch or ffi and ffi.arch or nil
else
-- For vanilla Lua, "bit"/"bit32" libraries are searched in global namespace only. No attempt is made to load a library if it's not loaded yet.
for _, libname in ipairs(_VERSION == "Lua 5.2" and {"bit32", "bit"} or {"bit", "bit32"}) do
if type(_G[libname]) == "table" and _G[libname].bxor then
b = _G[libname]
library_name = libname
break
end
end
end
--------------------------------------------------------------------------------
-- You can disable here some of your system's abilities (for testing purposes)
--------------------------------------------------------------------------------
-- is_LuaJIT = nil
-- is_LuaJIT_21 = nil
-- ffi = nil
-- Lua_has_int32 = nil
-- Lua_has_int64 = nil
-- b, library_name = nil
--------------------------------------------------------------------------------
if print_debug_messages then
-- Printing list of abilities of your system
print("Abilities:")
print(" Lua version: "..(is_LuaJIT and "LuaJIT "..(is_LuaJIT_21 and "2.1 " or "2.0 ")..(LuaJIT_arch or "")..(ffi and " with FFI" or " without FFI") or _VERSION))
print(" Integer bitwise operators: "..(Lua_has_int64 and "int64" or Lua_has_int32 and "int32" or "no"))
print(" 32-bit bitwise library: "..(library_name or "not found"))
end
-- Selecting the most suitable implementation for given set of abilities
local method, branch
if is_LuaJIT and ffi then
method = "Using 'ffi' library of LuaJIT"
branch = "FFI"
elseif is_LuaJIT then
method = "Using special code for sandboxed LuaJIT (no FFI)"
branch = "LJ"
elseif Lua_has_int64 then
method = "Using native int64 bitwise operators"
branch = "INT64"
elseif Lua_has_int32 then
method = "Using native int32 bitwise operators"
branch = "INT32"
elseif library_name then -- when bitwise library is available (Lua 5.2 with native library "bit32" or Lua 5.1 with external library "bit")
method = "Using '"..library_name.."' library"
branch = "LIB32"
else
method = "Emulating bitwise operators using look-up table"
branch = "EMUL"
end
if print_debug_messages then
-- Printing the implementation selected to be used on your system
print("Implementation selected:")
print(" "..method)
end
--------------------------------------------------------------------------------
-- BASIC 32-BIT BITWISE FUNCTIONS
--------------------------------------------------------------------------------
local AND, OR, XOR, SHL, SHR, ROL, ROR, NOT, NORM, HEX, XOR_BYTE
-- Only low 32 bits of function arguments matter, high bits are ignored
-- The result of all functions (except HEX) is an integer inside "correct range":
-- for "bit" library: (-2^31)..(2^31-1)
-- for "bit32" library: 0..(2^32-1)
if branch == "FFI" or branch == "LJ" or branch == "LIB32" then
-- Your system has 32-bit bitwise library (either "bit" or "bit32")
AND = b.band -- 2 arguments
OR = b.bor -- 2 arguments
XOR = b.bxor -- 2..5 arguments
SHL = b.lshift -- second argument is integer 0..31
SHR = b.rshift -- second argument is integer 0..31
ROL = b.rol or b.lrotate -- second argument is integer 0..31
ROR = b.ror or b.rrotate -- second argument is integer 0..31
NOT = b.bnot -- only for LuaJIT
NORM = b.tobit -- only for LuaJIT
HEX = b.tohex -- returns string of 8 lowercase hexadecimal digits
assert(AND and OR and XOR and SHL and SHR and ROL and ROR and NOT, "Library '"..library_name.."' is incomplete")
XOR_BYTE = XOR -- XOR of two bytes (0..255)
elseif branch == "EMUL" then
-- Emulating 32-bit bitwise operations using 53-bit floating point arithmetic
function SHL(x, n)
return (x * 2^n) % 2^32
end
function SHR(x, n)
x = x % 2^32 / 2^n
return x - x % 1
end
function ROL(x, n)
x = x % 2^32 * 2^n
local r = x % 2^32
return r + (x - r) / 2^32
end
function ROR(x, n)
x = x % 2^32 / 2^n
local r = x % 1
return r * 2^32 + (x - r)
end
local AND_of_two_bytes = {[0] = 0} -- look-up table (256*256 entries)
local idx = 0
for y = 0, 127 * 256, 256 do
for x = y, y + 127 do
x = AND_of_two_bytes[x] * 2
AND_of_two_bytes[idx] = x
AND_of_two_bytes[idx + 1] = x
AND_of_two_bytes[idx + 256] = x
AND_of_two_bytes[idx + 257] = x + 1
idx = idx + 2
end
idx = idx + 256
end
local function and_or_xor(x, y, operation)
-- operation: nil = AND, 1 = OR, 2 = XOR
local x0 = x % 2^32
local y0 = y % 2^32
local rx = x0 % 256
local ry = y0 % 256
local res = AND_of_two_bytes[rx + ry * 256]
x = x0 - rx
y = (y0 - ry) / 256
rx = x % 65536
ry = y % 256
res = res + AND_of_two_bytes[rx + ry] * 256
x = (x - rx) / 256
y = (y - ry) / 256
rx = x % 65536 + y % 256
res = res + AND_of_two_bytes[rx] * 65536
res = res + AND_of_two_bytes[(x + y - rx) / 256] * 16777216
if operation then
res = x0 + y0 - operation * res
end
return res
end
function AND(x, y)
return and_or_xor(x, y)
end
function OR(x, y)
return and_or_xor(x, y, 1)
end
function XOR(x, y, z, t, u) -- 2..5 arguments
if z then
if t then
if u then
t = and_or_xor(t, u, 2)
end
z = and_or_xor(z, t, 2)
end
y = and_or_xor(y, z, 2)
end
return and_or_xor(x, y, 2)
end
function XOR_BYTE(x, y)
return x + y - 2 * AND_of_two_bytes[x + y * 256]
end
end
HEX = HEX
or
pcall(string_format, "%x", 2^31) and
function (x) -- returns string of 8 lowercase hexadecimal digits
return string_format("%08x", x % 4294967296)
end
or
function (x) -- for OpenWrt's dialect of Lua
return string_format("%08x", (x + 2^31) % 2^32 - 2^31)
end
local function XORA5(x, y)
return XOR(x, y or 0xA5A5A5A5) % 4294967296
end
local function create_array_of_lanes()
return {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
end
--------------------------------------------------------------------------------
-- CREATING OPTIMIZED INNER LOOP
--------------------------------------------------------------------------------
-- Inner loop functions
local sha256_feed_64, sha512_feed_128, md5_feed_64, sha1_feed_64, keccak_feed, blake2s_feed_64, blake2b_feed_128, blake3_feed_64
-- Arrays of SHA-2 "magic numbers" (in "INT64" and "FFI" branches "*_lo" arrays contain 64-bit values)
local sha2_K_lo, sha2_K_hi, sha2_H_lo, sha2_H_hi, sha3_RC_lo, sha3_RC_hi = {}, {}, {}, {}, {}, {}
local sha2_H_ext256 = {[224] = {}, [256] = sha2_H_hi}
local sha2_H_ext512_lo, sha2_H_ext512_hi = {[384] = {}, [512] = sha2_H_lo}, {[384] = {}, [512] = sha2_H_hi}
local md5_K, md5_sha1_H = {}, {0x67452301, 0xEFCDAB89, 0x98BADCFE, 0x10325476, 0xC3D2E1F0}
local md5_next_shift = {0, 0, 0, 0, 0, 0, 0, 0, 28, 25, 26, 27, 0, 0, 10, 9, 11, 12, 0, 15, 16, 17, 18, 0, 20, 22, 23, 21}
local HEX64, lanes_index_base -- defined only for branches that internally use 64-bit integers: "INT64" and "FFI"
local common_W = {} -- temporary table shared between all calculations (to avoid creating new temporary table every time)
local common_W_blake2b, common_W_blake2s, v_for_blake2s_feed_64 = common_W, common_W, {}
local K_lo_modulo, hi_factor, hi_factor_keccak = 4294967296, 0, 0
local sigma = {
{ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 },
{ 15, 11, 5, 9, 10, 16, 14, 7, 2, 13, 1, 3, 12, 8, 6, 4 },
{ 12, 9, 13, 1, 6, 3, 16, 14, 11, 15, 4, 7, 8, 2, 10, 5 },
{ 8, 10, 4, 2, 14, 13, 12, 15, 3, 7, 6, 11, 5, 1, 16, 9 },
{ 10, 1, 6, 8, 3, 5, 11, 16, 15, 2, 12, 13, 7, 9, 4, 14 },
{ 3, 13, 7, 11, 1, 12, 9, 4, 5, 14, 8, 6, 16, 15, 2, 10 },
{ 13, 6, 2, 16, 15, 14, 5, 11, 1, 8, 7, 4, 10, 3, 9, 12 },
{ 14, 12, 8, 15, 13, 2, 4, 10, 6, 1, 16, 5, 9, 7, 3, 11 },
{ 7, 16, 15, 10, 12, 4, 1, 9, 13, 3, 14, 8, 2, 5, 11, 6 },
{ 11, 3, 9, 5, 8, 7, 2, 6, 16, 12, 10, 15, 4, 13, 14, 1 },
}; sigma[11], sigma[12] = sigma[1], sigma[2]
local perm_blake3 = {
1, 3, 4, 11, 13, 10, 12, 6,
1, 3, 4, 11, 13, 10,
2, 7, 5, 8, 14, 15, 16, 9,
2, 7, 5, 8, 14, 15,
}
local function build_keccak_format(elem)
local keccak_format = {}
for _, size in ipairs{1, 9, 13, 17, 18, 21} do
keccak_format[size] = "<"..string_rep(elem, size)
end
return keccak_format
end
if branch == "FFI" then
local common_W_FFI_int32 = ffi.new("int32_t[?]", 80) -- 64 is enough for SHA256, but 80 is needed for SHA-1
common_W_blake2s = common_W_FFI_int32
v_for_blake2s_feed_64 = ffi.new("int32_t[?]", 16)
perm_blake3 = ffi.new("uint8_t[?]", #perm_blake3 + 1, 0, unpack(perm_blake3))
for j = 1, 10 do
sigma[j] = ffi.new("uint8_t[?]", #sigma[j] + 1, 0, unpack(sigma[j]))
end; sigma[11], sigma[12] = sigma[1], sigma[2]
-- SHA256 implementation for "LuaJIT with FFI" branch
function sha256_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W, K = common_W_FFI_int32, sha2_K_hi
for pos = offs, offs + size - 1, 64 do
for j = 0, 15 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos) -- slow, but doesn't depend on endianness
W[j] = OR(SHL(a, 24), SHL(b, 16), SHL(c, 8), d)
end
for j = 16, 63 do
local a, b = W[j-15], W[j-2]
W[j] = NORM( XOR(ROR(a, 7), ROL(a, 14), SHR(a, 3)) + XOR(ROL(b, 15), ROL(b, 13), SHR(b, 10)) + W[j-7] + W[j-16] )
end
local a, b, c, d, e, f, g, h = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for j = 0, 63, 8 do -- Thanks to Peter Cawley for this workaround (unroll the loop to avoid "PHI shuffling too complex" due to PHIs overlap)
local z = NORM( XOR(g, AND(e, XOR(f, g))) + XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + (W[j] + K[j+1] + h) )
h, g, f, e = g, f, e, NORM( d + z )
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(g, AND(e, XOR(f, g))) + XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + (W[j+1] + K[j+2] + h) )
h, g, f, e = g, f, e, NORM( d + z )
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(g, AND(e, XOR(f, g))) + XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + (W[j+2] + K[j+3] + h) )
h, g, f, e = g, f, e, NORM( d + z )
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(g, AND(e, XOR(f, g))) + XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + (W[j+3] + K[j+4] + h) )
h, g, f, e = g, f, e, NORM( d + z )
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(g, AND(e, XOR(f, g))) + XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + (W[j+4] + K[j+5] + h) )
h, g, f, e = g, f, e, NORM( d + z )
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(g, AND(e, XOR(f, g))) + XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + (W[j+5] + K[j+6] + h) )
h, g, f, e = g, f, e, NORM( d + z )
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(g, AND(e, XOR(f, g))) + XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + (W[j+6] + K[j+7] + h) )
h, g, f, e = g, f, e, NORM( d + z )
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(g, AND(e, XOR(f, g))) + XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + (W[j+7] + K[j+8] + h) )
h, g, f, e = g, f, e, NORM( d + z )
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
end
H[1], H[2], H[3], H[4] = NORM(a + H[1]), NORM(b + H[2]), NORM(c + H[3]), NORM(d + H[4])
H[5], H[6], H[7], H[8] = NORM(e + H[5]), NORM(f + H[6]), NORM(g + H[7]), NORM(h + H[8])
end
end
local common_W_FFI_int64 = ffi.new("int64_t[?]", 80)
common_W_blake2b = common_W_FFI_int64
local int64 = ffi.typeof"int64_t"
local int32 = ffi.typeof"int32_t"
local uint32 = ffi.typeof"uint32_t"
hi_factor = int64(2^32)
if is_LuaJIT_21 then -- LuaJIT 2.1 supports bitwise 64-bit operations
local AND64, OR64, XOR64, NOT64, SHL64, SHR64, ROL64, ROR64 -- introducing synonyms for better code readability
= AND, OR, XOR, NOT, SHL, SHR, ROL, ROR
HEX64 = HEX
-- BLAKE2b implementation for "LuaJIT 2.1 + FFI" branch
do
local v = ffi.new("int64_t[?]", 16)
local W = common_W_blake2b
local function G(a, b, c, d, k1, k2)
local va, vb, vc, vd = v[a], v[b], v[c], v[d]
va = W[k1] + (va + vb)
vd = ROR64(XOR64(vd, va), 32)
vc = vc + vd
vb = ROR64(XOR64(vb, vc), 24)
va = W[k2] + (va + vb)
vd = ROR64(XOR64(vd, va), 16)
vc = vc + vd
vb = ROL64(XOR64(vb, vc), 1)
v[a], v[b], v[c], v[d] = va, vb, vc, vd
end
function blake2b_feed_128(H, _, str, offs, size, bytes_compressed, last_block_size, is_last_node)
-- offs >= 0, size >= 0, size is multiple of 128
local h1, h2, h3, h4, h5, h6, h7, h8 = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for pos = offs, offs + size - 1, 128 do
if str then
for j = 1, 16 do
pos = pos + 8
local a, b, c, d, e, f, g, h = byte(str, pos - 7, pos)
W[j] = XOR64(OR(SHL(h, 24), SHL(g, 16), SHL(f, 8), e) * int64(2^32), uint32(int32(OR(SHL(d, 24), SHL(c, 16), SHL(b, 8), a))))
end
end
v[0x0], v[0x1], v[0x2], v[0x3], v[0x4], v[0x5], v[0x6], v[0x7] = h1, h2, h3, h4, h5, h6, h7, h8
v[0x8], v[0x9], v[0xA], v[0xB], v[0xD], v[0xE], v[0xF] = sha2_H_lo[1], sha2_H_lo[2], sha2_H_lo[3], sha2_H_lo[4], sha2_H_lo[6], sha2_H_lo[7], sha2_H_lo[8]
bytes_compressed = bytes_compressed + (last_block_size or 128)
v[0xC] = XOR64(sha2_H_lo[5], bytes_compressed) -- t0 = low_8_bytes(bytes_compressed)
-- t1 = high_8_bytes(bytes_compressed) = 0, message length is always below 2^53 bytes
if last_block_size then -- flag f0
v[0xE] = NOT64(v[0xE])
end
if is_last_node then -- flag f1
v[0xF] = NOT64(v[0xF])
end
for j = 1, 12 do
local row = sigma[j]
G(0, 4, 8, 12, row[ 1], row[ 2])
G(1, 5, 9, 13, row[ 3], row[ 4])
G(2, 6, 10, 14, row[ 5], row[ 6])
G(3, 7, 11, 15, row[ 7], row[ 8])
G(0, 5, 10, 15, row[ 9], row[10])
G(1, 6, 11, 12, row[11], row[12])
G(2, 7, 8, 13, row[13], row[14])
G(3, 4, 9, 14, row[15], row[16])
end
h1 = XOR64(h1, v[0x0], v[0x8])
h2 = XOR64(h2, v[0x1], v[0x9])
h3 = XOR64(h3, v[0x2], v[0xA])
h4 = XOR64(h4, v[0x3], v[0xB])
h5 = XOR64(h5, v[0x4], v[0xC])
h6 = XOR64(h6, v[0x5], v[0xD])
h7 = XOR64(h7, v[0x6], v[0xE])
h8 = XOR64(h8, v[0x7], v[0xF])
end
H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8] = h1, h2, h3, h4, h5, h6, h7, h8
return bytes_compressed
end
end
-- SHA-3 implementation for "LuaJIT 2.1 + FFI" branch
local arr64_t = ffi.typeof"int64_t[?]"
-- lanes array is indexed from 0
lanes_index_base = 0
hi_factor_keccak = int64(2^32)
function create_array_of_lanes()
return arr64_t(30) -- 25 + 5 for temporary usage
end
function keccak_feed(lanes, _, str, offs, size, block_size_in_bytes)
-- offs >= 0, size >= 0, size is multiple of block_size_in_bytes, block_size_in_bytes is positive multiple of 8
local RC = sha3_RC_lo
local qwords_qty = SHR(block_size_in_bytes, 3)
for pos = offs, offs + size - 1, block_size_in_bytes do
for j = 0, qwords_qty - 1 do
pos = pos + 8
local h, g, f, e, d, c, b, a = byte(str, pos - 7, pos) -- slow, but doesn't depend on endianness
lanes[j] = XOR64(lanes[j], OR64(OR(SHL(a, 24), SHL(b, 16), SHL(c, 8), d) * int64(2^32), uint32(int32(OR(SHL(e, 24), SHL(f, 16), SHL(g, 8), h)))))
end
for round_idx = 1, 24 do
for j = 0, 4 do
lanes[25 + j] = XOR64(lanes[j], lanes[j+5], lanes[j+10], lanes[j+15], lanes[j+20])
end
local D = XOR64(lanes[25], ROL64(lanes[27], 1))
lanes[1], lanes[6], lanes[11], lanes[16] = ROL64(XOR64(D, lanes[6]), 44), ROL64(XOR64(D, lanes[16]), 45), ROL64(XOR64(D, lanes[1]), 1), ROL64(XOR64(D, lanes[11]), 10)
lanes[21] = ROL64(XOR64(D, lanes[21]), 2)
D = XOR64(lanes[26], ROL64(lanes[28], 1))
lanes[2], lanes[7], lanes[12], lanes[22] = ROL64(XOR64(D, lanes[12]), 43), ROL64(XOR64(D, lanes[22]), 61), ROL64(XOR64(D, lanes[7]), 6), ROL64(XOR64(D, lanes[2]), 62)
lanes[17] = ROL64(XOR64(D, lanes[17]), 15)
D = XOR64(lanes[27], ROL64(lanes[29], 1))
lanes[3], lanes[8], lanes[18], lanes[23] = ROL64(XOR64(D, lanes[18]), 21), ROL64(XOR64(D, lanes[3]), 28), ROL64(XOR64(D, lanes[23]), 56), ROL64(XOR64(D, lanes[8]), 55)
lanes[13] = ROL64(XOR64(D, lanes[13]), 25)
D = XOR64(lanes[28], ROL64(lanes[25], 1))
lanes[4], lanes[14], lanes[19], lanes[24] = ROL64(XOR64(D, lanes[24]), 14), ROL64(XOR64(D, lanes[19]), 8), ROL64(XOR64(D, lanes[4]), 27), ROL64(XOR64(D, lanes[14]), 39)
lanes[9] = ROL64(XOR64(D, lanes[9]), 20)
D = XOR64(lanes[29], ROL64(lanes[26], 1))
lanes[5], lanes[10], lanes[15], lanes[20] = ROL64(XOR64(D, lanes[10]), 3), ROL64(XOR64(D, lanes[20]), 18), ROL64(XOR64(D, lanes[5]), 36), ROL64(XOR64(D, lanes[15]), 41)
lanes[0] = XOR64(D, lanes[0])
lanes[0], lanes[1], lanes[2], lanes[3], lanes[4] = XOR64(lanes[0], AND64(NOT64(lanes[1]), lanes[2]), RC[round_idx]), XOR64(lanes[1], AND64(NOT64(lanes[2]), lanes[3])), XOR64(lanes[2], AND64(NOT64(lanes[3]), lanes[4])), XOR64(lanes[3], AND64(NOT64(lanes[4]), lanes[0])), XOR64(lanes[4], AND64(NOT64(lanes[0]), lanes[1]))
lanes[5], lanes[6], lanes[7], lanes[8], lanes[9] = XOR64(lanes[8], AND64(NOT64(lanes[9]), lanes[5])), XOR64(lanes[9], AND64(NOT64(lanes[5]), lanes[6])), XOR64(lanes[5], AND64(NOT64(lanes[6]), lanes[7])), XOR64(lanes[6], AND64(NOT64(lanes[7]), lanes[8])), XOR64(lanes[7], AND64(NOT64(lanes[8]), lanes[9]))
lanes[10], lanes[11], lanes[12], lanes[13], lanes[14] = XOR64(lanes[11], AND64(NOT64(lanes[12]), lanes[13])), XOR64(lanes[12], AND64(NOT64(lanes[13]), lanes[14])), XOR64(lanes[13], AND64(NOT64(lanes[14]), lanes[10])), XOR64(lanes[14], AND64(NOT64(lanes[10]), lanes[11])), XOR64(lanes[10], AND64(NOT64(lanes[11]), lanes[12]))
lanes[15], lanes[16], lanes[17], lanes[18], lanes[19] = XOR64(lanes[19], AND64(NOT64(lanes[15]), lanes[16])), XOR64(lanes[15], AND64(NOT64(lanes[16]), lanes[17])), XOR64(lanes[16], AND64(NOT64(lanes[17]), lanes[18])), XOR64(lanes[17], AND64(NOT64(lanes[18]), lanes[19])), XOR64(lanes[18], AND64(NOT64(lanes[19]), lanes[15]))
lanes[20], lanes[21], lanes[22], lanes[23], lanes[24] = XOR64(lanes[22], AND64(NOT64(lanes[23]), lanes[24])), XOR64(lanes[23], AND64(NOT64(lanes[24]), lanes[20])), XOR64(lanes[24], AND64(NOT64(lanes[20]), lanes[21])), XOR64(lanes[20], AND64(NOT64(lanes[21]), lanes[22])), XOR64(lanes[21], AND64(NOT64(lanes[22]), lanes[23]))
end
end
end
local A5_long = 0xA5A5A5A5 * int64(2^32 + 1) -- It's impossible to use constant 0xA5A5A5A5A5A5A5A5LL because it will raise syntax error on other Lua versions
function XORA5(long, long2)
return XOR64(long, long2 or A5_long)
end
-- SHA512 implementation for "LuaJIT 2.1 + FFI" branch
function sha512_feed_128(H, _, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 128
local W, K = common_W_FFI_int64, sha2_K_lo
for pos = offs, offs + size - 1, 128 do
for j = 0, 15 do
pos = pos + 8
local a, b, c, d, e, f, g, h = byte(str, pos - 7, pos) -- slow, but doesn't depend on endianness
W[j] = OR64(OR(SHL(a, 24), SHL(b, 16), SHL(c, 8), d) * int64(2^32), uint32(int32(OR(SHL(e, 24), SHL(f, 16), SHL(g, 8), h))))
end
for j = 16, 79 do
local a, b = W[j-15], W[j-2]
W[j] = XOR64(ROR64(a, 1), ROR64(a, 8), SHR64(a, 7)) + XOR64(ROR64(b, 19), ROL64(b, 3), SHR64(b, 6)) + W[j-7] + W[j-16]
end
local a, b, c, d, e, f, g, h = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for j = 0, 79, 8 do
local z = XOR64(ROR64(e, 14), ROR64(e, 18), ROL64(e, 23)) + XOR64(g, AND64(e, XOR64(f, g))) + h + K[j+1] + W[j]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XOR64(AND64(XOR64(a, b), c), AND64(a, b)) + XOR64(ROR64(a, 28), ROL64(a, 25), ROL64(a, 30)) + z
z = XOR64(ROR64(e, 14), ROR64(e, 18), ROL64(e, 23)) + XOR64(g, AND64(e, XOR64(f, g))) + h + K[j+2] + W[j+1]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XOR64(AND64(XOR64(a, b), c), AND64(a, b)) + XOR64(ROR64(a, 28), ROL64(a, 25), ROL64(a, 30)) + z
z = XOR64(ROR64(e, 14), ROR64(e, 18), ROL64(e, 23)) + XOR64(g, AND64(e, XOR64(f, g))) + h + K[j+3] + W[j+2]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XOR64(AND64(XOR64(a, b), c), AND64(a, b)) + XOR64(ROR64(a, 28), ROL64(a, 25), ROL64(a, 30)) + z
z = XOR64(ROR64(e, 14), ROR64(e, 18), ROL64(e, 23)) + XOR64(g, AND64(e, XOR64(f, g))) + h + K[j+4] + W[j+3]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XOR64(AND64(XOR64(a, b), c), AND64(a, b)) + XOR64(ROR64(a, 28), ROL64(a, 25), ROL64(a, 30)) + z
z = XOR64(ROR64(e, 14), ROR64(e, 18), ROL64(e, 23)) + XOR64(g, AND64(e, XOR64(f, g))) + h + K[j+5] + W[j+4]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XOR64(AND64(XOR64(a, b), c), AND64(a, b)) + XOR64(ROR64(a, 28), ROL64(a, 25), ROL64(a, 30)) + z
z = XOR64(ROR64(e, 14), ROR64(e, 18), ROL64(e, 23)) + XOR64(g, AND64(e, XOR64(f, g))) + h + K[j+6] + W[j+5]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XOR64(AND64(XOR64(a, b), c), AND64(a, b)) + XOR64(ROR64(a, 28), ROL64(a, 25), ROL64(a, 30)) + z
z = XOR64(ROR64(e, 14), ROR64(e, 18), ROL64(e, 23)) + XOR64(g, AND64(e, XOR64(f, g))) + h + K[j+7] + W[j+6]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XOR64(AND64(XOR64(a, b), c), AND64(a, b)) + XOR64(ROR64(a, 28), ROL64(a, 25), ROL64(a, 30)) + z
z = XOR64(ROR64(e, 14), ROR64(e, 18), ROL64(e, 23)) + XOR64(g, AND64(e, XOR64(f, g))) + h + K[j+8] + W[j+7]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XOR64(AND64(XOR64(a, b), c), AND64(a, b)) + XOR64(ROR64(a, 28), ROL64(a, 25), ROL64(a, 30)) + z
end
H[1] = a + H[1]
H[2] = b + H[2]
H[3] = c + H[3]
H[4] = d + H[4]
H[5] = e + H[5]
H[6] = f + H[6]
H[7] = g + H[7]
H[8] = h + H[8]
end
end
else -- LuaJIT 2.0 doesn't support 64-bit bitwise operations
local U = ffi.new("union{int64_t i64; struct{int32_t "..(ffi.abi("le") and "lo, hi" or "hi, lo")..";} i32;}[3]")
-- this array of unions is used for fast splitting int64 into int32_high and int32_low
-- "xorrific" 64-bit functions :-)
-- int64 input is splitted into two int32 parts, some bitwise 32-bit operations are performed, finally the result is converted to int64
-- these functions are needed because bit.* functions in LuaJIT 2.0 don't work with int64_t
local function XORROR64_1(a)
-- return XOR64(ROR64(a, 1), ROR64(a, 8), SHR64(a, 7))
U[0].i64 = a
local a_lo, a_hi = U[0].i32.lo, U[0].i32.hi
local t_lo = XOR(SHR(a_lo, 1), SHL(a_hi, 31), SHR(a_lo, 8), SHL(a_hi, 24), SHR(a_lo, 7), SHL(a_hi, 25))
local t_hi = XOR(SHR(a_hi, 1), SHL(a_lo, 31), SHR(a_hi, 8), SHL(a_lo, 24), SHR(a_hi, 7))
return t_hi * int64(2^32) + uint32(int32(t_lo))
end
local function XORROR64_2(b)
-- return XOR64(ROR64(b, 19), ROL64(b, 3), SHR64(b, 6))
U[0].i64 = b
local b_lo, b_hi = U[0].i32.lo, U[0].i32.hi
local u_lo = XOR(SHR(b_lo, 19), SHL(b_hi, 13), SHL(b_lo, 3), SHR(b_hi, 29), SHR(b_lo, 6), SHL(b_hi, 26))
local u_hi = XOR(SHR(b_hi, 19), SHL(b_lo, 13), SHL(b_hi, 3), SHR(b_lo, 29), SHR(b_hi, 6))
return u_hi * int64(2^32) + uint32(int32(u_lo))
end
local function XORROR64_3(e)
-- return XOR64(ROR64(e, 14), ROR64(e, 18), ROL64(e, 23))
U[0].i64 = e
local e_lo, e_hi = U[0].i32.lo, U[0].i32.hi
local u_lo = XOR(SHR(e_lo, 14), SHL(e_hi, 18), SHR(e_lo, 18), SHL(e_hi, 14), SHL(e_lo, 23), SHR(e_hi, 9))
local u_hi = XOR(SHR(e_hi, 14), SHL(e_lo, 18), SHR(e_hi, 18), SHL(e_lo, 14), SHL(e_hi, 23), SHR(e_lo, 9))
return u_hi * int64(2^32) + uint32(int32(u_lo))
end
local function XORROR64_6(a)
-- return XOR64(ROR64(a, 28), ROL64(a, 25), ROL64(a, 30))
U[0].i64 = a
local b_lo, b_hi = U[0].i32.lo, U[0].i32.hi
local u_lo = XOR(SHR(b_lo, 28), SHL(b_hi, 4), SHL(b_lo, 30), SHR(b_hi, 2), SHL(b_lo, 25), SHR(b_hi, 7))
local u_hi = XOR(SHR(b_hi, 28), SHL(b_lo, 4), SHL(b_hi, 30), SHR(b_lo, 2), SHL(b_hi, 25), SHR(b_lo, 7))
return u_hi * int64(2^32) + uint32(int32(u_lo))
end
local function XORROR64_4(e, f, g)
-- return XOR64(g, AND64(e, XOR64(f, g)))
U[0].i64 = f
U[1].i64 = g
U[2].i64 = e
local f_lo, f_hi = U[0].i32.lo, U[0].i32.hi
local g_lo, g_hi = U[1].i32.lo, U[1].i32.hi
local e_lo, e_hi = U[2].i32.lo, U[2].i32.hi
local result_lo = XOR(g_lo, AND(e_lo, XOR(f_lo, g_lo)))
local result_hi = XOR(g_hi, AND(e_hi, XOR(f_hi, g_hi)))
return result_hi * int64(2^32) + uint32(int32(result_lo))
end
local function XORROR64_5(a, b, c)
-- return XOR64(AND64(XOR64(a, b), c), AND64(a, b))
U[0].i64 = a
U[1].i64 = b
U[2].i64 = c
local a_lo, a_hi = U[0].i32.lo, U[0].i32.hi
local b_lo, b_hi = U[1].i32.lo, U[1].i32.hi
local c_lo, c_hi = U[2].i32.lo, U[2].i32.hi
local result_lo = XOR(AND(XOR(a_lo, b_lo), c_lo), AND(a_lo, b_lo))
local result_hi = XOR(AND(XOR(a_hi, b_hi), c_hi), AND(a_hi, b_hi))
return result_hi * int64(2^32) + uint32(int32(result_lo))
end
local function XORROR64_7(a, b, m)
-- return ROR64(XOR64(a, b), m), m = 1..31
U[0].i64 = a
U[1].i64 = b
local a_lo, a_hi = U[0].i32.lo, U[0].i32.hi
local b_lo, b_hi = U[1].i32.lo, U[1].i32.hi
local c_lo, c_hi = XOR(a_lo, b_lo), XOR(a_hi, b_hi)
local t_lo = XOR(SHR(c_lo, m), SHL(c_hi, -m))
local t_hi = XOR(SHR(c_hi, m), SHL(c_lo, -m))
return t_hi * int64(2^32) + uint32(int32(t_lo))
end
local function XORROR64_8(a, b)
-- return ROL64(XOR64(a, b), 1)
U[0].i64 = a
U[1].i64 = b
local a_lo, a_hi = U[0].i32.lo, U[0].i32.hi
local b_lo, b_hi = U[1].i32.lo, U[1].i32.hi
local c_lo, c_hi = XOR(a_lo, b_lo), XOR(a_hi, b_hi)
local t_lo = XOR(SHL(c_lo, 1), SHR(c_hi, 31))
local t_hi = XOR(SHL(c_hi, 1), SHR(c_lo, 31))
return t_hi * int64(2^32) + uint32(int32(t_lo))
end
local function XORROR64_9(a, b)
-- return ROR64(XOR64(a, b), 32)
U[0].i64 = a
U[1].i64 = b
local a_lo, a_hi = U[0].i32.lo, U[0].i32.hi
local b_lo, b_hi = U[1].i32.lo, U[1].i32.hi
local t_hi, t_lo = XOR(a_lo, b_lo), XOR(a_hi, b_hi)
return t_hi * int64(2^32) + uint32(int32(t_lo))
end
local function XOR64(a, b)
-- return XOR64(a, b)
U[0].i64 = a
U[1].i64 = b
local a_lo, a_hi = U[0].i32.lo, U[0].i32.hi
local b_lo, b_hi = U[1].i32.lo, U[1].i32.hi
local t_lo, t_hi = XOR(a_lo, b_lo), XOR(a_hi, b_hi)
return t_hi * int64(2^32) + uint32(int32(t_lo))
end
local function XORROR64_11(a, b, c)
-- return XOR64(a, b, c)
U[0].i64 = a
U[1].i64 = b
U[2].i64 = c
local a_lo, a_hi = U[0].i32.lo, U[0].i32.hi
local b_lo, b_hi = U[1].i32.lo, U[1].i32.hi
local c_lo, c_hi = U[2].i32.lo, U[2].i32.hi
local t_lo, t_hi = XOR(a_lo, b_lo, c_lo), XOR(a_hi, b_hi, c_hi)
return t_hi * int64(2^32) + uint32(int32(t_lo))
end
function XORA5(long, long2)
-- return XOR64(long, long2 or 0xA5A5A5A5A5A5A5A5)
U[0].i64 = long
local lo32, hi32 = U[0].i32.lo, U[0].i32.hi
local long2_lo, long2_hi = 0xA5A5A5A5, 0xA5A5A5A5
if long2 then
U[1].i64 = long2
long2_lo, long2_hi = U[1].i32.lo, U[1].i32.hi
end
lo32 = XOR(lo32, long2_lo)
hi32 = XOR(hi32, long2_hi)
return hi32 * int64(2^32) + uint32(int32(lo32))
end
function HEX64(long)
U[0].i64 = long
return HEX(U[0].i32.hi)..HEX(U[0].i32.lo)
end
-- SHA512 implementation for "LuaJIT 2.0 + FFI" branch
function sha512_feed_128(H, _, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 128
local W, K = common_W_FFI_int64, sha2_K_lo
for pos = offs, offs + size - 1, 128 do
for j = 0, 15 do
pos = pos + 8
local a, b, c, d, e, f, g, h = byte(str, pos - 7, pos) -- slow, but doesn't depend on endianness
W[j] = OR(SHL(a, 24), SHL(b, 16), SHL(c, 8), d) * int64(2^32) + uint32(int32(OR(SHL(e, 24), SHL(f, 16), SHL(g, 8), h)))
end
for j = 16, 79 do
W[j] = XORROR64_1(W[j-15]) + XORROR64_2(W[j-2]) + W[j-7] + W[j-16]
end
local a, b, c, d, e, f, g, h = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for j = 0, 79, 8 do
local z = XORROR64_3(e) + XORROR64_4(e, f, g) + h + K[j+1] + W[j]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XORROR64_5(a, b, c) + XORROR64_6(a) + z
z = XORROR64_3(e) + XORROR64_4(e, f, g) + h + K[j+2] + W[j+1]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XORROR64_5(a, b, c) + XORROR64_6(a) + z
z = XORROR64_3(e) + XORROR64_4(e, f, g) + h + K[j+3] + W[j+2]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XORROR64_5(a, b, c) + XORROR64_6(a) + z
z = XORROR64_3(e) + XORROR64_4(e, f, g) + h + K[j+4] + W[j+3]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XORROR64_5(a, b, c) + XORROR64_6(a) + z
z = XORROR64_3(e) + XORROR64_4(e, f, g) + h + K[j+5] + W[j+4]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XORROR64_5(a, b, c) + XORROR64_6(a) + z
z = XORROR64_3(e) + XORROR64_4(e, f, g) + h + K[j+6] + W[j+5]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XORROR64_5(a, b, c) + XORROR64_6(a) + z
z = XORROR64_3(e) + XORROR64_4(e, f, g) + h + K[j+7] + W[j+6]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XORROR64_5(a, b, c) + XORROR64_6(a) + z
z = XORROR64_3(e) + XORROR64_4(e, f, g) + h + K[j+8] + W[j+7]
h, g, f, e = g, f, e, z + d
d, c, b, a = c, b, a, XORROR64_5(a, b, c) + XORROR64_6(a) + z
end
H[1] = a + H[1]
H[2] = b + H[2]
H[3] = c + H[3]
H[4] = d + H[4]
H[5] = e + H[5]
H[6] = f + H[6]
H[7] = g + H[7]
H[8] = h + H[8]
end
end
-- BLAKE2b implementation for "LuaJIT 2.0 + FFI" branch
do
local v = ffi.new("int64_t[?]", 16)
local W = common_W_blake2b
local function G(a, b, c, d, k1, k2)
local va, vb, vc, vd = v[a], v[b], v[c], v[d]
va = W[k1] + (va + vb)
vd = XORROR64_9(vd, va)
vc = vc + vd
vb = XORROR64_7(vb, vc, 24)
va = W[k2] + (va + vb)
vd = XORROR64_7(vd, va, 16)
vc = vc + vd
vb = XORROR64_8(vb, vc)
v[a], v[b], v[c], v[d] = va, vb, vc, vd
end
function blake2b_feed_128(H, _, str, offs, size, bytes_compressed, last_block_size, is_last_node)
-- offs >= 0, size >= 0, size is multiple of 128
local h1, h2, h3, h4, h5, h6, h7, h8 = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for pos = offs, offs + size - 1, 128 do
if str then
for j = 1, 16 do
pos = pos + 8
local a, b, c, d, e, f, g, h = byte(str, pos - 7, pos)
W[j] = XOR64(OR(SHL(h, 24), SHL(g, 16), SHL(f, 8), e) * int64(2^32), uint32(int32(OR(SHL(d, 24), SHL(c, 16), SHL(b, 8), a))))
end
end
v[0x0], v[0x1], v[0x2], v[0x3], v[0x4], v[0x5], v[0x6], v[0x7] = h1, h2, h3, h4, h5, h6, h7, h8
v[0x8], v[0x9], v[0xA], v[0xB], v[0xD], v[0xE], v[0xF] = sha2_H_lo[1], sha2_H_lo[2], sha2_H_lo[3], sha2_H_lo[4], sha2_H_lo[6], sha2_H_lo[7], sha2_H_lo[8]
bytes_compressed = bytes_compressed + (last_block_size or 128)
v[0xC] = XOR64(sha2_H_lo[5], bytes_compressed) -- t0 = low_8_bytes(bytes_compressed)
-- t1 = high_8_bytes(bytes_compressed) = 0, message length is always below 2^53 bytes
if last_block_size then -- flag f0
v[0xE] = -1 - v[0xE]
end
if is_last_node then -- flag f1
v[0xF] = -1 - v[0xF]
end
for j = 1, 12 do
local row = sigma[j]
G(0, 4, 8, 12, row[ 1], row[ 2])
G(1, 5, 9, 13, row[ 3], row[ 4])
G(2, 6, 10, 14, row[ 5], row[ 6])
G(3, 7, 11, 15, row[ 7], row[ 8])
G(0, 5, 10, 15, row[ 9], row[10])
G(1, 6, 11, 12, row[11], row[12])
G(2, 7, 8, 13, row[13], row[14])
G(3, 4, 9, 14, row[15], row[16])
end
h1 = XORROR64_11(h1, v[0x0], v[0x8])
h2 = XORROR64_11(h2, v[0x1], v[0x9])
h3 = XORROR64_11(h3, v[0x2], v[0xA])
h4 = XORROR64_11(h4, v[0x3], v[0xB])
h5 = XORROR64_11(h5, v[0x4], v[0xC])
h6 = XORROR64_11(h6, v[0x5], v[0xD])
h7 = XORROR64_11(h7, v[0x6], v[0xE])
h8 = XORROR64_11(h8, v[0x7], v[0xF])
end
H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8] = h1, h2, h3, h4, h5, h6, h7, h8
return bytes_compressed
end
end
end
-- MD5 implementation for "LuaJIT with FFI" branch
function md5_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W, K = common_W_FFI_int32, md5_K
for pos = offs, offs + size - 1, 64 do
for j = 0, 15 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos) -- slow, but doesn't depend on endianness
W[j] = OR(SHL(d, 24), SHL(c, 16), SHL(b, 8), a)
end
local a, b, c, d = H[1], H[2], H[3], H[4]
for j = 0, 15, 4 do
a, d, c, b = d, c, b, NORM(ROL(XOR(d, AND(b, XOR(c, d))) + (K[j+1] + W[j ] + a), 7) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(d, AND(b, XOR(c, d))) + (K[j+2] + W[j+1] + a), 12) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(d, AND(b, XOR(c, d))) + (K[j+3] + W[j+2] + a), 17) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(d, AND(b, XOR(c, d))) + (K[j+4] + W[j+3] + a), 22) + b)
end
for j = 16, 31, 4 do
local g = 5*j
a, d, c, b = d, c, b, NORM(ROL(XOR(c, AND(d, XOR(b, c))) + (K[j+1] + W[AND(g + 1, 15)] + a), 5) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, AND(d, XOR(b, c))) + (K[j+2] + W[AND(g + 6, 15)] + a), 9) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, AND(d, XOR(b, c))) + (K[j+3] + W[AND(g - 5, 15)] + a), 14) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, AND(d, XOR(b, c))) + (K[j+4] + W[AND(g , 15)] + a), 20) + b)
end
for j = 32, 47, 4 do
local g = 3*j
a, d, c, b = d, c, b, NORM(ROL(XOR(b, c, d) + (K[j+1] + W[AND(g + 5, 15)] + a), 4) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(b, c, d) + (K[j+2] + W[AND(g + 8, 15)] + a), 11) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(b, c, d) + (K[j+3] + W[AND(g - 5, 15)] + a), 16) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(b, c, d) + (K[j+4] + W[AND(g - 2, 15)] + a), 23) + b)
end
for j = 48, 63, 4 do
local g = 7*j
a, d, c, b = d, c, b, NORM(ROL(XOR(c, OR(b, NOT(d))) + (K[j+1] + W[AND(g , 15)] + a), 6) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, OR(b, NOT(d))) + (K[j+2] + W[AND(g + 7, 15)] + a), 10) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, OR(b, NOT(d))) + (K[j+3] + W[AND(g - 2, 15)] + a), 15) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, OR(b, NOT(d))) + (K[j+4] + W[AND(g + 5, 15)] + a), 21) + b)
end
H[1], H[2], H[3], H[4] = NORM(a + H[1]), NORM(b + H[2]), NORM(c + H[3]), NORM(d + H[4])
end
end
-- SHA-1 implementation for "LuaJIT with FFI" branch
function sha1_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W = common_W_FFI_int32
for pos = offs, offs + size - 1, 64 do
for j = 0, 15 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos) -- slow, but doesn't depend on endianness
W[j] = OR(SHL(a, 24), SHL(b, 16), SHL(c, 8), d)
end
for j = 16, 79 do
W[j] = ROL(XOR(W[j-3], W[j-8], W[j-14], W[j-16]), 1)
end
local a, b, c, d, e = H[1], H[2], H[3], H[4], H[5]
for j = 0, 19, 5 do
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(d, AND(b, XOR(d, c))) + (W[j] + 0x5A827999 + e)) -- constant = floor(2^30 * sqrt(2))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(d, AND(b, XOR(d, c))) + (W[j+1] + 0x5A827999 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(d, AND(b, XOR(d, c))) + (W[j+2] + 0x5A827999 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(d, AND(b, XOR(d, c))) + (W[j+3] + 0x5A827999 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(d, AND(b, XOR(d, c))) + (W[j+4] + 0x5A827999 + e))
end
for j = 20, 39, 5 do
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j] + 0x6ED9EBA1 + e)) -- 2^30 * sqrt(3)
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+1] + 0x6ED9EBA1 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+2] + 0x6ED9EBA1 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+3] + 0x6ED9EBA1 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+4] + 0x6ED9EBA1 + e))
end
for j = 40, 59, 5 do
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(AND(d, XOR(b, c)), AND(b, c)) + (W[j] + 0x8F1BBCDC + e)) -- 2^30 * sqrt(5)
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(AND(d, XOR(b, c)), AND(b, c)) + (W[j+1] + 0x8F1BBCDC + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(AND(d, XOR(b, c)), AND(b, c)) + (W[j+2] + 0x8F1BBCDC + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(AND(d, XOR(b, c)), AND(b, c)) + (W[j+3] + 0x8F1BBCDC + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(AND(d, XOR(b, c)), AND(b, c)) + (W[j+4] + 0x8F1BBCDC + e))
end
for j = 60, 79, 5 do
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j] + 0xCA62C1D6 + e)) -- 2^30 * sqrt(10)
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+1] + 0xCA62C1D6 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+2] + 0xCA62C1D6 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+3] + 0xCA62C1D6 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+4] + 0xCA62C1D6 + e))
end
H[1], H[2], H[3], H[4], H[5] = NORM(a + H[1]), NORM(b + H[2]), NORM(c + H[3]), NORM(d + H[4]), NORM(e + H[5])
end
end
end
if branch == "FFI" and not is_LuaJIT_21 or branch == "LJ" then
if branch == "FFI" then
local arr32_t = ffi.typeof"int32_t[?]"
function create_array_of_lanes()
return arr32_t(31) -- 25 + 5 + 1 (due to 1-based indexing)
end
end
-- SHA-3 implementation for "LuaJIT 2.0 + FFI" and "LuaJIT without FFI" branches
function keccak_feed(lanes_lo, lanes_hi, str, offs, size, block_size_in_bytes)
-- offs >= 0, size >= 0, size is multiple of block_size_in_bytes, block_size_in_bytes is positive multiple of 8
local RC_lo, RC_hi = sha3_RC_lo, sha3_RC_hi
local qwords_qty = SHR(block_size_in_bytes, 3)
for pos = offs, offs + size - 1, block_size_in_bytes do
for j = 1, qwords_qty do
local a, b, c, d = byte(str, pos + 1, pos + 4)
lanes_lo[j] = XOR(lanes_lo[j], OR(SHL(d, 24), SHL(c, 16), SHL(b, 8), a))
pos = pos + 8
a, b, c, d = byte(str, pos - 3, pos)
lanes_hi[j] = XOR(lanes_hi[j], OR(SHL(d, 24), SHL(c, 16), SHL(b, 8), a))
end
for round_idx = 1, 24 do
for j = 1, 5 do
lanes_lo[25 + j] = XOR(lanes_lo[j], lanes_lo[j + 5], lanes_lo[j + 10], lanes_lo[j + 15], lanes_lo[j + 20])
end
for j = 1, 5 do
lanes_hi[25 + j] = XOR(lanes_hi[j], lanes_hi[j + 5], lanes_hi[j + 10], lanes_hi[j + 15], lanes_hi[j + 20])
end
local D_lo = XOR(lanes_lo[26], SHL(lanes_lo[28], 1), SHR(lanes_hi[28], 31))
local D_hi = XOR(lanes_hi[26], SHL(lanes_hi[28], 1), SHR(lanes_lo[28], 31))
lanes_lo[2], lanes_hi[2], lanes_lo[7], lanes_hi[7], lanes_lo[12], lanes_hi[12], lanes_lo[17], lanes_hi[17] = XOR(SHR(XOR(D_lo, lanes_lo[7]), 20), SHL(XOR(D_hi, lanes_hi[7]), 12)), XOR(SHR(XOR(D_hi, lanes_hi[7]), 20), SHL(XOR(D_lo, lanes_lo[7]), 12)), XOR(SHR(XOR(D_lo, lanes_lo[17]), 19), SHL(XOR(D_hi, lanes_hi[17]), 13)), XOR(SHR(XOR(D_hi, lanes_hi[17]), 19), SHL(XOR(D_lo, lanes_lo[17]), 13)), XOR(SHL(XOR(D_lo, lanes_lo[2]), 1), SHR(XOR(D_hi, lanes_hi[2]), 31)), XOR(SHL(XOR(D_hi, lanes_hi[2]), 1), SHR(XOR(D_lo, lanes_lo[2]), 31)), XOR(SHL(XOR(D_lo, lanes_lo[12]), 10), SHR(XOR(D_hi, lanes_hi[12]), 22)), XOR(SHL(XOR(D_hi, lanes_hi[12]), 10), SHR(XOR(D_lo, lanes_lo[12]), 22))
local L, H = XOR(D_lo, lanes_lo[22]), XOR(D_hi, lanes_hi[22])
lanes_lo[22], lanes_hi[22] = XOR(SHL(L, 2), SHR(H, 30)), XOR(SHL(H, 2), SHR(L, 30))
D_lo = XOR(lanes_lo[27], SHL(lanes_lo[29], 1), SHR(lanes_hi[29], 31))
D_hi = XOR(lanes_hi[27], SHL(lanes_hi[29], 1), SHR(lanes_lo[29], 31))
lanes_lo[3], lanes_hi[3], lanes_lo[8], lanes_hi[8], lanes_lo[13], lanes_hi[13], lanes_lo[23], lanes_hi[23] = XOR(SHR(XOR(D_lo, lanes_lo[13]), 21), SHL(XOR(D_hi, lanes_hi[13]), 11)), XOR(SHR(XOR(D_hi, lanes_hi[13]), 21), SHL(XOR(D_lo, lanes_lo[13]), 11)), XOR(SHR(XOR(D_lo, lanes_lo[23]), 3), SHL(XOR(D_hi, lanes_hi[23]), 29)), XOR(SHR(XOR(D_hi, lanes_hi[23]), 3), SHL(XOR(D_lo, lanes_lo[23]), 29)), XOR(SHL(XOR(D_lo, lanes_lo[8]), 6), SHR(XOR(D_hi, lanes_hi[8]), 26)), XOR(SHL(XOR(D_hi, lanes_hi[8]), 6), SHR(XOR(D_lo, lanes_lo[8]), 26)), XOR(SHR(XOR(D_lo, lanes_lo[3]), 2), SHL(XOR(D_hi, lanes_hi[3]), 30)), XOR(SHR(XOR(D_hi, lanes_hi[3]), 2), SHL(XOR(D_lo, lanes_lo[3]), 30))
L, H = XOR(D_lo, lanes_lo[18]), XOR(D_hi, lanes_hi[18])
lanes_lo[18], lanes_hi[18] = XOR(SHL(L, 15), SHR(H, 17)), XOR(SHL(H, 15), SHR(L, 17))
D_lo = XOR(lanes_lo[28], SHL(lanes_lo[30], 1), SHR(lanes_hi[30], 31))
D_hi = XOR(lanes_hi[28], SHL(lanes_hi[30], 1), SHR(lanes_lo[30], 31))
lanes_lo[4], lanes_hi[4], lanes_lo[9], lanes_hi[9], lanes_lo[19], lanes_hi[19], lanes_lo[24], lanes_hi[24] = XOR(SHL(XOR(D_lo, lanes_lo[19]), 21), SHR(XOR(D_hi, lanes_hi[19]), 11)), XOR(SHL(XOR(D_hi, lanes_hi[19]), 21), SHR(XOR(D_lo, lanes_lo[19]), 11)), XOR(SHL(XOR(D_lo, lanes_lo[4]), 28), SHR(XOR(D_hi, lanes_hi[4]), 4)), XOR(SHL(XOR(D_hi, lanes_hi[4]), 28), SHR(XOR(D_lo, lanes_lo[4]), 4)), XOR(SHR(XOR(D_lo, lanes_lo[24]), 8), SHL(XOR(D_hi, lanes_hi[24]), 24)), XOR(SHR(XOR(D_hi, lanes_hi[24]), 8), SHL(XOR(D_lo, lanes_lo[24]), 24)), XOR(SHR(XOR(D_lo, lanes_lo[9]), 9), SHL(XOR(D_hi, lanes_hi[9]), 23)), XOR(SHR(XOR(D_hi, lanes_hi[9]), 9), SHL(XOR(D_lo, lanes_lo[9]), 23))
L, H = XOR(D_lo, lanes_lo[14]), XOR(D_hi, lanes_hi[14])
lanes_lo[14], lanes_hi[14] = XOR(SHL(L, 25), SHR(H, 7)), XOR(SHL(H, 25), SHR(L, 7))
D_lo = XOR(lanes_lo[29], SHL(lanes_lo[26], 1), SHR(lanes_hi[26], 31))
D_hi = XOR(lanes_hi[29], SHL(lanes_hi[26], 1), SHR(lanes_lo[26], 31))
lanes_lo[5], lanes_hi[5], lanes_lo[15], lanes_hi[15], lanes_lo[20], lanes_hi[20], lanes_lo[25], lanes_hi[25] = XOR(SHL(XOR(D_lo, lanes_lo[25]), 14), SHR(XOR(D_hi, lanes_hi[25]), 18)), XOR(SHL(XOR(D_hi, lanes_hi[25]), 14), SHR(XOR(D_lo, lanes_lo[25]), 18)), XOR(SHL(XOR(D_lo, lanes_lo[20]), 8), SHR(XOR(D_hi, lanes_hi[20]), 24)), XOR(SHL(XOR(D_hi, lanes_hi[20]), 8), SHR(XOR(D_lo, lanes_lo[20]), 24)), XOR(SHL(XOR(D_lo, lanes_lo[5]), 27), SHR(XOR(D_hi, lanes_hi[5]), 5)), XOR(SHL(XOR(D_hi, lanes_hi[5]), 27), SHR(XOR(D_lo, lanes_lo[5]), 5)), XOR(SHR(XOR(D_lo, lanes_lo[15]), 25), SHL(XOR(D_hi, lanes_hi[15]), 7)), XOR(SHR(XOR(D_hi, lanes_hi[15]), 25), SHL(XOR(D_lo, lanes_lo[15]), 7))
L, H = XOR(D_lo, lanes_lo[10]), XOR(D_hi, lanes_hi[10])
lanes_lo[10], lanes_hi[10] = XOR(SHL(L, 20), SHR(H, 12)), XOR(SHL(H, 20), SHR(L, 12))
D_lo = XOR(lanes_lo[30], SHL(lanes_lo[27], 1), SHR(lanes_hi[27], 31))
D_hi = XOR(lanes_hi[30], SHL(lanes_hi[27], 1), SHR(lanes_lo[27], 31))
lanes_lo[6], lanes_hi[6], lanes_lo[11], lanes_hi[11], lanes_lo[16], lanes_hi[16], lanes_lo[21], lanes_hi[21] = XOR(SHL(XOR(D_lo, lanes_lo[11]), 3), SHR(XOR(D_hi, lanes_hi[11]), 29)), XOR(SHL(XOR(D_hi, lanes_hi[11]), 3), SHR(XOR(D_lo, lanes_lo[11]), 29)), XOR(SHL(XOR(D_lo, lanes_lo[21]), 18), SHR(XOR(D_hi, lanes_hi[21]), 14)), XOR(SHL(XOR(D_hi, lanes_hi[21]), 18), SHR(XOR(D_lo, lanes_lo[21]), 14)), XOR(SHR(XOR(D_lo, lanes_lo[6]), 28), SHL(XOR(D_hi, lanes_hi[6]), 4)), XOR(SHR(XOR(D_hi, lanes_hi[6]), 28), SHL(XOR(D_lo, lanes_lo[6]), 4)), XOR(SHR(XOR(D_lo, lanes_lo[16]), 23), SHL(XOR(D_hi, lanes_hi[16]), 9)), XOR(SHR(XOR(D_hi, lanes_hi[16]), 23), SHL(XOR(D_lo, lanes_lo[16]), 9))
lanes_lo[1], lanes_hi[1] = XOR(D_lo, lanes_lo[1]), XOR(D_hi, lanes_hi[1])
lanes_lo[1], lanes_lo[2], lanes_lo[3], lanes_lo[4], lanes_lo[5] = XOR(lanes_lo[1], AND(NOT(lanes_lo[2]), lanes_lo[3]), RC_lo[round_idx]), XOR(lanes_lo[2], AND(NOT(lanes_lo[3]), lanes_lo[4])), XOR(lanes_lo[3], AND(NOT(lanes_lo[4]), lanes_lo[5])), XOR(lanes_lo[4], AND(NOT(lanes_lo[5]), lanes_lo[1])), XOR(lanes_lo[5], AND(NOT(lanes_lo[1]), lanes_lo[2]))
lanes_lo[6], lanes_lo[7], lanes_lo[8], lanes_lo[9], lanes_lo[10] = XOR(lanes_lo[9], AND(NOT(lanes_lo[10]), lanes_lo[6])), XOR(lanes_lo[10], AND(NOT(lanes_lo[6]), lanes_lo[7])), XOR(lanes_lo[6], AND(NOT(lanes_lo[7]), lanes_lo[8])), XOR(lanes_lo[7], AND(NOT(lanes_lo[8]), lanes_lo[9])), XOR(lanes_lo[8], AND(NOT(lanes_lo[9]), lanes_lo[10]))
lanes_lo[11], lanes_lo[12], lanes_lo[13], lanes_lo[14], lanes_lo[15] = XOR(lanes_lo[12], AND(NOT(lanes_lo[13]), lanes_lo[14])), XOR(lanes_lo[13], AND(NOT(lanes_lo[14]), lanes_lo[15])), XOR(lanes_lo[14], AND(NOT(lanes_lo[15]), lanes_lo[11])), XOR(lanes_lo[15], AND(NOT(lanes_lo[11]), lanes_lo[12])), XOR(lanes_lo[11], AND(NOT(lanes_lo[12]), lanes_lo[13]))
lanes_lo[16], lanes_lo[17], lanes_lo[18], lanes_lo[19], lanes_lo[20] = XOR(lanes_lo[20], AND(NOT(lanes_lo[16]), lanes_lo[17])), XOR(lanes_lo[16], AND(NOT(lanes_lo[17]), lanes_lo[18])), XOR(lanes_lo[17], AND(NOT(lanes_lo[18]), lanes_lo[19])), XOR(lanes_lo[18], AND(NOT(lanes_lo[19]), lanes_lo[20])), XOR(lanes_lo[19], AND(NOT(lanes_lo[20]), lanes_lo[16]))
lanes_lo[21], lanes_lo[22], lanes_lo[23], lanes_lo[24], lanes_lo[25] = XOR(lanes_lo[23], AND(NOT(lanes_lo[24]), lanes_lo[25])), XOR(lanes_lo[24], AND(NOT(lanes_lo[25]), lanes_lo[21])), XOR(lanes_lo[25], AND(NOT(lanes_lo[21]), lanes_lo[22])), XOR(lanes_lo[21], AND(NOT(lanes_lo[22]), lanes_lo[23])), XOR(lanes_lo[22], AND(NOT(lanes_lo[23]), lanes_lo[24]))
lanes_hi[1], lanes_hi[2], lanes_hi[3], lanes_hi[4], lanes_hi[5] = XOR(lanes_hi[1], AND(NOT(lanes_hi[2]), lanes_hi[3]), RC_hi[round_idx]), XOR(lanes_hi[2], AND(NOT(lanes_hi[3]), lanes_hi[4])), XOR(lanes_hi[3], AND(NOT(lanes_hi[4]), lanes_hi[5])), XOR(lanes_hi[4], AND(NOT(lanes_hi[5]), lanes_hi[1])), XOR(lanes_hi[5], AND(NOT(lanes_hi[1]), lanes_hi[2]))
lanes_hi[6], lanes_hi[7], lanes_hi[8], lanes_hi[9], lanes_hi[10] = XOR(lanes_hi[9], AND(NOT(lanes_hi[10]), lanes_hi[6])), XOR(lanes_hi[10], AND(NOT(lanes_hi[6]), lanes_hi[7])), XOR(lanes_hi[6], AND(NOT(lanes_hi[7]), lanes_hi[8])), XOR(lanes_hi[7], AND(NOT(lanes_hi[8]), lanes_hi[9])), XOR(lanes_hi[8], AND(NOT(lanes_hi[9]), lanes_hi[10]))
lanes_hi[11], lanes_hi[12], lanes_hi[13], lanes_hi[14], lanes_hi[15] = XOR(lanes_hi[12], AND(NOT(lanes_hi[13]), lanes_hi[14])), XOR(lanes_hi[13], AND(NOT(lanes_hi[14]), lanes_hi[15])), XOR(lanes_hi[14], AND(NOT(lanes_hi[15]), lanes_hi[11])), XOR(lanes_hi[15], AND(NOT(lanes_hi[11]), lanes_hi[12])), XOR(lanes_hi[11], AND(NOT(lanes_hi[12]), lanes_hi[13]))
lanes_hi[16], lanes_hi[17], lanes_hi[18], lanes_hi[19], lanes_hi[20] = XOR(lanes_hi[20], AND(NOT(lanes_hi[16]), lanes_hi[17])), XOR(lanes_hi[16], AND(NOT(lanes_hi[17]), lanes_hi[18])), XOR(lanes_hi[17], AND(NOT(lanes_hi[18]), lanes_hi[19])), XOR(lanes_hi[18], AND(NOT(lanes_hi[19]), lanes_hi[20])), XOR(lanes_hi[19], AND(NOT(lanes_hi[20]), lanes_hi[16]))
lanes_hi[21], lanes_hi[22], lanes_hi[23], lanes_hi[24], lanes_hi[25] = XOR(lanes_hi[23], AND(NOT(lanes_hi[24]), lanes_hi[25])), XOR(lanes_hi[24], AND(NOT(lanes_hi[25]), lanes_hi[21])), XOR(lanes_hi[25], AND(NOT(lanes_hi[21]), lanes_hi[22])), XOR(lanes_hi[21], AND(NOT(lanes_hi[22]), lanes_hi[23])), XOR(lanes_hi[22], AND(NOT(lanes_hi[23]), lanes_hi[24]))
end
end
end
end
if branch == "LJ" then
-- SHA256 implementation for "LuaJIT without FFI" branch
function sha256_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W, K = common_W, sha2_K_hi
for pos = offs, offs + size - 1, 64 do
for j = 1, 16 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = OR(SHL(a, 24), SHL(b, 16), SHL(c, 8), d)
end
for j = 17, 64 do
local a, b = W[j-15], W[j-2]
W[j] = NORM( NORM( XOR(ROR(a, 7), ROL(a, 14), SHR(a, 3)) + XOR(ROL(b, 15), ROL(b, 13), SHR(b, 10)) ) + NORM( W[j-7] + W[j-16] ) )
end
local a, b, c, d, e, f, g, h = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for j = 1, 64, 8 do -- Thanks to Peter Cawley for this workaround (unroll the loop to avoid "PHI shuffling too complex" due to PHIs overlap)
local z = NORM( XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + XOR(g, AND(e, XOR(f, g))) + (K[j] + W[j] + h) )
h, g, f, e = g, f, e, NORM(d + z)
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + XOR(g, AND(e, XOR(f, g))) + (K[j+1] + W[j+1] + h) )
h, g, f, e = g, f, e, NORM(d + z)
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + XOR(g, AND(e, XOR(f, g))) + (K[j+2] + W[j+2] + h) )
h, g, f, e = g, f, e, NORM(d + z)
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + XOR(g, AND(e, XOR(f, g))) + (K[j+3] + W[j+3] + h) )
h, g, f, e = g, f, e, NORM(d + z)
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + XOR(g, AND(e, XOR(f, g))) + (K[j+4] + W[j+4] + h) )
h, g, f, e = g, f, e, NORM(d + z)
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + XOR(g, AND(e, XOR(f, g))) + (K[j+5] + W[j+5] + h) )
h, g, f, e = g, f, e, NORM(d + z)
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + XOR(g, AND(e, XOR(f, g))) + (K[j+6] + W[j+6] + h) )
h, g, f, e = g, f, e, NORM(d + z)
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
z = NORM( XOR(ROR(e, 6), ROR(e, 11), ROL(e, 7)) + XOR(g, AND(e, XOR(f, g))) + (K[j+7] + W[j+7] + h) )
h, g, f, e = g, f, e, NORM(d + z)
d, c, b, a = c, b, a, NORM( XOR(AND(a, XOR(b, c)), AND(b, c)) + XOR(ROR(a, 2), ROR(a, 13), ROL(a, 10)) + z )
end
H[1], H[2], H[3], H[4] = NORM(a + H[1]), NORM(b + H[2]), NORM(c + H[3]), NORM(d + H[4])
H[5], H[6], H[7], H[8] = NORM(e + H[5]), NORM(f + H[6]), NORM(g + H[7]), NORM(h + H[8])
end
end
local function ADD64_4(a_lo, a_hi, b_lo, b_hi, c_lo, c_hi, d_lo, d_hi)
local sum_lo = a_lo % 2^32 + b_lo % 2^32 + c_lo % 2^32 + d_lo % 2^32
local sum_hi = a_hi + b_hi + c_hi + d_hi
local result_lo = NORM( sum_lo )
local result_hi = NORM( sum_hi + floor(sum_lo / 2^32) )
return result_lo, result_hi
end
if LuaJIT_arch == "x86" then -- Special trick is required to avoid "PHI shuffling too complex" on x86 platform
-- SHA512 implementation for "LuaJIT x86 without FFI" branch
function sha512_feed_128(H_lo, H_hi, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 128
-- W1_hi, W1_lo, W2_hi, W2_lo, ... Wk_hi = W[2*k-1], Wk_lo = W[2*k]
local W, K_lo, K_hi = common_W, sha2_K_lo, sha2_K_hi
for pos = offs, offs + size - 1, 128 do
for j = 1, 16*2 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = OR(SHL(a, 24), SHL(b, 16), SHL(c, 8), d)
end
for jj = 17*2, 80*2, 2 do
local a_lo, a_hi = W[jj-30], W[jj-31]
local t_lo = XOR(OR(SHR(a_lo, 1), SHL(a_hi, 31)), OR(SHR(a_lo, 8), SHL(a_hi, 24)), OR(SHR(a_lo, 7), SHL(a_hi, 25)))
local t_hi = XOR(OR(SHR(a_hi, 1), SHL(a_lo, 31)), OR(SHR(a_hi, 8), SHL(a_lo, 24)), SHR(a_hi, 7))
local b_lo, b_hi = W[jj-4], W[jj-5]
local u_lo = XOR(OR(SHR(b_lo, 19), SHL(b_hi, 13)), OR(SHL(b_lo, 3), SHR(b_hi, 29)), OR(SHR(b_lo, 6), SHL(b_hi, 26)))
local u_hi = XOR(OR(SHR(b_hi, 19), SHL(b_lo, 13)), OR(SHL(b_hi, 3), SHR(b_lo, 29)), SHR(b_hi, 6))
W[jj], W[jj-1] = ADD64_4(t_lo, t_hi, u_lo, u_hi, W[jj-14], W[jj-15], W[jj-32], W[jj-33])
end
local a_lo, b_lo, c_lo, d_lo, e_lo, f_lo, g_lo, h_lo = H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8]
local a_hi, b_hi, c_hi, d_hi, e_hi, f_hi, g_hi, h_hi = H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8]
local zero = 0
for j = 1, 80 do
local t_lo = XOR(g_lo, AND(e_lo, XOR(f_lo, g_lo)))
local t_hi = XOR(g_hi, AND(e_hi, XOR(f_hi, g_hi)))
local u_lo = XOR(OR(SHR(e_lo, 14), SHL(e_hi, 18)), OR(SHR(e_lo, 18), SHL(e_hi, 14)), OR(SHL(e_lo, 23), SHR(e_hi, 9)))
local u_hi = XOR(OR(SHR(e_hi, 14), SHL(e_lo, 18)), OR(SHR(e_hi, 18), SHL(e_lo, 14)), OR(SHL(e_hi, 23), SHR(e_lo, 9)))
local sum_lo = u_lo % 2^32 + t_lo % 2^32 + h_lo % 2^32 + K_lo[j] + W[2*j] % 2^32
local z_lo, z_hi = NORM( sum_lo ), NORM( u_hi + t_hi + h_hi + K_hi[j] + W[2*j-1] + floor(sum_lo / 2^32) )
zero = zero + zero -- this thick is needed to avoid "PHI shuffling too complex" due to PHIs overlap
h_lo, h_hi, g_lo, g_hi, f_lo, f_hi = OR(zero, g_lo), OR(zero, g_hi), OR(zero, f_lo), OR(zero, f_hi), OR(zero, e_lo), OR(zero, e_hi)
local sum_lo = z_lo % 2^32 + d_lo % 2^32
e_lo, e_hi = NORM( sum_lo ), NORM( z_hi + d_hi + floor(sum_lo / 2^32) )
d_lo, d_hi, c_lo, c_hi, b_lo, b_hi = OR(zero, c_lo), OR(zero, c_hi), OR(zero, b_lo), OR(zero, b_hi), OR(zero, a_lo), OR(zero, a_hi)
u_lo = XOR(OR(SHR(b_lo, 28), SHL(b_hi, 4)), OR(SHL(b_lo, 30), SHR(b_hi, 2)), OR(SHL(b_lo, 25), SHR(b_hi, 7)))
u_hi = XOR(OR(SHR(b_hi, 28), SHL(b_lo, 4)), OR(SHL(b_hi, 30), SHR(b_lo, 2)), OR(SHL(b_hi, 25), SHR(b_lo, 7)))
t_lo = OR(AND(d_lo, c_lo), AND(b_lo, XOR(d_lo, c_lo)))
t_hi = OR(AND(d_hi, c_hi), AND(b_hi, XOR(d_hi, c_hi)))
local sum_lo = z_lo % 2^32 + t_lo % 2^32 + u_lo % 2^32
a_lo, a_hi = NORM( sum_lo ), NORM( z_hi + t_hi + u_hi + floor(sum_lo / 2^32) )
end
H_lo[1], H_hi[1] = ADD64_4(H_lo[1], H_hi[1], a_lo, a_hi, 0, 0, 0, 0)
H_lo[2], H_hi[2] = ADD64_4(H_lo[2], H_hi[2], b_lo, b_hi, 0, 0, 0, 0)
H_lo[3], H_hi[3] = ADD64_4(H_lo[3], H_hi[3], c_lo, c_hi, 0, 0, 0, 0)
H_lo[4], H_hi[4] = ADD64_4(H_lo[4], H_hi[4], d_lo, d_hi, 0, 0, 0, 0)
H_lo[5], H_hi[5] = ADD64_4(H_lo[5], H_hi[5], e_lo, e_hi, 0, 0, 0, 0)
H_lo[6], H_hi[6] = ADD64_4(H_lo[6], H_hi[6], f_lo, f_hi, 0, 0, 0, 0)
H_lo[7], H_hi[7] = ADD64_4(H_lo[7], H_hi[7], g_lo, g_hi, 0, 0, 0, 0)
H_lo[8], H_hi[8] = ADD64_4(H_lo[8], H_hi[8], h_lo, h_hi, 0, 0, 0, 0)
end
end
else -- all platforms except x86
-- SHA512 implementation for "LuaJIT non-x86 without FFI" branch
function sha512_feed_128(H_lo, H_hi, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 128
-- W1_hi, W1_lo, W2_hi, W2_lo, ... Wk_hi = W[2*k-1], Wk_lo = W[2*k]
local W, K_lo, K_hi = common_W, sha2_K_lo, sha2_K_hi
for pos = offs, offs + size - 1, 128 do
for j = 1, 16*2 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = OR(SHL(a, 24), SHL(b, 16), SHL(c, 8), d)
end
for jj = 17*2, 80*2, 2 do
local a_lo, a_hi = W[jj-30], W[jj-31]
local t_lo = XOR(OR(SHR(a_lo, 1), SHL(a_hi, 31)), OR(SHR(a_lo, 8), SHL(a_hi, 24)), OR(SHR(a_lo, 7), SHL(a_hi, 25)))
local t_hi = XOR(OR(SHR(a_hi, 1), SHL(a_lo, 31)), OR(SHR(a_hi, 8), SHL(a_lo, 24)), SHR(a_hi, 7))
local b_lo, b_hi = W[jj-4], W[jj-5]
local u_lo = XOR(OR(SHR(b_lo, 19), SHL(b_hi, 13)), OR(SHL(b_lo, 3), SHR(b_hi, 29)), OR(SHR(b_lo, 6), SHL(b_hi, 26)))
local u_hi = XOR(OR(SHR(b_hi, 19), SHL(b_lo, 13)), OR(SHL(b_hi, 3), SHR(b_lo, 29)), SHR(b_hi, 6))
W[jj], W[jj-1] = ADD64_4(t_lo, t_hi, u_lo, u_hi, W[jj-14], W[jj-15], W[jj-32], W[jj-33])
end
local a_lo, b_lo, c_lo, d_lo, e_lo, f_lo, g_lo, h_lo = H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8]
local a_hi, b_hi, c_hi, d_hi, e_hi, f_hi, g_hi, h_hi = H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8]
for j = 1, 80 do
local t_lo = XOR(g_lo, AND(e_lo, XOR(f_lo, g_lo)))
local t_hi = XOR(g_hi, AND(e_hi, XOR(f_hi, g_hi)))
local u_lo = XOR(OR(SHR(e_lo, 14), SHL(e_hi, 18)), OR(SHR(e_lo, 18), SHL(e_hi, 14)), OR(SHL(e_lo, 23), SHR(e_hi, 9)))
local u_hi = XOR(OR(SHR(e_hi, 14), SHL(e_lo, 18)), OR(SHR(e_hi, 18), SHL(e_lo, 14)), OR(SHL(e_hi, 23), SHR(e_lo, 9)))
local sum_lo = u_lo % 2^32 + t_lo % 2^32 + h_lo % 2^32 + K_lo[j] + W[2*j] % 2^32
local z_lo, z_hi = NORM( sum_lo ), NORM( u_hi + t_hi + h_hi + K_hi[j] + W[2*j-1] + floor(sum_lo / 2^32) )
h_lo, h_hi, g_lo, g_hi, f_lo, f_hi = g_lo, g_hi, f_lo, f_hi, e_lo, e_hi
local sum_lo = z_lo % 2^32 + d_lo % 2^32
e_lo, e_hi = NORM( sum_lo ), NORM( z_hi + d_hi + floor(sum_lo / 2^32) )
d_lo, d_hi, c_lo, c_hi, b_lo, b_hi = c_lo, c_hi, b_lo, b_hi, a_lo, a_hi
u_lo = XOR(OR(SHR(b_lo, 28), SHL(b_hi, 4)), OR(SHL(b_lo, 30), SHR(b_hi, 2)), OR(SHL(b_lo, 25), SHR(b_hi, 7)))
u_hi = XOR(OR(SHR(b_hi, 28), SHL(b_lo, 4)), OR(SHL(b_hi, 30), SHR(b_lo, 2)), OR(SHL(b_hi, 25), SHR(b_lo, 7)))
t_lo = OR(AND(d_lo, c_lo), AND(b_lo, XOR(d_lo, c_lo)))
t_hi = OR(AND(d_hi, c_hi), AND(b_hi, XOR(d_hi, c_hi)))
local sum_lo = z_lo % 2^32 + u_lo % 2^32 + t_lo % 2^32
a_lo, a_hi = NORM( sum_lo ), NORM( z_hi + u_hi + t_hi + floor(sum_lo / 2^32) )
end
H_lo[1], H_hi[1] = ADD64_4(H_lo[1], H_hi[1], a_lo, a_hi, 0, 0, 0, 0)
H_lo[2], H_hi[2] = ADD64_4(H_lo[2], H_hi[2], b_lo, b_hi, 0, 0, 0, 0)
H_lo[3], H_hi[3] = ADD64_4(H_lo[3], H_hi[3], c_lo, c_hi, 0, 0, 0, 0)
H_lo[4], H_hi[4] = ADD64_4(H_lo[4], H_hi[4], d_lo, d_hi, 0, 0, 0, 0)
H_lo[5], H_hi[5] = ADD64_4(H_lo[5], H_hi[5], e_lo, e_hi, 0, 0, 0, 0)
H_lo[6], H_hi[6] = ADD64_4(H_lo[6], H_hi[6], f_lo, f_hi, 0, 0, 0, 0)
H_lo[7], H_hi[7] = ADD64_4(H_lo[7], H_hi[7], g_lo, g_hi, 0, 0, 0, 0)
H_lo[8], H_hi[8] = ADD64_4(H_lo[8], H_hi[8], h_lo, h_hi, 0, 0, 0, 0)
end
end
end
-- MD5 implementation for "LuaJIT without FFI" branch
function md5_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W, K = common_W, md5_K
for pos = offs, offs + size - 1, 64 do
for j = 1, 16 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = OR(SHL(d, 24), SHL(c, 16), SHL(b, 8), a)
end
local a, b, c, d = H[1], H[2], H[3], H[4]
for j = 1, 16, 4 do
a, d, c, b = d, c, b, NORM(ROL(XOR(d, AND(b, XOR(c, d))) + (K[j ] + W[j ] + a), 7) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(d, AND(b, XOR(c, d))) + (K[j+1] + W[j+1] + a), 12) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(d, AND(b, XOR(c, d))) + (K[j+2] + W[j+2] + a), 17) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(d, AND(b, XOR(c, d))) + (K[j+3] + W[j+3] + a), 22) + b)
end
for j = 17, 32, 4 do
local g = 5*j-4
a, d, c, b = d, c, b, NORM(ROL(XOR(c, AND(d, XOR(b, c))) + (K[j ] + W[AND(g , 15) + 1] + a), 5) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, AND(d, XOR(b, c))) + (K[j+1] + W[AND(g + 5, 15) + 1] + a), 9) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, AND(d, XOR(b, c))) + (K[j+2] + W[AND(g + 10, 15) + 1] + a), 14) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, AND(d, XOR(b, c))) + (K[j+3] + W[AND(g - 1, 15) + 1] + a), 20) + b)
end
for j = 33, 48, 4 do
local g = 3*j+2
a, d, c, b = d, c, b, NORM(ROL(XOR(b, c, d) + (K[j ] + W[AND(g , 15) + 1] + a), 4) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(b, c, d) + (K[j+1] + W[AND(g + 3, 15) + 1] + a), 11) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(b, c, d) + (K[j+2] + W[AND(g + 6, 15) + 1] + a), 16) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(b, c, d) + (K[j+3] + W[AND(g - 7, 15) + 1] + a), 23) + b)
end
for j = 49, 64, 4 do
local g = j*7
a, d, c, b = d, c, b, NORM(ROL(XOR(c, OR(b, NOT(d))) + (K[j ] + W[AND(g - 7, 15) + 1] + a), 6) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, OR(b, NOT(d))) + (K[j+1] + W[AND(g , 15) + 1] + a), 10) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, OR(b, NOT(d))) + (K[j+2] + W[AND(g + 7, 15) + 1] + a), 15) + b)
a, d, c, b = d, c, b, NORM(ROL(XOR(c, OR(b, NOT(d))) + (K[j+3] + W[AND(g - 2, 15) + 1] + a), 21) + b)
end
H[1], H[2], H[3], H[4] = NORM(a + H[1]), NORM(b + H[2]), NORM(c + H[3]), NORM(d + H[4])
end
end
-- SHA-1 implementation for "LuaJIT without FFI" branch
function sha1_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W = common_W
for pos = offs, offs + size - 1, 64 do
for j = 1, 16 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = OR(SHL(a, 24), SHL(b, 16), SHL(c, 8), d)
end
for j = 17, 80 do
W[j] = ROL(XOR(W[j-3], W[j-8], W[j-14], W[j-16]), 1)
end
local a, b, c, d, e = H[1], H[2], H[3], H[4], H[5]
for j = 1, 20, 5 do
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(d, AND(b, XOR(d, c))) + (W[j] + 0x5A827999 + e)) -- constant = floor(2^30 * sqrt(2))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(d, AND(b, XOR(d, c))) + (W[j+1] + 0x5A827999 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(d, AND(b, XOR(d, c))) + (W[j+2] + 0x5A827999 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(d, AND(b, XOR(d, c))) + (W[j+3] + 0x5A827999 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(d, AND(b, XOR(d, c))) + (W[j+4] + 0x5A827999 + e))
end
for j = 21, 40, 5 do
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j] + 0x6ED9EBA1 + e)) -- 2^30 * sqrt(3)
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+1] + 0x6ED9EBA1 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+2] + 0x6ED9EBA1 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+3] + 0x6ED9EBA1 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+4] + 0x6ED9EBA1 + e))
end
for j = 41, 60, 5 do
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(AND(d, XOR(b, c)), AND(b, c)) + (W[j] + 0x8F1BBCDC + e)) -- 2^30 * sqrt(5)
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(AND(d, XOR(b, c)), AND(b, c)) + (W[j+1] + 0x8F1BBCDC + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(AND(d, XOR(b, c)), AND(b, c)) + (W[j+2] + 0x8F1BBCDC + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(AND(d, XOR(b, c)), AND(b, c)) + (W[j+3] + 0x8F1BBCDC + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(AND(d, XOR(b, c)), AND(b, c)) + (W[j+4] + 0x8F1BBCDC + e))
end
for j = 61, 80, 5 do
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j] + 0xCA62C1D6 + e)) -- 2^30 * sqrt(10)
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+1] + 0xCA62C1D6 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+2] + 0xCA62C1D6 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+3] + 0xCA62C1D6 + e))
e, d, c, b, a = d, c, ROR(b, 2), a, NORM(ROL(a, 5) + XOR(b, c, d) + (W[j+4] + 0xCA62C1D6 + e))
end
H[1], H[2], H[3], H[4], H[5] = NORM(a + H[1]), NORM(b + H[2]), NORM(c + H[3]), NORM(d + H[4]), NORM(e + H[5])
end
end
-- BLAKE2b implementation for "LuaJIT without FFI" branch
do
local v_lo, v_hi = {}, {}
local function G(a, b, c, d, k1, k2)
local W = common_W
local va_lo, vb_lo, vc_lo, vd_lo = v_lo[a], v_lo[b], v_lo[c], v_lo[d]
local va_hi, vb_hi, vc_hi, vd_hi = v_hi[a], v_hi[b], v_hi[c], v_hi[d]
local z = W[2*k1-1] + (va_lo % 2^32 + vb_lo % 2^32)
va_lo = NORM(z)
va_hi = NORM(W[2*k1] + (va_hi + vb_hi + floor(z / 2^32)))
vd_lo, vd_hi = XOR(vd_hi, va_hi), XOR(vd_lo, va_lo)
z = vc_lo % 2^32 + vd_lo % 2^32
vc_lo = NORM(z)
vc_hi = NORM(vc_hi + vd_hi + floor(z / 2^32))
vb_lo, vb_hi = XOR(vb_lo, vc_lo), XOR(vb_hi, vc_hi)
vb_lo, vb_hi = XOR(SHR(vb_lo, 24), SHL(vb_hi, 8)), XOR(SHR(vb_hi, 24), SHL(vb_lo, 8))
z = W[2*k2-1] + (va_lo % 2^32 + vb_lo % 2^32)
va_lo = NORM(z)
va_hi = NORM(W[2*k2] + (va_hi + vb_hi + floor(z / 2^32)))
vd_lo, vd_hi = XOR(vd_lo, va_lo), XOR(vd_hi, va_hi)
vd_lo, vd_hi = XOR(SHR(vd_lo, 16), SHL(vd_hi, 16)), XOR(SHR(vd_hi, 16), SHL(vd_lo, 16))
z = vc_lo % 2^32 + vd_lo % 2^32
vc_lo = NORM(z)
vc_hi = NORM(vc_hi + vd_hi + floor(z / 2^32))
vb_lo, vb_hi = XOR(vb_lo, vc_lo), XOR(vb_hi, vc_hi)
vb_lo, vb_hi = XOR(SHL(vb_lo, 1), SHR(vb_hi, 31)), XOR(SHL(vb_hi, 1), SHR(vb_lo, 31))
v_lo[a], v_lo[b], v_lo[c], v_lo[d] = va_lo, vb_lo, vc_lo, vd_lo
v_hi[a], v_hi[b], v_hi[c], v_hi[d] = va_hi, vb_hi, vc_hi, vd_hi
end
function blake2b_feed_128(H_lo, H_hi, str, offs, size, bytes_compressed, last_block_size, is_last_node)
-- offs >= 0, size >= 0, size is multiple of 128
local W = common_W
local h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo = H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8]
local h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi = H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8]
for pos = offs, offs + size - 1, 128 do
if str then
for j = 1, 32 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = d * 2^24 + OR(SHL(c, 16), SHL(b, 8), a)
end
end
v_lo[0x0], v_lo[0x1], v_lo[0x2], v_lo[0x3], v_lo[0x4], v_lo[0x5], v_lo[0x6], v_lo[0x7] = h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo
v_lo[0x8], v_lo[0x9], v_lo[0xA], v_lo[0xB], v_lo[0xC], v_lo[0xD], v_lo[0xE], v_lo[0xF] = sha2_H_lo[1], sha2_H_lo[2], sha2_H_lo[3], sha2_H_lo[4], sha2_H_lo[5], sha2_H_lo[6], sha2_H_lo[7], sha2_H_lo[8]
v_hi[0x0], v_hi[0x1], v_hi[0x2], v_hi[0x3], v_hi[0x4], v_hi[0x5], v_hi[0x6], v_hi[0x7] = h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi
v_hi[0x8], v_hi[0x9], v_hi[0xA], v_hi[0xB], v_hi[0xC], v_hi[0xD], v_hi[0xE], v_hi[0xF] = sha2_H_hi[1], sha2_H_hi[2], sha2_H_hi[3], sha2_H_hi[4], sha2_H_hi[5], sha2_H_hi[6], sha2_H_hi[7], sha2_H_hi[8]
bytes_compressed = bytes_compressed + (last_block_size or 128)
local t0_lo = bytes_compressed % 2^32
local t0_hi = floor(bytes_compressed / 2^32)
v_lo[0xC] = XOR(v_lo[0xC], t0_lo) -- t0 = low_8_bytes(bytes_compressed)
v_hi[0xC] = XOR(v_hi[0xC], t0_hi)
-- t1 = high_8_bytes(bytes_compressed) = 0, message length is always below 2^53 bytes
if last_block_size then -- flag f0
v_lo[0xE] = NOT(v_lo[0xE])
v_hi[0xE] = NOT(v_hi[0xE])
end
if is_last_node then -- flag f1
v_lo[0xF] = NOT(v_lo[0xF])
v_hi[0xF] = NOT(v_hi[0xF])
end
for j = 1, 12 do
local row = sigma[j]
G(0, 4, 8, 12, row[ 1], row[ 2])
G(1, 5, 9, 13, row[ 3], row[ 4])
G(2, 6, 10, 14, row[ 5], row[ 6])
G(3, 7, 11, 15, row[ 7], row[ 8])
G(0, 5, 10, 15, row[ 9], row[10])
G(1, 6, 11, 12, row[11], row[12])
G(2, 7, 8, 13, row[13], row[14])
G(3, 4, 9, 14, row[15], row[16])
end
h1_lo = XOR(h1_lo, v_lo[0x0], v_lo[0x8])
h2_lo = XOR(h2_lo, v_lo[0x1], v_lo[0x9])
h3_lo = XOR(h3_lo, v_lo[0x2], v_lo[0xA])
h4_lo = XOR(h4_lo, v_lo[0x3], v_lo[0xB])
h5_lo = XOR(h5_lo, v_lo[0x4], v_lo[0xC])
h6_lo = XOR(h6_lo, v_lo[0x5], v_lo[0xD])
h7_lo = XOR(h7_lo, v_lo[0x6], v_lo[0xE])
h8_lo = XOR(h8_lo, v_lo[0x7], v_lo[0xF])
h1_hi = XOR(h1_hi, v_hi[0x0], v_hi[0x8])
h2_hi = XOR(h2_hi, v_hi[0x1], v_hi[0x9])
h3_hi = XOR(h3_hi, v_hi[0x2], v_hi[0xA])
h4_hi = XOR(h4_hi, v_hi[0x3], v_hi[0xB])
h5_hi = XOR(h5_hi, v_hi[0x4], v_hi[0xC])
h6_hi = XOR(h6_hi, v_hi[0x5], v_hi[0xD])
h7_hi = XOR(h7_hi, v_hi[0x6], v_hi[0xE])
h8_hi = XOR(h8_hi, v_hi[0x7], v_hi[0xF])
end
H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8] = h1_lo % 2^32, h2_lo % 2^32, h3_lo % 2^32, h4_lo % 2^32, h5_lo % 2^32, h6_lo % 2^32, h7_lo % 2^32, h8_lo % 2^32
H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8] = h1_hi % 2^32, h2_hi % 2^32, h3_hi % 2^32, h4_hi % 2^32, h5_hi % 2^32, h6_hi % 2^32, h7_hi % 2^32, h8_hi % 2^32
return bytes_compressed
end
end
end
if branch == "FFI" or branch == "LJ" then
-- BLAKE2s and BLAKE3 implementations for "LuaJIT with FFI" and "LuaJIT without FFI" branches
do
local W = common_W_blake2s
local v = v_for_blake2s_feed_64
local function G(a, b, c, d, k1, k2)
local va, vb, vc, vd = v[a], v[b], v[c], v[d]
va = NORM(W[k1] + (va + vb))
vd = ROR(XOR(vd, va), 16)
vc = NORM(vc + vd)
vb = ROR(XOR(vb, vc), 12)
va = NORM(W[k2] + (va + vb))
vd = ROR(XOR(vd, va), 8)
vc = NORM(vc + vd)
vb = ROR(XOR(vb, vc), 7)
v[a], v[b], v[c], v[d] = va, vb, vc, vd
end
function blake2s_feed_64(H, str, offs, size, bytes_compressed, last_block_size, is_last_node)
-- offs >= 0, size >= 0, size is multiple of 64
local h1, h2, h3, h4, h5, h6, h7, h8 = NORM(H[1]), NORM(H[2]), NORM(H[3]), NORM(H[4]), NORM(H[5]), NORM(H[6]), NORM(H[7]), NORM(H[8])
for pos = offs, offs + size - 1, 64 do
if str then
for j = 1, 16 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = OR(SHL(d, 24), SHL(c, 16), SHL(b, 8), a)
end
end
v[0x0], v[0x1], v[0x2], v[0x3], v[0x4], v[0x5], v[0x6], v[0x7] = h1, h2, h3, h4, h5, h6, h7, h8
v[0x8], v[0x9], v[0xA], v[0xB], v[0xE], v[0xF] = NORM(sha2_H_hi[1]), NORM(sha2_H_hi[2]), NORM(sha2_H_hi[3]), NORM(sha2_H_hi[4]), NORM(sha2_H_hi[7]), NORM(sha2_H_hi[8])
bytes_compressed = bytes_compressed + (last_block_size or 64)
local t0 = bytes_compressed % 2^32
local t1 = floor(bytes_compressed / 2^32)
v[0xC] = XOR(sha2_H_hi[5], t0) -- t0 = low_4_bytes(bytes_compressed)
v[0xD] = XOR(sha2_H_hi[6], t1) -- t1 = high_4_bytes(bytes_compressed
if last_block_size then -- flag f0
v[0xE] = NOT(v[0xE])
end
if is_last_node then -- flag f1
v[0xF] = NOT(v[0xF])
end
for j = 1, 10 do
local row = sigma[j]
G(0, 4, 8, 12, row[ 1], row[ 2])
G(1, 5, 9, 13, row[ 3], row[ 4])
G(2, 6, 10, 14, row[ 5], row[ 6])
G(3, 7, 11, 15, row[ 7], row[ 8])
G(0, 5, 10, 15, row[ 9], row[10])
G(1, 6, 11, 12, row[11], row[12])
G(2, 7, 8, 13, row[13], row[14])
G(3, 4, 9, 14, row[15], row[16])
end
h1 = XOR(h1, v[0x0], v[0x8])
h2 = XOR(h2, v[0x1], v[0x9])
h3 = XOR(h3, v[0x2], v[0xA])
h4 = XOR(h4, v[0x3], v[0xB])
h5 = XOR(h5, v[0x4], v[0xC])
h6 = XOR(h6, v[0x5], v[0xD])
h7 = XOR(h7, v[0x6], v[0xE])
h8 = XOR(h8, v[0x7], v[0xF])
end
H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8] = h1, h2, h3, h4, h5, h6, h7, h8
return bytes_compressed
end
function blake3_feed_64(str, offs, size, flags, chunk_index, H_in, H_out, wide_output, block_length)
-- offs >= 0, size >= 0, size is multiple of 64
block_length = block_length or 64
local h1, h2, h3, h4, h5, h6, h7, h8 = NORM(H_in[1]), NORM(H_in[2]), NORM(H_in[3]), NORM(H_in[4]), NORM(H_in[5]), NORM(H_in[6]), NORM(H_in[7]), NORM(H_in[8])
H_out = H_out or H_in
for pos = offs, offs + size - 1, 64 do
if str then
for j = 1, 16 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = OR(SHL(d, 24), SHL(c, 16), SHL(b, 8), a)
end
end
v[0x0], v[0x1], v[0x2], v[0x3], v[0x4], v[0x5], v[0x6], v[0x7] = h1, h2, h3, h4, h5, h6, h7, h8
v[0x8], v[0x9], v[0xA], v[0xB] = NORM(sha2_H_hi[1]), NORM(sha2_H_hi[2]), NORM(sha2_H_hi[3]), NORM(sha2_H_hi[4])
v[0xC] = NORM(chunk_index % 2^32) -- t0 = low_4_bytes(chunk_index)
v[0xD] = floor(chunk_index / 2^32) -- t1 = high_4_bytes(chunk_index)
v[0xE], v[0xF] = block_length, flags
for j = 1, 7 do
G(0, 4, 8, 12, perm_blake3[j], perm_blake3[j + 14])
G(1, 5, 9, 13, perm_blake3[j + 1], perm_blake3[j + 2])
G(2, 6, 10, 14, perm_blake3[j + 16], perm_blake3[j + 7])
G(3, 7, 11, 15, perm_blake3[j + 15], perm_blake3[j + 17])
G(0, 5, 10, 15, perm_blake3[j + 21], perm_blake3[j + 5])
G(1, 6, 11, 12, perm_blake3[j + 3], perm_blake3[j + 6])
G(2, 7, 8, 13, perm_blake3[j + 4], perm_blake3[j + 18])
G(3, 4, 9, 14, perm_blake3[j + 19], perm_blake3[j + 20])
end
if wide_output then
H_out[ 9] = XOR(h1, v[0x8])
H_out[10] = XOR(h2, v[0x9])
H_out[11] = XOR(h3, v[0xA])
H_out[12] = XOR(h4, v[0xB])
H_out[13] = XOR(h5, v[0xC])
H_out[14] = XOR(h6, v[0xD])
H_out[15] = XOR(h7, v[0xE])
H_out[16] = XOR(h8, v[0xF])
end
h1 = XOR(v[0x0], v[0x8])
h2 = XOR(v[0x1], v[0x9])
h3 = XOR(v[0x2], v[0xA])
h4 = XOR(v[0x3], v[0xB])
h5 = XOR(v[0x4], v[0xC])
h6 = XOR(v[0x5], v[0xD])
h7 = XOR(v[0x6], v[0xE])
h8 = XOR(v[0x7], v[0xF])
end
H_out[1], H_out[2], H_out[3], H_out[4], H_out[5], H_out[6], H_out[7], H_out[8] = h1, h2, h3, h4, h5, h6, h7, h8
end
end
end
if branch == "INT64" then
-- implementation for Lua 5.3/5.4
hi_factor = 4294967296
hi_factor_keccak = 4294967296
lanes_index_base = 1
HEX64, XORA5, XOR_BYTE, sha256_feed_64, sha512_feed_128, md5_feed_64, sha1_feed_64, keccak_feed, blake2s_feed_64, blake2b_feed_128, blake3_feed_64 = load[=[-- branch "INT64"
local md5_next_shift, md5_K, sha2_K_lo, sha2_K_hi, build_keccak_format, sha3_RC_lo, sigma, common_W, sha2_H_lo, sha2_H_hi, perm_blake3 = ...
local string_format, string_unpack = string.format, string.unpack
local function HEX64(x)
return string_format("%016x", x)
end
local function XORA5(x, y)
return x ~ (y or 0xa5a5a5a5a5a5a5a5)
end
local function XOR_BYTE(x, y)
return x ~ y
end
local function sha256_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W, K = common_W, sha2_K_hi
local h1, h2, h3, h4, h5, h6, h7, h8 = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for pos = offs + 1, offs + size, 64 do
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack(">I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4", str, pos)
for j = 17, 64 do
local a = W[j-15]
a = a<<32 | a
local b = W[j-2]
b = b<<32 | b
W[j] = (a>>7 ~ a>>18 ~ a>>35) + (b>>17 ~ b>>19 ~ b>>42) + W[j-7] + W[j-16] & (1<<32)-1
end
local a, b, c, d, e, f, g, h = h1, h2, h3, h4, h5, h6, h7, h8
for j = 1, 64 do
e = e<<32 | e & (1<<32)-1
local z = (e>>6 ~ e>>11 ~ e>>25) + (g ~ e & (f ~ g)) + h + K[j] + W[j]
h = g
g = f
f = e
e = z + d
d = c
c = b
b = a
a = a<<32 | a & (1<<32)-1
a = z + ((a ~ c) & d ~ a & c) + (a>>2 ~ a>>13 ~ a>>22)
end
h1 = a + h1
h2 = b + h2
h3 = c + h3
h4 = d + h4
h5 = e + h5
h6 = f + h6
h7 = g + h7
h8 = h + h8
end
H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8] = h1, h2, h3, h4, h5, h6, h7, h8
end
local function sha512_feed_128(H, _, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 128
local W, K = common_W, sha2_K_lo
local h1, h2, h3, h4, h5, h6, h7, h8 = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for pos = offs + 1, offs + size, 128 do
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack(">i8i8i8i8i8i8i8i8i8i8i8i8i8i8i8i8", str, pos)
for j = 17, 80 do
local a = W[j-15]
local b = W[j-2]
W[j] = (a >> 1 ~ a >> 7 ~ a >> 8 ~ a << 56 ~ a << 63) + (b >> 6 ~ b >> 19 ~ b >> 61 ~ b << 3 ~ b << 45) + W[j-7] + W[j-16]
end
local a, b, c, d, e, f, g, h = h1, h2, h3, h4, h5, h6, h7, h8
for j = 1, 80 do
local z = (e >> 14 ~ e >> 18 ~ e >> 41 ~ e << 23 ~ e << 46 ~ e << 50) + (g ~ e & (f ~ g)) + h + K[j] + W[j]
h = g
g = f
f = e
e = z + d
d = c
c = b
b = a
a = z + ((a ~ c) & d ~ a & c) + (a >> 28 ~ a >> 34 ~ a >> 39 ~ a << 25 ~ a << 30 ~ a << 36)
end
h1 = a + h1
h2 = b + h2
h3 = c + h3
h4 = d + h4
h5 = e + h5
h6 = f + h6
h7 = g + h7
h8 = h + h8
end
H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8] = h1, h2, h3, h4, h5, h6, h7, h8
end
local function md5_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W, K, md5_next_shift = common_W, md5_K, md5_next_shift
local h1, h2, h3, h4 = H[1], H[2], H[3], H[4]
for pos = offs + 1, offs + size, 64 do
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack("> s) + b
s = md5_next_shift[s]
end
s = 32-5
for j = 17, 32 do
local F = (c ~ d & (b ~ c)) + a + K[j] + W[(5*j-4 & 15) + 1]
a = d
d = c
c = b
b = ((F<<32 | F & (1<<32)-1) >> s) + b
s = md5_next_shift[s]
end
s = 32-4
for j = 33, 48 do
local F = (b ~ c ~ d) + a + K[j] + W[(3*j+2 & 15) + 1]
a = d
d = c
c = b
b = ((F<<32 | F & (1<<32)-1) >> s) + b
s = md5_next_shift[s]
end
s = 32-6
for j = 49, 64 do
local F = (c ~ (b | ~d)) + a + K[j] + W[(j*7-7 & 15) + 1]
a = d
d = c
c = b
b = ((F<<32 | F & (1<<32)-1) >> s) + b
s = md5_next_shift[s]
end
h1 = a + h1
h2 = b + h2
h3 = c + h3
h4 = d + h4
end
H[1], H[2], H[3], H[4] = h1, h2, h3, h4
end
local function sha1_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W = common_W
local h1, h2, h3, h4, h5 = H[1], H[2], H[3], H[4], H[5]
for pos = offs + 1, offs + size, 64 do
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack(">I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4", str, pos)
for j = 17, 80 do
local a = W[j-3] ~ W[j-8] ~ W[j-14] ~ W[j-16]
W[j] = (a<<32 | a) << 1 >> 32
end
local a, b, c, d, e = h1, h2, h3, h4, h5
for j = 1, 20 do
local z = ((a<<32 | a & (1<<32)-1) >> 27) + (d ~ b & (c ~ d)) + 0x5A827999 + W[j] + e -- constant = floor(2^30 * sqrt(2))
e = d
d = c
c = (b<<32 | b & (1<<32)-1) >> 2
b = a
a = z
end
for j = 21, 40 do
local z = ((a<<32 | a & (1<<32)-1) >> 27) + (b ~ c ~ d) + 0x6ED9EBA1 + W[j] + e -- 2^30 * sqrt(3)
e = d
d = c
c = (b<<32 | b & (1<<32)-1) >> 2
b = a
a = z
end
for j = 41, 60 do
local z = ((a<<32 | a & (1<<32)-1) >> 27) + ((b ~ c) & d ~ b & c) + 0x8F1BBCDC + W[j] + e -- 2^30 * sqrt(5)
e = d
d = c
c = (b<<32 | b & (1<<32)-1) >> 2
b = a
a = z
end
for j = 61, 80 do
local z = ((a<<32 | a & (1<<32)-1) >> 27) + (b ~ c ~ d) + 0xCA62C1D6 + W[j] + e -- 2^30 * sqrt(10)
e = d
d = c
c = (b<<32 | b & (1<<32)-1) >> 2
b = a
a = z
end
h1 = a + h1
h2 = b + h2
h3 = c + h3
h4 = d + h4
h5 = e + h5
end
H[1], H[2], H[3], H[4], H[5] = h1, h2, h3, h4, h5
end
local keccak_format_i8 = build_keccak_format("i8")
local function keccak_feed(lanes, _, str, offs, size, block_size_in_bytes)
-- offs >= 0, size >= 0, size is multiple of block_size_in_bytes, block_size_in_bytes is positive multiple of 8
local RC = sha3_RC_lo
local qwords_qty = block_size_in_bytes / 8
local keccak_format = keccak_format_i8[qwords_qty]
for pos = offs + 1, offs + size, block_size_in_bytes do
local qwords_from_message = {string_unpack(keccak_format, str, pos)}
for j = 1, qwords_qty do
lanes[j] = lanes[j] ~ qwords_from_message[j]
end
local L01, L02, L03, L04, L05, L06, L07, L08, L09, L10, L11, L12, L13, L14, L15, L16, L17, L18, L19, L20, L21, L22, L23, L24, L25 =
lanes[1], lanes[2], lanes[3], lanes[4], lanes[5], lanes[6], lanes[7], lanes[8], lanes[9], lanes[10], lanes[11], lanes[12], lanes[13],
lanes[14], lanes[15], lanes[16], lanes[17], lanes[18], lanes[19], lanes[20], lanes[21], lanes[22], lanes[23], lanes[24], lanes[25]
for round_idx = 1, 24 do
local C1 = L01 ~ L06 ~ L11 ~ L16 ~ L21
local C2 = L02 ~ L07 ~ L12 ~ L17 ~ L22
local C3 = L03 ~ L08 ~ L13 ~ L18 ~ L23
local C4 = L04 ~ L09 ~ L14 ~ L19 ~ L24
local C5 = L05 ~ L10 ~ L15 ~ L20 ~ L25
local D = C1 ~ C3<<1 ~ C3>>63
local T0 = D ~ L02
local T1 = D ~ L07
local T2 = D ~ L12
local T3 = D ~ L17
local T4 = D ~ L22
L02 = T1<<44 ~ T1>>20
L07 = T3<<45 ~ T3>>19
L12 = T0<<1 ~ T0>>63
L17 = T2<<10 ~ T2>>54
L22 = T4<<2 ~ T4>>62
D = C2 ~ C4<<1 ~ C4>>63
T0 = D ~ L03
T1 = D ~ L08
T2 = D ~ L13
T3 = D ~ L18
T4 = D ~ L23
L03 = T2<<43 ~ T2>>21
L08 = T4<<61 ~ T4>>3
L13 = T1<<6 ~ T1>>58
L18 = T3<<15 ~ T3>>49
L23 = T0<<62 ~ T0>>2
D = C3 ~ C5<<1 ~ C5>>63
T0 = D ~ L04
T1 = D ~ L09
T2 = D ~ L14
T3 = D ~ L19
T4 = D ~ L24
L04 = T3<<21 ~ T3>>43
L09 = T0<<28 ~ T0>>36
L14 = T2<<25 ~ T2>>39
L19 = T4<<56 ~ T4>>8
L24 = T1<<55 ~ T1>>9
D = C4 ~ C1<<1 ~ C1>>63
T0 = D ~ L05
T1 = D ~ L10
T2 = D ~ L15
T3 = D ~ L20
T4 = D ~ L25
L05 = T4<<14 ~ T4>>50
L10 = T1<<20 ~ T1>>44
L15 = T3<<8 ~ T3>>56
L20 = T0<<27 ~ T0>>37
L25 = T2<<39 ~ T2>>25
D = C5 ~ C2<<1 ~ C2>>63
T1 = D ~ L06
T2 = D ~ L11
T3 = D ~ L16
T4 = D ~ L21
L06 = T2<<3 ~ T2>>61
L11 = T4<<18 ~ T4>>46
L16 = T1<<36 ~ T1>>28
L21 = T3<<41 ~ T3>>23
L01 = D ~ L01
L01, L02, L03, L04, L05 = L01 ~ ~L02 & L03, L02 ~ ~L03 & L04, L03 ~ ~L04 & L05, L04 ~ ~L05 & L01, L05 ~ ~L01 & L02
L06, L07, L08, L09, L10 = L09 ~ ~L10 & L06, L10 ~ ~L06 & L07, L06 ~ ~L07 & L08, L07 ~ ~L08 & L09, L08 ~ ~L09 & L10
L11, L12, L13, L14, L15 = L12 ~ ~L13 & L14, L13 ~ ~L14 & L15, L14 ~ ~L15 & L11, L15 ~ ~L11 & L12, L11 ~ ~L12 & L13
L16, L17, L18, L19, L20 = L20 ~ ~L16 & L17, L16 ~ ~L17 & L18, L17 ~ ~L18 & L19, L18 ~ ~L19 & L20, L19 ~ ~L20 & L16
L21, L22, L23, L24, L25 = L23 ~ ~L24 & L25, L24 ~ ~L25 & L21, L25 ~ ~L21 & L22, L21 ~ ~L22 & L23, L22 ~ ~L23 & L24
L01 = L01 ~ RC[round_idx]
end
lanes[1] = L01
lanes[2] = L02
lanes[3] = L03
lanes[4] = L04
lanes[5] = L05
lanes[6] = L06
lanes[7] = L07
lanes[8] = L08
lanes[9] = L09
lanes[10] = L10
lanes[11] = L11
lanes[12] = L12
lanes[13] = L13
lanes[14] = L14
lanes[15] = L15
lanes[16] = L16
lanes[17] = L17
lanes[18] = L18
lanes[19] = L19
lanes[20] = L20
lanes[21] = L21
lanes[22] = L22
lanes[23] = L23
lanes[24] = L24
lanes[25] = L25
end
end
local function blake2s_feed_64(H, str, offs, size, bytes_compressed, last_block_size, is_last_node)
-- offs >= 0, size >= 0, size is multiple of 64
local W = common_W
local h1, h2, h3, h4, h5, h6, h7, h8 = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for pos = offs + 1, offs + size, 64 do
if str then
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack("> 32 -- t1 = high_4_bytes(bytes_compressed)
if last_block_size then -- flag f0
vE = ~vE
end
if is_last_node then -- flag f1
vF = ~vF
end
for j = 1, 10 do
local row = sigma[j]
v0 = v0 + v4 + W[row[1]]
vC = vC ~ v0
vC = (vC & (1<<32)-1) >> 16 | vC << 16
v8 = v8 + vC
v4 = v4 ~ v8
v4 = (v4 & (1<<32)-1) >> 12 | v4 << 20
v0 = v0 + v4 + W[row[2]]
vC = vC ~ v0
vC = (vC & (1<<32)-1) >> 8 | vC << 24
v8 = v8 + vC
v4 = v4 ~ v8
v4 = (v4 & (1<<32)-1) >> 7 | v4 << 25
v1 = v1 + v5 + W[row[3]]
vD = vD ~ v1
vD = (vD & (1<<32)-1) >> 16 | vD << 16
v9 = v9 + vD
v5 = v5 ~ v9
v5 = (v5 & (1<<32)-1) >> 12 | v5 << 20
v1 = v1 + v5 + W[row[4]]
vD = vD ~ v1
vD = (vD & (1<<32)-1) >> 8 | vD << 24
v9 = v9 + vD
v5 = v5 ~ v9
v5 = (v5 & (1<<32)-1) >> 7 | v5 << 25
v2 = v2 + v6 + W[row[5]]
vE = vE ~ v2
vE = (vE & (1<<32)-1) >> 16 | vE << 16
vA = vA + vE
v6 = v6 ~ vA
v6 = (v6 & (1<<32)-1) >> 12 | v6 << 20
v2 = v2 + v6 + W[row[6]]
vE = vE ~ v2
vE = (vE & (1<<32)-1) >> 8 | vE << 24
vA = vA + vE
v6 = v6 ~ vA
v6 = (v6 & (1<<32)-1) >> 7 | v6 << 25
v3 = v3 + v7 + W[row[7]]
vF = vF ~ v3
vF = (vF & (1<<32)-1) >> 16 | vF << 16
vB = vB + vF
v7 = v7 ~ vB
v7 = (v7 & (1<<32)-1) >> 12 | v7 << 20
v3 = v3 + v7 + W[row[8]]
vF = vF ~ v3
vF = (vF & (1<<32)-1) >> 8 | vF << 24
vB = vB + vF
v7 = v7 ~ vB
v7 = (v7 & (1<<32)-1) >> 7 | v7 << 25
v0 = v0 + v5 + W[row[9]]
vF = vF ~ v0
vF = (vF & (1<<32)-1) >> 16 | vF << 16
vA = vA + vF
v5 = v5 ~ vA
v5 = (v5 & (1<<32)-1) >> 12 | v5 << 20
v0 = v0 + v5 + W[row[10]]
vF = vF ~ v0
vF = (vF & (1<<32)-1) >> 8 | vF << 24
vA = vA + vF
v5 = v5 ~ vA
v5 = (v5 & (1<<32)-1) >> 7 | v5 << 25
v1 = v1 + v6 + W[row[11]]
vC = vC ~ v1
vC = (vC & (1<<32)-1) >> 16 | vC << 16
vB = vB + vC
v6 = v6 ~ vB
v6 = (v6 & (1<<32)-1) >> 12 | v6 << 20
v1 = v1 + v6 + W[row[12]]
vC = vC ~ v1
vC = (vC & (1<<32)-1) >> 8 | vC << 24
vB = vB + vC
v6 = v6 ~ vB
v6 = (v6 & (1<<32)-1) >> 7 | v6 << 25
v2 = v2 + v7 + W[row[13]]
vD = vD ~ v2
vD = (vD & (1<<32)-1) >> 16 | vD << 16
v8 = v8 + vD
v7 = v7 ~ v8
v7 = (v7 & (1<<32)-1) >> 12 | v7 << 20
v2 = v2 + v7 + W[row[14]]
vD = vD ~ v2
vD = (vD & (1<<32)-1) >> 8 | vD << 24
v8 = v8 + vD
v7 = v7 ~ v8
v7 = (v7 & (1<<32)-1) >> 7 | v7 << 25
v3 = v3 + v4 + W[row[15]]
vE = vE ~ v3
vE = (vE & (1<<32)-1) >> 16 | vE << 16
v9 = v9 + vE
v4 = v4 ~ v9
v4 = (v4 & (1<<32)-1) >> 12 | v4 << 20
v3 = v3 + v4 + W[row[16]]
vE = vE ~ v3
vE = (vE & (1<<32)-1) >> 8 | vE << 24
v9 = v9 + vE
v4 = v4 ~ v9
v4 = (v4 & (1<<32)-1) >> 7 | v4 << 25
end
h1 = h1 ~ v0 ~ v8
h2 = h2 ~ v1 ~ v9
h3 = h3 ~ v2 ~ vA
h4 = h4 ~ v3 ~ vB
h5 = h5 ~ v4 ~ vC
h6 = h6 ~ v5 ~ vD
h7 = h7 ~ v6 ~ vE
h8 = h8 ~ v7 ~ vF
end
H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8] = h1, h2, h3, h4, h5, h6, h7, h8
return bytes_compressed
end
local function blake2b_feed_128(H, _, str, offs, size, bytes_compressed, last_block_size, is_last_node)
-- offs >= 0, size >= 0, size is multiple of 128
local W = common_W
local h1, h2, h3, h4, h5, h6, h7, h8 = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for pos = offs + 1, offs + size, 128 do
if str then
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack("> 32 | vC << 32
v8 = v8 + vC
v4 = v4 ~ v8
v4 = v4 >> 24 | v4 << 40
v0 = v0 + v4 + W[row[2]]
vC = vC ~ v0
vC = vC >> 16 | vC << 48
v8 = v8 + vC
v4 = v4 ~ v8
v4 = v4 >> 63 | v4 << 1
v1 = v1 + v5 + W[row[3]]
vD = vD ~ v1
vD = vD >> 32 | vD << 32
v9 = v9 + vD
v5 = v5 ~ v9
v5 = v5 >> 24 | v5 << 40
v1 = v1 + v5 + W[row[4]]
vD = vD ~ v1
vD = vD >> 16 | vD << 48
v9 = v9 + vD
v5 = v5 ~ v9
v5 = v5 >> 63 | v5 << 1
v2 = v2 + v6 + W[row[5]]
vE = vE ~ v2
vE = vE >> 32 | vE << 32
vA = vA + vE
v6 = v6 ~ vA
v6 = v6 >> 24 | v6 << 40
v2 = v2 + v6 + W[row[6]]
vE = vE ~ v2
vE = vE >> 16 | vE << 48
vA = vA + vE
v6 = v6 ~ vA
v6 = v6 >> 63 | v6 << 1
v3 = v3 + v7 + W[row[7]]
vF = vF ~ v3
vF = vF >> 32 | vF << 32
vB = vB + vF
v7 = v7 ~ vB
v7 = v7 >> 24 | v7 << 40
v3 = v3 + v7 + W[row[8]]
vF = vF ~ v3
vF = vF >> 16 | vF << 48
vB = vB + vF
v7 = v7 ~ vB
v7 = v7 >> 63 | v7 << 1
v0 = v0 + v5 + W[row[9]]
vF = vF ~ v0
vF = vF >> 32 | vF << 32
vA = vA + vF
v5 = v5 ~ vA
v5 = v5 >> 24 | v5 << 40
v0 = v0 + v5 + W[row[10]]
vF = vF ~ v0
vF = vF >> 16 | vF << 48
vA = vA + vF
v5 = v5 ~ vA
v5 = v5 >> 63 | v5 << 1
v1 = v1 + v6 + W[row[11]]
vC = vC ~ v1
vC = vC >> 32 | vC << 32
vB = vB + vC
v6 = v6 ~ vB
v6 = v6 >> 24 | v6 << 40
v1 = v1 + v6 + W[row[12]]
vC = vC ~ v1
vC = vC >> 16 | vC << 48
vB = vB + vC
v6 = v6 ~ vB
v6 = v6 >> 63 | v6 << 1
v2 = v2 + v7 + W[row[13]]
vD = vD ~ v2
vD = vD >> 32 | vD << 32
v8 = v8 + vD
v7 = v7 ~ v8
v7 = v7 >> 24 | v7 << 40
v2 = v2 + v7 + W[row[14]]
vD = vD ~ v2
vD = vD >> 16 | vD << 48
v8 = v8 + vD
v7 = v7 ~ v8
v7 = v7 >> 63 | v7 << 1
v3 = v3 + v4 + W[row[15]]
vE = vE ~ v3
vE = vE >> 32 | vE << 32
v9 = v9 + vE
v4 = v4 ~ v9
v4 = v4 >> 24 | v4 << 40
v3 = v3 + v4 + W[row[16]]
vE = vE ~ v3
vE = vE >> 16 | vE << 48
v9 = v9 + vE
v4 = v4 ~ v9
v4 = v4 >> 63 | v4 << 1
end
h1 = h1 ~ v0 ~ v8
h2 = h2 ~ v1 ~ v9
h3 = h3 ~ v2 ~ vA
h4 = h4 ~ v3 ~ vB
h5 = h5 ~ v4 ~ vC
h6 = h6 ~ v5 ~ vD
h7 = h7 ~ v6 ~ vE
h8 = h8 ~ v7 ~ vF
end
H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8] = h1, h2, h3, h4, h5, h6, h7, h8
return bytes_compressed
end
local function blake3_feed_64(str, offs, size, flags, chunk_index, H_in, H_out, wide_output, block_length)
-- offs >= 0, size >= 0, size is multiple of 64
block_length = block_length or 64
local W = common_W
local h1, h2, h3, h4, h5, h6, h7, h8 = H_in[1], H_in[2], H_in[3], H_in[4], H_in[5], H_in[6], H_in[7], H_in[8]
H_out = H_out or H_in
for pos = offs + 1, offs + size, 64 do
if str then
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack("> 16 | vC << 16
v8 = v8 + vC
v4 = v4 ~ v8
v4 = (v4 & (1<<32)-1) >> 12 | v4 << 20
v0 = v0 + v4 + W[perm_blake3[j + 14]]
vC = vC ~ v0
vC = (vC & (1<<32)-1) >> 8 | vC << 24
v8 = v8 + vC
v4 = v4 ~ v8
v4 = (v4 & (1<<32)-1) >> 7 | v4 << 25
v1 = v1 + v5 + W[perm_blake3[j + 1]]
vD = vD ~ v1
vD = (vD & (1<<32)-1) >> 16 | vD << 16
v9 = v9 + vD
v5 = v5 ~ v9
v5 = (v5 & (1<<32)-1) >> 12 | v5 << 20
v1 = v1 + v5 + W[perm_blake3[j + 2]]
vD = vD ~ v1
vD = (vD & (1<<32)-1) >> 8 | vD << 24
v9 = v9 + vD
v5 = v5 ~ v9
v5 = (v5 & (1<<32)-1) >> 7 | v5 << 25
v2 = v2 + v6 + W[perm_blake3[j + 16]]
vE = vE ~ v2
vE = (vE & (1<<32)-1) >> 16 | vE << 16
vA = vA + vE
v6 = v6 ~ vA
v6 = (v6 & (1<<32)-1) >> 12 | v6 << 20
v2 = v2 + v6 + W[perm_blake3[j + 7]]
vE = vE ~ v2
vE = (vE & (1<<32)-1) >> 8 | vE << 24
vA = vA + vE
v6 = v6 ~ vA
v6 = (v6 & (1<<32)-1) >> 7 | v6 << 25
v3 = v3 + v7 + W[perm_blake3[j + 15]]
vF = vF ~ v3
vF = (vF & (1<<32)-1) >> 16 | vF << 16
vB = vB + vF
v7 = v7 ~ vB
v7 = (v7 & (1<<32)-1) >> 12 | v7 << 20
v3 = v3 + v7 + W[perm_blake3[j + 17]]
vF = vF ~ v3
vF = (vF & (1<<32)-1) >> 8 | vF << 24
vB = vB + vF
v7 = v7 ~ vB
v7 = (v7 & (1<<32)-1) >> 7 | v7 << 25
v0 = v0 + v5 + W[perm_blake3[j + 21]]
vF = vF ~ v0
vF = (vF & (1<<32)-1) >> 16 | vF << 16
vA = vA + vF
v5 = v5 ~ vA
v5 = (v5 & (1<<32)-1) >> 12 | v5 << 20
v0 = v0 + v5 + W[perm_blake3[j + 5]]
vF = vF ~ v0
vF = (vF & (1<<32)-1) >> 8 | vF << 24
vA = vA + vF
v5 = v5 ~ vA
v5 = (v5 & (1<<32)-1) >> 7 | v5 << 25
v1 = v1 + v6 + W[perm_blake3[j + 3]]
vC = vC ~ v1
vC = (vC & (1<<32)-1) >> 16 | vC << 16
vB = vB + vC
v6 = v6 ~ vB
v6 = (v6 & (1<<32)-1) >> 12 | v6 << 20
v1 = v1 + v6 + W[perm_blake3[j + 6]]
vC = vC ~ v1
vC = (vC & (1<<32)-1) >> 8 | vC << 24
vB = vB + vC
v6 = v6 ~ vB
v6 = (v6 & (1<<32)-1) >> 7 | v6 << 25
v2 = v2 + v7 + W[perm_blake3[j + 4]]
vD = vD ~ v2
vD = (vD & (1<<32)-1) >> 16 | vD << 16
v8 = v8 + vD
v7 = v7 ~ v8
v7 = (v7 & (1<<32)-1) >> 12 | v7 << 20
v2 = v2 + v7 + W[perm_blake3[j + 18]]
vD = vD ~ v2
vD = (vD & (1<<32)-1) >> 8 | vD << 24
v8 = v8 + vD
v7 = v7 ~ v8
v7 = (v7 & (1<<32)-1) >> 7 | v7 << 25
v3 = v3 + v4 + W[perm_blake3[j + 19]]
vE = vE ~ v3
vE = (vE & (1<<32)-1) >> 16 | vE << 16
v9 = v9 + vE
v4 = v4 ~ v9
v4 = (v4 & (1<<32)-1) >> 12 | v4 << 20
v3 = v3 + v4 + W[perm_blake3[j + 20]]
vE = vE ~ v3
vE = (vE & (1<<32)-1) >> 8 | vE << 24
v9 = v9 + vE
v4 = v4 ~ v9
v4 = (v4 & (1<<32)-1) >> 7 | v4 << 25
end
if wide_output then
H_out[ 9] = h1 ~ v8
H_out[10] = h2 ~ v9
H_out[11] = h3 ~ vA
H_out[12] = h4 ~ vB
H_out[13] = h5 ~ vC
H_out[14] = h6 ~ vD
H_out[15] = h7 ~ vE
H_out[16] = h8 ~ vF
end
h1 = v0 ~ v8
h2 = v1 ~ v9
h3 = v2 ~ vA
h4 = v3 ~ vB
h5 = v4 ~ vC
h6 = v5 ~ vD
h7 = v6 ~ vE
h8 = v7 ~ vF
end
H_out[1], H_out[2], H_out[3], H_out[4], H_out[5], H_out[6], H_out[7], H_out[8] = h1, h2, h3, h4, h5, h6, h7, h8
end
return HEX64, XORA5, XOR_BYTE, sha256_feed_64, sha512_feed_128, md5_feed_64, sha1_feed_64, keccak_feed, blake2s_feed_64, blake2b_feed_128, blake3_feed_64
]=](md5_next_shift, md5_K, sha2_K_lo, sha2_K_hi, build_keccak_format, sha3_RC_lo, sigma, common_W, sha2_H_lo, sha2_H_hi, perm_blake3)
end
if branch == "INT32" then
-- implementation for Lua 5.3/5.4 having non-standard numbers config "int32"+"double" (built with LUA_INT_TYPE=LUA_INT_INT)
K_lo_modulo = 2^32
function HEX(x) -- returns string of 8 lowercase hexadecimal digits
return string_format("%08x", x)
end
XORA5, XOR_BYTE, sha256_feed_64, sha512_feed_128, md5_feed_64, sha1_feed_64, keccak_feed, blake2s_feed_64, blake2b_feed_128, blake3_feed_64 = load[=[-- branch "INT32"
local md5_next_shift, md5_K, sha2_K_lo, sha2_K_hi, build_keccak_format, sha3_RC_lo, sha3_RC_hi, sigma, common_W, sha2_H_lo, sha2_H_hi, perm_blake3 = ...
local string_unpack, floor = string.unpack, math.floor
local function XORA5(x, y)
return x ~ (y and (y + 2^31) % 2^32 - 2^31 or 0xA5A5A5A5)
end
local function XOR_BYTE(x, y)
return x ~ y
end
local function sha256_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W, K = common_W, sha2_K_hi
local h1, h2, h3, h4, h5, h6, h7, h8 = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for pos = offs + 1, offs + size, 64 do
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack(">i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4", str, pos)
for j = 17, 64 do
local a, b = W[j-15], W[j-2]
W[j] = (a>>7 ~ a<<25 ~ a<<14 ~ a>>18 ~ a>>3) + (b<<15 ~ b>>17 ~ b<<13 ~ b>>19 ~ b>>10) + W[j-7] + W[j-16]
end
local a, b, c, d, e, f, g, h = h1, h2, h3, h4, h5, h6, h7, h8
for j = 1, 64 do
local z = (e>>6 ~ e<<26 ~ e>>11 ~ e<<21 ~ e>>25 ~ e<<7) + (g ~ e & (f ~ g)) + h + K[j] + W[j]
h = g
g = f
f = e
e = z + d
d = c
c = b
b = a
a = z + ((a ~ c) & d ~ a & c) + (a>>2 ~ a<<30 ~ a>>13 ~ a<<19 ~ a<<10 ~ a>>22)
end
h1 = a + h1
h2 = b + h2
h3 = c + h3
h4 = d + h4
h5 = e + h5
h6 = f + h6
h7 = g + h7
h8 = h + h8
end
H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8] = h1, h2, h3, h4, h5, h6, h7, h8
end
local function sha512_feed_128(H_lo, H_hi, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 128
-- W1_hi, W1_lo, W2_hi, W2_lo, ... Wk_hi = W[2*k-1], Wk_lo = W[2*k]
local floor, W, K_lo, K_hi = floor, common_W, sha2_K_lo, sha2_K_hi
local h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo = H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8]
local h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi = H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8]
for pos = offs + 1, offs + size, 128 do
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16],
W[17], W[18], W[19], W[20], W[21], W[22], W[23], W[24], W[25], W[26], W[27], W[28], W[29], W[30], W[31], W[32] =
string_unpack(">i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4", str, pos)
for jj = 17*2, 80*2, 2 do
local a_lo, a_hi, b_lo, b_hi = W[jj-30], W[jj-31], W[jj-4], W[jj-5]
local tmp =
(a_lo>>1 ~ a_hi<<31 ~ a_lo>>8 ~ a_hi<<24 ~ a_lo>>7 ~ a_hi<<25) % 2^32
+ (b_lo>>19 ~ b_hi<<13 ~ b_lo<<3 ~ b_hi>>29 ~ b_lo>>6 ~ b_hi<<26) % 2^32
+ W[jj-14] % 2^32 + W[jj-32] % 2^32
W[jj-1] =
(a_hi>>1 ~ a_lo<<31 ~ a_hi>>8 ~ a_lo<<24 ~ a_hi>>7)
+ (b_hi>>19 ~ b_lo<<13 ~ b_hi<<3 ~ b_lo>>29 ~ b_hi>>6)
+ W[jj-15] + W[jj-33] + floor(tmp / 2^32)
W[jj] = 0|((tmp + 2^31) % 2^32 - 2^31)
end
local a_lo, b_lo, c_lo, d_lo, e_lo, f_lo, g_lo, h_lo = h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo
local a_hi, b_hi, c_hi, d_hi, e_hi, f_hi, g_hi, h_hi = h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi
for j = 1, 80 do
local jj = 2*j
local z_lo = (e_lo>>14 ~ e_hi<<18 ~ e_lo>>18 ~ e_hi<<14 ~ e_lo<<23 ~ e_hi>>9) % 2^32 + (g_lo ~ e_lo & (f_lo ~ g_lo)) % 2^32 + h_lo % 2^32 + K_lo[j] + W[jj] % 2^32
local z_hi = (e_hi>>14 ~ e_lo<<18 ~ e_hi>>18 ~ e_lo<<14 ~ e_hi<<23 ~ e_lo>>9) + (g_hi ~ e_hi & (f_hi ~ g_hi)) + h_hi + K_hi[j] + W[jj-1] + floor(z_lo / 2^32)
z_lo = z_lo % 2^32
h_lo = g_lo; h_hi = g_hi
g_lo = f_lo; g_hi = f_hi
f_lo = e_lo; f_hi = e_hi
e_lo = z_lo + d_lo % 2^32
e_hi = z_hi + d_hi + floor(e_lo / 2^32)
e_lo = 0|((e_lo + 2^31) % 2^32 - 2^31)
d_lo = c_lo; d_hi = c_hi
c_lo = b_lo; c_hi = b_hi
b_lo = a_lo; b_hi = a_hi
z_lo = z_lo + (d_lo & c_lo ~ b_lo & (d_lo ~ c_lo)) % 2^32 + (b_lo>>28 ~ b_hi<<4 ~ b_lo<<30 ~ b_hi>>2 ~ b_lo<<25 ~ b_hi>>7) % 2^32
a_hi = z_hi + (d_hi & c_hi ~ b_hi & (d_hi ~ c_hi)) + (b_hi>>28 ~ b_lo<<4 ~ b_hi<<30 ~ b_lo>>2 ~ b_hi<<25 ~ b_lo>>7) + floor(z_lo / 2^32)
a_lo = 0|((z_lo + 2^31) % 2^32 - 2^31)
end
a_lo = h1_lo % 2^32 + a_lo % 2^32
h1_hi = h1_hi + a_hi + floor(a_lo / 2^32)
h1_lo = 0|((a_lo + 2^31) % 2^32 - 2^31)
a_lo = h2_lo % 2^32 + b_lo % 2^32
h2_hi = h2_hi + b_hi + floor(a_lo / 2^32)
h2_lo = 0|((a_lo + 2^31) % 2^32 - 2^31)
a_lo = h3_lo % 2^32 + c_lo % 2^32
h3_hi = h3_hi + c_hi + floor(a_lo / 2^32)
h3_lo = 0|((a_lo + 2^31) % 2^32 - 2^31)
a_lo = h4_lo % 2^32 + d_lo % 2^32
h4_hi = h4_hi + d_hi + floor(a_lo / 2^32)
h4_lo = 0|((a_lo + 2^31) % 2^32 - 2^31)
a_lo = h5_lo % 2^32 + e_lo % 2^32
h5_hi = h5_hi + e_hi + floor(a_lo / 2^32)
h5_lo = 0|((a_lo + 2^31) % 2^32 - 2^31)
a_lo = h6_lo % 2^32 + f_lo % 2^32
h6_hi = h6_hi + f_hi + floor(a_lo / 2^32)
h6_lo = 0|((a_lo + 2^31) % 2^32 - 2^31)
a_lo = h7_lo % 2^32 + g_lo % 2^32
h7_hi = h7_hi + g_hi + floor(a_lo / 2^32)
h7_lo = 0|((a_lo + 2^31) % 2^32 - 2^31)
a_lo = h8_lo % 2^32 + h_lo % 2^32
h8_hi = h8_hi + h_hi + floor(a_lo / 2^32)
h8_lo = 0|((a_lo + 2^31) % 2^32 - 2^31)
end
H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8] = h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo
H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8] = h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi
end
local function md5_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W, K, md5_next_shift = common_W, md5_K, md5_next_shift
local h1, h2, h3, h4 = H[1], H[2], H[3], H[4]
for pos = offs + 1, offs + size, 64 do
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack(">s) + b
s = md5_next_shift[s]
end
s = 32-5
for j = 17, 32 do
local F = (c ~ d & (b ~ c)) + a + K[j] + W[(5*j-4 & 15) + 1]
a = d
d = c
c = b
b = (F << 32-s | F>>s) + b
s = md5_next_shift[s]
end
s = 32-4
for j = 33, 48 do
local F = (b ~ c ~ d) + a + K[j] + W[(3*j+2 & 15) + 1]
a = d
d = c
c = b
b = (F << 32-s | F>>s) + b
s = md5_next_shift[s]
end
s = 32-6
for j = 49, 64 do
local F = (c ~ (b | ~d)) + a + K[j] + W[(j*7-7 & 15) + 1]
a = d
d = c
c = b
b = (F << 32-s | F>>s) + b
s = md5_next_shift[s]
end
h1 = a + h1
h2 = b + h2
h3 = c + h3
h4 = d + h4
end
H[1], H[2], H[3], H[4] = h1, h2, h3, h4
end
local function sha1_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W = common_W
local h1, h2, h3, h4, h5 = H[1], H[2], H[3], H[4], H[5]
for pos = offs + 1, offs + size, 64 do
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack(">i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4i4", str, pos)
for j = 17, 80 do
local a = W[j-3] ~ W[j-8] ~ W[j-14] ~ W[j-16]
W[j] = a << 1 ~ a >> 31
end
local a, b, c, d, e = h1, h2, h3, h4, h5
for j = 1, 20 do
local z = (a << 5 ~ a >> 27) + (d ~ b & (c ~ d)) + 0x5A827999 + W[j] + e -- constant = floor(2^30 * sqrt(2))
e = d
d = c
c = b << 30 ~ b >> 2
b = a
a = z
end
for j = 21, 40 do
local z = (a << 5 ~ a >> 27) + (b ~ c ~ d) + 0x6ED9EBA1 + W[j] + e -- 2^30 * sqrt(3)
e = d
d = c
c = b << 30 ~ b >> 2
b = a
a = z
end
for j = 41, 60 do
local z = (a << 5 ~ a >> 27) + ((b ~ c) & d ~ b & c) + 0x8F1BBCDC + W[j] + e -- 2^30 * sqrt(5)
e = d
d = c
c = b << 30 ~ b >> 2
b = a
a = z
end
for j = 61, 80 do
local z = (a << 5 ~ a >> 27) + (b ~ c ~ d) + 0xCA62C1D6 + W[j] + e -- 2^30 * sqrt(10)
e = d
d = c
c = b << 30 ~ b >> 2
b = a
a = z
end
h1 = a + h1
h2 = b + h2
h3 = c + h3
h4 = d + h4
h5 = e + h5
end
H[1], H[2], H[3], H[4], H[5] = h1, h2, h3, h4, h5
end
local keccak_format_i4i4 = build_keccak_format("i4i4")
local function keccak_feed(lanes_lo, lanes_hi, str, offs, size, block_size_in_bytes)
-- offs >= 0, size >= 0, size is multiple of block_size_in_bytes, block_size_in_bytes is positive multiple of 8
local RC_lo, RC_hi = sha3_RC_lo, sha3_RC_hi
local qwords_qty = block_size_in_bytes / 8
local keccak_format = keccak_format_i4i4[qwords_qty]
for pos = offs + 1, offs + size, block_size_in_bytes do
local dwords_from_message = {string_unpack(keccak_format, str, pos)}
for j = 1, qwords_qty do
lanes_lo[j] = lanes_lo[j] ~ dwords_from_message[2*j-1]
lanes_hi[j] = lanes_hi[j] ~ dwords_from_message[2*j]
end
local L01_lo, L01_hi, L02_lo, L02_hi, L03_lo, L03_hi, L04_lo, L04_hi, L05_lo, L05_hi, L06_lo, L06_hi, L07_lo, L07_hi, L08_lo, L08_hi,
L09_lo, L09_hi, L10_lo, L10_hi, L11_lo, L11_hi, L12_lo, L12_hi, L13_lo, L13_hi, L14_lo, L14_hi, L15_lo, L15_hi, L16_lo, L16_hi,
L17_lo, L17_hi, L18_lo, L18_hi, L19_lo, L19_hi, L20_lo, L20_hi, L21_lo, L21_hi, L22_lo, L22_hi, L23_lo, L23_hi, L24_lo, L24_hi, L25_lo, L25_hi =
lanes_lo[1], lanes_hi[1], lanes_lo[2], lanes_hi[2], lanes_lo[3], lanes_hi[3], lanes_lo[4], lanes_hi[4], lanes_lo[5], lanes_hi[5],
lanes_lo[6], lanes_hi[6], lanes_lo[7], lanes_hi[7], lanes_lo[8], lanes_hi[8], lanes_lo[9], lanes_hi[9], lanes_lo[10], lanes_hi[10],
lanes_lo[11], lanes_hi[11], lanes_lo[12], lanes_hi[12], lanes_lo[13], lanes_hi[13], lanes_lo[14], lanes_hi[14], lanes_lo[15], lanes_hi[15],
lanes_lo[16], lanes_hi[16], lanes_lo[17], lanes_hi[17], lanes_lo[18], lanes_hi[18], lanes_lo[19], lanes_hi[19], lanes_lo[20], lanes_hi[20],
lanes_lo[21], lanes_hi[21], lanes_lo[22], lanes_hi[22], lanes_lo[23], lanes_hi[23], lanes_lo[24], lanes_hi[24], lanes_lo[25], lanes_hi[25]
for round_idx = 1, 24 do
local C1_lo = L01_lo ~ L06_lo ~ L11_lo ~ L16_lo ~ L21_lo
local C1_hi = L01_hi ~ L06_hi ~ L11_hi ~ L16_hi ~ L21_hi
local C2_lo = L02_lo ~ L07_lo ~ L12_lo ~ L17_lo ~ L22_lo
local C2_hi = L02_hi ~ L07_hi ~ L12_hi ~ L17_hi ~ L22_hi
local C3_lo = L03_lo ~ L08_lo ~ L13_lo ~ L18_lo ~ L23_lo
local C3_hi = L03_hi ~ L08_hi ~ L13_hi ~ L18_hi ~ L23_hi
local C4_lo = L04_lo ~ L09_lo ~ L14_lo ~ L19_lo ~ L24_lo
local C4_hi = L04_hi ~ L09_hi ~ L14_hi ~ L19_hi ~ L24_hi
local C5_lo = L05_lo ~ L10_lo ~ L15_lo ~ L20_lo ~ L25_lo
local C5_hi = L05_hi ~ L10_hi ~ L15_hi ~ L20_hi ~ L25_hi
local D_lo = C1_lo ~ C3_lo<<1 ~ C3_hi>>31
local D_hi = C1_hi ~ C3_hi<<1 ~ C3_lo>>31
local T0_lo = D_lo ~ L02_lo
local T0_hi = D_hi ~ L02_hi
local T1_lo = D_lo ~ L07_lo
local T1_hi = D_hi ~ L07_hi
local T2_lo = D_lo ~ L12_lo
local T2_hi = D_hi ~ L12_hi
local T3_lo = D_lo ~ L17_lo
local T3_hi = D_hi ~ L17_hi
local T4_lo = D_lo ~ L22_lo
local T4_hi = D_hi ~ L22_hi
L02_lo = T1_lo>>20 ~ T1_hi<<12
L02_hi = T1_hi>>20 ~ T1_lo<<12
L07_lo = T3_lo>>19 ~ T3_hi<<13
L07_hi = T3_hi>>19 ~ T3_lo<<13
L12_lo = T0_lo<<1 ~ T0_hi>>31
L12_hi = T0_hi<<1 ~ T0_lo>>31
L17_lo = T2_lo<<10 ~ T2_hi>>22
L17_hi = T2_hi<<10 ~ T2_lo>>22
L22_lo = T4_lo<<2 ~ T4_hi>>30
L22_hi = T4_hi<<2 ~ T4_lo>>30
D_lo = C2_lo ~ C4_lo<<1 ~ C4_hi>>31
D_hi = C2_hi ~ C4_hi<<1 ~ C4_lo>>31
T0_lo = D_lo ~ L03_lo
T0_hi = D_hi ~ L03_hi
T1_lo = D_lo ~ L08_lo
T1_hi = D_hi ~ L08_hi
T2_lo = D_lo ~ L13_lo
T2_hi = D_hi ~ L13_hi
T3_lo = D_lo ~ L18_lo
T3_hi = D_hi ~ L18_hi
T4_lo = D_lo ~ L23_lo
T4_hi = D_hi ~ L23_hi
L03_lo = T2_lo>>21 ~ T2_hi<<11
L03_hi = T2_hi>>21 ~ T2_lo<<11
L08_lo = T4_lo>>3 ~ T4_hi<<29
L08_hi = T4_hi>>3 ~ T4_lo<<29
L13_lo = T1_lo<<6 ~ T1_hi>>26
L13_hi = T1_hi<<6 ~ T1_lo>>26
L18_lo = T3_lo<<15 ~ T3_hi>>17
L18_hi = T3_hi<<15 ~ T3_lo>>17
L23_lo = T0_lo>>2 ~ T0_hi<<30
L23_hi = T0_hi>>2 ~ T0_lo<<30
D_lo = C3_lo ~ C5_lo<<1 ~ C5_hi>>31
D_hi = C3_hi ~ C5_hi<<1 ~ C5_lo>>31
T0_lo = D_lo ~ L04_lo
T0_hi = D_hi ~ L04_hi
T1_lo = D_lo ~ L09_lo
T1_hi = D_hi ~ L09_hi
T2_lo = D_lo ~ L14_lo
T2_hi = D_hi ~ L14_hi
T3_lo = D_lo ~ L19_lo
T3_hi = D_hi ~ L19_hi
T4_lo = D_lo ~ L24_lo
T4_hi = D_hi ~ L24_hi
L04_lo = T3_lo<<21 ~ T3_hi>>11
L04_hi = T3_hi<<21 ~ T3_lo>>11
L09_lo = T0_lo<<28 ~ T0_hi>>4
L09_hi = T0_hi<<28 ~ T0_lo>>4
L14_lo = T2_lo<<25 ~ T2_hi>>7
L14_hi = T2_hi<<25 ~ T2_lo>>7
L19_lo = T4_lo>>8 ~ T4_hi<<24
L19_hi = T4_hi>>8 ~ T4_lo<<24
L24_lo = T1_lo>>9 ~ T1_hi<<23
L24_hi = T1_hi>>9 ~ T1_lo<<23
D_lo = C4_lo ~ C1_lo<<1 ~ C1_hi>>31
D_hi = C4_hi ~ C1_hi<<1 ~ C1_lo>>31
T0_lo = D_lo ~ L05_lo
T0_hi = D_hi ~ L05_hi
T1_lo = D_lo ~ L10_lo
T1_hi = D_hi ~ L10_hi
T2_lo = D_lo ~ L15_lo
T2_hi = D_hi ~ L15_hi
T3_lo = D_lo ~ L20_lo
T3_hi = D_hi ~ L20_hi
T4_lo = D_lo ~ L25_lo
T4_hi = D_hi ~ L25_hi
L05_lo = T4_lo<<14 ~ T4_hi>>18
L05_hi = T4_hi<<14 ~ T4_lo>>18
L10_lo = T1_lo<<20 ~ T1_hi>>12
L10_hi = T1_hi<<20 ~ T1_lo>>12
L15_lo = T3_lo<<8 ~ T3_hi>>24
L15_hi = T3_hi<<8 ~ T3_lo>>24
L20_lo = T0_lo<<27 ~ T0_hi>>5
L20_hi = T0_hi<<27 ~ T0_lo>>5
L25_lo = T2_lo>>25 ~ T2_hi<<7
L25_hi = T2_hi>>25 ~ T2_lo<<7
D_lo = C5_lo ~ C2_lo<<1 ~ C2_hi>>31
D_hi = C5_hi ~ C2_hi<<1 ~ C2_lo>>31
T1_lo = D_lo ~ L06_lo
T1_hi = D_hi ~ L06_hi
T2_lo = D_lo ~ L11_lo
T2_hi = D_hi ~ L11_hi
T3_lo = D_lo ~ L16_lo
T3_hi = D_hi ~ L16_hi
T4_lo = D_lo ~ L21_lo
T4_hi = D_hi ~ L21_hi
L06_lo = T2_lo<<3 ~ T2_hi>>29
L06_hi = T2_hi<<3 ~ T2_lo>>29
L11_lo = T4_lo<<18 ~ T4_hi>>14
L11_hi = T4_hi<<18 ~ T4_lo>>14
L16_lo = T1_lo>>28 ~ T1_hi<<4
L16_hi = T1_hi>>28 ~ T1_lo<<4
L21_lo = T3_lo>>23 ~ T3_hi<<9
L21_hi = T3_hi>>23 ~ T3_lo<<9
L01_lo = D_lo ~ L01_lo
L01_hi = D_hi ~ L01_hi
L01_lo, L02_lo, L03_lo, L04_lo, L05_lo = L01_lo ~ ~L02_lo & L03_lo, L02_lo ~ ~L03_lo & L04_lo, L03_lo ~ ~L04_lo & L05_lo, L04_lo ~ ~L05_lo & L01_lo, L05_lo ~ ~L01_lo & L02_lo
L01_hi, L02_hi, L03_hi, L04_hi, L05_hi = L01_hi ~ ~L02_hi & L03_hi, L02_hi ~ ~L03_hi & L04_hi, L03_hi ~ ~L04_hi & L05_hi, L04_hi ~ ~L05_hi & L01_hi, L05_hi ~ ~L01_hi & L02_hi
L06_lo, L07_lo, L08_lo, L09_lo, L10_lo = L09_lo ~ ~L10_lo & L06_lo, L10_lo ~ ~L06_lo & L07_lo, L06_lo ~ ~L07_lo & L08_lo, L07_lo ~ ~L08_lo & L09_lo, L08_lo ~ ~L09_lo & L10_lo
L06_hi, L07_hi, L08_hi, L09_hi, L10_hi = L09_hi ~ ~L10_hi & L06_hi, L10_hi ~ ~L06_hi & L07_hi, L06_hi ~ ~L07_hi & L08_hi, L07_hi ~ ~L08_hi & L09_hi, L08_hi ~ ~L09_hi & L10_hi
L11_lo, L12_lo, L13_lo, L14_lo, L15_lo = L12_lo ~ ~L13_lo & L14_lo, L13_lo ~ ~L14_lo & L15_lo, L14_lo ~ ~L15_lo & L11_lo, L15_lo ~ ~L11_lo & L12_lo, L11_lo ~ ~L12_lo & L13_lo
L11_hi, L12_hi, L13_hi, L14_hi, L15_hi = L12_hi ~ ~L13_hi & L14_hi, L13_hi ~ ~L14_hi & L15_hi, L14_hi ~ ~L15_hi & L11_hi, L15_hi ~ ~L11_hi & L12_hi, L11_hi ~ ~L12_hi & L13_hi
L16_lo, L17_lo, L18_lo, L19_lo, L20_lo = L20_lo ~ ~L16_lo & L17_lo, L16_lo ~ ~L17_lo & L18_lo, L17_lo ~ ~L18_lo & L19_lo, L18_lo ~ ~L19_lo & L20_lo, L19_lo ~ ~L20_lo & L16_lo
L16_hi, L17_hi, L18_hi, L19_hi, L20_hi = L20_hi ~ ~L16_hi & L17_hi, L16_hi ~ ~L17_hi & L18_hi, L17_hi ~ ~L18_hi & L19_hi, L18_hi ~ ~L19_hi & L20_hi, L19_hi ~ ~L20_hi & L16_hi
L21_lo, L22_lo, L23_lo, L24_lo, L25_lo = L23_lo ~ ~L24_lo & L25_lo, L24_lo ~ ~L25_lo & L21_lo, L25_lo ~ ~L21_lo & L22_lo, L21_lo ~ ~L22_lo & L23_lo, L22_lo ~ ~L23_lo & L24_lo
L21_hi, L22_hi, L23_hi, L24_hi, L25_hi = L23_hi ~ ~L24_hi & L25_hi, L24_hi ~ ~L25_hi & L21_hi, L25_hi ~ ~L21_hi & L22_hi, L21_hi ~ ~L22_hi & L23_hi, L22_hi ~ ~L23_hi & L24_hi
L01_lo = L01_lo ~ RC_lo[round_idx]
L01_hi = L01_hi ~ RC_hi[round_idx]
end
lanes_lo[1] = L01_lo; lanes_hi[1] = L01_hi
lanes_lo[2] = L02_lo; lanes_hi[2] = L02_hi
lanes_lo[3] = L03_lo; lanes_hi[3] = L03_hi
lanes_lo[4] = L04_lo; lanes_hi[4] = L04_hi
lanes_lo[5] = L05_lo; lanes_hi[5] = L05_hi
lanes_lo[6] = L06_lo; lanes_hi[6] = L06_hi
lanes_lo[7] = L07_lo; lanes_hi[7] = L07_hi
lanes_lo[8] = L08_lo; lanes_hi[8] = L08_hi
lanes_lo[9] = L09_lo; lanes_hi[9] = L09_hi
lanes_lo[10] = L10_lo; lanes_hi[10] = L10_hi
lanes_lo[11] = L11_lo; lanes_hi[11] = L11_hi
lanes_lo[12] = L12_lo; lanes_hi[12] = L12_hi
lanes_lo[13] = L13_lo; lanes_hi[13] = L13_hi
lanes_lo[14] = L14_lo; lanes_hi[14] = L14_hi
lanes_lo[15] = L15_lo; lanes_hi[15] = L15_hi
lanes_lo[16] = L16_lo; lanes_hi[16] = L16_hi
lanes_lo[17] = L17_lo; lanes_hi[17] = L17_hi
lanes_lo[18] = L18_lo; lanes_hi[18] = L18_hi
lanes_lo[19] = L19_lo; lanes_hi[19] = L19_hi
lanes_lo[20] = L20_lo; lanes_hi[20] = L20_hi
lanes_lo[21] = L21_lo; lanes_hi[21] = L21_hi
lanes_lo[22] = L22_lo; lanes_hi[22] = L22_hi
lanes_lo[23] = L23_lo; lanes_hi[23] = L23_hi
lanes_lo[24] = L24_lo; lanes_hi[24] = L24_hi
lanes_lo[25] = L25_lo; lanes_hi[25] = L25_hi
end
end
local function blake2s_feed_64(H, str, offs, size, bytes_compressed, last_block_size, is_last_node)
-- offs >= 0, size >= 0, size is multiple of 64
local W = common_W
local h1, h2, h3, h4, h5, h6, h7, h8 = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for pos = offs + 1, offs + size, 64 do
if str then
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack("> 16 | vC << 16
v8 = v8 + vC
v4 = v4 ~ v8
v4 = v4 >> 12 | v4 << 20
v0 = v0 + v4 + W[row[2]]
vC = vC ~ v0
vC = vC >> 8 | vC << 24
v8 = v8 + vC
v4 = v4 ~ v8
v4 = v4 >> 7 | v4 << 25
v1 = v1 + v5 + W[row[3]]
vD = vD ~ v1
vD = vD >> 16 | vD << 16
v9 = v9 + vD
v5 = v5 ~ v9
v5 = v5 >> 12 | v5 << 20
v1 = v1 + v5 + W[row[4]]
vD = vD ~ v1
vD = vD >> 8 | vD << 24
v9 = v9 + vD
v5 = v5 ~ v9
v5 = v5 >> 7 | v5 << 25
v2 = v2 + v6 + W[row[5]]
vE = vE ~ v2
vE = vE >> 16 | vE << 16
vA = vA + vE
v6 = v6 ~ vA
v6 = v6 >> 12 | v6 << 20
v2 = v2 + v6 + W[row[6]]
vE = vE ~ v2
vE = vE >> 8 | vE << 24
vA = vA + vE
v6 = v6 ~ vA
v6 = v6 >> 7 | v6 << 25
v3 = v3 + v7 + W[row[7]]
vF = vF ~ v3
vF = vF >> 16 | vF << 16
vB = vB + vF
v7 = v7 ~ vB
v7 = v7 >> 12 | v7 << 20
v3 = v3 + v7 + W[row[8]]
vF = vF ~ v3
vF = vF >> 8 | vF << 24
vB = vB + vF
v7 = v7 ~ vB
v7 = v7 >> 7 | v7 << 25
v0 = v0 + v5 + W[row[9]]
vF = vF ~ v0
vF = vF >> 16 | vF << 16
vA = vA + vF
v5 = v5 ~ vA
v5 = v5 >> 12 | v5 << 20
v0 = v0 + v5 + W[row[10]]
vF = vF ~ v0
vF = vF >> 8 | vF << 24
vA = vA + vF
v5 = v5 ~ vA
v5 = v5 >> 7 | v5 << 25
v1 = v1 + v6 + W[row[11]]
vC = vC ~ v1
vC = vC >> 16 | vC << 16
vB = vB + vC
v6 = v6 ~ vB
v6 = v6 >> 12 | v6 << 20
v1 = v1 + v6 + W[row[12]]
vC = vC ~ v1
vC = vC >> 8 | vC << 24
vB = vB + vC
v6 = v6 ~ vB
v6 = v6 >> 7 | v6 << 25
v2 = v2 + v7 + W[row[13]]
vD = vD ~ v2
vD = vD >> 16 | vD << 16
v8 = v8 + vD
v7 = v7 ~ v8
v7 = v7 >> 12 | v7 << 20
v2 = v2 + v7 + W[row[14]]
vD = vD ~ v2
vD = vD >> 8 | vD << 24
v8 = v8 + vD
v7 = v7 ~ v8
v7 = v7 >> 7 | v7 << 25
v3 = v3 + v4 + W[row[15]]
vE = vE ~ v3
vE = vE >> 16 | vE << 16
v9 = v9 + vE
v4 = v4 ~ v9
v4 = v4 >> 12 | v4 << 20
v3 = v3 + v4 + W[row[16]]
vE = vE ~ v3
vE = vE >> 8 | vE << 24
v9 = v9 + vE
v4 = v4 ~ v9
v4 = v4 >> 7 | v4 << 25
end
h1 = h1 ~ v0 ~ v8
h2 = h2 ~ v1 ~ v9
h3 = h3 ~ v2 ~ vA
h4 = h4 ~ v3 ~ vB
h5 = h5 ~ v4 ~ vC
h6 = h6 ~ v5 ~ vD
h7 = h7 ~ v6 ~ vE
h8 = h8 ~ v7 ~ vF
end
H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8] = h1, h2, h3, h4, h5, h6, h7, h8
return bytes_compressed
end
local function blake2b_feed_128(H_lo, H_hi, str, offs, size, bytes_compressed, last_block_size, is_last_node)
-- offs >= 0, size >= 0, size is multiple of 128
local W = common_W
local h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo = H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8]
local h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi = H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8]
for pos = offs + 1, offs + size, 128 do
if str then
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16],
W[17], W[18], W[19], W[20], W[21], W[22], W[23], W[24], W[25], W[26], W[27], W[28], W[29], W[30], W[31], W[32] =
string_unpack("> 24 | v4_hi << 8, v4_hi >> 24 | v4_lo << 8
k = row[2] * 2
v0_lo = v0_lo % 2^32 + v4_lo % 2^32 + W[k-1] % 2^32
v0_hi = v0_hi + v4_hi + floor(v0_lo / 2^32) + W[k]
v0_lo = 0|((v0_lo + 2^31) % 2^32 - 2^31)
vC_lo, vC_hi = vC_lo ~ v0_lo, vC_hi ~ v0_hi
vC_lo, vC_hi = vC_lo >> 16 | vC_hi << 16, vC_hi >> 16 | vC_lo << 16
v8_lo = v8_lo % 2^32 + vC_lo % 2^32
v8_hi = v8_hi + vC_hi + floor(v8_lo / 2^32)
v8_lo = 0|((v8_lo + 2^31) % 2^32 - 2^31)
v4_lo, v4_hi = v4_lo ~ v8_lo, v4_hi ~ v8_hi
v4_lo, v4_hi = v4_lo << 1 | v4_hi >> 31, v4_hi << 1 | v4_lo >> 31
k = row[3] * 2
v1_lo = v1_lo % 2^32 + v5_lo % 2^32 + W[k-1] % 2^32
v1_hi = v1_hi + v5_hi + floor(v1_lo / 2^32) + W[k]
v1_lo = 0|((v1_lo + 2^31) % 2^32 - 2^31)
vD_lo, vD_hi = vD_hi ~ v1_hi, vD_lo ~ v1_lo
v9_lo = v9_lo % 2^32 + vD_lo % 2^32
v9_hi = v9_hi + vD_hi + floor(v9_lo / 2^32)
v9_lo = 0|((v9_lo + 2^31) % 2^32 - 2^31)
v5_lo, v5_hi = v5_lo ~ v9_lo, v5_hi ~ v9_hi
v5_lo, v5_hi = v5_lo >> 24 | v5_hi << 8, v5_hi >> 24 | v5_lo << 8
k = row[4] * 2
v1_lo = v1_lo % 2^32 + v5_lo % 2^32 + W[k-1] % 2^32
v1_hi = v1_hi + v5_hi + floor(v1_lo / 2^32) + W[k]
v1_lo = 0|((v1_lo + 2^31) % 2^32 - 2^31)
vD_lo, vD_hi = vD_lo ~ v1_lo, vD_hi ~ v1_hi
vD_lo, vD_hi = vD_lo >> 16 | vD_hi << 16, vD_hi >> 16 | vD_lo << 16
v9_lo = v9_lo % 2^32 + vD_lo % 2^32
v9_hi = v9_hi + vD_hi + floor(v9_lo / 2^32)
v9_lo = 0|((v9_lo + 2^31) % 2^32 - 2^31)
v5_lo, v5_hi = v5_lo ~ v9_lo, v5_hi ~ v9_hi
v5_lo, v5_hi = v5_lo << 1 | v5_hi >> 31, v5_hi << 1 | v5_lo >> 31
k = row[5] * 2
v2_lo = v2_lo % 2^32 + v6_lo % 2^32 + W[k-1] % 2^32
v2_hi = v2_hi + v6_hi + floor(v2_lo / 2^32) + W[k]
v2_lo = 0|((v2_lo + 2^31) % 2^32 - 2^31)
vE_lo, vE_hi = vE_hi ~ v2_hi, vE_lo ~ v2_lo
vA_lo = vA_lo % 2^32 + vE_lo % 2^32
vA_hi = vA_hi + vE_hi + floor(vA_lo / 2^32)
vA_lo = 0|((vA_lo + 2^31) % 2^32 - 2^31)
v6_lo, v6_hi = v6_lo ~ vA_lo, v6_hi ~ vA_hi
v6_lo, v6_hi = v6_lo >> 24 | v6_hi << 8, v6_hi >> 24 | v6_lo << 8
k = row[6] * 2
v2_lo = v2_lo % 2^32 + v6_lo % 2^32 + W[k-1] % 2^32
v2_hi = v2_hi + v6_hi + floor(v2_lo / 2^32) + W[k]
v2_lo = 0|((v2_lo + 2^31) % 2^32 - 2^31)
vE_lo, vE_hi = vE_lo ~ v2_lo, vE_hi ~ v2_hi
vE_lo, vE_hi = vE_lo >> 16 | vE_hi << 16, vE_hi >> 16 | vE_lo << 16
vA_lo = vA_lo % 2^32 + vE_lo % 2^32
vA_hi = vA_hi + vE_hi + floor(vA_lo / 2^32)
vA_lo = 0|((vA_lo + 2^31) % 2^32 - 2^31)
v6_lo, v6_hi = v6_lo ~ vA_lo, v6_hi ~ vA_hi
v6_lo, v6_hi = v6_lo << 1 | v6_hi >> 31, v6_hi << 1 | v6_lo >> 31
k = row[7] * 2
v3_lo = v3_lo % 2^32 + v7_lo % 2^32 + W[k-1] % 2^32
v3_hi = v3_hi + v7_hi + floor(v3_lo / 2^32) + W[k]
v3_lo = 0|((v3_lo + 2^31) % 2^32 - 2^31)
vF_lo, vF_hi = vF_hi ~ v3_hi, vF_lo ~ v3_lo
vB_lo = vB_lo % 2^32 + vF_lo % 2^32
vB_hi = vB_hi + vF_hi + floor(vB_lo / 2^32)
vB_lo = 0|((vB_lo + 2^31) % 2^32 - 2^31)
v7_lo, v7_hi = v7_lo ~ vB_lo, v7_hi ~ vB_hi
v7_lo, v7_hi = v7_lo >> 24 | v7_hi << 8, v7_hi >> 24 | v7_lo << 8
k = row[8] * 2
v3_lo = v3_lo % 2^32 + v7_lo % 2^32 + W[k-1] % 2^32
v3_hi = v3_hi + v7_hi + floor(v3_lo / 2^32) + W[k]
v3_lo = 0|((v3_lo + 2^31) % 2^32 - 2^31)
vF_lo, vF_hi = vF_lo ~ v3_lo, vF_hi ~ v3_hi
vF_lo, vF_hi = vF_lo >> 16 | vF_hi << 16, vF_hi >> 16 | vF_lo << 16
vB_lo = vB_lo % 2^32 + vF_lo % 2^32
vB_hi = vB_hi + vF_hi + floor(vB_lo / 2^32)
vB_lo = 0|((vB_lo + 2^31) % 2^32 - 2^31)
v7_lo, v7_hi = v7_lo ~ vB_lo, v7_hi ~ vB_hi
v7_lo, v7_hi = v7_lo << 1 | v7_hi >> 31, v7_hi << 1 | v7_lo >> 31
k = row[9] * 2
v0_lo = v0_lo % 2^32 + v5_lo % 2^32 + W[k-1] % 2^32
v0_hi = v0_hi + v5_hi + floor(v0_lo / 2^32) + W[k]
v0_lo = 0|((v0_lo + 2^31) % 2^32 - 2^31)
vF_lo, vF_hi = vF_hi ~ v0_hi, vF_lo ~ v0_lo
vA_lo = vA_lo % 2^32 + vF_lo % 2^32
vA_hi = vA_hi + vF_hi + floor(vA_lo / 2^32)
vA_lo = 0|((vA_lo + 2^31) % 2^32 - 2^31)
v5_lo, v5_hi = v5_lo ~ vA_lo, v5_hi ~ vA_hi
v5_lo, v5_hi = v5_lo >> 24 | v5_hi << 8, v5_hi >> 24 | v5_lo << 8
k = row[10] * 2
v0_lo = v0_lo % 2^32 + v5_lo % 2^32 + W[k-1] % 2^32
v0_hi = v0_hi + v5_hi + floor(v0_lo / 2^32) + W[k]
v0_lo = 0|((v0_lo + 2^31) % 2^32 - 2^31)
vF_lo, vF_hi = vF_lo ~ v0_lo, vF_hi ~ v0_hi
vF_lo, vF_hi = vF_lo >> 16 | vF_hi << 16, vF_hi >> 16 | vF_lo << 16
vA_lo = vA_lo % 2^32 + vF_lo % 2^32
vA_hi = vA_hi + vF_hi + floor(vA_lo / 2^32)
vA_lo = 0|((vA_lo + 2^31) % 2^32 - 2^31)
v5_lo, v5_hi = v5_lo ~ vA_lo, v5_hi ~ vA_hi
v5_lo, v5_hi = v5_lo << 1 | v5_hi >> 31, v5_hi << 1 | v5_lo >> 31
k = row[11] * 2
v1_lo = v1_lo % 2^32 + v6_lo % 2^32 + W[k-1] % 2^32
v1_hi = v1_hi + v6_hi + floor(v1_lo / 2^32) + W[k]
v1_lo = 0|((v1_lo + 2^31) % 2^32 - 2^31)
vC_lo, vC_hi = vC_hi ~ v1_hi, vC_lo ~ v1_lo
vB_lo = vB_lo % 2^32 + vC_lo % 2^32
vB_hi = vB_hi + vC_hi + floor(vB_lo / 2^32)
vB_lo = 0|((vB_lo + 2^31) % 2^32 - 2^31)
v6_lo, v6_hi = v6_lo ~ vB_lo, v6_hi ~ vB_hi
v6_lo, v6_hi = v6_lo >> 24 | v6_hi << 8, v6_hi >> 24 | v6_lo << 8
k = row[12] * 2
v1_lo = v1_lo % 2^32 + v6_lo % 2^32 + W[k-1] % 2^32
v1_hi = v1_hi + v6_hi + floor(v1_lo / 2^32) + W[k]
v1_lo = 0|((v1_lo + 2^31) % 2^32 - 2^31)
vC_lo, vC_hi = vC_lo ~ v1_lo, vC_hi ~ v1_hi
vC_lo, vC_hi = vC_lo >> 16 | vC_hi << 16, vC_hi >> 16 | vC_lo << 16
vB_lo = vB_lo % 2^32 + vC_lo % 2^32
vB_hi = vB_hi + vC_hi + floor(vB_lo / 2^32)
vB_lo = 0|((vB_lo + 2^31) % 2^32 - 2^31)
v6_lo, v6_hi = v6_lo ~ vB_lo, v6_hi ~ vB_hi
v6_lo, v6_hi = v6_lo << 1 | v6_hi >> 31, v6_hi << 1 | v6_lo >> 31
k = row[13] * 2
v2_lo = v2_lo % 2^32 + v7_lo % 2^32 + W[k-1] % 2^32
v2_hi = v2_hi + v7_hi + floor(v2_lo / 2^32) + W[k]
v2_lo = 0|((v2_lo + 2^31) % 2^32 - 2^31)
vD_lo, vD_hi = vD_hi ~ v2_hi, vD_lo ~ v2_lo
v8_lo = v8_lo % 2^32 + vD_lo % 2^32
v8_hi = v8_hi + vD_hi + floor(v8_lo / 2^32)
v8_lo = 0|((v8_lo + 2^31) % 2^32 - 2^31)
v7_lo, v7_hi = v7_lo ~ v8_lo, v7_hi ~ v8_hi
v7_lo, v7_hi = v7_lo >> 24 | v7_hi << 8, v7_hi >> 24 | v7_lo << 8
k = row[14] * 2
v2_lo = v2_lo % 2^32 + v7_lo % 2^32 + W[k-1] % 2^32
v2_hi = v2_hi + v7_hi + floor(v2_lo / 2^32) + W[k]
v2_lo = 0|((v2_lo + 2^31) % 2^32 - 2^31)
vD_lo, vD_hi = vD_lo ~ v2_lo, vD_hi ~ v2_hi
vD_lo, vD_hi = vD_lo >> 16 | vD_hi << 16, vD_hi >> 16 | vD_lo << 16
v8_lo = v8_lo % 2^32 + vD_lo % 2^32
v8_hi = v8_hi + vD_hi + floor(v8_lo / 2^32)
v8_lo = 0|((v8_lo + 2^31) % 2^32 - 2^31)
v7_lo, v7_hi = v7_lo ~ v8_lo, v7_hi ~ v8_hi
v7_lo, v7_hi = v7_lo << 1 | v7_hi >> 31, v7_hi << 1 | v7_lo >> 31
k = row[15] * 2
v3_lo = v3_lo % 2^32 + v4_lo % 2^32 + W[k-1] % 2^32
v3_hi = v3_hi + v4_hi + floor(v3_lo / 2^32) + W[k]
v3_lo = 0|((v3_lo + 2^31) % 2^32 - 2^31)
vE_lo, vE_hi = vE_hi ~ v3_hi, vE_lo ~ v3_lo
v9_lo = v9_lo % 2^32 + vE_lo % 2^32
v9_hi = v9_hi + vE_hi + floor(v9_lo / 2^32)
v9_lo = 0|((v9_lo + 2^31) % 2^32 - 2^31)
v4_lo, v4_hi = v4_lo ~ v9_lo, v4_hi ~ v9_hi
v4_lo, v4_hi = v4_lo >> 24 | v4_hi << 8, v4_hi >> 24 | v4_lo << 8
k = row[16] * 2
v3_lo = v3_lo % 2^32 + v4_lo % 2^32 + W[k-1] % 2^32
v3_hi = v3_hi + v4_hi + floor(v3_lo / 2^32) + W[k]
v3_lo = 0|((v3_lo + 2^31) % 2^32 - 2^31)
vE_lo, vE_hi = vE_lo ~ v3_lo, vE_hi ~ v3_hi
vE_lo, vE_hi = vE_lo >> 16 | vE_hi << 16, vE_hi >> 16 | vE_lo << 16
v9_lo = v9_lo % 2^32 + vE_lo % 2^32
v9_hi = v9_hi + vE_hi + floor(v9_lo / 2^32)
v9_lo = 0|((v9_lo + 2^31) % 2^32 - 2^31)
v4_lo, v4_hi = v4_lo ~ v9_lo, v4_hi ~ v9_hi
v4_lo, v4_hi = v4_lo << 1 | v4_hi >> 31, v4_hi << 1 | v4_lo >> 31
end
h1_lo = h1_lo ~ v0_lo ~ v8_lo
h2_lo = h2_lo ~ v1_lo ~ v9_lo
h3_lo = h3_lo ~ v2_lo ~ vA_lo
h4_lo = h4_lo ~ v3_lo ~ vB_lo
h5_lo = h5_lo ~ v4_lo ~ vC_lo
h6_lo = h6_lo ~ v5_lo ~ vD_lo
h7_lo = h7_lo ~ v6_lo ~ vE_lo
h8_lo = h8_lo ~ v7_lo ~ vF_lo
h1_hi = h1_hi ~ v0_hi ~ v8_hi
h2_hi = h2_hi ~ v1_hi ~ v9_hi
h3_hi = h3_hi ~ v2_hi ~ vA_hi
h4_hi = h4_hi ~ v3_hi ~ vB_hi
h5_hi = h5_hi ~ v4_hi ~ vC_hi
h6_hi = h6_hi ~ v5_hi ~ vD_hi
h7_hi = h7_hi ~ v6_hi ~ vE_hi
h8_hi = h8_hi ~ v7_hi ~ vF_hi
end
H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8] = h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo
H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8] = h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi
return bytes_compressed
end
local function blake3_feed_64(str, offs, size, flags, chunk_index, H_in, H_out, wide_output, block_length)
-- offs >= 0, size >= 0, size is multiple of 64
block_length = block_length or 64
local W = common_W
local h1, h2, h3, h4, h5, h6, h7, h8 = H_in[1], H_in[2], H_in[3], H_in[4], H_in[5], H_in[6], H_in[7], H_in[8]
H_out = H_out or H_in
for pos = offs + 1, offs + size, 64 do
if str then
W[1], W[2], W[3], W[4], W[5], W[6], W[7], W[8], W[9], W[10], W[11], W[12], W[13], W[14], W[15], W[16] =
string_unpack("> 16 | vC << 16
v8 = v8 + vC
v4 = v4 ~ v8
v4 = v4 >> 12 | v4 << 20
v0 = v0 + v4 + W[perm_blake3[j + 14]]
vC = vC ~ v0
vC = vC >> 8 | vC << 24
v8 = v8 + vC
v4 = v4 ~ v8
v4 = v4 >> 7 | v4 << 25
v1 = v1 + v5 + W[perm_blake3[j + 1]]
vD = vD ~ v1
vD = vD >> 16 | vD << 16
v9 = v9 + vD
v5 = v5 ~ v9
v5 = v5 >> 12 | v5 << 20
v1 = v1 + v5 + W[perm_blake3[j + 2]]
vD = vD ~ v1
vD = vD >> 8 | vD << 24
v9 = v9 + vD
v5 = v5 ~ v9
v5 = v5 >> 7 | v5 << 25
v2 = v2 + v6 + W[perm_blake3[j + 16]]
vE = vE ~ v2
vE = vE >> 16 | vE << 16
vA = vA + vE
v6 = v6 ~ vA
v6 = v6 >> 12 | v6 << 20
v2 = v2 + v6 + W[perm_blake3[j + 7]]
vE = vE ~ v2
vE = vE >> 8 | vE << 24
vA = vA + vE
v6 = v6 ~ vA
v6 = v6 >> 7 | v6 << 25
v3 = v3 + v7 + W[perm_blake3[j + 15]]
vF = vF ~ v3
vF = vF >> 16 | vF << 16
vB = vB + vF
v7 = v7 ~ vB
v7 = v7 >> 12 | v7 << 20
v3 = v3 + v7 + W[perm_blake3[j + 17]]
vF = vF ~ v3
vF = vF >> 8 | vF << 24
vB = vB + vF
v7 = v7 ~ vB
v7 = v7 >> 7 | v7 << 25
v0 = v0 + v5 + W[perm_blake3[j + 21]]
vF = vF ~ v0
vF = vF >> 16 | vF << 16
vA = vA + vF
v5 = v5 ~ vA
v5 = v5 >> 12 | v5 << 20
v0 = v0 + v5 + W[perm_blake3[j + 5]]
vF = vF ~ v0
vF = vF >> 8 | vF << 24
vA = vA + vF
v5 = v5 ~ vA
v5 = v5 >> 7 | v5 << 25
v1 = v1 + v6 + W[perm_blake3[j + 3]]
vC = vC ~ v1
vC = vC >> 16 | vC << 16
vB = vB + vC
v6 = v6 ~ vB
v6 = v6 >> 12 | v6 << 20
v1 = v1 + v6 + W[perm_blake3[j + 6]]
vC = vC ~ v1
vC = vC >> 8 | vC << 24
vB = vB + vC
v6 = v6 ~ vB
v6 = v6 >> 7 | v6 << 25
v2 = v2 + v7 + W[perm_blake3[j + 4]]
vD = vD ~ v2
vD = vD >> 16 | vD << 16
v8 = v8 + vD
v7 = v7 ~ v8
v7 = v7 >> 12 | v7 << 20
v2 = v2 + v7 + W[perm_blake3[j + 18]]
vD = vD ~ v2
vD = vD >> 8 | vD << 24
v8 = v8 + vD
v7 = v7 ~ v8
v7 = v7 >> 7 | v7 << 25
v3 = v3 + v4 + W[perm_blake3[j + 19]]
vE = vE ~ v3
vE = vE >> 16 | vE << 16
v9 = v9 + vE
v4 = v4 ~ v9
v4 = v4 >> 12 | v4 << 20
v3 = v3 + v4 + W[perm_blake3[j + 20]]
vE = vE ~ v3
vE = vE >> 8 | vE << 24
v9 = v9 + vE
v4 = v4 ~ v9
v4 = v4 >> 7 | v4 << 25
end
if wide_output then
H_out[ 9] = h1 ~ v8
H_out[10] = h2 ~ v9
H_out[11] = h3 ~ vA
H_out[12] = h4 ~ vB
H_out[13] = h5 ~ vC
H_out[14] = h6 ~ vD
H_out[15] = h7 ~ vE
H_out[16] = h8 ~ vF
end
h1 = v0 ~ v8
h2 = v1 ~ v9
h3 = v2 ~ vA
h4 = v3 ~ vB
h5 = v4 ~ vC
h6 = v5 ~ vD
h7 = v6 ~ vE
h8 = v7 ~ vF
end
H_out[1], H_out[2], H_out[3], H_out[4], H_out[5], H_out[6], H_out[7], H_out[8] = h1, h2, h3, h4, h5, h6, h7, h8
end
return XORA5, XOR_BYTE, sha256_feed_64, sha512_feed_128, md5_feed_64, sha1_feed_64, keccak_feed, blake2s_feed_64, blake2b_feed_128, blake3_feed_64
]=](md5_next_shift, md5_K, sha2_K_lo, sha2_K_hi, build_keccak_format, sha3_RC_lo, sha3_RC_hi, sigma, common_W, sha2_H_lo, sha2_H_hi, perm_blake3)
end
XOR = XOR or XORA5
if branch == "LIB32" or branch == "EMUL" then
-- implementation for Lua 5.1/5.2 (with or without bitwise library available)
function sha256_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W, K = common_W, sha2_K_hi
local h1, h2, h3, h4, h5, h6, h7, h8 = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for pos = offs, offs + size - 1, 64 do
for j = 1, 16 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = ((a * 256 + b) * 256 + c) * 256 + d
end
for j = 17, 64 do
local a, b = W[j-15], W[j-2]
local a7, a18, b17, b19 = a / 2^7, a / 2^18, b / 2^17, b / 2^19
W[j] = (XOR(a7 % 1 * (2^32 - 1) + a7, a18 % 1 * (2^32 - 1) + a18, (a - a % 2^3) / 2^3) + W[j-16] + W[j-7]
+ XOR(b17 % 1 * (2^32 - 1) + b17, b19 % 1 * (2^32 - 1) + b19, (b - b % 2^10) / 2^10)) % 2^32
end
local a, b, c, d, e, f, g, h = h1, h2, h3, h4, h5, h6, h7, h8
for j = 1, 64 do
e = e % 2^32
local e6, e11, e7 = e / 2^6, e / 2^11, e * 2^7
local e7_lo = e7 % 2^32
local z = AND(e, f) + AND(-1-e, g) + h + K[j] + W[j]
+ XOR(e6 % 1 * (2^32 - 1) + e6, e11 % 1 * (2^32 - 1) + e11, e7_lo + (e7 - e7_lo) / 2^32)
h = g
g = f
f = e
e = z + d
d = c
c = b
b = a % 2^32
local b2, b13, b10 = b / 2^2, b / 2^13, b * 2^10
local b10_lo = b10 % 2^32
a = z + AND(d, c) + AND(b, XOR(d, c)) +
XOR(b2 % 1 * (2^32 - 1) + b2, b13 % 1 * (2^32 - 1) + b13, b10_lo + (b10 - b10_lo) / 2^32)
end
h1, h2, h3, h4 = (a + h1) % 2^32, (b + h2) % 2^32, (c + h3) % 2^32, (d + h4) % 2^32
h5, h6, h7, h8 = (e + h5) % 2^32, (f + h6) % 2^32, (g + h7) % 2^32, (h + h8) % 2^32
end
H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8] = h1, h2, h3, h4, h5, h6, h7, h8
end
function sha512_feed_128(H_lo, H_hi, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 128
-- W1_hi, W1_lo, W2_hi, W2_lo, ... Wk_hi = W[2*k-1], Wk_lo = W[2*k]
local W, K_lo, K_hi = common_W, sha2_K_lo, sha2_K_hi
local h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo = H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8]
local h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi = H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8]
for pos = offs, offs + size - 1, 128 do
for j = 1, 16*2 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = ((a * 256 + b) * 256 + c) * 256 + d
end
for jj = 17*2, 80*2, 2 do
local a_hi, a_lo, b_hi, b_lo = W[jj-31], W[jj-30], W[jj-5], W[jj-4]
local b_hi_6, b_hi_19, b_hi_29, b_lo_19, b_lo_29, a_hi_1, a_hi_7, a_hi_8, a_lo_1, a_lo_8 =
b_hi % 2^6, b_hi % 2^19, b_hi % 2^29, b_lo % 2^19, b_lo % 2^29, a_hi % 2^1, a_hi % 2^7, a_hi % 2^8, a_lo % 2^1, a_lo % 2^8
local tmp1 = XOR((a_lo - a_lo_1) / 2^1 + a_hi_1 * 2^31, (a_lo - a_lo_8) / 2^8 + a_hi_8 * 2^24, (a_lo - a_lo % 2^7) / 2^7 + a_hi_7 * 2^25) % 2^32
+ XOR((b_lo - b_lo_19) / 2^19 + b_hi_19 * 2^13, b_lo_29 * 2^3 + (b_hi - b_hi_29) / 2^29, (b_lo - b_lo % 2^6) / 2^6 + b_hi_6 * 2^26) % 2^32
+ W[jj-14] + W[jj-32]
local tmp2 = tmp1 % 2^32
W[jj-1] = (XOR((a_hi - a_hi_1) / 2^1 + a_lo_1 * 2^31, (a_hi - a_hi_8) / 2^8 + a_lo_8 * 2^24, (a_hi - a_hi_7) / 2^7)
+ XOR((b_hi - b_hi_19) / 2^19 + b_lo_19 * 2^13, b_hi_29 * 2^3 + (b_lo - b_lo_29) / 2^29, (b_hi - b_hi_6) / 2^6)
+ W[jj-15] + W[jj-33] + (tmp1 - tmp2) / 2^32) % 2^32
W[jj] = tmp2
end
local a_lo, b_lo, c_lo, d_lo, e_lo, f_lo, g_lo, h_lo = h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo
local a_hi, b_hi, c_hi, d_hi, e_hi, f_hi, g_hi, h_hi = h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi
for j = 1, 80 do
local jj = 2*j
local e_lo_9, e_lo_14, e_lo_18, e_hi_9, e_hi_14, e_hi_18 = e_lo % 2^9, e_lo % 2^14, e_lo % 2^18, e_hi % 2^9, e_hi % 2^14, e_hi % 2^18
local tmp1 = (AND(e_lo, f_lo) + AND(-1-e_lo, g_lo)) % 2^32 + h_lo + K_lo[j] + W[jj]
+ XOR((e_lo - e_lo_14) / 2^14 + e_hi_14 * 2^18, (e_lo - e_lo_18) / 2^18 + e_hi_18 * 2^14, e_lo_9 * 2^23 + (e_hi - e_hi_9) / 2^9) % 2^32
local z_lo = tmp1 % 2^32
local z_hi = AND(e_hi, f_hi) + AND(-1-e_hi, g_hi) + h_hi + K_hi[j] + W[jj-1] + (tmp1 - z_lo) / 2^32
+ XOR((e_hi - e_hi_14) / 2^14 + e_lo_14 * 2^18, (e_hi - e_hi_18) / 2^18 + e_lo_18 * 2^14, e_hi_9 * 2^23 + (e_lo - e_lo_9) / 2^9)
h_lo = g_lo; h_hi = g_hi
g_lo = f_lo; g_hi = f_hi
f_lo = e_lo; f_hi = e_hi
tmp1 = z_lo + d_lo
e_lo = tmp1 % 2^32
e_hi = (z_hi + d_hi + (tmp1 - e_lo) / 2^32) % 2^32
d_lo = c_lo; d_hi = c_hi
c_lo = b_lo; c_hi = b_hi
b_lo = a_lo; b_hi = a_hi
local b_lo_2, b_lo_7, b_lo_28, b_hi_2, b_hi_7, b_hi_28 = b_lo % 2^2, b_lo % 2^7, b_lo % 2^28, b_hi % 2^2, b_hi % 2^7, b_hi % 2^28
tmp1 = z_lo + (AND(d_lo, c_lo) + AND(b_lo, XOR(d_lo, c_lo))) % 2^32
+ XOR((b_lo - b_lo_28) / 2^28 + b_hi_28 * 2^4, b_lo_2 * 2^30 + (b_hi - b_hi_2) / 2^2, b_lo_7 * 2^25 + (b_hi - b_hi_7) / 2^7) % 2^32
a_lo = tmp1 % 2^32
a_hi = (z_hi + AND(d_hi, c_hi) + AND(b_hi, XOR(d_hi, c_hi)) + (tmp1 - a_lo) / 2^32
+ XOR((b_hi - b_hi_28) / 2^28 + b_lo_28 * 2^4, b_hi_2 * 2^30 + (b_lo - b_lo_2) / 2^2, b_hi_7 * 2^25 + (b_lo - b_lo_7) / 2^7)) % 2^32
end
a_lo = h1_lo + a_lo
h1_lo = a_lo % 2^32
h1_hi = (h1_hi + a_hi + (a_lo - h1_lo) / 2^32) % 2^32
a_lo = h2_lo + b_lo
h2_lo = a_lo % 2^32
h2_hi = (h2_hi + b_hi + (a_lo - h2_lo) / 2^32) % 2^32
a_lo = h3_lo + c_lo
h3_lo = a_lo % 2^32
h3_hi = (h3_hi + c_hi + (a_lo - h3_lo) / 2^32) % 2^32
a_lo = h4_lo + d_lo
h4_lo = a_lo % 2^32
h4_hi = (h4_hi + d_hi + (a_lo - h4_lo) / 2^32) % 2^32
a_lo = h5_lo + e_lo
h5_lo = a_lo % 2^32
h5_hi = (h5_hi + e_hi + (a_lo - h5_lo) / 2^32) % 2^32
a_lo = h6_lo + f_lo
h6_lo = a_lo % 2^32
h6_hi = (h6_hi + f_hi + (a_lo - h6_lo) / 2^32) % 2^32
a_lo = h7_lo + g_lo
h7_lo = a_lo % 2^32
h7_hi = (h7_hi + g_hi + (a_lo - h7_lo) / 2^32) % 2^32
a_lo = h8_lo + h_lo
h8_lo = a_lo % 2^32
h8_hi = (h8_hi + h_hi + (a_lo - h8_lo) / 2^32) % 2^32
end
H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8] = h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo
H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8] = h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi
end
if branch == "LIB32" then
function md5_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W, K, md5_next_shift = common_W, md5_K, md5_next_shift
local h1, h2, h3, h4 = H[1], H[2], H[3], H[4]
for pos = offs, offs + size - 1, 64 do
for j = 1, 16 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = ((d * 256 + c) * 256 + b) * 256 + a
end
local a, b, c, d = h1, h2, h3, h4
local s = 25
for j = 1, 16 do
local F = ROR(AND(b, c) + AND(-1-b, d) + a + K[j] + W[j], s) + b
s = md5_next_shift[s]
a = d
d = c
c = b
b = F
end
s = 27
for j = 17, 32 do
local F = ROR(AND(d, b) + AND(-1-d, c) + a + K[j] + W[(5*j-4) % 16 + 1], s) + b
s = md5_next_shift[s]
a = d
d = c
c = b
b = F
end
s = 28
for j = 33, 48 do
local F = ROR(XOR(XOR(b, c), d) + a + K[j] + W[(3*j+2) % 16 + 1], s) + b
s = md5_next_shift[s]
a = d
d = c
c = b
b = F
end
s = 26
for j = 49, 64 do
local F = ROR(XOR(c, OR(b, -1-d)) + a + K[j] + W[(j*7-7) % 16 + 1], s) + b
s = md5_next_shift[s]
a = d
d = c
c = b
b = F
end
h1 = (a + h1) % 2^32
h2 = (b + h2) % 2^32
h3 = (c + h3) % 2^32
h4 = (d + h4) % 2^32
end
H[1], H[2], H[3], H[4] = h1, h2, h3, h4
end
elseif branch == "EMUL" then
function md5_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W, K, md5_next_shift = common_W, md5_K, md5_next_shift
local h1, h2, h3, h4 = H[1], H[2], H[3], H[4]
for pos = offs, offs + size - 1, 64 do
for j = 1, 16 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = ((d * 256 + c) * 256 + b) * 256 + a
end
local a, b, c, d = h1, h2, h3, h4
local s = 25
for j = 1, 16 do
local z = (AND(b, c) + AND(-1-b, d) + a + K[j] + W[j]) % 2^32 / 2^s
local y = z % 1
s = md5_next_shift[s]
a = d
d = c
c = b
b = y * 2^32 + (z - y) + b
end
s = 27
for j = 17, 32 do
local z = (AND(d, b) + AND(-1-d, c) + a + K[j] + W[(5*j-4) % 16 + 1]) % 2^32 / 2^s
local y = z % 1
s = md5_next_shift[s]
a = d
d = c
c = b
b = y * 2^32 + (z - y) + b
end
s = 28
for j = 33, 48 do
local z = (XOR(XOR(b, c), d) + a + K[j] + W[(3*j+2) % 16 + 1]) % 2^32 / 2^s
local y = z % 1
s = md5_next_shift[s]
a = d
d = c
c = b
b = y * 2^32 + (z - y) + b
end
s = 26
for j = 49, 64 do
local z = (XOR(c, OR(b, -1-d)) + a + K[j] + W[(j*7-7) % 16 + 1]) % 2^32 / 2^s
local y = z % 1
s = md5_next_shift[s]
a = d
d = c
c = b
b = y * 2^32 + (z - y) + b
end
h1 = (a + h1) % 2^32
h2 = (b + h2) % 2^32
h3 = (c + h3) % 2^32
h4 = (d + h4) % 2^32
end
H[1], H[2], H[3], H[4] = h1, h2, h3, h4
end
end
function sha1_feed_64(H, str, offs, size)
-- offs >= 0, size >= 0, size is multiple of 64
local W = common_W
local h1, h2, h3, h4, h5 = H[1], H[2], H[3], H[4], H[5]
for pos = offs, offs + size - 1, 64 do
for j = 1, 16 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = ((a * 256 + b) * 256 + c) * 256 + d
end
for j = 17, 80 do
local a = XOR(W[j-3], W[j-8], W[j-14], W[j-16]) % 2^32 * 2
local b = a % 2^32
W[j] = b + (a - b) / 2^32
end
local a, b, c, d, e = h1, h2, h3, h4, h5
for j = 1, 20 do
local a5 = a * 2^5
local z = a5 % 2^32
z = z + (a5 - z) / 2^32 + AND(b, c) + AND(-1-b, d) + 0x5A827999 + W[j] + e -- constant = floor(2^30 * sqrt(2))
e = d
d = c
c = b / 2^2
c = c % 1 * (2^32 - 1) + c
b = a
a = z % 2^32
end
for j = 21, 40 do
local a5 = a * 2^5
local z = a5 % 2^32
z = z + (a5 - z) / 2^32 + XOR(b, c, d) + 0x6ED9EBA1 + W[j] + e -- 2^30 * sqrt(3)
e = d
d = c
c = b / 2^2
c = c % 1 * (2^32 - 1) + c
b = a
a = z % 2^32
end
for j = 41, 60 do
local a5 = a * 2^5
local z = a5 % 2^32
z = z + (a5 - z) / 2^32 + AND(d, c) + AND(b, XOR(d, c)) + 0x8F1BBCDC + W[j] + e -- 2^30 * sqrt(5)
e = d
d = c
c = b / 2^2
c = c % 1 * (2^32 - 1) + c
b = a
a = z % 2^32
end
for j = 61, 80 do
local a5 = a * 2^5
local z = a5 % 2^32
z = z + (a5 - z) / 2^32 + XOR(b, c, d) + 0xCA62C1D6 + W[j] + e -- 2^30 * sqrt(10)
e = d
d = c
c = b / 2^2
c = c % 1 * (2^32 - 1) + c
b = a
a = z % 2^32
end
h1 = (a + h1) % 2^32
h2 = (b + h2) % 2^32
h3 = (c + h3) % 2^32
h4 = (d + h4) % 2^32
h5 = (e + h5) % 2^32
end
H[1], H[2], H[3], H[4], H[5] = h1, h2, h3, h4, h5
end
function keccak_feed(lanes_lo, lanes_hi, str, offs, size, block_size_in_bytes)
-- This is an example of a Lua function having 79 local variables :-)
-- offs >= 0, size >= 0, size is multiple of block_size_in_bytes, block_size_in_bytes is positive multiple of 8
local RC_lo, RC_hi = sha3_RC_lo, sha3_RC_hi
local qwords_qty = block_size_in_bytes / 8
for pos = offs, offs + size - 1, block_size_in_bytes do
for j = 1, qwords_qty do
local a, b, c, d = byte(str, pos + 1, pos + 4)
lanes_lo[j] = XOR(lanes_lo[j], ((d * 256 + c) * 256 + b) * 256 + a)
pos = pos + 8
a, b, c, d = byte(str, pos - 3, pos)
lanes_hi[j] = XOR(lanes_hi[j], ((d * 256 + c) * 256 + b) * 256 + a)
end
local L01_lo, L01_hi, L02_lo, L02_hi, L03_lo, L03_hi, L04_lo, L04_hi, L05_lo, L05_hi, L06_lo, L06_hi, L07_lo, L07_hi, L08_lo, L08_hi,
L09_lo, L09_hi, L10_lo, L10_hi, L11_lo, L11_hi, L12_lo, L12_hi, L13_lo, L13_hi, L14_lo, L14_hi, L15_lo, L15_hi, L16_lo, L16_hi,
L17_lo, L17_hi, L18_lo, L18_hi, L19_lo, L19_hi, L20_lo, L20_hi, L21_lo, L21_hi, L22_lo, L22_hi, L23_lo, L23_hi, L24_lo, L24_hi, L25_lo, L25_hi =
lanes_lo[1], lanes_hi[1], lanes_lo[2], lanes_hi[2], lanes_lo[3], lanes_hi[3], lanes_lo[4], lanes_hi[4], lanes_lo[5], lanes_hi[5],
lanes_lo[6], lanes_hi[6], lanes_lo[7], lanes_hi[7], lanes_lo[8], lanes_hi[8], lanes_lo[9], lanes_hi[9], lanes_lo[10], lanes_hi[10],
lanes_lo[11], lanes_hi[11], lanes_lo[12], lanes_hi[12], lanes_lo[13], lanes_hi[13], lanes_lo[14], lanes_hi[14], lanes_lo[15], lanes_hi[15],
lanes_lo[16], lanes_hi[16], lanes_lo[17], lanes_hi[17], lanes_lo[18], lanes_hi[18], lanes_lo[19], lanes_hi[19], lanes_lo[20], lanes_hi[20],
lanes_lo[21], lanes_hi[21], lanes_lo[22], lanes_hi[22], lanes_lo[23], lanes_hi[23], lanes_lo[24], lanes_hi[24], lanes_lo[25], lanes_hi[25]
for round_idx = 1, 24 do
local C1_lo = XOR(L01_lo, L06_lo, L11_lo, L16_lo, L21_lo)
local C1_hi = XOR(L01_hi, L06_hi, L11_hi, L16_hi, L21_hi)
local C2_lo = XOR(L02_lo, L07_lo, L12_lo, L17_lo, L22_lo)
local C2_hi = XOR(L02_hi, L07_hi, L12_hi, L17_hi, L22_hi)
local C3_lo = XOR(L03_lo, L08_lo, L13_lo, L18_lo, L23_lo)
local C3_hi = XOR(L03_hi, L08_hi, L13_hi, L18_hi, L23_hi)
local C4_lo = XOR(L04_lo, L09_lo, L14_lo, L19_lo, L24_lo)
local C4_hi = XOR(L04_hi, L09_hi, L14_hi, L19_hi, L24_hi)
local C5_lo = XOR(L05_lo, L10_lo, L15_lo, L20_lo, L25_lo)
local C5_hi = XOR(L05_hi, L10_hi, L15_hi, L20_hi, L25_hi)
local D_lo = XOR(C1_lo, C3_lo * 2 + (C3_hi % 2^32 - C3_hi % 2^31) / 2^31)
local D_hi = XOR(C1_hi, C3_hi * 2 + (C3_lo % 2^32 - C3_lo % 2^31) / 2^31)
local T0_lo = XOR(D_lo, L02_lo)
local T0_hi = XOR(D_hi, L02_hi)
local T1_lo = XOR(D_lo, L07_lo)
local T1_hi = XOR(D_hi, L07_hi)
local T2_lo = XOR(D_lo, L12_lo)
local T2_hi = XOR(D_hi, L12_hi)
local T3_lo = XOR(D_lo, L17_lo)
local T3_hi = XOR(D_hi, L17_hi)
local T4_lo = XOR(D_lo, L22_lo)
local T4_hi = XOR(D_hi, L22_hi)
L02_lo = (T1_lo % 2^32 - T1_lo % 2^20) / 2^20 + T1_hi * 2^12
L02_hi = (T1_hi % 2^32 - T1_hi % 2^20) / 2^20 + T1_lo * 2^12
L07_lo = (T3_lo % 2^32 - T3_lo % 2^19) / 2^19 + T3_hi * 2^13
L07_hi = (T3_hi % 2^32 - T3_hi % 2^19) / 2^19 + T3_lo * 2^13
L12_lo = T0_lo * 2 + (T0_hi % 2^32 - T0_hi % 2^31) / 2^31
L12_hi = T0_hi * 2 + (T0_lo % 2^32 - T0_lo % 2^31) / 2^31
L17_lo = T2_lo * 2^10 + (T2_hi % 2^32 - T2_hi % 2^22) / 2^22
L17_hi = T2_hi * 2^10 + (T2_lo % 2^32 - T2_lo % 2^22) / 2^22
L22_lo = T4_lo * 2^2 + (T4_hi % 2^32 - T4_hi % 2^30) / 2^30
L22_hi = T4_hi * 2^2 + (T4_lo % 2^32 - T4_lo % 2^30) / 2^30
D_lo = XOR(C2_lo, C4_lo * 2 + (C4_hi % 2^32 - C4_hi % 2^31) / 2^31)
D_hi = XOR(C2_hi, C4_hi * 2 + (C4_lo % 2^32 - C4_lo % 2^31) / 2^31)
T0_lo = XOR(D_lo, L03_lo)
T0_hi = XOR(D_hi, L03_hi)
T1_lo = XOR(D_lo, L08_lo)
T1_hi = XOR(D_hi, L08_hi)
T2_lo = XOR(D_lo, L13_lo)
T2_hi = XOR(D_hi, L13_hi)
T3_lo = XOR(D_lo, L18_lo)
T3_hi = XOR(D_hi, L18_hi)
T4_lo = XOR(D_lo, L23_lo)
T4_hi = XOR(D_hi, L23_hi)
L03_lo = (T2_lo % 2^32 - T2_lo % 2^21) / 2^21 + T2_hi * 2^11
L03_hi = (T2_hi % 2^32 - T2_hi % 2^21) / 2^21 + T2_lo * 2^11
L08_lo = (T4_lo % 2^32 - T4_lo % 2^3) / 2^3 + T4_hi * 2^29 % 2^32
L08_hi = (T4_hi % 2^32 - T4_hi % 2^3) / 2^3 + T4_lo * 2^29 % 2^32
L13_lo = T1_lo * 2^6 + (T1_hi % 2^32 - T1_hi % 2^26) / 2^26
L13_hi = T1_hi * 2^6 + (T1_lo % 2^32 - T1_lo % 2^26) / 2^26
L18_lo = T3_lo * 2^15 + (T3_hi % 2^32 - T3_hi % 2^17) / 2^17
L18_hi = T3_hi * 2^15 + (T3_lo % 2^32 - T3_lo % 2^17) / 2^17
L23_lo = (T0_lo % 2^32 - T0_lo % 2^2) / 2^2 + T0_hi * 2^30 % 2^32
L23_hi = (T0_hi % 2^32 - T0_hi % 2^2) / 2^2 + T0_lo * 2^30 % 2^32
D_lo = XOR(C3_lo, C5_lo * 2 + (C5_hi % 2^32 - C5_hi % 2^31) / 2^31)
D_hi = XOR(C3_hi, C5_hi * 2 + (C5_lo % 2^32 - C5_lo % 2^31) / 2^31)
T0_lo = XOR(D_lo, L04_lo)
T0_hi = XOR(D_hi, L04_hi)
T1_lo = XOR(D_lo, L09_lo)
T1_hi = XOR(D_hi, L09_hi)
T2_lo = XOR(D_lo, L14_lo)
T2_hi = XOR(D_hi, L14_hi)
T3_lo = XOR(D_lo, L19_lo)
T3_hi = XOR(D_hi, L19_hi)
T4_lo = XOR(D_lo, L24_lo)
T4_hi = XOR(D_hi, L24_hi)
L04_lo = T3_lo * 2^21 % 2^32 + (T3_hi % 2^32 - T3_hi % 2^11) / 2^11
L04_hi = T3_hi * 2^21 % 2^32 + (T3_lo % 2^32 - T3_lo % 2^11) / 2^11
L09_lo = T0_lo * 2^28 % 2^32 + (T0_hi % 2^32 - T0_hi % 2^4) / 2^4
L09_hi = T0_hi * 2^28 % 2^32 + (T0_lo % 2^32 - T0_lo % 2^4) / 2^4
L14_lo = T2_lo * 2^25 % 2^32 + (T2_hi % 2^32 - T2_hi % 2^7) / 2^7
L14_hi = T2_hi * 2^25 % 2^32 + (T2_lo % 2^32 - T2_lo % 2^7) / 2^7
L19_lo = (T4_lo % 2^32 - T4_lo % 2^8) / 2^8 + T4_hi * 2^24 % 2^32
L19_hi = (T4_hi % 2^32 - T4_hi % 2^8) / 2^8 + T4_lo * 2^24 % 2^32
L24_lo = (T1_lo % 2^32 - T1_lo % 2^9) / 2^9 + T1_hi * 2^23 % 2^32
L24_hi = (T1_hi % 2^32 - T1_hi % 2^9) / 2^9 + T1_lo * 2^23 % 2^32
D_lo = XOR(C4_lo, C1_lo * 2 + (C1_hi % 2^32 - C1_hi % 2^31) / 2^31)
D_hi = XOR(C4_hi, C1_hi * 2 + (C1_lo % 2^32 - C1_lo % 2^31) / 2^31)
T0_lo = XOR(D_lo, L05_lo)
T0_hi = XOR(D_hi, L05_hi)
T1_lo = XOR(D_lo, L10_lo)
T1_hi = XOR(D_hi, L10_hi)
T2_lo = XOR(D_lo, L15_lo)
T2_hi = XOR(D_hi, L15_hi)
T3_lo = XOR(D_lo, L20_lo)
T3_hi = XOR(D_hi, L20_hi)
T4_lo = XOR(D_lo, L25_lo)
T4_hi = XOR(D_hi, L25_hi)
L05_lo = T4_lo * 2^14 + (T4_hi % 2^32 - T4_hi % 2^18) / 2^18
L05_hi = T4_hi * 2^14 + (T4_lo % 2^32 - T4_lo % 2^18) / 2^18
L10_lo = T1_lo * 2^20 % 2^32 + (T1_hi % 2^32 - T1_hi % 2^12) / 2^12
L10_hi = T1_hi * 2^20 % 2^32 + (T1_lo % 2^32 - T1_lo % 2^12) / 2^12
L15_lo = T3_lo * 2^8 + (T3_hi % 2^32 - T3_hi % 2^24) / 2^24
L15_hi = T3_hi * 2^8 + (T3_lo % 2^32 - T3_lo % 2^24) / 2^24
L20_lo = T0_lo * 2^27 % 2^32 + (T0_hi % 2^32 - T0_hi % 2^5) / 2^5
L20_hi = T0_hi * 2^27 % 2^32 + (T0_lo % 2^32 - T0_lo % 2^5) / 2^5
L25_lo = (T2_lo % 2^32 - T2_lo % 2^25) / 2^25 + T2_hi * 2^7
L25_hi = (T2_hi % 2^32 - T2_hi % 2^25) / 2^25 + T2_lo * 2^7
D_lo = XOR(C5_lo, C2_lo * 2 + (C2_hi % 2^32 - C2_hi % 2^31) / 2^31)
D_hi = XOR(C5_hi, C2_hi * 2 + (C2_lo % 2^32 - C2_lo % 2^31) / 2^31)
T1_lo = XOR(D_lo, L06_lo)
T1_hi = XOR(D_hi, L06_hi)
T2_lo = XOR(D_lo, L11_lo)
T2_hi = XOR(D_hi, L11_hi)
T3_lo = XOR(D_lo, L16_lo)
T3_hi = XOR(D_hi, L16_hi)
T4_lo = XOR(D_lo, L21_lo)
T4_hi = XOR(D_hi, L21_hi)
L06_lo = T2_lo * 2^3 + (T2_hi % 2^32 - T2_hi % 2^29) / 2^29
L06_hi = T2_hi * 2^3 + (T2_lo % 2^32 - T2_lo % 2^29) / 2^29
L11_lo = T4_lo * 2^18 + (T4_hi % 2^32 - T4_hi % 2^14) / 2^14
L11_hi = T4_hi * 2^18 + (T4_lo % 2^32 - T4_lo % 2^14) / 2^14
L16_lo = (T1_lo % 2^32 - T1_lo % 2^28) / 2^28 + T1_hi * 2^4
L16_hi = (T1_hi % 2^32 - T1_hi % 2^28) / 2^28 + T1_lo * 2^4
L21_lo = (T3_lo % 2^32 - T3_lo % 2^23) / 2^23 + T3_hi * 2^9
L21_hi = (T3_hi % 2^32 - T3_hi % 2^23) / 2^23 + T3_lo * 2^9
L01_lo = XOR(D_lo, L01_lo)
L01_hi = XOR(D_hi, L01_hi)
L01_lo, L02_lo, L03_lo, L04_lo, L05_lo = XOR(L01_lo, AND(-1-L02_lo, L03_lo)), XOR(L02_lo, AND(-1-L03_lo, L04_lo)), XOR(L03_lo, AND(-1-L04_lo, L05_lo)), XOR(L04_lo, AND(-1-L05_lo, L01_lo)), XOR(L05_lo, AND(-1-L01_lo, L02_lo))
L01_hi, L02_hi, L03_hi, L04_hi, L05_hi = XOR(L01_hi, AND(-1-L02_hi, L03_hi)), XOR(L02_hi, AND(-1-L03_hi, L04_hi)), XOR(L03_hi, AND(-1-L04_hi, L05_hi)), XOR(L04_hi, AND(-1-L05_hi, L01_hi)), XOR(L05_hi, AND(-1-L01_hi, L02_hi))
L06_lo, L07_lo, L08_lo, L09_lo, L10_lo = XOR(L09_lo, AND(-1-L10_lo, L06_lo)), XOR(L10_lo, AND(-1-L06_lo, L07_lo)), XOR(L06_lo, AND(-1-L07_lo, L08_lo)), XOR(L07_lo, AND(-1-L08_lo, L09_lo)), XOR(L08_lo, AND(-1-L09_lo, L10_lo))
L06_hi, L07_hi, L08_hi, L09_hi, L10_hi = XOR(L09_hi, AND(-1-L10_hi, L06_hi)), XOR(L10_hi, AND(-1-L06_hi, L07_hi)), XOR(L06_hi, AND(-1-L07_hi, L08_hi)), XOR(L07_hi, AND(-1-L08_hi, L09_hi)), XOR(L08_hi, AND(-1-L09_hi, L10_hi))
L11_lo, L12_lo, L13_lo, L14_lo, L15_lo = XOR(L12_lo, AND(-1-L13_lo, L14_lo)), XOR(L13_lo, AND(-1-L14_lo, L15_lo)), XOR(L14_lo, AND(-1-L15_lo, L11_lo)), XOR(L15_lo, AND(-1-L11_lo, L12_lo)), XOR(L11_lo, AND(-1-L12_lo, L13_lo))
L11_hi, L12_hi, L13_hi, L14_hi, L15_hi = XOR(L12_hi, AND(-1-L13_hi, L14_hi)), XOR(L13_hi, AND(-1-L14_hi, L15_hi)), XOR(L14_hi, AND(-1-L15_hi, L11_hi)), XOR(L15_hi, AND(-1-L11_hi, L12_hi)), XOR(L11_hi, AND(-1-L12_hi, L13_hi))
L16_lo, L17_lo, L18_lo, L19_lo, L20_lo = XOR(L20_lo, AND(-1-L16_lo, L17_lo)), XOR(L16_lo, AND(-1-L17_lo, L18_lo)), XOR(L17_lo, AND(-1-L18_lo, L19_lo)), XOR(L18_lo, AND(-1-L19_lo, L20_lo)), XOR(L19_lo, AND(-1-L20_lo, L16_lo))
L16_hi, L17_hi, L18_hi, L19_hi, L20_hi = XOR(L20_hi, AND(-1-L16_hi, L17_hi)), XOR(L16_hi, AND(-1-L17_hi, L18_hi)), XOR(L17_hi, AND(-1-L18_hi, L19_hi)), XOR(L18_hi, AND(-1-L19_hi, L20_hi)), XOR(L19_hi, AND(-1-L20_hi, L16_hi))
L21_lo, L22_lo, L23_lo, L24_lo, L25_lo = XOR(L23_lo, AND(-1-L24_lo, L25_lo)), XOR(L24_lo, AND(-1-L25_lo, L21_lo)), XOR(L25_lo, AND(-1-L21_lo, L22_lo)), XOR(L21_lo, AND(-1-L22_lo, L23_lo)), XOR(L22_lo, AND(-1-L23_lo, L24_lo))
L21_hi, L22_hi, L23_hi, L24_hi, L25_hi = XOR(L23_hi, AND(-1-L24_hi, L25_hi)), XOR(L24_hi, AND(-1-L25_hi, L21_hi)), XOR(L25_hi, AND(-1-L21_hi, L22_hi)), XOR(L21_hi, AND(-1-L22_hi, L23_hi)), XOR(L22_hi, AND(-1-L23_hi, L24_hi))
L01_lo = XOR(L01_lo, RC_lo[round_idx])
L01_hi = L01_hi + RC_hi[round_idx] -- RC_hi[] is either 0 or 0x80000000, so we could use fast addition instead of slow XOR
end
lanes_lo[1] = L01_lo; lanes_hi[1] = L01_hi
lanes_lo[2] = L02_lo; lanes_hi[2] = L02_hi
lanes_lo[3] = L03_lo; lanes_hi[3] = L03_hi
lanes_lo[4] = L04_lo; lanes_hi[4] = L04_hi
lanes_lo[5] = L05_lo; lanes_hi[5] = L05_hi
lanes_lo[6] = L06_lo; lanes_hi[6] = L06_hi
lanes_lo[7] = L07_lo; lanes_hi[7] = L07_hi
lanes_lo[8] = L08_lo; lanes_hi[8] = L08_hi
lanes_lo[9] = L09_lo; lanes_hi[9] = L09_hi
lanes_lo[10] = L10_lo; lanes_hi[10] = L10_hi
lanes_lo[11] = L11_lo; lanes_hi[11] = L11_hi
lanes_lo[12] = L12_lo; lanes_hi[12] = L12_hi
lanes_lo[13] = L13_lo; lanes_hi[13] = L13_hi
lanes_lo[14] = L14_lo; lanes_hi[14] = L14_hi
lanes_lo[15] = L15_lo; lanes_hi[15] = L15_hi
lanes_lo[16] = L16_lo; lanes_hi[16] = L16_hi
lanes_lo[17] = L17_lo; lanes_hi[17] = L17_hi
lanes_lo[18] = L18_lo; lanes_hi[18] = L18_hi
lanes_lo[19] = L19_lo; lanes_hi[19] = L19_hi
lanes_lo[20] = L20_lo; lanes_hi[20] = L20_hi
lanes_lo[21] = L21_lo; lanes_hi[21] = L21_hi
lanes_lo[22] = L22_lo; lanes_hi[22] = L22_hi
lanes_lo[23] = L23_lo; lanes_hi[23] = L23_hi
lanes_lo[24] = L24_lo; lanes_hi[24] = L24_hi
lanes_lo[25] = L25_lo; lanes_hi[25] = L25_hi
end
end
function blake2s_feed_64(H, str, offs, size, bytes_compressed, last_block_size, is_last_node)
-- offs >= 0, size >= 0, size is multiple of 64
local W = common_W
local h1, h2, h3, h4, h5, h6, h7, h8 = H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8]
for pos = offs, offs + size - 1, 64 do
if str then
for j = 1, 16 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = ((d * 256 + c) * 256 + b) * 256 + a
end
end
local v0, v1, v2, v3, v4, v5, v6, v7 = h1, h2, h3, h4, h5, h6, h7, h8
local v8, v9, vA, vB, vC, vD, vE, vF = sha2_H_hi[1], sha2_H_hi[2], sha2_H_hi[3], sha2_H_hi[4], sha2_H_hi[5], sha2_H_hi[6], sha2_H_hi[7], sha2_H_hi[8]
bytes_compressed = bytes_compressed + (last_block_size or 64)
local t0 = bytes_compressed % 2^32
local t1 = (bytes_compressed - t0) / 2^32
vC = XOR(vC, t0) -- t0 = low_4_bytes(bytes_compressed)
vD = XOR(vD, t1) -- t1 = high_4_bytes(bytes_compressed)
if last_block_size then -- flag f0
vE = -1 - vE
end
if is_last_node then -- flag f1
vF = -1 - vF
end
for j = 1, 10 do
local row = sigma[j]
v0 = v0 + v4 + W[row[1]]
vC = XOR(vC, v0) % 2^32 / 2^16
vC = vC % 1 * (2^32 - 1) + vC
v8 = v8 + vC
v4 = XOR(v4, v8) % 2^32 / 2^12
v4 = v4 % 1 * (2^32 - 1) + v4
v0 = v0 + v4 + W[row[2]]
vC = XOR(vC, v0) % 2^32 / 2^8
vC = vC % 1 * (2^32 - 1) + vC
v8 = v8 + vC
v4 = XOR(v4, v8) % 2^32 / 2^7
v4 = v4 % 1 * (2^32 - 1) + v4
v1 = v1 + v5 + W[row[3]]
vD = XOR(vD, v1) % 2^32 / 2^16
vD = vD % 1 * (2^32 - 1) + vD
v9 = v9 + vD
v5 = XOR(v5, v9) % 2^32 / 2^12
v5 = v5 % 1 * (2^32 - 1) + v5
v1 = v1 + v5 + W[row[4]]
vD = XOR(vD, v1) % 2^32 / 2^8
vD = vD % 1 * (2^32 - 1) + vD
v9 = v9 + vD
v5 = XOR(v5, v9) % 2^32 / 2^7
v5 = v5 % 1 * (2^32 - 1) + v5
v2 = v2 + v6 + W[row[5]]
vE = XOR(vE, v2) % 2^32 / 2^16
vE = vE % 1 * (2^32 - 1) + vE
vA = vA + vE
v6 = XOR(v6, vA) % 2^32 / 2^12
v6 = v6 % 1 * (2^32 - 1) + v6
v2 = v2 + v6 + W[row[6]]
vE = XOR(vE, v2) % 2^32 / 2^8
vE = vE % 1 * (2^32 - 1) + vE
vA = vA + vE
v6 = XOR(v6, vA) % 2^32 / 2^7
v6 = v6 % 1 * (2^32 - 1) + v6
v3 = v3 + v7 + W[row[7]]
vF = XOR(vF, v3) % 2^32 / 2^16
vF = vF % 1 * (2^32 - 1) + vF
vB = vB + vF
v7 = XOR(v7, vB) % 2^32 / 2^12
v7 = v7 % 1 * (2^32 - 1) + v7
v3 = v3 + v7 + W[row[8]]
vF = XOR(vF, v3) % 2^32 / 2^8
vF = vF % 1 * (2^32 - 1) + vF
vB = vB + vF
v7 = XOR(v7, vB) % 2^32 / 2^7
v7 = v7 % 1 * (2^32 - 1) + v7
v0 = v0 + v5 + W[row[9]]
vF = XOR(vF, v0) % 2^32 / 2^16
vF = vF % 1 * (2^32 - 1) + vF
vA = vA + vF
v5 = XOR(v5, vA) % 2^32 / 2^12
v5 = v5 % 1 * (2^32 - 1) + v5
v0 = v0 + v5 + W[row[10]]
vF = XOR(vF, v0) % 2^32 / 2^8
vF = vF % 1 * (2^32 - 1) + vF
vA = vA + vF
v5 = XOR(v5, vA) % 2^32 / 2^7
v5 = v5 % 1 * (2^32 - 1) + v5
v1 = v1 + v6 + W[row[11]]
vC = XOR(vC, v1) % 2^32 / 2^16
vC = vC % 1 * (2^32 - 1) + vC
vB = vB + vC
v6 = XOR(v6, vB) % 2^32 / 2^12
v6 = v6 % 1 * (2^32 - 1) + v6
v1 = v1 + v6 + W[row[12]]
vC = XOR(vC, v1) % 2^32 / 2^8
vC = vC % 1 * (2^32 - 1) + vC
vB = vB + vC
v6 = XOR(v6, vB) % 2^32 / 2^7
v6 = v6 % 1 * (2^32 - 1) + v6
v2 = v2 + v7 + W[row[13]]
vD = XOR(vD, v2) % 2^32 / 2^16
vD = vD % 1 * (2^32 - 1) + vD
v8 = v8 + vD
v7 = XOR(v7, v8) % 2^32 / 2^12
v7 = v7 % 1 * (2^32 - 1) + v7
v2 = v2 + v7 + W[row[14]]
vD = XOR(vD, v2) % 2^32 / 2^8
vD = vD % 1 * (2^32 - 1) + vD
v8 = v8 + vD
v7 = XOR(v7, v8) % 2^32 / 2^7
v7 = v7 % 1 * (2^32 - 1) + v7
v3 = v3 + v4 + W[row[15]]
vE = XOR(vE, v3) % 2^32 / 2^16
vE = vE % 1 * (2^32 - 1) + vE
v9 = v9 + vE
v4 = XOR(v4, v9) % 2^32 / 2^12
v4 = v4 % 1 * (2^32 - 1) + v4
v3 = v3 + v4 + W[row[16]]
vE = XOR(vE, v3) % 2^32 / 2^8
vE = vE % 1 * (2^32 - 1) + vE
v9 = v9 + vE
v4 = XOR(v4, v9) % 2^32 / 2^7
v4 = v4 % 1 * (2^32 - 1) + v4
end
h1 = XOR(h1, v0, v8)
h2 = XOR(h2, v1, v9)
h3 = XOR(h3, v2, vA)
h4 = XOR(h4, v3, vB)
h5 = XOR(h5, v4, vC)
h6 = XOR(h6, v5, vD)
h7 = XOR(h7, v6, vE)
h8 = XOR(h8, v7, vF)
end
H[1], H[2], H[3], H[4], H[5], H[6], H[7], H[8] = h1, h2, h3, h4, h5, h6, h7, h8
return bytes_compressed
end
function blake2b_feed_128(H_lo, H_hi, str, offs, size, bytes_compressed, last_block_size, is_last_node)
-- offs >= 0, size >= 0, size is multiple of 128
local W = common_W
local h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo = H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8]
local h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi = H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8]
for pos = offs, offs + size - 1, 128 do
if str then
for j = 1, 32 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = ((d * 256 + c) * 256 + b) * 256 + a
end
end
local v0_lo, v1_lo, v2_lo, v3_lo, v4_lo, v5_lo, v6_lo, v7_lo = h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo
local v0_hi, v1_hi, v2_hi, v3_hi, v4_hi, v5_hi, v6_hi, v7_hi = h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi
local v8_lo, v9_lo, vA_lo, vB_lo, vC_lo, vD_lo, vE_lo, vF_lo = sha2_H_lo[1], sha2_H_lo[2], sha2_H_lo[3], sha2_H_lo[4], sha2_H_lo[5], sha2_H_lo[6], sha2_H_lo[7], sha2_H_lo[8]
local v8_hi, v9_hi, vA_hi, vB_hi, vC_hi, vD_hi, vE_hi, vF_hi = sha2_H_hi[1], sha2_H_hi[2], sha2_H_hi[3], sha2_H_hi[4], sha2_H_hi[5], sha2_H_hi[6], sha2_H_hi[7], sha2_H_hi[8]
bytes_compressed = bytes_compressed + (last_block_size or 128)
local t0_lo = bytes_compressed % 2^32
local t0_hi = (bytes_compressed - t0_lo) / 2^32
vC_lo = XOR(vC_lo, t0_lo) -- t0 = low_8_bytes(bytes_compressed)
vC_hi = XOR(vC_hi, t0_hi)
-- t1 = high_8_bytes(bytes_compressed) = 0, message length is always below 2^53 bytes
if last_block_size then -- flag f0
vE_lo = -1 - vE_lo
vE_hi = -1 - vE_hi
end
if is_last_node then -- flag f1
vF_lo = -1 - vF_lo
vF_hi = -1 - vF_hi
end
for j = 1, 12 do
local row = sigma[j]
local k = row[1] * 2
local z = v0_lo % 2^32 + v4_lo % 2^32 + W[k-1]
v0_lo = z % 2^32
v0_hi = v0_hi + v4_hi + (z - v0_lo) / 2^32 + W[k]
vC_lo, vC_hi = XOR(vC_hi, v0_hi), XOR(vC_lo, v0_lo)
z = v8_lo % 2^32 + vC_lo % 2^32
v8_lo = z % 2^32
v8_hi = v8_hi + vC_hi + (z - v8_lo) / 2^32
v4_lo, v4_hi = XOR(v4_lo, v8_lo), XOR(v4_hi, v8_hi)
local z_lo, z_hi = v4_lo % 2^24, v4_hi % 2^24
v4_lo, v4_hi = (v4_lo - z_lo) / 2^24 % 2^8 + z_hi * 2^8, (v4_hi - z_hi) / 2^24 % 2^8 + z_lo * 2^8
k = row[2] * 2
z = v0_lo % 2^32 + v4_lo % 2^32 + W[k-1]
v0_lo = z % 2^32
v0_hi = v0_hi + v4_hi + (z - v0_lo) / 2^32 + W[k]
vC_lo, vC_hi = XOR(vC_lo, v0_lo), XOR(vC_hi, v0_hi)
z_lo, z_hi = vC_lo % 2^16, vC_hi % 2^16
vC_lo, vC_hi = (vC_lo - z_lo) / 2^16 % 2^16 + z_hi * 2^16, (vC_hi - z_hi) / 2^16 % 2^16 + z_lo * 2^16
z = v8_lo % 2^32 + vC_lo % 2^32
v8_lo = z % 2^32
v8_hi = v8_hi + vC_hi + (z - v8_lo) / 2^32
v4_lo, v4_hi = XOR(v4_lo, v8_lo), XOR(v4_hi, v8_hi)
z_lo, z_hi = v4_lo % 2^31, v4_hi % 2^31
v4_lo, v4_hi = z_lo * 2^1 + (v4_hi - z_hi) / 2^31 % 2^1, z_hi * 2^1 + (v4_lo - z_lo) / 2^31 % 2^1
k = row[3] * 2
z = v1_lo % 2^32 + v5_lo % 2^32 + W[k-1]
v1_lo = z % 2^32
v1_hi = v1_hi + v5_hi + (z - v1_lo) / 2^32 + W[k]
vD_lo, vD_hi = XOR(vD_hi, v1_hi), XOR(vD_lo, v1_lo)
z = v9_lo % 2^32 + vD_lo % 2^32
v9_lo = z % 2^32
v9_hi = v9_hi + vD_hi + (z - v9_lo) / 2^32
v5_lo, v5_hi = XOR(v5_lo, v9_lo), XOR(v5_hi, v9_hi)
z_lo, z_hi = v5_lo % 2^24, v5_hi % 2^24
v5_lo, v5_hi = (v5_lo - z_lo) / 2^24 % 2^8 + z_hi * 2^8, (v5_hi - z_hi) / 2^24 % 2^8 + z_lo * 2^8
k = row[4] * 2
z = v1_lo % 2^32 + v5_lo % 2^32 + W[k-1]
v1_lo = z % 2^32
v1_hi = v1_hi + v5_hi + (z - v1_lo) / 2^32 + W[k]
vD_lo, vD_hi = XOR(vD_lo, v1_lo), XOR(vD_hi, v1_hi)
z_lo, z_hi = vD_lo % 2^16, vD_hi % 2^16
vD_lo, vD_hi = (vD_lo - z_lo) / 2^16 % 2^16 + z_hi * 2^16, (vD_hi - z_hi) / 2^16 % 2^16 + z_lo * 2^16
z = v9_lo % 2^32 + vD_lo % 2^32
v9_lo = z % 2^32
v9_hi = v9_hi + vD_hi + (z - v9_lo) / 2^32
v5_lo, v5_hi = XOR(v5_lo, v9_lo), XOR(v5_hi, v9_hi)
z_lo, z_hi = v5_lo % 2^31, v5_hi % 2^31
v5_lo, v5_hi = z_lo * 2^1 + (v5_hi - z_hi) / 2^31 % 2^1, z_hi * 2^1 + (v5_lo - z_lo) / 2^31 % 2^1
k = row[5] * 2
z = v2_lo % 2^32 + v6_lo % 2^32 + W[k-1]
v2_lo = z % 2^32
v2_hi = v2_hi + v6_hi + (z - v2_lo) / 2^32 + W[k]
vE_lo, vE_hi = XOR(vE_hi, v2_hi), XOR(vE_lo, v2_lo)
z = vA_lo % 2^32 + vE_lo % 2^32
vA_lo = z % 2^32
vA_hi = vA_hi + vE_hi + (z - vA_lo) / 2^32
v6_lo, v6_hi = XOR(v6_lo, vA_lo), XOR(v6_hi, vA_hi)
z_lo, z_hi = v6_lo % 2^24, v6_hi % 2^24
v6_lo, v6_hi = (v6_lo - z_lo) / 2^24 % 2^8 + z_hi * 2^8, (v6_hi - z_hi) / 2^24 % 2^8 + z_lo * 2^8
k = row[6] * 2
z = v2_lo % 2^32 + v6_lo % 2^32 + W[k-1]
v2_lo = z % 2^32
v2_hi = v2_hi + v6_hi + (z - v2_lo) / 2^32 + W[k]
vE_lo, vE_hi = XOR(vE_lo, v2_lo), XOR(vE_hi, v2_hi)
z_lo, z_hi = vE_lo % 2^16, vE_hi % 2^16
vE_lo, vE_hi = (vE_lo - z_lo) / 2^16 % 2^16 + z_hi * 2^16, (vE_hi - z_hi) / 2^16 % 2^16 + z_lo * 2^16
z = vA_lo % 2^32 + vE_lo % 2^32
vA_lo = z % 2^32
vA_hi = vA_hi + vE_hi + (z - vA_lo) / 2^32
v6_lo, v6_hi = XOR(v6_lo, vA_lo), XOR(v6_hi, vA_hi)
z_lo, z_hi = v6_lo % 2^31, v6_hi % 2^31
v6_lo, v6_hi = z_lo * 2^1 + (v6_hi - z_hi) / 2^31 % 2^1, z_hi * 2^1 + (v6_lo - z_lo) / 2^31 % 2^1
k = row[7] * 2
z = v3_lo % 2^32 + v7_lo % 2^32 + W[k-1]
v3_lo = z % 2^32
v3_hi = v3_hi + v7_hi + (z - v3_lo) / 2^32 + W[k]
vF_lo, vF_hi = XOR(vF_hi, v3_hi), XOR(vF_lo, v3_lo)
z = vB_lo % 2^32 + vF_lo % 2^32
vB_lo = z % 2^32
vB_hi = vB_hi + vF_hi + (z - vB_lo) / 2^32
v7_lo, v7_hi = XOR(v7_lo, vB_lo), XOR(v7_hi, vB_hi)
z_lo, z_hi = v7_lo % 2^24, v7_hi % 2^24
v7_lo, v7_hi = (v7_lo - z_lo) / 2^24 % 2^8 + z_hi * 2^8, (v7_hi - z_hi) / 2^24 % 2^8 + z_lo * 2^8
k = row[8] * 2
z = v3_lo % 2^32 + v7_lo % 2^32 + W[k-1]
v3_lo = z % 2^32
v3_hi = v3_hi + v7_hi + (z - v3_lo) / 2^32 + W[k]
vF_lo, vF_hi = XOR(vF_lo, v3_lo), XOR(vF_hi, v3_hi)
z_lo, z_hi = vF_lo % 2^16, vF_hi % 2^16
vF_lo, vF_hi = (vF_lo - z_lo) / 2^16 % 2^16 + z_hi * 2^16, (vF_hi - z_hi) / 2^16 % 2^16 + z_lo * 2^16
z = vB_lo % 2^32 + vF_lo % 2^32
vB_lo = z % 2^32
vB_hi = vB_hi + vF_hi + (z - vB_lo) / 2^32
v7_lo, v7_hi = XOR(v7_lo, vB_lo), XOR(v7_hi, vB_hi)
z_lo, z_hi = v7_lo % 2^31, v7_hi % 2^31
v7_lo, v7_hi = z_lo * 2^1 + (v7_hi - z_hi) / 2^31 % 2^1, z_hi * 2^1 + (v7_lo - z_lo) / 2^31 % 2^1
k = row[9] * 2
z = v0_lo % 2^32 + v5_lo % 2^32 + W[k-1]
v0_lo = z % 2^32
v0_hi = v0_hi + v5_hi + (z - v0_lo) / 2^32 + W[k]
vF_lo, vF_hi = XOR(vF_hi, v0_hi), XOR(vF_lo, v0_lo)
z = vA_lo % 2^32 + vF_lo % 2^32
vA_lo = z % 2^32
vA_hi = vA_hi + vF_hi + (z - vA_lo) / 2^32
v5_lo, v5_hi = XOR(v5_lo, vA_lo), XOR(v5_hi, vA_hi)
z_lo, z_hi = v5_lo % 2^24, v5_hi % 2^24
v5_lo, v5_hi = (v5_lo - z_lo) / 2^24 % 2^8 + z_hi * 2^8, (v5_hi - z_hi) / 2^24 % 2^8 + z_lo * 2^8
k = row[10] * 2
z = v0_lo % 2^32 + v5_lo % 2^32 + W[k-1]
v0_lo = z % 2^32
v0_hi = v0_hi + v5_hi + (z - v0_lo) / 2^32 + W[k]
vF_lo, vF_hi = XOR(vF_lo, v0_lo), XOR(vF_hi, v0_hi)
z_lo, z_hi = vF_lo % 2^16, vF_hi % 2^16
vF_lo, vF_hi = (vF_lo - z_lo) / 2^16 % 2^16 + z_hi * 2^16, (vF_hi - z_hi) / 2^16 % 2^16 + z_lo * 2^16
z = vA_lo % 2^32 + vF_lo % 2^32
vA_lo = z % 2^32
vA_hi = vA_hi + vF_hi + (z - vA_lo) / 2^32
v5_lo, v5_hi = XOR(v5_lo, vA_lo), XOR(v5_hi, vA_hi)
z_lo, z_hi = v5_lo % 2^31, v5_hi % 2^31
v5_lo, v5_hi = z_lo * 2^1 + (v5_hi - z_hi) / 2^31 % 2^1, z_hi * 2^1 + (v5_lo - z_lo) / 2^31 % 2^1
k = row[11] * 2
z = v1_lo % 2^32 + v6_lo % 2^32 + W[k-1]
v1_lo = z % 2^32
v1_hi = v1_hi + v6_hi + (z - v1_lo) / 2^32 + W[k]
vC_lo, vC_hi = XOR(vC_hi, v1_hi), XOR(vC_lo, v1_lo)
z = vB_lo % 2^32 + vC_lo % 2^32
vB_lo = z % 2^32
vB_hi = vB_hi + vC_hi + (z - vB_lo) / 2^32
v6_lo, v6_hi = XOR(v6_lo, vB_lo), XOR(v6_hi, vB_hi)
z_lo, z_hi = v6_lo % 2^24, v6_hi % 2^24
v6_lo, v6_hi = (v6_lo - z_lo) / 2^24 % 2^8 + z_hi * 2^8, (v6_hi - z_hi) / 2^24 % 2^8 + z_lo * 2^8
k = row[12] * 2
z = v1_lo % 2^32 + v6_lo % 2^32 + W[k-1]
v1_lo = z % 2^32
v1_hi = v1_hi + v6_hi + (z - v1_lo) / 2^32 + W[k]
vC_lo, vC_hi = XOR(vC_lo, v1_lo), XOR(vC_hi, v1_hi)
z_lo, z_hi = vC_lo % 2^16, vC_hi % 2^16
vC_lo, vC_hi = (vC_lo - z_lo) / 2^16 % 2^16 + z_hi * 2^16, (vC_hi - z_hi) / 2^16 % 2^16 + z_lo * 2^16
z = vB_lo % 2^32 + vC_lo % 2^32
vB_lo = z % 2^32
vB_hi = vB_hi + vC_hi + (z - vB_lo) / 2^32
v6_lo, v6_hi = XOR(v6_lo, vB_lo), XOR(v6_hi, vB_hi)
z_lo, z_hi = v6_lo % 2^31, v6_hi % 2^31
v6_lo, v6_hi = z_lo * 2^1 + (v6_hi - z_hi) / 2^31 % 2^1, z_hi * 2^1 + (v6_lo - z_lo) / 2^31 % 2^1
k = row[13] * 2
z = v2_lo % 2^32 + v7_lo % 2^32 + W[k-1]
v2_lo = z % 2^32
v2_hi = v2_hi + v7_hi + (z - v2_lo) / 2^32 + W[k]
vD_lo, vD_hi = XOR(vD_hi, v2_hi), XOR(vD_lo, v2_lo)
z = v8_lo % 2^32 + vD_lo % 2^32
v8_lo = z % 2^32
v8_hi = v8_hi + vD_hi + (z - v8_lo) / 2^32
v7_lo, v7_hi = XOR(v7_lo, v8_lo), XOR(v7_hi, v8_hi)
z_lo, z_hi = v7_lo % 2^24, v7_hi % 2^24
v7_lo, v7_hi = (v7_lo - z_lo) / 2^24 % 2^8 + z_hi * 2^8, (v7_hi - z_hi) / 2^24 % 2^8 + z_lo * 2^8
k = row[14] * 2
z = v2_lo % 2^32 + v7_lo % 2^32 + W[k-1]
v2_lo = z % 2^32
v2_hi = v2_hi + v7_hi + (z - v2_lo) / 2^32 + W[k]
vD_lo, vD_hi = XOR(vD_lo, v2_lo), XOR(vD_hi, v2_hi)
z_lo, z_hi = vD_lo % 2^16, vD_hi % 2^16
vD_lo, vD_hi = (vD_lo - z_lo) / 2^16 % 2^16 + z_hi * 2^16, (vD_hi - z_hi) / 2^16 % 2^16 + z_lo * 2^16
z = v8_lo % 2^32 + vD_lo % 2^32
v8_lo = z % 2^32
v8_hi = v8_hi + vD_hi + (z - v8_lo) / 2^32
v7_lo, v7_hi = XOR(v7_lo, v8_lo), XOR(v7_hi, v8_hi)
z_lo, z_hi = v7_lo % 2^31, v7_hi % 2^31
v7_lo, v7_hi = z_lo * 2^1 + (v7_hi - z_hi) / 2^31 % 2^1, z_hi * 2^1 + (v7_lo - z_lo) / 2^31 % 2^1
k = row[15] * 2
z = v3_lo % 2^32 + v4_lo % 2^32 + W[k-1]
v3_lo = z % 2^32
v3_hi = v3_hi + v4_hi + (z - v3_lo) / 2^32 + W[k]
vE_lo, vE_hi = XOR(vE_hi, v3_hi), XOR(vE_lo, v3_lo)
z = v9_lo % 2^32 + vE_lo % 2^32
v9_lo = z % 2^32
v9_hi = v9_hi + vE_hi + (z - v9_lo) / 2^32
v4_lo, v4_hi = XOR(v4_lo, v9_lo), XOR(v4_hi, v9_hi)
z_lo, z_hi = v4_lo % 2^24, v4_hi % 2^24
v4_lo, v4_hi = (v4_lo - z_lo) / 2^24 % 2^8 + z_hi * 2^8, (v4_hi - z_hi) / 2^24 % 2^8 + z_lo * 2^8
k = row[16] * 2
z = v3_lo % 2^32 + v4_lo % 2^32 + W[k-1]
v3_lo = z % 2^32
v3_hi = v3_hi + v4_hi + (z - v3_lo) / 2^32 + W[k]
vE_lo, vE_hi = XOR(vE_lo, v3_lo), XOR(vE_hi, v3_hi)
z_lo, z_hi = vE_lo % 2^16, vE_hi % 2^16
vE_lo, vE_hi = (vE_lo - z_lo) / 2^16 % 2^16 + z_hi * 2^16, (vE_hi - z_hi) / 2^16 % 2^16 + z_lo * 2^16
z = v9_lo % 2^32 + vE_lo % 2^32
v9_lo = z % 2^32
v9_hi = v9_hi + vE_hi + (z - v9_lo) / 2^32
v4_lo, v4_hi = XOR(v4_lo, v9_lo), XOR(v4_hi, v9_hi)
z_lo, z_hi = v4_lo % 2^31, v4_hi % 2^31
v4_lo, v4_hi = z_lo * 2^1 + (v4_hi - z_hi) / 2^31 % 2^1, z_hi * 2^1 + (v4_lo - z_lo) / 2^31 % 2^1
end
h1_lo = XOR(h1_lo, v0_lo, v8_lo) % 2^32
h2_lo = XOR(h2_lo, v1_lo, v9_lo) % 2^32
h3_lo = XOR(h3_lo, v2_lo, vA_lo) % 2^32
h4_lo = XOR(h4_lo, v3_lo, vB_lo) % 2^32
h5_lo = XOR(h5_lo, v4_lo, vC_lo) % 2^32
h6_lo = XOR(h6_lo, v5_lo, vD_lo) % 2^32
h7_lo = XOR(h7_lo, v6_lo, vE_lo) % 2^32
h8_lo = XOR(h8_lo, v7_lo, vF_lo) % 2^32
h1_hi = XOR(h1_hi, v0_hi, v8_hi) % 2^32
h2_hi = XOR(h2_hi, v1_hi, v9_hi) % 2^32
h3_hi = XOR(h3_hi, v2_hi, vA_hi) % 2^32
h4_hi = XOR(h4_hi, v3_hi, vB_hi) % 2^32
h5_hi = XOR(h5_hi, v4_hi, vC_hi) % 2^32
h6_hi = XOR(h6_hi, v5_hi, vD_hi) % 2^32
h7_hi = XOR(h7_hi, v6_hi, vE_hi) % 2^32
h8_hi = XOR(h8_hi, v7_hi, vF_hi) % 2^32
end
H_lo[1], H_lo[2], H_lo[3], H_lo[4], H_lo[5], H_lo[6], H_lo[7], H_lo[8] = h1_lo, h2_lo, h3_lo, h4_lo, h5_lo, h6_lo, h7_lo, h8_lo
H_hi[1], H_hi[2], H_hi[3], H_hi[4], H_hi[5], H_hi[6], H_hi[7], H_hi[8] = h1_hi, h2_hi, h3_hi, h4_hi, h5_hi, h6_hi, h7_hi, h8_hi
return bytes_compressed
end
function blake3_feed_64(str, offs, size, flags, chunk_index, H_in, H_out, wide_output, block_length)
-- offs >= 0, size >= 0, size is multiple of 64
block_length = block_length or 64
local W = common_W
local h1, h2, h3, h4, h5, h6, h7, h8 = H_in[1], H_in[2], H_in[3], H_in[4], H_in[5], H_in[6], H_in[7], H_in[8]
H_out = H_out or H_in
for pos = offs, offs + size - 1, 64 do
if str then
for j = 1, 16 do
pos = pos + 4
local a, b, c, d = byte(str, pos - 3, pos)
W[j] = ((d * 256 + c) * 256 + b) * 256 + a
end
end
local v0, v1, v2, v3, v4, v5, v6, v7 = h1, h2, h3, h4, h5, h6, h7, h8
local v8, v9, vA, vB = sha2_H_hi[1], sha2_H_hi[2], sha2_H_hi[3], sha2_H_hi[4]
local vC = chunk_index % 2^32 -- t0 = low_4_bytes(chunk_index)
local vD = (chunk_index - vC) / 2^32 -- t1 = high_4_bytes(chunk_index)
local vE, vF = block_length, flags
for j = 1, 7 do
v0 = v0 + v4 + W[perm_blake3[j]]
vC = XOR(vC, v0) % 2^32 / 2^16
vC = vC % 1 * (2^32 - 1) + vC
v8 = v8 + vC
v4 = XOR(v4, v8) % 2^32 / 2^12
v4 = v4 % 1 * (2^32 - 1) + v4
v0 = v0 + v4 + W[perm_blake3[j + 14]]
vC = XOR(vC, v0) % 2^32 / 2^8
vC = vC % 1 * (2^32 - 1) + vC
v8 = v8 + vC
v4 = XOR(v4, v8) % 2^32 / 2^7
v4 = v4 % 1 * (2^32 - 1) + v4
v1 = v1 + v5 + W[perm_blake3[j + 1]]
vD = XOR(vD, v1) % 2^32 / 2^16
vD = vD % 1 * (2^32 - 1) + vD
v9 = v9 + vD
v5 = XOR(v5, v9) % 2^32 / 2^12
v5 = v5 % 1 * (2^32 - 1) + v5
v1 = v1 + v5 + W[perm_blake3[j + 2]]
vD = XOR(vD, v1) % 2^32 / 2^8
vD = vD % 1 * (2^32 - 1) + vD
v9 = v9 + vD
v5 = XOR(v5, v9) % 2^32 / 2^7
v5 = v5 % 1 * (2^32 - 1) + v5
v2 = v2 + v6 + W[perm_blake3[j + 16]]
vE = XOR(vE, v2) % 2^32 / 2^16
vE = vE % 1 * (2^32 - 1) + vE
vA = vA + vE
v6 = XOR(v6, vA) % 2^32 / 2^12
v6 = v6 % 1 * (2^32 - 1) + v6
v2 = v2 + v6 + W[perm_blake3[j + 7]]
vE = XOR(vE, v2) % 2^32 / 2^8
vE = vE % 1 * (2^32 - 1) + vE
vA = vA + vE
v6 = XOR(v6, vA) % 2^32 / 2^7
v6 = v6 % 1 * (2^32 - 1) + v6
v3 = v3 + v7 + W[perm_blake3[j + 15]]
vF = XOR(vF, v3) % 2^32 / 2^16
vF = vF % 1 * (2^32 - 1) + vF
vB = vB + vF
v7 = XOR(v7, vB) % 2^32 / 2^12
v7 = v7 % 1 * (2^32 - 1) + v7
v3 = v3 + v7 + W[perm_blake3[j + 17]]
vF = XOR(vF, v3) % 2^32 / 2^8
vF = vF % 1 * (2^32 - 1) + vF
vB = vB + vF
v7 = XOR(v7, vB) % 2^32 / 2^7
v7 = v7 % 1 * (2^32 - 1) + v7
v0 = v0 + v5 + W[perm_blake3[j + 21]]
vF = XOR(vF, v0) % 2^32 / 2^16
vF = vF % 1 * (2^32 - 1) + vF
vA = vA + vF
v5 = XOR(v5, vA) % 2^32 / 2^12
v5 = v5 % 1 * (2^32 - 1) + v5
v0 = v0 + v5 + W[perm_blake3[j + 5]]
vF = XOR(vF, v0) % 2^32 / 2^8
vF = vF % 1 * (2^32 - 1) + vF
vA = vA + vF
v5 = XOR(v5, vA) % 2^32 / 2^7
v5 = v5 % 1 * (2^32 - 1) + v5
v1 = v1 + v6 + W[perm_blake3[j + 3]]
vC = XOR(vC, v1) % 2^32 / 2^16
vC = vC % 1 * (2^32 - 1) + vC
vB = vB + vC
v6 = XOR(v6, vB) % 2^32 / 2^12
v6 = v6 % 1 * (2^32 - 1) + v6
v1 = v1 + v6 + W[perm_blake3[j + 6]]
vC = XOR(vC, v1) % 2^32 / 2^8
vC = vC % 1 * (2^32 - 1) + vC
vB = vB + vC
v6 = XOR(v6, vB) % 2^32 / 2^7
v6 = v6 % 1 * (2^32 - 1) + v6
v2 = v2 + v7 + W[perm_blake3[j + 4]]
vD = XOR(vD, v2) % 2^32 / 2^16
vD = vD % 1 * (2^32 - 1) + vD
v8 = v8 + vD
v7 = XOR(v7, v8) % 2^32 / 2^12
v7 = v7 % 1 * (2^32 - 1) + v7
v2 = v2 + v7 + W[perm_blake3[j + 18]]
vD = XOR(vD, v2) % 2^32 / 2^8
vD = vD % 1 * (2^32 - 1) + vD
v8 = v8 + vD
v7 = XOR(v7, v8) % 2^32 / 2^7
v7 = v7 % 1 * (2^32 - 1) + v7
v3 = v3 + v4 + W[perm_blake3[j + 19]]
vE = XOR(vE, v3) % 2^32 / 2^16
vE = vE % 1 * (2^32 - 1) + vE
v9 = v9 + vE
v4 = XOR(v4, v9) % 2^32 / 2^12
v4 = v4 % 1 * (2^32 - 1) + v4
v3 = v3 + v4 + W[perm_blake3[j + 20]]
vE = XOR(vE, v3) % 2^32 / 2^8
vE = vE % 1 * (2^32 - 1) + vE
v9 = v9 + vE
v4 = XOR(v4, v9) % 2^32 / 2^7
v4 = v4 % 1 * (2^32 - 1) + v4
end
if wide_output then
H_out[ 9] = XOR(h1, v8)
H_out[10] = XOR(h2, v9)
H_out[11] = XOR(h3, vA)
H_out[12] = XOR(h4, vB)
H_out[13] = XOR(h5, vC)
H_out[14] = XOR(h6, vD)
H_out[15] = XOR(h7, vE)
H_out[16] = XOR(h8, vF)
end
h1 = XOR(v0, v8)
h2 = XOR(v1, v9)
h3 = XOR(v2, vA)
h4 = XOR(v3, vB)
h5 = XOR(v4, vC)
h6 = XOR(v5, vD)
h7 = XOR(v6, vE)
h8 = XOR(v7, vF)
end
H_out[1], H_out[2], H_out[3], H_out[4], H_out[5], H_out[6], H_out[7], H_out[8] = h1, h2, h3, h4, h5, h6, h7, h8
end
end
--------------------------------------------------------------------------------
-- MAGIC NUMBERS CALCULATOR
--------------------------------------------------------------------------------
-- Q:
-- Is 53-bit "double" math enough to calculate square roots and cube roots of primes with 64 correct bits after decimal point?
-- A:
-- Yes, 53-bit "double" arithmetic is enough.
-- We could obtain first 40 bits by direct calculation of p^(1/3) and next 40 bits by one step of Newton's method.
do
local function mul(src1, src2, factor, result_length)
-- src1, src2 - long integers (arrays of digits in base 2^24)
-- factor - small integer
-- returns long integer result (src1 * src2 * factor) and its floating point approximation
local result, carry, value, weight = {}, 0.0, 0.0, 1.0
for j = 1, result_length do
for k = math_max(1, j + 1 - #src2), math_min(j, #src1) do
carry = carry + factor * src1[k] * src2[j + 1 - k] -- "int32" is not enough for multiplication result, that's why "factor" must be of type "double"
end
local digit = carry % 2^24
result[j] = floor(digit)
carry = (carry - digit) / 2^24
value = value + digit * weight
weight = weight * 2^24
end
return result, value
end
local idx, step, p, one, sqrt_hi, sqrt_lo = 0, {4, 1, 2, -2, 2}, 4, {1}, sha2_H_hi, sha2_H_lo
repeat
p = p + step[p % 6]
local d = 1
repeat
d = d + step[d % 6]
if d*d > p then -- next prime number is found
local root = p^(1/3)
local R = root * 2^40
R = mul({R - R % 1}, one, 1.0, 2)
local _, delta = mul(R, mul(R, R, 1.0, 4), -1.0, 4)
local hi = R[2] % 65536 * 65536 + floor(R[1] / 256)
local lo = R[1] % 256 * 16777216 + floor(delta * (2^-56 / 3) * root / p)
if idx < 16 then
root = p^(1/2)
R = root * 2^40
R = mul({R - R % 1}, one, 1.0, 2)
_, delta = mul(R, R, -1.0, 2)
local hi = R[2] % 65536 * 65536 + floor(R[1] / 256)
local lo = R[1] % 256 * 16777216 + floor(delta * 2^-17 / root)
local idx = idx % 8 + 1
sha2_H_ext256[224][idx] = lo
sqrt_hi[idx], sqrt_lo[idx] = hi, lo + hi * hi_factor
if idx > 7 then
sqrt_hi, sqrt_lo = sha2_H_ext512_hi[384], sha2_H_ext512_lo[384]
end
end
idx = idx + 1
sha2_K_hi[idx], sha2_K_lo[idx] = hi, lo % K_lo_modulo + hi * hi_factor
break
end
until p % d == 0
until idx > 79
end
-- Calculating IVs for SHA512/224 and SHA512/256
for width = 224, 256, 32 do
local H_lo, H_hi = {}
if HEX64 then
for j = 1, 8 do
H_lo[j] = XORA5(sha2_H_lo[j])
end
else
H_hi = {}
for j = 1, 8 do
H_lo[j] = XORA5(sha2_H_lo[j])
H_hi[j] = XORA5(sha2_H_hi[j])
end
end
sha512_feed_128(H_lo, H_hi, "SHA-512/"..tostring(width).."\128"..string_rep("\0", 115).."\88", 0, 128)
sha2_H_ext512_lo[width] = H_lo
sha2_H_ext512_hi[width] = H_hi
end
-- Constants for MD5
do
local sin, abs, modf = math.sin, math.abs, math.modf
for idx = 1, 64 do
-- we can't use formula floor(abs(sin(idx))*2^32) because its result may be beyond integer range on Lua built with 32-bit integers
local hi, lo = modf(abs(sin(idx)) * 2^16)
md5_K[idx] = hi * 65536 + floor(lo * 2^16)
end
end
-- Constants for SHA-3
do
local sh_reg = 29
local function next_bit()
local r = sh_reg % 2
sh_reg = XOR_BYTE((sh_reg - r) / 2, 142 * r)
return r
end
for idx = 1, 24 do
local lo, m = 0
for _ = 1, 6 do
m = m and m * m * 2 or 1
lo = lo + next_bit() * m
end
local hi = next_bit() * m
sha3_RC_hi[idx], sha3_RC_lo[idx] = hi, lo + hi * hi_factor_keccak
end
end
if branch == "FFI" then
sha2_K_hi = ffi.new("uint32_t[?]", #sha2_K_hi + 1, 0, unpack(sha2_K_hi))
sha2_K_lo = ffi.new("int64_t[?]", #sha2_K_lo + 1, 0, unpack(sha2_K_lo))
--md5_K = ffi.new("uint32_t[?]", #md5_K + 1, 0, unpack(md5_K))
if hi_factor_keccak == 0 then
sha3_RC_lo = ffi.new("uint32_t[?]", #sha3_RC_lo + 1, 0, unpack(sha3_RC_lo))
sha3_RC_hi = ffi.new("uint32_t[?]", #sha3_RC_hi + 1, 0, unpack(sha3_RC_hi))
else
sha3_RC_lo = ffi.new("int64_t[?]", #sha3_RC_lo + 1, 0, unpack(sha3_RC_lo))
end
end
--------------------------------------------------------------------------------
-- MAIN FUNCTIONS
--------------------------------------------------------------------------------
local function sha256ext(width, message)
-- Create an instance (private objects for current calculation)
local H, length, tail = {unpack(sha2_H_ext256[width])}, 0.0, ""
local function partial(message_part)
if message_part then
if tail then
length = length + #message_part
local offs = 0
if tail ~= "" and #tail + #message_part >= 64 then
offs = 64 - #tail
sha256_feed_64(H, tail..sub(message_part, 1, offs), 0, 64)
tail = ""
end
local size = #message_part - offs
local size_tail = size % 64
sha256_feed_64(H, message_part, offs, size - size_tail)
tail = tail..sub(message_part, #message_part + 1 - size_tail)
return partial
else
error("Adding more chunks is not allowed after receiving the result", 2)
end
else
if tail then
local final_blocks = {tail, "\128", string_rep("\0", (-9 - length) % 64 + 1)}
tail = nil
-- Assuming user data length is shorter than (2^53)-9 bytes
-- Anyway, it looks very unrealistic that someone would spend more than a year of calculations to process 2^53 bytes of data by using this Lua script :-)
-- 2^53 bytes = 2^56 bits, so "bit-counter" fits in 7 bytes
length = length * (8 / 256^7) -- convert "byte-counter" to "bit-counter" and move decimal point to the left
for j = 4, 10 do
length = length % 1 * 256
final_blocks[j] = char(floor(length))
end
final_blocks = table_concat(final_blocks)
sha256_feed_64(H, final_blocks, 0, #final_blocks)
local max_reg = width / 32
for j = 1, max_reg do
H[j] = HEX(H[j])
end
H = table_concat(H, "", 1, max_reg)
end
return H
end
end
if message then
-- Actually perform calculations and return the SHA256 digest of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading
-- User should feed every chunk of input data as single argument to this function and finally get SHA256 digest by invoking this function without an argument
return partial
end
end
local function sha512ext(width, message)
-- Create an instance (private objects for current calculation)
local length, tail, H_lo, H_hi = 0.0, "", {unpack(sha2_H_ext512_lo[width])}, not HEX64 and {unpack(sha2_H_ext512_hi[width])}
local function partial(message_part)
if message_part then
if tail then
length = length + #message_part
local offs = 0
if tail ~= "" and #tail + #message_part >= 128 then
offs = 128 - #tail
sha512_feed_128(H_lo, H_hi, tail..sub(message_part, 1, offs), 0, 128)
tail = ""
end
local size = #message_part - offs
local size_tail = size % 128
sha512_feed_128(H_lo, H_hi, message_part, offs, size - size_tail)
tail = tail..sub(message_part, #message_part + 1 - size_tail)
return partial
else
error("Adding more chunks is not allowed after receiving the result", 2)
end
else
if tail then
local final_blocks = {tail, "\128", string_rep("\0", (-17-length) % 128 + 9)}
tail = nil
-- Assuming user data length is shorter than (2^53)-17 bytes
-- 2^53 bytes = 2^56 bits, so "bit-counter" fits in 7 bytes
length = length * (8 / 256^7) -- convert "byte-counter" to "bit-counter" and move floating point to the left
for j = 4, 10 do
length = length % 1 * 256
final_blocks[j] = char(floor(length))
end
final_blocks = table_concat(final_blocks)
sha512_feed_128(H_lo, H_hi, final_blocks, 0, #final_blocks)
local max_reg = ceil(width / 64)
if HEX64 then
for j = 1, max_reg do
H_lo[j] = HEX64(H_lo[j])
end
else
for j = 1, max_reg do
H_lo[j] = HEX(H_hi[j])..HEX(H_lo[j])
end
H_hi = nil
end
H_lo = sub(table_concat(H_lo, "", 1, max_reg), 1, width / 4)
end
return H_lo
end
end
if message then
-- Actually perform calculations and return the SHA512 digest of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading
-- User should feed every chunk of input data as single argument to this function and finally get SHA512 digest by invoking this function without an argument
return partial
end
end
local function md5(message)
-- Create an instance (private objects for current calculation)
local H, length, tail = {unpack(md5_sha1_H, 1, 4)}, 0.0, ""
local function partial(message_part)
if message_part then
if tail then
length = length + #message_part
local offs = 0
if tail ~= "" and #tail + #message_part >= 64 then
offs = 64 - #tail
md5_feed_64(H, tail..sub(message_part, 1, offs), 0, 64)
tail = ""
end
local size = #message_part - offs
local size_tail = size % 64
md5_feed_64(H, message_part, offs, size - size_tail)
tail = tail..sub(message_part, #message_part + 1 - size_tail)
return partial
else
error("Adding more chunks is not allowed after receiving the result", 2)
end
else
if tail then
local final_blocks = {tail, "\128", string_rep("\0", (-9 - length) % 64)}
tail = nil
length = length * 8 -- convert "byte-counter" to "bit-counter"
for j = 4, 11 do
local low_byte = length % 256
final_blocks[j] = char(low_byte)
length = (length - low_byte) / 256
end
final_blocks = table_concat(final_blocks)
md5_feed_64(H, final_blocks, 0, #final_blocks)
for j = 1, 4 do
H[j] = HEX(H[j])
end
H = gsub(table_concat(H), "(..)(..)(..)(..)", "%4%3%2%1")
end
return H
end
end
if message then
-- Actually perform calculations and return the MD5 digest of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading
-- User should feed every chunk of input data as single argument to this function and finally get MD5 digest by invoking this function without an argument
return partial
end
end
local function sha1(message)
-- Create an instance (private objects for current calculation)
local H, length, tail = {unpack(md5_sha1_H)}, 0.0, ""
local function partial(message_part)
if message_part then
if tail then
length = length + #message_part
local offs = 0
if tail ~= "" and #tail + #message_part >= 64 then
offs = 64 - #tail
sha1_feed_64(H, tail..sub(message_part, 1, offs), 0, 64)
tail = ""
end
local size = #message_part - offs
local size_tail = size % 64
sha1_feed_64(H, message_part, offs, size - size_tail)
tail = tail..sub(message_part, #message_part + 1 - size_tail)
return partial
else
error("Adding more chunks is not allowed after receiving the result", 2)
end
else
if tail then
local final_blocks = {tail, "\128", string_rep("\0", (-9 - length) % 64 + 1)}
tail = nil
-- Assuming user data length is shorter than (2^53)-9 bytes
-- 2^53 bytes = 2^56 bits, so "bit-counter" fits in 7 bytes
length = length * (8 / 256^7) -- convert "byte-counter" to "bit-counter" and move decimal point to the left
for j = 4, 10 do
length = length % 1 * 256
final_blocks[j] = char(floor(length))
end
final_blocks = table_concat(final_blocks)
sha1_feed_64(H, final_blocks, 0, #final_blocks)
for j = 1, 5 do
H[j] = HEX(H[j])
end
H = table_concat(H)
end
return H
end
end
if message then
-- Actually perform calculations and return the SHA-1 digest of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading
-- User should feed every chunk of input data as single argument to this function and finally get SHA-1 digest by invoking this function without an argument
return partial
end
end
local function keccak(block_size_in_bytes, digest_size_in_bytes, is_SHAKE, message)
-- "block_size_in_bytes" is multiple of 8
if type(digest_size_in_bytes) ~= "number" then
-- arguments in SHAKE are swapped:
-- NIST FIPS 202 defines SHAKE(message,num_bits)
-- this module defines SHAKE(num_bytes,message)
-- it's easy to forget about this swap, hence the check
error("Argument 'digest_size_in_bytes' must be a number", 2)
end
-- Create an instance (private objects for current calculation)
local tail, lanes_lo, lanes_hi = "", create_array_of_lanes(), hi_factor_keccak == 0 and create_array_of_lanes()
local result
local function partial(message_part)
if message_part then
if tail then
local offs = 0
if tail ~= "" and #tail + #message_part >= block_size_in_bytes then
offs = block_size_in_bytes - #tail
keccak_feed(lanes_lo, lanes_hi, tail..sub(message_part, 1, offs), 0, block_size_in_bytes, block_size_in_bytes)
tail = ""
end
local size = #message_part - offs
local size_tail = size % block_size_in_bytes
keccak_feed(lanes_lo, lanes_hi, message_part, offs, size - size_tail, block_size_in_bytes)
tail = tail..sub(message_part, #message_part + 1 - size_tail)
return partial
else
error("Adding more chunks is not allowed after receiving the result", 2)
end
else
if tail then
-- append the following bits to the message: for usual SHA-3: 011(0*)1, for SHAKE: 11111(0*)1
local gap_start = is_SHAKE and 31 or 6
tail = tail..(#tail + 1 == block_size_in_bytes and char(gap_start + 128) or char(gap_start)..string_rep("\0", (-2 - #tail) % block_size_in_bytes).."\128")
keccak_feed(lanes_lo, lanes_hi, tail, 0, #tail, block_size_in_bytes)
tail = nil
local lanes_used = 0
local total_lanes = floor(block_size_in_bytes / 8)
local qwords = {}
local function get_next_qwords_of_digest(qwords_qty)
-- returns not more than 'qwords_qty' qwords ('qwords_qty' might be non-integer)
-- doesn't go across keccak-buffer boundary
-- block_size_in_bytes is a multiple of 8, so, keccak-buffer contains integer number of qwords
if lanes_used >= total_lanes then
keccak_feed(lanes_lo, lanes_hi, "\0\0\0\0\0\0\0\0", 0, 8, 8)
lanes_used = 0
end
qwords_qty = floor(math_min(qwords_qty, total_lanes - lanes_used))
if hi_factor_keccak ~= 0 then
for j = 1, qwords_qty do
qwords[j] = HEX64(lanes_lo[lanes_used + j - 1 + lanes_index_base])
end
else
for j = 1, qwords_qty do
qwords[j] = HEX(lanes_hi[lanes_used + j])..HEX(lanes_lo[lanes_used + j])
end
end
lanes_used = lanes_used + qwords_qty
return
gsub(table_concat(qwords, "", 1, qwords_qty), "(..)(..)(..)(..)(..)(..)(..)(..)", "%8%7%6%5%4%3%2%1"),
qwords_qty * 8
end
local parts = {} -- digest parts
local last_part, last_part_size = "", 0
local function get_next_part_of_digest(bytes_needed)
-- returns 'bytes_needed' bytes, for arbitrary integer 'bytes_needed'
bytes_needed = bytes_needed or 1
if bytes_needed <= last_part_size then
last_part_size = last_part_size - bytes_needed
local part_size_in_nibbles = bytes_needed * 2
local result = sub(last_part, 1, part_size_in_nibbles)
last_part = sub(last_part, part_size_in_nibbles + 1)
return result
end
local parts_qty = 0
if last_part_size > 0 then
parts_qty = 1
parts[parts_qty] = last_part
bytes_needed = bytes_needed - last_part_size
end
-- repeats until the length is enough
while bytes_needed >= 8 do
local next_part, next_part_size = get_next_qwords_of_digest(bytes_needed / 8)
parts_qty = parts_qty + 1
parts[parts_qty] = next_part
bytes_needed = bytes_needed - next_part_size
end
if bytes_needed > 0 then
last_part, last_part_size = get_next_qwords_of_digest(1)
parts_qty = parts_qty + 1
parts[parts_qty] = get_next_part_of_digest(bytes_needed)
else
last_part, last_part_size = "", 0
end
return table_concat(parts, "", 1, parts_qty)
end
if digest_size_in_bytes < 0 then
result = get_next_part_of_digest
else
result = get_next_part_of_digest(digest_size_in_bytes)
end
end
return result
end
end
if message then
-- Actually perform calculations and return the SHA-3 digest of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading
-- User should feed every chunk of input data as single argument to this function and finally get SHA-3 digest by invoking this function without an argument
return partial
end
end
local hex_to_bin, bin_to_hex, bin_to_base64, base64_to_bin
do
function hex_to_bin(hex_string)
return (gsub(hex_string, "%x%x",
function (hh)
return char(tonumber(hh, 16))
end
))
end
function bin_to_hex(binary_string)
return (gsub(binary_string, ".",
function (c)
return string_format("%02x", byte(c))
end
))
end
local base64_symbols = {
['+'] = 62, ['-'] = 62, [62] = '+',
['/'] = 63, ['_'] = 63, [63] = '/',
['='] = -1, ['.'] = -1, [-1] = '='
}
local symbol_index = 0
for j, pair in ipairs{'AZ', 'az', '09'} do
for ascii = byte(pair), byte(pair, 2) do
local ch = char(ascii)
base64_symbols[ch] = symbol_index
base64_symbols[symbol_index] = ch
symbol_index = symbol_index + 1
end
end
function bin_to_base64(binary_string)
local result = {}
for pos = 1, #binary_string, 3 do
local c1, c2, c3, c4 = byte(sub(binary_string, pos, pos + 2)..'\0', 1, -1)
result[#result + 1] =
base64_symbols[floor(c1 / 4)]
..base64_symbols[c1 % 4 * 16 + floor(c2 / 16)]
..base64_symbols[c3 and c2 % 16 * 4 + floor(c3 / 64) or -1]
..base64_symbols[c4 and c3 % 64 or -1]
end
return table_concat(result)
end
function base64_to_bin(base64_string)
local result, chars_qty = {}, 3
for pos, ch in gmatch(gsub(base64_string, '%s+', ''), '()(.)') do
local code = base64_symbols[ch]
if code < 0 then
chars_qty = chars_qty - 1
code = 0
end
local idx = pos % 4
if idx > 0 then
result[-idx] = code
else
local c1 = result[-1] * 4 + floor(result[-2] / 16)
local c2 = (result[-2] % 16) * 16 + floor(result[-3] / 4)
local c3 = (result[-3] % 4) * 64 + code
result[#result + 1] = sub(char(c1, c2, c3), 1, chars_qty)
end
end
return table_concat(result)
end
end
local block_size_for_HMAC -- this table will be initialized at the end of the module
local function pad_and_xor(str, result_length, byte_for_xor)
return gsub(str, ".",
function(c)
return char(XOR_BYTE(byte(c), byte_for_xor))
end
)..string_rep(char(byte_for_xor), result_length - #str)
end
local function hmac(hash_func, key, message)
-- Create an instance (private objects for current calculation)
local block_size = block_size_for_HMAC[hash_func]
if not block_size then
error("Unknown hash function", 2)
end
if #key > block_size then
key = hex_to_bin(hash_func(key))
end
local append = hash_func()(pad_and_xor(key, block_size, 0x36))
local result
local function partial(message_part)
if not message_part then
result = result or hash_func(pad_and_xor(key, block_size, 0x5C)..hex_to_bin(append()))
return result
elseif result then
error("Adding more chunks is not allowed after receiving the result", 2)
else
append(message_part)
return partial
end
end
if message then
-- Actually perform calculations and return the HMAC of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading of a message
-- User should feed every chunk of the message as single argument to this function and finally get HMAC by invoking this function without an argument
return partial
end
end
local function xor_blake2_salt(salt, letter, H_lo, H_hi)
-- salt: concatenation of "Salt"+"Personalization" fields
local max_size = letter == "s" and 16 or 32
local salt_size = #salt
if salt_size > max_size then
error(string_format("For BLAKE2%s/BLAKE2%sp/BLAKE2X%s the 'salt' parameter length must not exceed %d bytes", letter, letter, letter, max_size), 2)
end
if H_lo then
local offset, blake2_word_size, xor = 0, letter == "s" and 4 or 8, letter == "s" and XOR or XORA5
for j = 5, 4 + ceil(salt_size / blake2_word_size) do
local prev, last
for _ = 1, blake2_word_size, 4 do
offset = offset + 4
local a, b, c, d = byte(salt, offset - 3, offset)
local four_bytes = (((d or 0) * 256 + (c or 0)) * 256 + (b or 0)) * 256 + (a or 0)
prev, last = last, four_bytes
end
H_lo[j] = xor(H_lo[j], prev and last * hi_factor + prev or last)
if H_hi then
H_hi[j] = xor(H_hi[j], last)
end
end
end
end
local function blake2s(message, key, salt, digest_size_in_bytes, XOF_length, B2_offset)
-- message: binary string to be hashed (or nil for "chunk-by-chunk" input mode)
-- key: (optional) binary string up to 32 bytes, by default empty string
-- salt: (optional) binary string up to 16 bytes, by default empty string
-- digest_size_in_bytes: (optional) integer from 1 to 32, by default 32
-- The last two parameters "XOF_length" and "B2_offset" are for internal use only, user must omit them (or pass nil)
digest_size_in_bytes = digest_size_in_bytes or 32
if digest_size_in_bytes < 1 or digest_size_in_bytes > 32 then
error("BLAKE2s digest length must be from 1 to 32 bytes", 2)
end
key = key or ""
local key_length = #key
if key_length > 32 then
error("BLAKE2s key length must not exceed 32 bytes", 2)
end
salt = salt or ""
local bytes_compressed, tail, H = 0.0, "", {unpack(sha2_H_hi)}
if B2_offset then
H[1] = XOR(H[1], digest_size_in_bytes)
H[2] = XOR(H[2], 0x20)
H[3] = XOR(H[3], B2_offset)
H[4] = XOR(H[4], 0x20000000 + XOF_length)
else
H[1] = XOR(H[1], 0x01010000 + key_length * 256 + digest_size_in_bytes)
if XOF_length then
H[4] = XOR(H[4], XOF_length)
end
end
if salt ~= "" then
xor_blake2_salt(salt, "s", H)
end
local function partial(message_part)
if message_part then
if tail then
local offs = 0
if tail ~= "" and #tail + #message_part > 64 then
offs = 64 - #tail
bytes_compressed = blake2s_feed_64(H, tail..sub(message_part, 1, offs), 0, 64, bytes_compressed)
tail = ""
end
local size = #message_part - offs
local size_tail = size > 0 and (size - 1) % 64 + 1 or 0
bytes_compressed = blake2s_feed_64(H, message_part, offs, size - size_tail, bytes_compressed)
tail = tail..sub(message_part, #message_part + 1 - size_tail)
return partial
else
error("Adding more chunks is not allowed after receiving the result", 2)
end
else
if tail then
if B2_offset then
blake2s_feed_64(H, nil, 0, 64, 0, 32)
else
blake2s_feed_64(H, tail..string_rep("\0", 64 - #tail), 0, 64, bytes_compressed, #tail)
end
tail = nil
if not XOF_length or B2_offset then
local max_reg = ceil(digest_size_in_bytes / 4)
for j = 1, max_reg do
H[j] = HEX(H[j])
end
H = sub(gsub(table_concat(H, "", 1, max_reg), "(..)(..)(..)(..)", "%4%3%2%1"), 1, digest_size_in_bytes * 2)
end
end
return H
end
end
if key_length > 0 then
partial(key..string_rep("\0", 64 - key_length))
end
if B2_offset then
return partial()
elseif message then
-- Actually perform calculations and return the BLAKE2s digest of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading
-- User should feed every chunk of input data as single argument to this function and finally get BLAKE2s digest by invoking this function without an argument
return partial
end
end
local function blake2b(message, key, salt, digest_size_in_bytes, XOF_length, B2_offset)
-- message: binary string to be hashed (or nil for "chunk-by-chunk" input mode)
-- key: (optional) binary string up to 64 bytes, by default empty string
-- salt: (optional) binary string up to 32 bytes, by default empty string
-- digest_size_in_bytes: (optional) integer from 1 to 64, by default 64
-- The last two parameters "XOF_length" and "B2_offset" are for internal use only, user must omit them (or pass nil)
digest_size_in_bytes = floor(digest_size_in_bytes or 64)
if digest_size_in_bytes < 1 or digest_size_in_bytes > 64 then
error("BLAKE2b digest length must be from 1 to 64 bytes", 2)
end
key = key or ""
local key_length = #key
if key_length > 64 then
error("BLAKE2b key length must not exceed 64 bytes", 2)
end
salt = salt or ""
local bytes_compressed, tail, H_lo, H_hi = 0.0, "", {unpack(sha2_H_lo)}, not HEX64 and {unpack(sha2_H_hi)}
if B2_offset then
if H_hi then
H_lo[1] = XORA5(H_lo[1], digest_size_in_bytes)
H_hi[1] = XORA5(H_hi[1], 0x40)
H_lo[2] = XORA5(H_lo[2], B2_offset)
H_hi[2] = XORA5(H_hi[2], XOF_length)
else
H_lo[1] = XORA5(H_lo[1], 0x40 * hi_factor + digest_size_in_bytes)
H_lo[2] = XORA5(H_lo[2], XOF_length * hi_factor + B2_offset)
end
H_lo[3] = XORA5(H_lo[3], 0x4000)
else
H_lo[1] = XORA5(H_lo[1], 0x01010000 + key_length * 256 + digest_size_in_bytes)
if XOF_length then
if H_hi then
H_hi[2] = XORA5(H_hi[2], XOF_length)
else
H_lo[2] = XORA5(H_lo[2], XOF_length * hi_factor)
end
end
end
if salt ~= "" then
xor_blake2_salt(salt, "b", H_lo, H_hi)
end
local function partial(message_part)
if message_part then
if tail then
local offs = 0
if tail ~= "" and #tail + #message_part > 128 then
offs = 128 - #tail
bytes_compressed = blake2b_feed_128(H_lo, H_hi, tail..sub(message_part, 1, offs), 0, 128, bytes_compressed)
tail = ""
end
local size = #message_part - offs
local size_tail = size > 0 and (size - 1) % 128 + 1 or 0
bytes_compressed = blake2b_feed_128(H_lo, H_hi, message_part, offs, size - size_tail, bytes_compressed)
tail = tail..sub(message_part, #message_part + 1 - size_tail)
return partial
else
error("Adding more chunks is not allowed after receiving the result", 2)
end
else
if tail then
if B2_offset then
blake2b_feed_128(H_lo, H_hi, nil, 0, 128, 0, 64)
else
blake2b_feed_128(H_lo, H_hi, tail..string_rep("\0", 128 - #tail), 0, 128, bytes_compressed, #tail)
end
tail = nil
if XOF_length and not B2_offset then
if H_hi then
for j = 8, 1, -1 do
H_lo[j*2] = H_hi[j]
H_lo[j*2-1] = H_lo[j]
end
return H_lo, 16
end
else
local max_reg = ceil(digest_size_in_bytes / 8)
if H_hi then
for j = 1, max_reg do
H_lo[j] = HEX(H_hi[j])..HEX(H_lo[j])
end
else
for j = 1, max_reg do
H_lo[j] = HEX64(H_lo[j])
end
end
H_lo = sub(gsub(table_concat(H_lo, "", 1, max_reg), "(..)(..)(..)(..)(..)(..)(..)(..)", "%8%7%6%5%4%3%2%1"), 1, digest_size_in_bytes * 2)
end
H_hi = nil
end
return H_lo
end
end
if key_length > 0 then
partial(key..string_rep("\0", 128 - key_length))
end
if B2_offset then
return partial()
elseif message then
-- Actually perform calculations and return the BLAKE2b digest of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading
-- User should feed every chunk of input data as single argument to this function and finally get BLAKE2b digest by invoking this function without an argument
return partial
end
end
local function blake2sp(message, key, salt, digest_size_in_bytes)
-- message: binary string to be hashed (or nil for "chunk-by-chunk" input mode)
-- key: (optional) binary string up to 32 bytes, by default empty string
-- salt: (optional) binary string up to 16 bytes, by default empty string
-- digest_size_in_bytes: (optional) integer from 1 to 32, by default 32
digest_size_in_bytes = digest_size_in_bytes or 32
if digest_size_in_bytes < 1 or digest_size_in_bytes > 32 then
error("BLAKE2sp digest length must be from 1 to 32 bytes", 2)
end
key = key or ""
local key_length = #key
if key_length > 32 then
error("BLAKE2sp key length must not exceed 32 bytes", 2)
end
salt = salt or ""
local instances, length, first_dword_of_parameter_block, result = {}, 0.0, 0x02080000 + key_length * 256 + digest_size_in_bytes
for j = 1, 8 do
local bytes_compressed, tail, H = 0.0, "", {unpack(sha2_H_hi)}
instances[j] = {bytes_compressed, tail, H}
H[1] = XOR(H[1], first_dword_of_parameter_block)
H[3] = XOR(H[3], j-1)
H[4] = XOR(H[4], 0x20000000)
if salt ~= "" then
xor_blake2_salt(salt, "s", H)
end
end
local function partial(message_part)
if message_part then
if instances then
local from = 0
while true do
local to = math_min(from + 64 - length % 64, #message_part)
if to > from then
local inst = instances[floor(length / 64) % 8 + 1]
local part = sub(message_part, from + 1, to)
length, from = length + to - from, to
local bytes_compressed, tail = inst[1], inst[2]
if #tail < 64 then
tail = tail..part
else
local H = inst[3]
bytes_compressed = blake2s_feed_64(H, tail, 0, 64, bytes_compressed)
tail = part
end
inst[1], inst[2] = bytes_compressed, tail
else
break
end
end
return partial
else
error("Adding more chunks is not allowed after receiving the result", 2)
end
else
if instances then
local root_H = {unpack(sha2_H_hi)}
root_H[1] = XOR(root_H[1], first_dword_of_parameter_block)
root_H[4] = XOR(root_H[4], 0x20010000)
if salt ~= "" then
xor_blake2_salt(salt, "s", root_H)
end
for j = 1, 8 do
local inst = instances[j]
local bytes_compressed, tail, H = inst[1], inst[2], inst[3]
blake2s_feed_64(H, tail..string_rep("\0", 64 - #tail), 0, 64, bytes_compressed, #tail, j == 8)
if j % 2 == 0 then
local index = 0
for k = j - 1, j do
local inst = instances[k]
local H = inst[3]
for i = 1, 8 do
index = index + 1
common_W_blake2s[index] = H[i]
end
end
blake2s_feed_64(root_H, nil, 0, 64, 64 * (j/2 - 1), j == 8 and 64, j == 8)
end
end
instances = nil
local max_reg = ceil(digest_size_in_bytes / 4)
for j = 1, max_reg do
root_H[j] = HEX(root_H[j])
end
result = sub(gsub(table_concat(root_H, "", 1, max_reg), "(..)(..)(..)(..)", "%4%3%2%1"), 1, digest_size_in_bytes * 2)
end
return result
end
end
if key_length > 0 then
key = key..string_rep("\0", 64 - key_length)
for j = 1, 8 do
partial(key)
end
end
if message then
-- Actually perform calculations and return the BLAKE2sp digest of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading
-- User should feed every chunk of input data as single argument to this function and finally get BLAKE2sp digest by invoking this function without an argument
return partial
end
end
local function blake2bp(message, key, salt, digest_size_in_bytes)
-- message: binary string to be hashed (or nil for "chunk-by-chunk" input mode)
-- key: (optional) binary string up to 64 bytes, by default empty string
-- salt: (optional) binary string up to 32 bytes, by default empty string
-- digest_size_in_bytes: (optional) integer from 1 to 64, by default 64
digest_size_in_bytes = digest_size_in_bytes or 64
if digest_size_in_bytes < 1 or digest_size_in_bytes > 64 then
error("BLAKE2bp digest length must be from 1 to 64 bytes", 2)
end
key = key or ""
local key_length = #key
if key_length > 64 then
error("BLAKE2bp key length must not exceed 64 bytes", 2)
end
salt = salt or ""
local instances, length, first_dword_of_parameter_block, result = {}, 0.0, 0x02040000 + key_length * 256 + digest_size_in_bytes
for j = 1, 4 do
local bytes_compressed, tail, H_lo, H_hi = 0.0, "", {unpack(sha2_H_lo)}, not HEX64 and {unpack(sha2_H_hi)}
instances[j] = {bytes_compressed, tail, H_lo, H_hi}
H_lo[1] = XORA5(H_lo[1], first_dword_of_parameter_block)
H_lo[2] = XORA5(H_lo[2], j-1)
H_lo[3] = XORA5(H_lo[3], 0x4000)
if salt ~= "" then
xor_blake2_salt(salt, "b", H_lo, H_hi)
end
end
local function partial(message_part)
if message_part then
if instances then
local from = 0
while true do
local to = math_min(from + 128 - length % 128, #message_part)
if to > from then
local inst = instances[floor(length / 128) % 4 + 1]
local part = sub(message_part, from + 1, to)
length, from = length + to - from, to
local bytes_compressed, tail = inst[1], inst[2]
if #tail < 128 then
tail = tail..part
else
local H_lo, H_hi = inst[3], inst[4]
bytes_compressed = blake2b_feed_128(H_lo, H_hi, tail, 0, 128, bytes_compressed)
tail = part
end
inst[1], inst[2] = bytes_compressed, tail
else
break
end
end
return partial
else
error("Adding more chunks is not allowed after receiving the result", 2)
end
else
if instances then
local root_H_lo, root_H_hi = {unpack(sha2_H_lo)}, not HEX64 and {unpack(sha2_H_hi)}
root_H_lo[1] = XORA5(root_H_lo[1], first_dword_of_parameter_block)
root_H_lo[3] = XORA5(root_H_lo[3], 0x4001)
if salt ~= "" then
xor_blake2_salt(salt, "b", root_H_lo, root_H_hi)
end
for j = 1, 4 do
local inst = instances[j]
local bytes_compressed, tail, H_lo, H_hi = inst[1], inst[2], inst[3], inst[4]
blake2b_feed_128(H_lo, H_hi, tail..string_rep("\0", 128 - #tail), 0, 128, bytes_compressed, #tail, j == 4)
if j % 2 == 0 then
local index = 0
for k = j - 1, j do
local inst = instances[k]
local H_lo, H_hi = inst[3], inst[4]
for i = 1, 8 do
index = index + 1
common_W_blake2b[index] = H_lo[i]
if H_hi then
index = index + 1
common_W_blake2b[index] = H_hi[i]
end
end
end
blake2b_feed_128(root_H_lo, root_H_hi, nil, 0, 128, 128 * (j/2 - 1), j == 4 and 128, j == 4)
end
end
instances = nil
local max_reg = ceil(digest_size_in_bytes / 8)
if HEX64 then
for j = 1, max_reg do
root_H_lo[j] = HEX64(root_H_lo[j])
end
else
for j = 1, max_reg do
root_H_lo[j] = HEX(root_H_hi[j])..HEX(root_H_lo[j])
end
end
result = sub(gsub(table_concat(root_H_lo, "", 1, max_reg), "(..)(..)(..)(..)(..)(..)(..)(..)", "%8%7%6%5%4%3%2%1"), 1, digest_size_in_bytes * 2)
end
return result
end
end
if key_length > 0 then
key = key..string_rep("\0", 128 - key_length)
for j = 1, 4 do
partial(key)
end
end
if message then
-- Actually perform calculations and return the BLAKE2bp digest of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading
-- User should feed every chunk of input data as single argument to this function and finally get BLAKE2bp digest by invoking this function without an argument
return partial
end
end
local function blake2x(inner_func, inner_func_letter, common_W_blake2, block_size, digest_size_in_bytes, message, key, salt)
local XOF_digest_length_limit, XOF_digest_length, chunk_by_chunk_output = 2^(block_size / 2) - 1
if digest_size_in_bytes == -1 then -- infinite digest
digest_size_in_bytes = math_huge
XOF_digest_length = floor(XOF_digest_length_limit)
chunk_by_chunk_output = true
else
if digest_size_in_bytes < 0 then
digest_size_in_bytes = -1.0 * digest_size_in_bytes
chunk_by_chunk_output = true
end
XOF_digest_length = floor(digest_size_in_bytes)
if XOF_digest_length >= XOF_digest_length_limit then
error("Requested digest is too long. BLAKE2X"..inner_func_letter.." finite digest is limited by (2^"..floor(block_size / 2)..")-2 bytes. Hint: you can generate infinite digest.", 2)
end
end
salt = salt or ""
if salt ~= "" then
xor_blake2_salt(salt, inner_func_letter) -- don't xor, only check the size of salt
end
local inner_partial = inner_func(nil, key, salt, nil, XOF_digest_length)
local result
local function partial(message_part)
if message_part then
if inner_partial then
inner_partial(message_part)
return partial
else
error("Adding more chunks is not allowed after receiving the result", 2)
end
else
if inner_partial then
local half_W, half_W_size = inner_partial()
half_W_size, inner_partial = half_W_size or 8
local function get_hash_block(block_no)
-- block_no = 0...(2^32-1)
local size = math_min(block_size, digest_size_in_bytes - block_no * block_size)
if size <= 0 then
return ""
end
for j = 1, half_W_size do
common_W_blake2[j] = half_W[j]
end
for j = half_W_size + 1, 2 * half_W_size do
common_W_blake2[j] = 0
end
return inner_func(nil, nil, salt, size, XOF_digest_length, floor(block_no))
end
local hash = {}
if chunk_by_chunk_output then
local pos, period, cached_block_no, cached_block = 0, block_size * 2^32
local function get_next_part_of_digest(arg1, arg2)
if arg1 == "seek" then
-- Usage #1: get_next_part_of_digest("seek", new_pos)
pos = arg2 % period
else
-- Usage #2: hex_string = get_next_part_of_digest(size)
local size, index = arg1 or 1, 0
while size > 0 do
local block_offset = pos % block_size
local block_no = (pos - block_offset) / block_size
local part_size = math_min(size, block_size - block_offset)
if cached_block_no ~= block_no then
cached_block_no = block_no
cached_block = get_hash_block(block_no)
end
index = index + 1
hash[index] = sub(cached_block, block_offset * 2 + 1, (block_offset + part_size) * 2)
size = size - part_size
pos = (pos + part_size) % period
end
return table_concat(hash, "", 1, index)
end
end
result = get_next_part_of_digest
else
for j = 1.0, ceil(digest_size_in_bytes / block_size) do
hash[j] = get_hash_block(j - 1.0)
end
result = table_concat(hash)
end
end
return result
end
end
if message then
-- Actually perform calculations and return the BLAKE2X digest of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading
-- User should feed every chunk of input data as single argument to this function and finally get BLAKE2X digest by invoking this function without an argument
return partial
end
end
local function blake2xs(digest_size_in_bytes, message, key, salt)
-- digest_size_in_bytes:
-- 0..65534 = get finite digest as single Lua string
-- (-1) = get infinite digest in "chunk-by-chunk" output mode
-- (-2)..(-65534) = get finite digest in "chunk-by-chunk" output mode
-- message: binary string to be hashed (or nil for "chunk-by-chunk" input mode)
-- key: (optional) binary string up to 32 bytes, by default empty string
-- salt: (optional) binary string up to 16 bytes, by default empty string
return blake2x(blake2s, "s", common_W_blake2s, 32, digest_size_in_bytes, message, key, salt)
end
local function blake2xb(digest_size_in_bytes, message, key, salt)
-- digest_size_in_bytes:
-- 0..4294967294 = get finite digest as single Lua string
-- (-1) = get infinite digest in "chunk-by-chunk" output mode
-- (-2)..(-4294967294) = get finite digest in "chunk-by-chunk" output mode
-- message: binary string to be hashed (or nil for "chunk-by-chunk" input mode)
-- key: (optional) binary string up to 64 bytes, by default empty string
-- salt: (optional) binary string up to 32 bytes, by default empty string
return blake2x(blake2b, "b", common_W_blake2b, 64, digest_size_in_bytes, message, key, salt)
end
local function blake3(message, key, digest_size_in_bytes, message_flags, K, return_array)
-- message: binary string to be hashed (or nil for "chunk-by-chunk" input mode)
-- key: (optional) binary string up to 32 bytes, by default empty string
-- digest_size_in_bytes: (optional) by default 32
-- 0,1,2,3,4,... = get finite digest as single Lua string
-- (-1) = get infinite digest in "chunk-by-chunk" output mode
-- -2,-3,-4,... = get finite digest in "chunk-by-chunk" output mode
-- The last three parameters "message_flags", "K" and "return_array" are for internal use only, user must omit them (or pass nil)
key = key or ""
digest_size_in_bytes = digest_size_in_bytes or 32
message_flags = message_flags or 0
if key == "" then
K = K or sha2_H_hi
else
local key_length = #key
if key_length > 32 then
error("BLAKE3 key length must not exceed 32 bytes", 2)
end
key = key..string_rep("\0", 32 - key_length)
K = {}
for j = 1, 8 do
local a, b, c, d = byte(key, 4*j-3, 4*j)
K[j] = ((d * 256 + c) * 256 + b) * 256 + a
end
message_flags = message_flags + 16 -- flag:KEYED_HASH
end
local tail, H, chunk_index, blocks_in_chunk, stack_size, stack = "", {}, 0, 0, 0, {}
local final_H_in, final_block_length, chunk_by_chunk_output, result, wide_output = K
local final_compression_flags = 3 -- flags:CHUNK_START,CHUNK_END
local function feed_blocks(str, offs, size)
-- size >= 0, size is multiple of 64
while size > 0 do
local part_size_in_blocks, block_flags, H_in = 1, 0, H
if blocks_in_chunk == 0 then
block_flags = 1 -- flag:CHUNK_START
H_in, final_H_in = K, H
final_compression_flags = 2 -- flag:CHUNK_END
elseif blocks_in_chunk == 15 then
block_flags = 2 -- flag:CHUNK_END
final_compression_flags = 3 -- flags:CHUNK_START,CHUNK_END
final_H_in = K
else
part_size_in_blocks = math_min(size / 64, 15 - blocks_in_chunk)
end
local part_size = part_size_in_blocks * 64
blake3_feed_64(str, offs, part_size, message_flags + block_flags, chunk_index, H_in, H)
offs, size = offs + part_size, size - part_size
blocks_in_chunk = (blocks_in_chunk + part_size_in_blocks) % 16
if blocks_in_chunk == 0 then
-- completing the currect chunk
chunk_index = chunk_index + 1.0
local divider = 2.0
while chunk_index % divider == 0 do
divider = divider * 2.0
stack_size = stack_size - 8
for j = 1, 8 do
common_W_blake2s[j] = stack[stack_size + j]
end
for j = 1, 8 do
common_W_blake2s[j + 8] = H[j]
end
blake3_feed_64(nil, 0, 64, message_flags + 4, 0, K, H) -- flag:PARENT
end
for j = 1, 8 do
stack[stack_size + j] = H[j]
end
stack_size = stack_size + 8
end
end
end
local function get_hash_block(block_no)
local size = math_min(64, digest_size_in_bytes - block_no * 64)
if block_no < 0 or size <= 0 then
return ""
end
if chunk_by_chunk_output then
for j = 1, 16 do
common_W_blake2s[j] = stack[j + 16]
end
end
blake3_feed_64(nil, 0, 64, final_compression_flags, block_no, final_H_in, stack, wide_output, final_block_length)
if return_array then
return stack
end
local max_reg = ceil(size / 4)
for j = 1, max_reg do
stack[j] = HEX(stack[j])
end
return sub(gsub(table_concat(stack, "", 1, max_reg), "(..)(..)(..)(..)", "%4%3%2%1"), 1, size * 2)
end
local function partial(message_part)
if message_part then
if tail then
local offs = 0
if tail ~= "" and #tail + #message_part > 64 then
offs = 64 - #tail
feed_blocks(tail..sub(message_part, 1, offs), 0, 64)
tail = ""
end
local size = #message_part - offs
local size_tail = size > 0 and (size - 1) % 64 + 1 or 0
feed_blocks(message_part, offs, size - size_tail)
tail = tail..sub(message_part, #message_part + 1 - size_tail)
return partial
else
error("Adding more chunks is not allowed after receiving the result", 2)
end
else
if tail then
final_block_length = #tail
tail = tail..string_rep("\0", 64 - #tail)
if common_W_blake2s[0] then
for j = 1, 16 do
local a, b, c, d = byte(tail, 4*j-3, 4*j)
common_W_blake2s[j] = OR(SHL(d, 24), SHL(c, 16), SHL(b, 8), a)
end
else
for j = 1, 16 do
local a, b, c, d = byte(tail, 4*j-3, 4*j)
common_W_blake2s[j] = ((d * 256 + c) * 256 + b) * 256 + a
end
end
tail = nil
for stack_size = stack_size - 8, 0, -8 do
blake3_feed_64(nil, 0, 64, message_flags + final_compression_flags, chunk_index, final_H_in, H, nil, final_block_length)
chunk_index, final_block_length, final_H_in, final_compression_flags = 0, 64, K, 4 -- flag:PARENT
for j = 1, 8 do
common_W_blake2s[j] = stack[stack_size + j]
end
for j = 1, 8 do
common_W_blake2s[j + 8] = H[j]
end
end
final_compression_flags = message_flags + final_compression_flags + 8 -- flag:ROOT
if digest_size_in_bytes < 0 then
if digest_size_in_bytes == -1 then -- infinite digest
digest_size_in_bytes = math_huge
else
digest_size_in_bytes = -1.0 * digest_size_in_bytes
end
chunk_by_chunk_output = true
for j = 1, 16 do
stack[j + 16] = common_W_blake2s[j]
end
end
digest_size_in_bytes = math_min(2^53, digest_size_in_bytes)
wide_output = digest_size_in_bytes > 32
if chunk_by_chunk_output then
local pos, cached_block_no, cached_block = 0.0
local function get_next_part_of_digest(arg1, arg2)
if arg1 == "seek" then
-- Usage #1: get_next_part_of_digest("seek", new_pos)
pos = arg2 * 1.0
else
-- Usage #2: hex_string = get_next_part_of_digest(size)
local size, index = arg1 or 1, 32
while size > 0 do
local block_offset = pos % 64
local block_no = (pos - block_offset) / 64
local part_size = math_min(size, 64 - block_offset)
if cached_block_no ~= block_no then
cached_block_no = block_no
cached_block = get_hash_block(block_no)
end
index = index + 1
stack[index] = sub(cached_block, block_offset * 2 + 1, (block_offset + part_size) * 2)
size = size - part_size
pos = pos + part_size
end
return table_concat(stack, "", 33, index)
end
end
result = get_next_part_of_digest
elseif digest_size_in_bytes <= 64 then
result = get_hash_block(0)
else
local last_block_no = ceil(digest_size_in_bytes / 64) - 1
for block_no = 0.0, last_block_no do
stack[33 + block_no] = get_hash_block(block_no)
end
result = table_concat(stack, "", 33, 33 + last_block_no)
end
end
return result
end
end
if message then
-- Actually perform calculations and return the BLAKE3 digest of a message
return partial(message)()
else
-- Return function for chunk-by-chunk loading
-- User should feed every chunk of input data as single argument to this function and finally get BLAKE3 digest by invoking this function without an argument
return partial
end
end
local function blake3_derive_key(key_material, context_string, derived_key_size_in_bytes)
-- key_material: (string) your source of entropy to derive a key from (for example, it can be a master password)
-- set to nil for feeding the key material in "chunk-by-chunk" input mode
-- context_string: (string) unique description of the derived key
-- digest_size_in_bytes: (optional) by default 32
-- 0,1,2,3,4,... = get finite derived key as single Lua string
-- (-1) = get infinite derived key in "chunk-by-chunk" output mode
-- -2,-3,-4,... = get finite derived key in "chunk-by-chunk" output mode
if type(context_string) ~= "string" then
error("'context_string' parameter must be a Lua string", 2)
end
local K = blake3(context_string, nil, nil, 32, nil, true) -- flag:DERIVE_KEY_CONTEXT
return blake3(key_material, nil, derived_key_size_in_bytes, 64, K) -- flag:DERIVE_KEY_MATERIAL
end
local sha = {
md5 = md5, -- MD5
sha1 = sha1, -- SHA-1
-- SHA-2 hash functions:
sha224 = function (message) return sha256ext(224, message) end, -- SHA-224
sha256 = function (message) return sha256ext(256, message) end, -- SHA-256
sha512_224 = function (message) return sha512ext(224, message) end, -- SHA-512/224
sha512_256 = function (message) return sha512ext(256, message) end, -- SHA-512/256
sha384 = function (message) return sha512ext(384, message) end, -- SHA-384
sha512 = function (message) return sha512ext(512, message) end, -- SHA-512
-- SHA-3 hash functions:
sha3_224 = function (message) return keccak((1600 - 2 * 224) / 8, 224 / 8, false, message) end, -- SHA3-224
sha3_256 = function (message) return keccak((1600 - 2 * 256) / 8, 256 / 8, false, message) end, -- SHA3-256
sha3_384 = function (message) return keccak((1600 - 2 * 384) / 8, 384 / 8, false, message) end, -- SHA3-384
sha3_512 = function (message) return keccak((1600 - 2 * 512) / 8, 512 / 8, false, message) end, -- SHA3-512
shake128 = function (digest_size_in_bytes, message) return keccak((1600 - 2 * 128) / 8, digest_size_in_bytes, true, message) end, -- SHAKE128
shake256 = function (digest_size_in_bytes, message) return keccak((1600 - 2 * 256) / 8, digest_size_in_bytes, true, message) end, -- SHAKE256
-- HMAC:
hmac = hmac, -- HMAC(hash_func, key, message) is applicable to any hash function from this module except SHAKE* and BLAKE*
-- misc utilities:
hex_to_bin = hex_to_bin, -- converts hexadecimal representation to binary string
bin_to_hex = bin_to_hex, -- converts binary string to hexadecimal representation
base64_to_bin = base64_to_bin, -- converts base64 representation to binary string
bin_to_base64 = bin_to_base64, -- converts binary string to base64 representation
-- old style names for backward compatibility:
hex2bin = hex_to_bin,
bin2hex = bin_to_hex,
base642bin = base64_to_bin,
bin2base64 = bin_to_base64,
-- BLAKE2 hash functions:
blake2b = blake2b, -- BLAKE2b (message, key, salt, digest_size_in_bytes)
blake2s = blake2s, -- BLAKE2s (message, key, salt, digest_size_in_bytes)
blake2bp = blake2bp, -- BLAKE2bp(message, key, salt, digest_size_in_bytes)
blake2sp = blake2sp, -- BLAKE2sp(message, key, salt, digest_size_in_bytes)
blake2xb = blake2xb, -- BLAKE2Xb(digest_size_in_bytes, message, key, salt)
blake2xs = blake2xs, -- BLAKE2Xs(digest_size_in_bytes, message, key, salt)
-- BLAKE2 aliases:
blake2 = blake2b,
blake2b_160 = function (message, key, salt) return blake2b(message, key, salt, 20) end, -- BLAKE2b-160
blake2b_256 = function (message, key, salt) return blake2b(message, key, salt, 32) end, -- BLAKE2b-256
blake2b_384 = function (message, key, salt) return blake2b(message, key, salt, 48) end, -- BLAKE2b-384
blake2b_512 = blake2b, -- 64 -- BLAKE2b-512
blake2s_128 = function (message, key, salt) return blake2s(message, key, salt, 16) end, -- BLAKE2s-128
blake2s_160 = function (message, key, salt) return blake2s(message, key, salt, 20) end, -- BLAKE2s-160
blake2s_224 = function (message, key, salt) return blake2s(message, key, salt, 28) end, -- BLAKE2s-224
blake2s_256 = blake2s, -- 32 -- BLAKE2s-256
-- BLAKE3 hash function
blake3 = blake3, -- BLAKE3 (message, key, digest_size_in_bytes)
blake3_derive_key = blake3_derive_key, -- BLAKE3_KDF(key_material, context_string, derived_key_size_in_bytes)
}
block_size_for_HMAC = {
[sha.md5] = 64,
[sha.sha1] = 64,
[sha.sha224] = 64,
[sha.sha256] = 64,
[sha.sha512_224] = 128,
[sha.sha512_256] = 128,
[sha.sha384] = 128,
[sha.sha512] = 128,
[sha.sha3_224] = 144, -- (1600 - 2 * 224) / 8
[sha.sha3_256] = 136, -- (1600 - 2 * 256) / 8
[sha.sha3_384] = 104, -- (1600 - 2 * 384) / 8
[sha.sha3_512] = 72, -- (1600 - 2 * 512) / 8
}
return sha]]>
{A60F2E59-B422-42FD-B1D7-9F0E0CB50A27}
0
false
SHA256
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{F0267BB5-5A78-451E-BC73-532536BA1902}
0
false
Base64
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{CAD5E8F2-F868-4EDC-977E-9E97D8BC1CA1}
0
false
XOR
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{6D01652D-7990-439D-A9D5-2B8500362AEA}
0
false
OriginalSystemPromptWORKING
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
Vibe Coder v1.1.0
0
false
VersionTitle
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
0
0
0
false
Imgs
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
rbxassetid://129168231652243
1
1
1
0
0
0
0
0
0
0
0
0
0
0
1
1
0
1
0
false
0
0
0
1
1
1
0
0
0
0
0
0
false
false
true
0
[null]
[null]
[null]
[null]
0
0
0
0
0
false
[null]
0
0
0
100
0
100
0
true
1
true
[null]
0
0
0
0
false
0
false
ImageLabel
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
0
0
0
false
App
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{50255a2a-3d89-4600-815d-7bd8f8aa7276}
0
false
State
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
(boolean, string?)): (boolean, string?)
local recording = ChangeHistoryService:TryBeginRecording(name)
local ok, result, err = pcall(fn)
if not ok then
if recording then
ChangeHistoryService:FinishRecording(recording, Enum.FinishRecordingOperation.Cancel)
end
return false, tostring(result)
end
if recording then
ChangeHistoryService:FinishRecording(
recording,
result and Enum.FinishRecordingOperation.Commit or Enum.FinishRecordingOperation.Cancel
)
end
return result, err
end
function DiffReview.applyEdit(edit): (boolean, string?)
if typeof(edit) ~= "table" then
return false, "invalid edit"
end
if edit.applied then
return false, "already applied"
end
if edit.action == "edit" then
local inst = edit.instance
if not inst or not inst.Parent then
local resolved, err = PathUtil.resolveScript(edit.path)
if not resolved then
return false, err
end
inst = resolved
edit.instance = inst
end
-- Refuse rather than overwrite if the script moved on since we proposed this.
local drifted = false
local ok, err = pcall(function()
ScriptEditorService:UpdateSourceAsync(inst, function(old)
if typeof(edit.oldSource) == "string" and old ~= edit.oldSource then
drifted = true
return nil
end
return edit.newSource
end)
end)
if drifted then
return false,
"changed since this edit was proposed — reject it and ask again so your own changes are not lost"
end
if not ok then
return false, tostring(err)
end
edit.applied = true
return true, nil
end
if edit.action == "create" then
local parent, err = PathUtil.resolve(edit.parentPath)
if not parent then
return false, err
end
local className = tostring(edit.className or "")
if className ~= "Script" and className ~= "LocalScript" and className ~= "ModuleScript" then
return false, "invalid className"
end
local childName = tostring(edit.name or "")
if childName == "" then
return false, "missing name"
end
if parent:FindFirstChild(childName) then
return false, "child already exists"
end
return withWaypoint("Vibe Coder: create " .. childName, function()
local scr = Instance.new(className)
scr.Name = childName
scr.Source = tostring(edit.newSource or "")
scr.Parent = parent
edit.instance = scr
edit.applied = true
return true, nil
end)
end
return false, "unknown action"
end
function DiffReview.revertEdit(edit): (boolean, string?)
if typeof(edit) ~= "table" then
return false, "invalid edit"
end
if not edit.applied then
return false, "not applied"
end
if edit.action == "edit" then
local inst = edit.instance
if not inst then
return false, "missing instance"
end
local ok, err = pcall(function()
ScriptEditorService:UpdateSourceAsync(inst, function(_old)
return edit.oldSource or ""
end)
end)
if not ok then
return false, tostring(err)
end
edit.applied = false
return true, nil
end
if edit.action == "create" then
return withWaypoint("Vibe Coder: undo create " .. tostring(edit.name or ""), function()
if edit.instance and edit.instance.Parent then
edit.instance:Destroy()
end
edit.applied = false
return true, nil
end)
end
return false, "unknown action"
end
function DiffReview.acceptAll(pd)
local applied, failed = 0, 0
local firstError = nil
for _, edit in ipairs(pd.pending_edits or {}) do
if not edit.applied then
local ok, err = DiffReview.applyEdit(edit)
if ok then
applied += 1
else
failed += 1
firstError = firstError or string.format("%s: %s", tostring(edit.path), tostring(err))
end
end
end
return applied, failed, firstError
end
function DiffReview.rejectAll(pd)
-- Drop unapplied edits; keep already-applied ones (matches DiffReviewUI).
local kept = {}
local dropped = {}
for _, edit in ipairs(pd.pending_edits or {}) do
if edit.applied then
table.insert(kept, edit)
else
table.insert(dropped, edit)
end
end
pd.pending_edits = kept
return dropped
end
function DiffReview.summary(pd): string
local pending, applied = 0, 0
for _, e in ipairs(pd.pending_edits or {}) do
if e.applied then
applied += 1
else
pending += 1
end
end
return string.format("%d pending, %d applied", pending, applied)
end
return DiffReview
]]>
{7f7a1161-3b1e-4e19-a4f5-cc1446758c72}
0
false
DiffReview
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{5d1a84a6-e13d-429a-8a64-44fb81669388}
0
false
ComposerControls
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
3 then
table.insert(lines, "… +" .. (shown - 3) .. " more")
end
return table.concat(lines, " | ")
end
function DiffReviewUI.setup(pd)
local frame = pd.widget.App:FindFirstChild("DiffReviewFrame")
if not frame then
warn("Vibe Coder: DiffReviewFrame missing")
return
end
pd._diff_frame = frame
local accept = frame:FindFirstChild("AcceptAllButton")
local reject = frame:FindFirstChild("RejectAllButton")
if accept then
accept.MouseButton1Click:Connect(function()
local applied, failed, reason = DiffReview.acceptAll(pd)
chatUI().syncAfterAcceptAll(pd)
if pd.spawn_system_bubble then
if applied == 0 and failed == 0 then
pd.spawn_system_bubble("No pending edits to accept.")
else
pd.spawn_system_bubble(string.format(
"Accepted %d edit(s)%s",
applied,
if failed > 0 then string.format(" — %d failed. %s", failed, tostring(reason or "")) else ""
))
end
end
DiffReviewUI.update(pd)
end)
end
if reject then
reject.MouseButton1Click:Connect(function()
local dropped = {}
local kept = {}
for _, e in ipairs(pd.pending_edits or {}) do
if e.applied then
table.insert(kept, e)
else
table.insert(dropped, e)
end
end
pd.pending_edits = kept
chatUI().syncAfterRejectAll(pd, dropped)
if pd.spawn_system_bubble then
if #dropped == 0 then
pd.spawn_system_bubble("No pending edits to reject.")
else
pd.spawn_system_bubble("Rejected " .. #dropped .. " pending edit(s).")
end
end
DiffReviewUI.update(pd)
end)
end
end
function DiffReviewUI.update(pd)
local frame = pd._diff_frame or (pd.widget and pd.widget.App:FindFirstChild("DiffReviewFrame"))
if not frame then
return
end
local n = pendingCount(pd)
frame.Visible = n > 0
local title = frame:FindFirstChild("Title")
if title then
local detail = summarize(pd)
title.Text = if n > 0
then string.format("Pending edits: %d — %s", n, detail)
else "Pending edits: 0"
end
end
function DiffReviewUI.notifyQueued(pd, path)
-- Prefer per-edit cards; fall back to a system line
local matched = false
for i = #(pd.pending_edits or {}), 1, -1 do
local e = pd.pending_edits[i]
if e.path == path and not e._card then
chatUI().spawnEditCard(pd, e)
matched = true
break
end
end
if not matched and pd.spawn_system_bubble then
pd.spawn_system_bubble("Queued edit: " .. tostring(path))
end
DiffReviewUI.update(pd)
end
return DiffReviewUI
]]>
{a0f92de7-bf5b-4b3c-875f-55c555944180}
0
false
DiffReviewUI
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{7f86beef-852e-434e-ad5f-4abd9df174f0}
0
false
ChatUI
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
= 3 then
local found = findByName(name)
if found then
pushUnique(mentions, found)
end
end
end
pd.at_mentions = mentions
return mentions
end
function AtMentions.summary(pd)
local n = #(pd.at_mentions or {})
if n == 0 then
return ""
end
local parts = {}
for i, inst in ipairs(pd.at_mentions) do
if i > 5 then
table.insert(parts, "…")
break
end
table.insert(parts, PathUtil.getFullName(inst))
end
return string.format("@ context (%d): %s", n, table.concat(parts, ", "))
end
return AtMentions
]]>
{609ad719-491c-4f0a-b40b-296ca906d32f}
0
false
AtMentions
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
0
0
0
false
Agent
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{7fc13935-27e0-4bd8-a7e1-433ca9bb9682}
0
false
Types
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0 then
table.insert(parts, "\n@-mentioned instances:")
for _, inst in ipairs(pd.at_mentions) do
if inst and inst.Parent then
table.insert(parts, "- " .. PathUtil.getFullName(inst) .. " (" .. inst.ClassName .. ")")
end
end
end
-- Naming what is already queued stops the agent re-proposing the same change.
if typeof(pd.pending_edits) == "table" then
local lines = {}
for _, edit in ipairs(pd.pending_edits) do
if typeof(edit) == "table" then
table.insert(
lines,
string.format(
"- %s (%s, %s)",
tostring(edit.path),
tostring(edit.action),
edit.applied and "accepted by user" or "awaiting review"
)
)
end
end
if #lines > 0 then
table.insert(parts, "\n\nChanges from this session:\n" .. table.concat(lines, "\n"))
end
end
return table.concat(parts, "")
end
return SystemPrompt
]]>
{14ef3f82-a7a3-436a-8ea0-abd7204900e9}
0
false
SystemPrompt
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0 then
for _, part in ipairs(content) do
if typeof(part) == "string" then
table.insert(parts, part)
elseif typeof(part) == "table" then
if typeof(part.text) == "string" then
table.insert(parts, part.text)
elseif typeof(part.content) == "string" then
table.insert(parts, part.content)
end
end
end
else
for _, part in pairs(content) do
if typeof(part) == "string" then
table.insert(parts, part)
elseif typeof(part) == "table" and typeof(part.text) == "string" then
table.insert(parts, part.text)
end
end
end
return table.concat(parts, "\n")
end
local function normalizeToolCalls(toolCalls)
if typeof(toolCalls) ~= "table" then
return {}
end
if #toolCalls > 0 then
return toolCalls
end
local out = {}
for _, call in pairs(toolCalls) do
if typeof(call) == "table" then
table.insert(out, call)
end
end
return out
end
local function ensureCallId(call)
if typeof(call.id) == "string" and call.id ~= "" then
return call.id
end
local id = string.format("call_%d_%d", os.time(), math.random(1000, 9999))
call.id = id
return id
end
-- Compact without splitting assistant(tool_calls) from their tool results.
local function compactHistory(pd)
local msgs = pd.agent_messages
if #msgs <= MAX_AGENT_MESSAGES then
return
end
local blocks = {}
local i = 1
while i <= #msgs do
local m = msgs[i]
if m.role == "assistant" then
local tools = normalizeToolCalls(m.tool_calls)
if #tools > 0 then
local block = { m }
i += 1
while i <= #msgs and msgs[i].role == "tool" do
table.insert(block, msgs[i])
i += 1
end
table.insert(blocks, block)
else
table.insert(blocks, { m })
i += 1
end
else
table.insert(blocks, { m })
i += 1
end
end
local keepBlocks = {}
if blocks[1] and blocks[1][1] and blocks[1][1].role == "system" then
table.insert(keepBlocks, blocks[1])
end
local start = math.max(#blocks - 12, (#keepBlocks > 0) and 2 or 1)
for b = start, #blocks do
table.insert(keepBlocks, blocks[b])
end
local keep = {}
for _, block in ipairs(keepBlocks) do
for _, m in ipairs(block) do
table.insert(keep, m)
end
end
-- If still too long, drop oldest non-system blocks
while #keep > MAX_AGENT_MESSAGES and #keepBlocks > 1 do
local removeIdx = (keepBlocks[1][1].role == "system") and 2 or 1
if removeIdx > #keepBlocks then
break
end
table.remove(keepBlocks, removeIdx)
keep = {}
for _, block in ipairs(keepBlocks) do
for _, m in ipairs(block) do
table.insert(keep, m)
end
end
end
pd.agent_messages = keep
end
local function pushSystem(pd)
local sys = SystemPrompt.build(pd)
if #pd.agent_messages == 0 or (pd.agent_messages[1] and pd.agent_messages[1].role ~= "system") then
table.insert(pd.agent_messages, 1, { role = "system", content = sys })
else
pd.agent_messages[1].content = sys
end
end
local function onStep(pd, kind, text, extra)
if pd.on_agent_step then
pcall(pd.on_agent_step, kind, text, extra)
end
end
local function fillMissingToolResults(pd, toolCalls, fromIndex, reason)
for idx = fromIndex, #toolCalls do
local call = toolCalls[idx]
local callId = ensureCallId(call)
local fn = call["function"]
local name = tostring((fn and fn.name) or call.name or "unknown")
table.insert(pd.agent_messages, {
role = "tool",
tool_call_id = callId,
name = name,
content = reason,
})
end
end
function Loop.run(pd, userText)
ensureSession(pd)
userText = tostring(userText or ""):match("^%s*(.-)%s*$") or ""
if userText == "" then
return nil, "Empty prompt"
end
-- Block re-entry only for the *current* run. Clear/Stop bumps run_id so a
-- superseded in-flight loop does not block the next send.
if pd._loop_owner ~= nil and pd._loop_owner == pd.run_id then
return nil, "Agent is already running"
end
local runId = State.beginRun(pd)
pd._loop_owner = runId
local finalText = nil
local errMsg = nil
local okOuter, outerErr = pcall(function()
pushSystem(pd)
table.insert(pd.agent_messages, { role = "user", content = userText })
for step = 1, MAX_STEPS do
if pd.abort_requested or pd.run_id ~= runId then
errMsg = "Stopped"
break
end
compactHistory(pd)
onStep(pd, "thinking", string.format("Step %d/%d…", step, MAX_STEPS))
local message, err
local model = tostring(pd.current_model or ""):lower()
if model == "auto" or model == "" then
local okChat, a, b = pcall(function()
return Auto.chat(pd, pd.agent_messages)
end)
if not okChat then
err = "Auto.chat crashed: " .. tostring(a)
else
message, err = a, b
end
else
err = "Agent loop supports Auto only. Pick Auto, or use a BYOK model."
end
if err then
errMsg = err
break
end
if pd.abort_requested or pd.run_id ~= runId then
errMsg = "Stopped"
break
end
if typeof(message) ~= "table" then
errMsg = "Empty model response"
break
end
local content = normalizeContent(message.content)
local toolCalls = normalizeToolCalls(message.tool_calls)
for _, call in ipairs(toolCalls) do
ensureCallId(call)
end
local assistantMsg = {
role = "assistant",
content = content,
tool_calls = (#toolCalls > 0) and toolCalls or nil,
}
table.insert(pd.agent_messages, assistantMsg)
if #toolCalls == 0 then
finalText = content
break
end
local stoppedAt = nil
for ti, call in ipairs(toolCalls) do
if pd.abort_requested or pd.run_id ~= runId then
stoppedAt = ti
errMsg = "Stopped"
break
end
local fn = call["function"]
local name = tostring((fn and fn.name) or call.name or "unknown")
local args = (fn and fn.arguments) or call.arguments or "{}"
local callId = ensureCallId(call)
local detail = ""
do
local okArgs, decoded = pcall(function()
if typeof(args) == "string" then
return HttpService:JSONDecode(args)
end
return args
end)
if okArgs and typeof(decoded) == "table" then
detail = tostring(decoded.path or decoded.parent_path or decoded.query or decoded.className or decoded.class_name or decoded.name or "")
end
end
onStep(pd, "tool", name, { detail = detail })
local result = ToolsRegistry.run(name, args, pd)
if typeof(result) ~= "table" then
result = { ok = false, content = "Tool returned invalid result" }
end
local toolContent = truncate(result.content or "")
if result.pendingEdit then
onStep(pd, "diff", tostring(result.pendingEdit.path), {
path = result.pendingEdit.path,
edit = result.pendingEdit,
})
end
table.insert(pd.agent_messages, {
role = "tool",
tool_call_id = callId,
name = name,
content = toolContent,
})
end
if stoppedAt then
fillMissingToolResults(pd, toolCalls, stoppedAt, "Stopped by user before tool completed.")
break
end
end
-- Budget spent mid-task: ask for a wrap-up so the user gets an account of
-- the work and what is waiting in Diff Review, not a bare error.
if not finalText and not errMsg and not pd.abort_requested and pd.run_id == runId then
table.insert(pd.agent_messages, {
role = "user",
content = "You have used the tool-call budget for this turn. Do not call any more tools. "
.. "Summarise what you changed, what is queued for review, and what still needs doing.",
})
onStep(pd, "thinking", "Wrapping up…")
local okWrap, message = pcall(function()
return Auto.chat(pd, pd.agent_messages, { noTools = true })
end)
if okWrap and typeof(message) == "table" then
local wrapText = normalizeContent(message.content)
if wrapText ~= "" then
table.insert(pd.agent_messages, { role = "assistant", content = wrapText })
finalText = wrapText
end
end
end
if not finalText and not errMsg then
errMsg = string.format(
"Hit the %d-step limit for this turn. Anything already queued is in Diff Review — send another message to continue.",
MAX_STEPS
)
end
end)
if not okOuter then
errMsg = "Agent loop crashed: " .. tostring(outerErr)
end
-- Only the owning run may clear busy / owner. Clear + quick resend must not
-- let a superseded loop wipe the new run's busy flag.
if pd._loop_owner == runId then
pd._loop_owner = nil
end
if pd.run_id == runId then
State.endRun(pd)
end
return finalText, errMsg
end
function Loop.stop(pd)
if typeof(pd) == "table" then
State.requestAbort(pd)
end
end
return Loop
]]>
{6e086474-a693-40b3-835e-015d0930b062}
0
false
Loop
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
0
0
0
false
Providers
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
= 500 then
return "The Vibe Coder service is having trouble. Try again in a moment."
end
if lastErr and lastErr ~= "" then
return lastErr
end
return "Auto could not complete this request. Try again in a moment."
end
local function isRetryableFailure(status, bodyText)
bodyText = tostring(bodyText or "")
local code = tonumber(status) or 0
if code == 0 or code == 404 or code == 408 or code == 429 or code == 500 or code == 502 or code == 503 or code == 504 or code == 529 then
return true
end
if string.find(bodyText, "DEPLOYMENT_NOT_FOUND", 1, true) then
return true
end
if string.find(bodyText, "No upstream", 1, true) then
return true
end
if string.find(bodyText, "All providers", 1, true) then
return true
end
if string.find(bodyText, "HttpError", 1, true) then
return true
end
return false
end
local function normalizeContent(content)
if typeof(content) == "string" then
return content
end
if typeof(content) ~= "table" then
return content == nil and "" or tostring(content)
end
-- OpenAI-style multipart content
local parts = {}
if #content > 0 then
for _, part in ipairs(content) do
if typeof(part) == "string" then
table.insert(parts, part)
elseif typeof(part) == "table" then
if typeof(part.text) == "string" then
table.insert(parts, part.text)
elseif typeof(part.content) == "string" then
table.insert(parts, part.content)
end
end
end
else
for _, part in pairs(content) do
if typeof(part) == "string" then
table.insert(parts, part)
elseif typeof(part) == "table" and typeof(part.text) == "string" then
table.insert(parts, part.text)
end
end
end
return table.concat(parts, "\n")
end
local function normalizeMessage(message)
if typeof(message) ~= "table" then
return nil
end
message.content = normalizeContent(message.content)
return message
end
function Auto.chat(pd, messages, opts)
if typeof(messages) ~= "table" or #messages == 0 then
return nil, "No messages to send"
end
-- noTools forces a prose answer, used for the end-of-budget wrap-up.
local tools = {}
if not (opts and opts.noTools) then
tools = ToolsRegistry.toOpenAITools("agent")
end
local bodyTable = {
model = "auto",
messages = messages,
stream = false,
}
if #tools > 0 then
bodyTable.tools = tools
bodyTable.tool_choice = "auto"
end
local okEncode, bodyJson = pcall(function()
return HttpService:JSONEncode(bodyTable)
end)
if not okEncode then
return nil, "Failed to encode Auto request: " .. tostring(bodyJson)
end
-- Localhost is only probed for developers who opted in; end users would just
-- pay a connection timeout for a host that is never there.
local bases = { Auto.BASE_URL }
local useLocal = false
if pd and pd.plugin then
local okSetting, value = pcall(function()
return pd.plugin:GetSetting("vibe-coder-auto-local")
end)
useLocal = okSetting and value == true
end
if useLocal then
bases = { Auto.LOCAL_URL, Auto.BASE_URL }
end
-- Returns: message, err, retryable, statusCode
local function attempt(base: string)
local okReq, response = pcall(function()
return HttpService:RequestAsync({
Url = chatUrl(base),
Method = "POST",
Headers = { ["Content-Type"] = "application/json" },
Body = bodyJson,
})
end)
if not okReq then
return nil, "HTTP failed (" .. base .. "): " .. tostring(response), true
end
if typeof(response) ~= "table" then
return nil, "Invalid HTTP response from " .. base, true
end
if not response.Success then
local bodyText = tostring(response.Body or "")
return nil,
string.format("Auto %s -> %s: %s", base, tostring(response.StatusCode), bodyText:sub(1, 300)),
isRetryableFailure(response.StatusCode, bodyText),
response.StatusCode
end
local okJson, data = pcall(function()
return HttpService:JSONDecode(response.Body)
end)
if not okJson or typeof(data) ~= "table" then
return nil, "Bad JSON from " .. base, true
end
local errField = data.error
if typeof(errField) == "table" and errField.message then
local text = tostring(errField.message)
return nil, text, isRetryableFailure(200, text)
end
if typeof(errField) == "string" then
return nil, errField, isRetryableFailure(200, errField)
end
local choice = data.choices and data.choices[1]
local message = normalizeMessage(choice and choice.message)
if message then
return message, nil, false
end
return nil, "Auto response missing choices[1].message from " .. base, true
end
local lastErr, lastStatus = nil, nil
for _, base in ipairs(bases) do
for try = 1, #RETRY_DELAYS + 1 do
if pd and pd.abort_requested then
return nil, "Stopped"
end
local message, err, retryable, status = attempt(base)
if message then
return message, nil
end
lastErr, lastStatus = err, status or lastStatus
warn("[Vibe Coder] Auto attempt failed: " .. tostring(err))
if not retryable or try > #RETRY_DELAYS then
break
end
task.wait(RETRY_DELAYS[try])
end
end
return nil, userFacingError(lastErr, lastStatus)
end
return Auto
]]>
{105adc1e-e0d9-4039-8459-275ef64d11b3}
0
false
Auto
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
0
0
0
false
Tools
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{60658367-5ee7-4ba7-ad97-941bc7cde16f}
0
false
PathUtil
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
total then
return total
end
return n
end
return {
name = "read_script",
description = "Read the Luau source of a Script, LocalScript, or ModuleScript by dotted path (e.g. game.ServerScriptService.Main). Output is line-numbered as 'N | text'. Reflects any edit you have already queued for this script. For long scripts, read a range with start_line/end_line.",
parameters = {
type = "object",
properties = {
path = {
type = "string",
description = "Dotted instance path starting with game. or a service name",
},
start_line = {
type = "integer",
description = "First line to read, 1-based. Defaults to 1.",
},
end_line = {
type = "integer",
description = "Last line to read, inclusive. Defaults to the end of the script.",
},
},
required = { "path" },
},
read_only = true,
run = function(args, pd)
local resolved, err = PendingSource.resolve(pd, args.path)
if not resolved then
return { ok = false, content = err or "not found" }
end
local source = resolved.source
local lines = splitLines(source)
local total = #lines
if total == 0 then
return {
ok = true,
content = string.format("path: %s\n(empty script)", resolved.canonicalPath),
}
end
local startLine = clampLine(args.start_line, 1, total)
local endLine = clampLine(args.end_line, total, total)
if endLine < startLine then
startLine, endLine = endLine, startLine
end
local out, used = {}, 0
local lastShown = startLine - 1
for n = startLine, endLine do
local rendered = string.format("%d | %s", n, lines[n])
if used + #rendered + 1 > MAX_CHARS then
break
end
table.insert(out, rendered)
used += #rendered + 1
lastShown = n
end
local header = {
"path: " .. resolved.canonicalPath,
"class: " .. (resolved.instance and resolved.instance.ClassName or tostring(resolved.pending and resolved.pending.className or "Script")),
string.format("lines: %d total, showing %d-%d", total, startLine, lastShown),
}
if not resolved.instance then
table.insert(header, "state: queued create, not yet accepted by the user")
elseif resolved.pending then
table.insert(header, "state: includes your queued edit, not yet accepted by the user")
end
if lastShown < endLine then
table.insert(
header,
string.format("note: output limit reached. Read the rest with start_line=%d.", lastShown + 1)
)
end
return {
ok = true,
content = table.concat(header, "\n") .. "\n\n" .. table.concat(out, "\n"),
}
end,
}
]]>
{7b48eb7b-fbfb-455e-a974-521143996d06}
0
false
ReadScript
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
= MAX_SCRIPTS then
return
end
if d:IsA("LuaSourceContainer") then
table.insert(out, d)
end
end
end
return {
name = "search_scripts",
description = "Search script names and source for a case-insensitive query. Returns matching paths and short snippets.",
parameters = {
type = "object",
properties = {
query = {
type = "string",
description = "Text to search for in script names and source",
},
scope = {
type = "string",
description = "Optional dotted path to limit search (default: common services)",
},
},
required = { "query" },
},
read_only = true,
run = function(args, _pd)
local query = tostring(args.query or "")
if query == "" then
return { ok = false, content = "query is required" }
end
local q = string.lower(query)
local scripts: { LuaSourceContainer } = {}
if args.scope and args.scope ~= "" then
local root, err = PathUtil.resolve(args.scope)
if not root then
return { ok = false, content = err or "bad scope" }
end
if root:IsA("LuaSourceContainer") then
table.insert(scripts, root)
end
collectScripts(root, scripts)
else
for _, serviceName in ipairs(SERVICES) do
local ok, service = pcall(function()
return game:GetService(serviceName)
end)
if ok and service then
collectScripts(service, scripts)
end
end
end
local lines = {}
local count = 0
for _, scr in ipairs(scripts) do
if count >= MAX_MATCHES then
break
end
local nameHit = string.find(string.lower(scr.Name), q, 1, true)
local source = scr.Source
local sourceLower = string.lower(source)
local idx = string.find(sourceLower, q, 1, true)
if nameHit or idx then
count += 1
local snippet = ""
if idx then
local startAt = math.max(1, idx - 40)
snippet = string.sub(source, startAt, startAt + MAX_SNIPPET - 1)
snippet = string.gsub(snippet, "\n", " ")
end
table.insert(
lines,
string.format(
"%d. %s (%s)%s",
count,
PathUtil.getFullName(scr),
scr.ClassName,
if snippet ~= "" then "\n … " .. snippet else " (name match)"
)
)
end
end
if count == 0 then
return { ok = true, content = "No matches for: " .. query }
end
local header = string.format("Found %d match(es) for %q:\n", count, query)
if count >= MAX_MATCHES then
header ..= "(truncated to " .. MAX_MATCHES .. ")\n"
end
return { ok = true, content = header .. table.concat(lines, "\n") }
end,
}
]]>
{3a6f6635-019b-4552-8c3a-ab2158bf8433}
0
false
SearchScripts
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
200 then
table.insert(lines, "... truncated")
break
end
table.insert(lines, string.format("- %s (%s)", child.Name, child.ClassName))
end
return { ok = true, content = table.concat(lines, "\n") }
end,
}
]]>
{8c44ae4b-5e16-440e-ab40-391335cbf0b9}
0
false
ListChildren
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{a78a7036-1ab4-4e11-bb7a-38c572ccff9a}
0
false
GetHierarchy
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
50 then
table.insert(lines, "... truncated")
break
end
local extra = ""
if inst:IsA("LuaSourceContainer") then
extra = string.format(" [script, %d chars]", #inst.Source)
end
table.insert(lines, string.format("- %s (%s)%s", PathUtil.getFullName(inst), inst.ClassName, extra))
end
if pd.selected_script then
table.insert(lines, "\nPlugin active script: " .. PathUtil.getFullName(pd.selected_script))
end
return { ok = true, content = table.concat(lines, "\n") }
end,
}
]]>
{422371ad-42d4-41ba-8fdd-87abc1e6e67c}
0
false
GetSelection
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{8d496dd2-fa63-4cf3-9d27-8acae426592d}
0
false
EditScript
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{c432cf42-5dcf-4e7c-894d-d580cb195ede}
0
false
CreateScript
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0 then
cleaned.required = parameters.required
end
return cleaned
end
-- OpenAI tools format
function Registry.toOpenAITools(mode: string)
local out = {}
for _, tool in ipairs(Registry.getForMode(mode)) do
table.insert(out, {
type = "function",
["function"] = {
name = tool.name,
description = tool.description,
parameters = sanitiseParameters(tool.parameters),
},
})
end
return out
end
function Registry.run(name: string, argsJsonOrTable: any, pd): any
local tool = byName[name]
if not tool then
return { ok = false, content = "Unknown tool: " .. tostring(name) }
end
pd = pd or {}
local args = argsJsonOrTable
if typeof(args) == "string" then
local ok, decoded = pcall(function()
return HttpService:JSONDecode(args)
end)
if not ok then
return { ok = false, content = "Invalid tool arguments JSON: " .. tostring(decoded) }
end
args = decoded
end
args = args or {}
local ok, result = pcall(function()
return tool.run(args, pd)
end)
if not ok then
return { ok = false, content = "Tool error: " .. tostring(result) }
end
return result
end
return Registry
]]>
{308a39b0-019d-49c8-9a7e-1def0f347088}
0
false
Registry
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{7899246e-7bd2-4ec6-b970-23931439b7d5}
0
false
OpenScript
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
{1f3352f1-225d-435d-ad0b-b7822ae77c7b}
0
false
PendingSource
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
MAX_EDITS then
return nil, string.format("too many edits (%d), max %d per call", #raw, MAX_EDITS)
end
return raw, nil
end
return {
name = "patch_script",
description = "Edit a script by exact string replacement — the preferred way to change existing code. Each edit's old_string must appear exactly once (or set replace_all). Much cheaper and safer than restating the whole file with edit_script. Queues a pending edit for the user to Accept/Reject.",
parameters = {
type = "object",
properties = {
path = {
type = "string",
description = "Dotted path to the script to edit",
},
edits = {
type = "array",
description = "Replacements applied in order, each to the result of the previous one",
items = {
type = "object",
properties = {
old_string = {
type = "string",
description = "Exact text to find, including indentation. Include enough surrounding lines to be unique.",
},
new_string = {
type = "string",
description = "Replacement text. Use an empty string to delete.",
},
replace_all = {
type = "boolean",
description = "Replace every occurrence instead of requiring exactly one",
},
},
required = { "old_string", "new_string" },
},
},
},
required = { "path", "edits" },
},
read_only = false,
run = function(args, pd)
local resolved, err = PendingSource.resolve(pd, args.path)
if not resolved then
return { ok = false, content = err or "not found" }
end
local edits, editsErr = coerceEdits(args.edits)
if not edits then
return { ok = false, content = editsErr or "invalid edits" }
end
local source = resolved.source
local applied = 0
for index, entry in ipairs(edits) do
if typeof(entry) ~= "table" then
return { ok = false, content = string.format("edit %d is not an object", index) }
end
local old = entry.old_string
local new = entry.new_string
if typeof(old) ~= "string" or old == "" then
return {
ok = false,
content = string.format("edit %d: old_string must be a non-empty string. To create a new script use create_script.", index),
}
end
if typeof(new) ~= "string" then
return { ok = false, content = string.format("edit %d: new_string must be a string", index) }
end
if old == new then
return { ok = false, content = string.format("edit %d: old_string and new_string are identical", index) }
end
local hits = countOccurrences(source, old)
if hits == 0 then
return {
ok = false,
content = string.format(
"edit %d of %d: old_string not found in %s. Nothing was queued — re-read the script, whitespace and indentation must match exactly.",
index,
#edits,
resolved.canonicalPath
),
}
end
if hits > 1 and entry.replace_all ~= true then
return {
ok = false,
content = string.format(
"edit %d: old_string appears %d times in %s. Add surrounding lines to make it unique, or set replace_all.",
index,
hits,
resolved.canonicalPath
),
}
end
if entry.replace_all == true then
source = replaceEvery(source, old, new)
applied += hits
else
source = replaceFirst(source, old, new)
applied += 1
end
end
if source == resolved.source then
return { ok = true, content = "No net change to " .. resolved.canonicalPath }
end
-- Patching a script that only exists as a queued create keeps it a create,
-- so Accept still makes the instance instead of editing a missing one.
local existingCreate = resolved.pending
if existingCreate and existingCreate.action == "create" then
existingCreate.newSource = source
return {
ok = true,
content = string.format(
"Applied %d replacement(s) to the queued new script %s (%d chars). Still waiting for user Accept.",
applied,
existingCreate.path,
#source
),
pendingEdit = existingCreate,
}
end
local scriptInst = resolved.instance
if not scriptInst then
return { ok = false, content = "cannot patch " .. resolved.canonicalPath .. " — no such script" }
end
-- oldSource stays the live text so Reject after Accept restores what the
-- user actually had, even across several patches in one run.
local edit = Types.makePendingEdit({
id = Types.newId("edit"),
path = resolved.canonicalPath,
instance = scriptInst,
oldSource = resolved.liveSource or "",
newSource = source,
action = "edit",
applied = false,
})
State.addPendingEdit(pd, edit)
return {
ok = true,
content = string.format(
"Applied %d replacement(s) to %s (%d → %d chars). Queued for user Accept in Diff Review.",
applied,
edit.path,
#(edit.oldSource or ""),
#edit.newSource
),
pendingEdit = edit,
}
end,
}
]]>
{ff89654e-64ad-4dbf-ac5e-98b278fce546}
0
false
PatchScript
-1
yuZpQdnvvUBOTYh1jqZ2cA==
-
0
0
0
0
false
Services
-1
yuZpQdnvvUBOTYh1jqZ2cA==