Remove dependency on string.pack

aead.lua

@@ -4,9 +4,12 @@
 --
 
 local expect   = require "cc.expect".expect
+local packing  = require "ccryptolib.internal.packing"
 local chacha20 = require "ccryptolib.chacha20"
 local poly1305 = require "ccryptolib.poly1305"
 
+local p8x1, fmt8x1 = packing.compilePack("<I8")
+local u4x4, fmt4x4 = packing.compileUnpack("<I4I4I4I4")
 local bxor = bit32.bxor
 local bor = bit32.bor
 
@@ -39,8 +42,8 @@ local function encrypt(key, nonce, message, aad, rounds)
     -- Authenticate.
     local pad1 = ("\0"):rep(-#aad % 16)
     local pad2 = ("\0"):rep(-#ciphertext % 16)
-    local aadLen = ("<I8"):pack(#aad)
+    local aadLen = p8x1("<I8", #aad)
-    local ctxLen = ("<I8"):pack(#ciphertext)
+    local ctxLen = p8x1("<I8", #ciphertext)
     local combined = aad .. pad1 .. ciphertext .. pad2 .. aadLen .. ctxLen
     local tag = poly1305.mac(authKey, combined)
 
@@ -77,11 +80,11 @@ local function decrypt(key, nonce, tag, ciphertext, aad, rounds)
     -- Check tag.
     local pad1 = ("\0"):rep(-#aad % 16)
     local pad2 = ("\0"):rep(-#ciphertext % 16)
-    local aadLen = ("<I8"):pack(#aad)
+    local aadLen = p8x1(fmt8x1, #aad)
-    local ctxLen = ("<I8"):pack(#ciphertext)
+    local ctxLen = p8x1(fmt8x1, #ciphertext)
     local combined = aad .. pad1 .. ciphertext .. pad2 .. aadLen .. ctxLen
-    local t1, t2, t3, t4 = ("<I4I4I4I4"):unpack(tag)
+    local t1, t2, t3, t4 = u4x4(fmt4x4, tag, 1)
-    local u1, u2, u3, u4 = ("<I4I4I4I4"):unpack(poly1305.mac(authKey, combined))
+    local u1, u2, u3, u4 = u4x4(fmt4x4, poly1305.mac(authKey, combined), 1)
     local eq = bor(bxor(t1, u1), bxor(t2, u2), bxor(t3, u3), bxor(t4, u4))
     if eq == 0 then return message end
 end

blake3.lua

@@ -3,11 +3,15 @@
 -- @module blake3
 --
 
-local expect = require "cc.expect".expect
+local expect  = require "cc.expect".expect
+local packing = require "ccryptolib.internal.packing"
 
 local unpack = unpack or table.unpack
 local bxor = bit32.bxor
 local rol = bit32.lrotate
+local p16x4, fmt16x4 = packing.compilePack("<I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4")
+local u16x4 = packing.compileUnpack(fmt16x4)
+local u8x4, fmt8x4 = packing.compileUnpack("<I4I4I4I4I4I4I4I4")
 
 local CHUNK_START         = 0x01
 local CHUNK_END           = 0x02
@@ -129,7 +133,7 @@ local function expand(state, len, offset)
     for i = 0, len / 64 do
         local n = offset + i
         local md = compress(state.cv, state.m, n, state.n, state.f, true)
-        out[i + 1] = ("<I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4"):pack(unpack(md))
+        out[i + 1] = p16x4(fmt16x4, unpack(md))
     end
 
     return table.concat(out):sub(1, len)
@@ -150,7 +154,7 @@ local function update(state, message)
     -- Digest complete blocks.
     for i = 1, #blocks, 64 do
         -- Compress the block.
-        local block = {("<I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4"):unpack(blocks, i)}
+        local block = {u16x4(fmt16x4, blocks, i)}
         local stateFlags = state.f + state.s + state.e
         state.cv = compress(state.cv, block, state.t, 64, stateFlags)
         state.s = 0
@@ -188,7 +192,7 @@ local function finalize(state)
     -- Pad the last message block.
     local lastLen = #state.m
     local padded = state.m .. ("\0"):rep(64)
-    local last = {("<I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4"):unpack(padded)}
+    local last = {u16x4(fmt16x4, padded, 1)}
 
     -- Prepare output expansion state.
     if state.t > 0 then
@@ -267,13 +271,13 @@ end
 function mod.newKeyed(key)
     expect(1, key, "string")
     assert(#key == 32, "key length must be 32")
-    return new({("<I4I4I4I4I4I4I4I4"):unpack(key)}, KEYED_HASH)
+    return new({u8x4(fmt8x4, key, 1)}, KEYED_HASH)
 end
 
 function mod.newDk(context)
     expect(1, context, "string")
     local iv = new(IV, DERIVE_KEY_CONTEXT):update(context):finalize():expand(32)
-    return new({("<I4I4I4I4I4I4I4I4"):unpack(iv)}, DERIVE_KEY_MATERIAL)
+    return new({u8x4(fmt8x4, iv, 1)}, DERIVE_KEY_MATERIAL)
 end
 
 --- Hashes data using BLAKE3.
@@ -304,7 +308,7 @@ function mod.digestKeyed(key, message, len)
     expect(3, len, "number", "nil")
     len = len or 32
     assert(len % 1 == 0 and len >= 1, "length must be a positive integer")
-    local h = new({("<I4I4I4I4I4I4I4I4"):unpack(key)}, KEYED_HASH)
+    local h = new({u8x4(fmt8x4, key, 1)}, KEYED_HASH)
     return h:update(message):finalize():expand(len)
 end
 
@@ -322,7 +326,7 @@ function mod.deriveKey(context)
         expect(2, len, "number", "nil")
         len = len or 32
         assert(len % 1 == 0 and len >= 1, "length must be a positive integer")
-        local h = new({("<I4I4I4I4I4I4I4I4"):unpack(iv)}, DERIVE_KEY_MATERIAL)
+        local h = new({u8x4(fmt8x4, iv, 1)}, DERIVE_KEY_MATERIAL)
         return h:update(material):finalize():expand(len)
     end
 end

chacha20.lua

@@ -3,10 +3,15 @@
 -- @module chacha20
 --
 
-local expect = require "cc.expect".expect
+local expect  = require "cc.expect".expect
+local packing = require "ccryptolib.internal.packing"
 
 local bxor = bit32.bxor
 local rol = bit32.lrotate
+local u8x4, fmt8x4 = packing.compileUnpack("<I4I4I4I4I4I4I4I4")
+local u3x4, fmt3x4 = packing.compileUnpack("<I4I4I4")
+local p16x4, fmt16x4 = packing.compilePack("<I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4")
+local u16x4 = packing.compileUnpack(fmt16x4)
 
 local mod = {}
 
@@ -36,8 +41,8 @@ function mod.crypt(key, nonce, message, rounds, offset)
 
     -- Build the state block.
     local i0, i1, i2, i3 = 0x61707865, 0x3320646e, 0x79622d32, 0x6b206574
-    local k0, k1, k2, k3, k4, k5, k6, k7 = ("<I4I4I4I4I4I4I4I4"):unpack(key)
+    local k0, k1, k2, k3, k4, k5, k6, k7 = u8x4(fmt8x4, key, 1)
-    local cr, n0, n1, n2 = offset, ("<I4I4I4"):unpack(nonce)
+    local cr, n0, n1, n2 = offset, u3x4(fmt3x4, nonce, 1)
 
     -- Pad the message.
     local padded = message .. ("\0"):rep(-#message % 64)
@@ -101,10 +106,10 @@ function mod.crypt(key, nonce, message, rounds, offset)
 
         m00, m01, m02, m03, m04, m05, m06, m07,
         m08, m09, m10, m11, m12, m13, m14, m15, idx =
-            ("<I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4"):unpack(padded, idx)
+            u16x4(fmt16x4, padded, idx)
 
         -- Feed-forward and combine.
-        out[i] = ("<I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4"):pack(
+        out[i] = p16x4(fmt16x4,
             bxor(m00, s00 + i0), bxor(m01, s01 + i1),
             bxor(m02, s02 + i2), bxor(m03, s03 + i3),
             bxor(m04, s04 + k0), bxor(m05, s05 + k1),

internal/fp.lua

@@ -10,7 +10,10 @@
 -- @module[kind=internal] internal.fp
 --
 
+local packing = require "ccryptolib.inernal.packing"
+
 local unpack = unpack or table.unpack
+local ufp, fmtfp = packing.compileUnpack("<I3I3I2I3I3I2I3I3I2I3I3I2")
 
 --- The modular square root of -1.
 local I = {
@@ -751,7 +754,7 @@ end
 --
 local function decode(b)
     local w00, w01, w02, w03, w04, w05, w06, w07, w08, w09, w10, w11 =
-        ("<I3I3I2I3I3I2I3I3I2I3I3I2"):unpack(b)
+        ufp(fmtfp, b, 1)
 
     w11 = w11 % 2 ^ 15
 

internal/fq.lua

@@ -10,10 +10,15 @@
 -- @module[kind=internal] internal.fq
 --
 
-local mp   = require "ccryptolib.internal.mp"
+local mp      = require "ccryptolib.internal.mp"
-local util = require "ccryptolib.internal.util"
+local util    = require "ccryptolib.internal.util"
+local packing = require "ccryptolib.internal.packing"
 
 local unpack = unpack or table.unpack
+local pfq, fmtfq = packing.compilePack("<I3I3I3I3I3I3I3I3I3I3I2")
+local ufq = packing.compileUnpack(fmtfq)
+local ufql, fmtfql = packing.compileUnpack("<I3I3I3I3I3I3I3I3I3I3I3")
+local ufqh, fmtfqh = packing.compileUnpack("<I3I3I3I3I3I3I3I3I3I3I1")
 
 --- The scalar field's order, q.
 local Q = {
@@ -180,7 +185,7 @@ end
 -- @treturn string The 32-byte string encoding of a.
 --
 local function encode(a)
-    return ("<I3I3I3I3I3I3I3I3I3I3I2"):pack(unpack(demontgomery(a)))
+    return pfq(fmtfq, unpack(demontgomery(a)))
 end
 
 --- Decodes a scalar.
@@ -189,7 +194,7 @@ end
 -- @treturn {number...} 2²⁶⁴ × a mod q as 11 limbs in [0..2²⁴).
 --
 local function decode(str)
-    local dec = {("<I3I3I3I3I3I3I3I3I3I3I2"):unpack(str)} dec[12] = nil
+    local dec = {ufq(fmtfq, str, 1)} dec[12] = nil
     return montgomery(dec)
 end
 
@@ -199,8 +204,8 @@ end
 -- @treturn {number...} 2²⁶⁴ × a mod q as 11 limbs in [0..2²⁴).
 --
 local function decodeWide(str)
-    local low = {("<I3I3I3I3I3I3I3I3I3I3I3"):unpack(str)} low[12] = nil
+    local low = {ufql(fmtfql, str, 1)} low[12] = nil
-    local high = {("<I3I3I3I3I3I3I3I3I3I3I1"):unpack(str, 34)} high[12] = nil
+    local high = {ufqh(fmtfqh, str, 34)} high[12] = nil
     return add(montgomery(low), montgomery(montgomery(high)))
 end
 
@@ -211,7 +216,7 @@ end
 --
 local function decodeClamped(str)
     -- Decode.
-    local words = {("<I3I3I3I3I3I3I3I3I3I3I2"):unpack(str)} words[12] = nil
+    local words = {ufq(fmtfq, str, 1)} words[12] = nil
 
     -- Clamp.
     words[1] = bit32.band(words[1], 0xfffff8)
@@ -229,7 +234,7 @@ end
 --
 local function decodeClamped8(str)
     -- Decode.
-    local words = {("<I3I3I3I3I3I3I3I3I3I3I2"):unpack(str)} words[12] = nil
+    local words = {ufq(fmtfq, str, 1)} words[12] = nil
 
     -- Clamp.
     words[1] = bit32.band(words[1], 0xfffff8)

internal/packing.lua

@@ -0,0 +1,166 @@
+--- High-performance binary packing of integers.
+--
+-- :::note Internal Module
+-- This module is meant for internal use within the library. Its API is unstable
+-- and subject to change without major version bumps.
+-- :::
+--
+-- <br />
+--
+-- :::warning
+-- For performance reasons, **the generated functions do not check types,
+-- lengths, nor ranges**. You must ensure that the passed arguments are
+-- well-formed and respect the format string yourself.
+-- :::
+--
+-- <br />
+--
+-- @module[kind=internal] internal.packing
+--
+
+local fmt = string.format
+
+local function mkPack(words, BE)
+    local out = "local C=string.char return function(_,"
+    local nb = 0
+    for i = 1, #words do
+        out = out .. fmt("n%d,", i)
+        nb = nb + words[i]
+    end
+    out = out:sub(1, -2) .. ")local "
+    for i = 1, nb do
+        out = out .. fmt("b%d,", i)
+    end
+    out = out:sub(1, -2) .. " "
+    local bi = 1
+    for i = 1, #words do
+        for _ = 1, words[i] - 1 do
+            out = out .. fmt("b%d=n%d%%2^8 n%d=(n%d-b%d)*2^-8 ", bi, i, i, i, bi)
+            bi = bi + 1
+        end
+        bi = bi + 1
+    end
+    out = out .. "return C("
+    bi = 1
+    if not BE then
+        for i = 1, #words do
+            for _ = 1, words[i] - 1 do
+                out = out .. fmt("b%d,", bi)
+                bi = bi + 1
+            end
+            out = out .. fmt("n%d%%2^8,", i)
+            bi = bi + 1
+        end
+    else
+        for i = 1, #words do
+            out = out .. fmt("n%d%%2^8,", i)
+            bi = bi + words[i] - 2
+            for _ = 1, words[i] - 1 do
+                out = out .. fmt("b%d,", bi)
+                bi = bi - 1
+            end
+            bi = bi + words[i] + 1
+        end
+    end
+    out = out:sub(1, -2) .. ")end"
+    return load(out)()
+end
+
+local function mkUnpack(words, BE)
+    local out = "local B=string.byte return function(_,s,i)local "
+    local bi = 1
+    if not BE then
+        for i = 1, #words do
+            for _ = 1, words[i] do
+                out = out .. fmt("b%d,", bi)
+                bi = bi + 1
+            end
+        end
+    else
+        for i = 1, #words do
+            bi = bi + words[i] - 1
+            for _ = 1, words[i] do
+                out = out .. fmt("b%d,", bi)
+                bi = bi - 1
+            end
+            bi = bi + words[i] + 1
+        end
+    end
+    out = out:sub(1, -2) .. fmt("=B(s,i,i+%d)return ", bi - 2)
+    bi = 1
+    for i = 1, #words do
+        out = out .. fmt("b%d", bi)
+        bi = bi + 1
+        for j = 2, words[i] do
+            out = out .. fmt("+b%d*2^%d", bi, 8 * j - 8)
+            bi = bi + 1
+        end
+        out = out .. ","
+    end
+    out = out .. fmt("i+%d end", bi - 1)
+    return load(out)()
+end
+
+local mod = {}
+
+-- Check whether string.pack is implemented in a high-speed language.
+if not string.pack or pcall(string.dump, string.pack) then
+    local function compile(fmt, fn)
+        local e = assert(fmt:match("^([><])I[I%d]+$"), "invalid format string")
+        local w = {}
+        for i in fmt:gmatch("I([%d]+)") do
+            local n = tonumber(i) or 4
+            assert(n > 0 and n <= 4, "integral size out of limits")
+            w[#w + 1] = n
+        end
+        return fn(w, e == ">")
+    end
+
+    local packCache = {}
+    local unpackCache = {}
+
+    --- (`string.pack == nil`) Compiles a binary packing function.
+    -- @tparam string fmt A string matched by `^([><])I[I%d]+$`.
+    -- @treturn function A high-performance function that behaves like an unsafe
+    -- version of `string.pack` for the given format string. Note that the third
+    -- argument isn't optional.
+    -- @treturn string fmt
+    -- @throws If the string is invalid or has an invalid integral size.
+    -- @throws If the compiled function is too large.
+    function mod.compilePack(fmt)
+        if not packCache[fmt] then
+            packCache[fmt] = compile(fmt, mkPack)
+        end
+        return packCache[fmt], fmt
+    end
+
+    --- (`string.pack == nil`) Compiles a binary unpacking function.
+    -- @tparam string fmt A string matched by `^([><])I[I%d]+$`.
+    -- @treturn function A high-performance function that behaves like an unsafe
+    -- version of `string.unpack` for the given format string.
+    -- @treturn string fmt
+    -- @throws If the string is invalid or has an invalid integral size.
+    -- @throws If the compiled function is too large.
+    function mod.compileUnpack(fmt)
+        if not unpackCache[fmt] then
+            unpackCache[fmt] = compile(fmt, mkUnpack)
+        end
+        return unpackCache[fmt], fmt
+    end
+
+    return mod
+else
+    --- (`string.pack ~= nil`) Compiles a binary packing function.
+    -- @tparam string fmt
+    -- @treturn function `string.pack`
+    -- @treturn string fmt
+    mod.compilePack = function(fmt) return string.pack, fmt end
+
+    --- (`string.pack ~= nil`) Compiles a binary unpacking function.
+    -- @tparam string fmt
+    -- @treturn function `string.unpack`
+    -- @treturn string fmt
+    mod.compileUnpack = function(fmt) return string.unpack, fmt end
+end
+
+return mod

internal/sha512.lua

@@ -10,13 +10,17 @@
 -- @module[kind=internal] internal.sha512
 --
 
-local expect = require "cc.expect".expect
+local expect  = require "cc.expect".expect
+local packing = require "ccryptolib.internal.packing"
 
 local shl = bit32.lshift
 local shr = bit32.rshift
 local bxor = bit32.bxor
 local bnot = bit32.bnot
 local band = bit32.band
+local p1x16, fmt1x16 = packing.compilePack(">I16")
+local p16x4, fmt16x4 = packing.compilePack(">I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4")
+local u32x4, fmt32x4 = packing.compileUnpack(">I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4")
 
 local function carry64(a1, a0)
     local r0 = a0 % 2 ^ 32
@@ -65,9 +69,7 @@ local function digest(data)
     -- Pad input.
     local bitlen = #data * 8
     local padlen = -(#data + 17) % 128
-    data = data .. "\x80" .. ("\0"):rep(padlen) .. (">I16"):pack(bitlen)
+    data = data .. "\x80" .. ("\0"):rep(padlen) .. p1x16(fmt1x16, bitlen)
-    local dataWords = {(">" .. ("I4"):rep(#data / 4)):unpack(data)}
-    dataWords[#dataWords] = nil
 
     -- Initialize state.
     local h01, h00 = 0x6a09e667, 0xf3bcc908
@@ -80,12 +82,8 @@ local function digest(data)
     local h71, h70 = 0x5be0cd19, 0x137e2179
 
     -- Digest.
-    local w = {}
+    for i = 1, #data, 128 do
-    for i = 1, #dataWords, 32 do
+        local w = {u32x4(fmt32x4, data, i)}
-        local index = i - 1
-        for j = 1, 32 do
-            w[j] = dataWords[index + j]
-        end
 
         -- Message schedule.
         for j = 33, 160, 2 do
@@ -172,7 +170,7 @@ local function digest(data)
         h71, h70 = carry64(h71 + h1, h70 + h0)
     end
 
-    return (">I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4"):pack(
+    return p16x4(fmt16x4,
         h01, h00, h11, h10, h21, h20, h31, h30,
         h41, h40, h51, h50, h61, h60, h71, h70
     )

poly1305.lua

@@ -3,8 +3,12 @@
 -- @module poly1305
 --
 
-local expect = require "cc.expect".expect
+local expect  = require "cc.expect".expect
-local random = require "ccryptolib.random"
+local random  = require "ccryptolib.random"
+local packing = require "ccryptolib.internal.packing"
+
+local u4x4, fmt4x4 = packing.compileUnpack("<I4I4I4I4")
+local p4x4 = packing.compilePack(fmt4x4)
 
 local mod = {}
 
@@ -27,7 +31,7 @@ function mod.mac(key, message)
     end
 
     -- Decode r.
-    local R0, R1, R2, R3 = ("<I4I4I4I4"):unpack(key)
+    local R0, R1, R2, R3 = u4x4(fmt4x4, key, 1)
 
     -- Clamp and shift.
     R0 = R0 % 2 ^ 28
@@ -55,7 +59,7 @@ function mod.mac(key, message)
 
     for i = 1, #message, 16 do
         -- Decode message block.
-        local m0, m1, m2, m3 = ("<I4I4I4I4"):unpack(message, i)
+        local m0, m1, m2, m3 = u4x4(fmt4x4, message, i)
 
         -- Shift message and add.
         local x0 = h0 + h1 + m0
@@ -119,7 +123,7 @@ function mod.mac(key, message)
     end
 
     -- Decode s.
-    local s0, s1, s2, s3 = ("<I4I4I4I4"):unpack(key, 17)
+    local s0, s1, s2, s3 = u4x4(fmt4x4, key, 17)
 
     -- Add.
     local t0 =           s0          + c0 + c1  local u0 = t0 % 2 ^ 32
@@ -128,7 +132,7 @@ function mod.mac(key, message)
     local t3 = t2 - u2 + s3 * 2 ^ 96 + c6 + c7  local u3 = t3 % 2 ^ 128
 
     -- Encode.
-    return ("<I4I4I4I4"):pack(u0, u1 / 2 ^ 32, u2 / 2 ^ 64, u3 / 2 ^ 96)
+    return p4x4(fmt4x4, u0, u1 / 2 ^ 32, u2 / 2 ^ 64, u3 / 2 ^ 96)
 end
 
 local mac = mod.mac

random.lua

@@ -1,5 +1,8 @@
 local blake3   = require "ccryptolib.blake3"
 local chacha20 = require "ccryptolib.chacha20"
+local packing  = require "ccryptolib.internal.packing"
+
+local u1x4, fmt1x4 = packing.compileUnpack("<I4")
 
 -- Initialize from local context.
 local ctx = {
@@ -55,7 +58,7 @@ end
 stir(512)
 
 -- Save.
-math.randomseed(("I4"):unpack(random(4)))
+math.randomseed(u1x4(fmt1x4, random(4), 1))
 save()
 
 return {

sha256.lua

@@ -3,7 +3,8 @@
 -- @module sha256
 --
 
-local expect = require "cc.expect".expect
+local expect  = require "cc.expect".expect
+local packing = require "ccryptolib.internal.packing"
 
 local rol = bit32.lrotate
 local shr = bit32.rshift
@@ -11,6 +12,11 @@ local bxor = bit32.bxor
 local bnot = bit32.bnot
 local band = bit32.band
 local unpack = unpack or table.unpack
+local p1x8, fmt1x8 = packing.compilePack(">I8")
+local p16x4, fmt16x4 = packing.compilePack(">I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4")
+local u16x4 = packing.compileUnpack(fmt16x4)
+local p8x4, fmt8x4 = packing.compilePack(">I4I4I4I4I4I4I4I4")
+local u8x4 = packing.compileUnpack(fmt8x4)
 
 local function primes(n, exp)
     local out = {}
@@ -82,15 +88,15 @@ local function digest(data)
     -- Pad input.
     local bitlen = #data * 8
     local padlen = -(#data + 9) % 64
-    data = data .. "\x80" .. ("\0"):rep(padlen) .. (">I8"):pack(bitlen)
+    data = data .. "\x80" .. ("\0"):rep(padlen) .. p1x8(fmt1x8, bitlen)
 
     -- Digest.
     local h = h0
     for i = 1, #data, 64 do
-        h = compress(h, {(">I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4"):unpack(data, i)})
+        h = compress(h, {u16x4(fmt16x4, data, i)})
     end
 
-    return (">I4I4I4I4I4I4I4I4"):pack(unpack(h))
+    return p8x4(fmt8x4, unpack(h))
 end
 
 --- Hashes a password using PBKDF2-HMAC-SHA256.
@@ -110,7 +116,7 @@ local function pbkdf2(password, salt, iter)
     -- Pad password.
     if #password > 64 then password = digest(password) end
     password = password .. ("\0"):rep(-#password % 64)
-    password = {(">I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4"):unpack(password)}
+    password = {u16x4(fmt16x4, password, 1)}
 
     -- Compute password blocks.
     local ikp = {}
@@ -127,8 +133,8 @@ local function pbkdf2(password, salt, iter)
     local pad96 = {2 ^ 31, 0, 0, 0, 0, 0, 0, 0x300}
 
     -- First iteration.
-    local pre = (">I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4I4"):pack(unpack(ikp))
+    local pre = p16x4(fmt16x4, unpack(ikp))
-    local hs = {(">I4I4I4I4I4I4I4I4"):unpack(digest(pre .. salt .. "\0\0\0\1"))}
+    local hs = {u8x4(fmt8x4, digest(pre .. salt .. "\0\0\0\1"), 1)}
     for i = 1, 8 do hs[i + 8] = pad96[i] end
     hs = compress(hokp, hs)
 
@@ -142,7 +148,7 @@ local function pbkdf2(password, salt, iter)
         for i = 1, 8 do out[i] = bxor(out[i], hs[i]) end
     end
 
-    return (">I4I4I4I4I4I4I4I4"):pack(unpack(out))
+    return p8x4(fmt8x4, unpack(out))
 end
 
 return {