Optimize SHA1 implementation. (#47)

Author: dcjones

src/common.jl

@@ -13,9 +13,7 @@ function update!(context::T, data::U) where {T<:SHA_CTX,
     usedspace = context.bytecount % blocklen(T)
     while len - data_idx + usedspace >= blocklen(T)
         # Fill up as much of the buffer as we can with the data given us
-        for i in 1:(blocklen(T) - usedspace)
-            context.buffer[usedspace + i] = data[data_idx + i]
-        end
+        copy!(context.buffer, usedspace + 1, data, data_idx + 1, blocklen(T) - usedspace)
 
         transform!(context)
         context.bytecount += blocklen(T) - usedspace
@@ -25,9 +23,7 @@ function update!(context::T, data::U) where {T<:SHA_CTX,
 
     # There is less than a complete block left, but we need to save the leftovers into context.buffer:
     if len > data_idx
-        for i = 1:(len - data_idx)
-            context.buffer[usedspace + i] = data[data_idx + i]
-        end
+        copy!(context.buffer, usedspace + 1, data, data_idx + 1, len - data_idx)
         context.bytecount += len - data_idx
     end
 end

src/constants.jl

@@ -123,7 +123,7 @@ const SHA3_ROTC = UInt64[
 ]
 
 # Permutation indices for SHA3 rounds (+1'ed so as to work with julia's 1-based indexing)
-const SHA3_PILN = UInt64[
+const SHA3_PILN = Int[
     11, 8,  12, 18, 19, 4, 6,  17, 9,  22, 25, 5,
     16, 24, 20, 14, 13, 3, 21, 15, 23, 10,  7,  2
 ]

src/sha3.jl

@@ -4,43 +4,48 @@ function transform!(context::T) where {T<:SHA3_CTX}
     for idx in 1:div(blocklen(T),8)
         context.state[idx] = context.state[idx] ⊻ unsafe_load(pbuf, idx)
     end
-    bc = Vector{UInt64}(uninitialized, 5)
+    bc = context.bc
+    state = context.state
 
     # We always assume 24 rounds
-    for round in 0:23
+    @inbounds for round in 0:23
         # Theta function
         for i in 1:5
-            bc[i] = context.state[i] ⊻ context.state[i + 5] ⊻ context.state[i + 10] ⊻ context.state[i + 15] ⊻ context.state[i + 20]
+            bc[i] = state[i] ⊻ state[i + 5] ⊻ state[i + 10] ⊻ state[i + 15] ⊻ state[i + 20]
         end
 
-        for i in 1:5
-            temp = bc[mod1(i + 4, 5)] ⊻ L64(1, bc[mod1(i + 1, 5)])
-            for j in 0:5:20
-                context.state[i + j] = context.state[i + j] ⊻ temp
+        for i in 0:4
+            temp = bc[rem(i + 4, 5) + 1] ⊻ L64(1, bc[rem(i + 1, 5) + 1])
+            j = 0
+            while j <= 20
+                state[Int(i + j + 1)] = state[i + j + 1] ⊻ temp
+                j += 5
             end
         end
 
         # Rho Pi
-        temp = context.state[2]
+        temp = state[2]
         for i in 1:24
             j = SHA3_PILN[i]
-            bc[1] = context.state[j]
-            context.state[j] = L64(SHA3_ROTC[i], temp)
+            bc[1] = state[j]
+            state[j] = L64(SHA3_ROTC[i], temp)
             temp = bc[1]
         end
 
         # Chi
-        for j in 0:5:20
+        j = 0
+        while j <= 20
             for i in 1:5
-                bc[i] = context.state[i + j]
+                bc[i] = state[i + j]
             end
-            for i in 1:5
-                context.state[j + i] = context.state[j + i] ⊻ (~bc[mod1(i + 1, 5)] & bc[mod1(i + 2, 5)])
+            for i in 0:4
+                state[j + i + 1] = state[j + i + 1] ⊻ (~bc[rem(i + 1, 5) + 1] & bc[rem(i + 2, 5) + 1])
             end
+            j += 5
         end
 
         # Iota
-        context.state[1] = context.state[1] ⊻ SHA3_ROUND_CONSTS[round+1]
+        state[1] = state[1] ⊻ SHA3_ROUND_CONSTS[round+1]
     end
 
     return context.state

src/types.jl

@@ -51,21 +51,25 @@ mutable struct SHA3_224_CTX <: SHA3_CTX
     state::Array{UInt64,1}
     bytecount::UInt128
     buffer::Array{UInt8,1}
+    bc::Array{UInt64,1}
 end
 mutable struct SHA3_256_CTX <: SHA3_CTX
     state::Array{UInt64,1}
     bytecount::UInt128
     buffer::Array{UInt8,1}
+    bc::Array{UInt64,1}
 end
 mutable struct SHA3_384_CTX <: SHA3_CTX
     state::Array{UInt64,1}
     bytecount::UInt128
     buffer::Array{UInt8,1}
+    bc::Array{UInt64,1}
 end
 mutable struct SHA3_512_CTX <: SHA3_CTX
     state::Array{UInt64,1}
     bytecount::UInt128
     buffer::Array{UInt8,1}
+    bc::Array{UInt64,1}
 end
 
 # Define constants via functions so as not to bloat context objects.  Yay dispatch!
@@ -111,10 +115,10 @@ SHA2_256_CTX() = SHA2_256_CTX(copy(SHA2_256_initial_hash_value), 0, zeros(UInt8,
 SHA2_384_CTX() = SHA2_384_CTX(copy(SHA2_384_initial_hash_value), 0, zeros(UInt8, blocklen(SHA2_384_CTX)))
 SHA2_512_CTX() = SHA2_512_CTX(copy(SHA2_512_initial_hash_value), 0, zeros(UInt8, blocklen(SHA2_512_CTX)))
 
-SHA3_224_CTX() = SHA3_224_CTX(zeros(UInt64, 25), 0, zeros(UInt8, blocklen(SHA3_224_CTX)))
-SHA3_256_CTX() = SHA3_256_CTX(zeros(UInt64, 25), 0, zeros(UInt8, blocklen(SHA3_256_CTX)))
-SHA3_384_CTX() = SHA3_384_CTX(zeros(UInt64, 25), 0, zeros(UInt8, blocklen(SHA3_384_CTX)))
-SHA3_512_CTX() = SHA3_512_CTX(zeros(UInt64, 25), 0, zeros(UInt8, blocklen(SHA3_512_CTX)))
+SHA3_224_CTX() = SHA3_224_CTX(zeros(UInt64, 25), 0, zeros(UInt8, blocklen(SHA3_224_CTX)), Vector{UInt64}(5))
+SHA3_256_CTX() = SHA3_256_CTX(zeros(UInt64, 25), 0, zeros(UInt8, blocklen(SHA3_256_CTX)), Vector{UInt64}(5))
+SHA3_384_CTX() = SHA3_384_CTX(zeros(UInt64, 25), 0, zeros(UInt8, blocklen(SHA3_384_CTX)), Vector{UInt64}(5))
+SHA3_512_CTX() = SHA3_512_CTX(zeros(UInt64, 25), 0, zeros(UInt8, blocklen(SHA3_512_CTX)), Vector{UInt64}(5))
 
 # Nickname'd outer constructor methods for SHA2
 const SHA224_CTX = SHA2_224_CTX
@@ -129,7 +133,7 @@ SHA1_CTX() = SHA1_CTX(copy(SHA1_initial_hash_value), 0, zeros(UInt8, blocklen(SH
 # Copy functions
 copy(ctx::T) where {T<:SHA1_CTX} = T(copy(ctx.state), ctx.bytecount, copy(ctx.buffer), copy(ctx.W))
 copy(ctx::T) where {T<:SHA2_CTX} = T(copy(ctx.state), ctx.bytecount, copy(ctx.buffer))
-copy(ctx::T) where {T<:SHA3_CTX} = T(copy(ctx.state), ctx.bytecount, copy(ctx.buffer))
+copy(ctx::T) where {T<:SHA3_CTX} = T(copy(ctx.state), ctx.bytecount, copy(ctx.buffer), Array{UInt64}(5))
 
 
 # Make printing these types a little friendlier

test/perf.jl

@@ -12,7 +12,7 @@ function do_tests(filepath)
     print("read:    ")
     @time begin
         const fh = open(filepath, "r")
-        const bytes = readbytes(fh)
+        const bytes = read(fh)
     end
     gc()