mirrors/zig
1
0
Fork 0
mirror of https://codeberg.org/ziglang/zig.git synced 2025-12-06 13:54:21 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

Unroll Sha3 inner loop

Browse source 
Issue #699 since fixed. Nearly a x3 perf improvement.

Using --release-fast.

Sha3_256 (before): 96 Mb/s
Sha3_256  (after): 267 Mb/s

Sha3_512 (before): 53 Mb/s
Sha3_512  (after): 142 Mb/s

No real gains from unrolling other initialization loops in crypto
functions so have been left as is.
This commit is contained in:
Marc Tiehuis 2018-03-10 10:00:07 +13:00
parent 5a7a0e8518
commit 7a893691c0
3 changed files with 10 additions and 14 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1
std/crypto/md5.zig
View file

@ -108,7 +108,6 @@ pub const Md5 = struct {
var s: [16]u32 = undefined; var s: [16]u32 = undefined;
// ERROR: cannot unroll this at comptime
var i: usize = 0; var i: usize = 0;
while (i < 16) : (i += 1) { while (i < 16) : (i += 1) {
// NOTE: Performing or's separately improves perf by ~10% // NOTE: Performing or's separately improves perf by ~10%

2
std/crypto/sha2.zig
View file

@ -156,7 +156,6 @@ fn Sha2_32(comptime params: Sha2Params32) type { return struct {
var s: [64]u32 = undefined; var s: [64]u32 = undefined;
// ERROR: Cannot unroll at compile-time.
var i: usize = 0; var i: usize = 0;
while (i < 16) : (i += 1) { while (i < 16) : (i += 1) {
s[i] = 0; s[i] = 0;
@ -472,7 +471,6 @@ fn Sha2_64(comptime params: Sha2Params64) type { return struct {
var s: [80]u64 = undefined; var s: [80]u64 = undefined;
// ERROR: Cannot unroll at compile-time.
var i: usize = 0; var i: usize = 0;
while (i < 16) : (i += 1) { while (i < 16) : (i += 1) {
s[i] = 0; s[i] = 0;

21
std/crypto/sha3.zig
View file

@ -123,35 +123,34 @@ fn keccak_f(comptime F: usize, d: []u8) void {
*r = mem.readIntLE(u64, d[8*i .. 8*i + 8]); *r = mem.readIntLE(u64, d[8*i .. 8*i + 8]);
} }
var x: usize = 0; comptime var x: usize = 0;
var y: usize = 0; comptime var y: usize = 0;
// TODO: Cannot unroll all loops here due to comptime differences. for (RC[0..no_rounds]) |round| {
inline for (RC[0..no_rounds]) |round| {
// theta // theta
x = 0; while (x < 5) : (x += 1) { x = 0; inline while (x < 5) : (x += 1) {
c[x] = s[x] ^ s[x+5] ^ s[x+10] ^ s[x+15] ^ s[x+20]; c[x] = s[x] ^ s[x+5] ^ s[x+10] ^ s[x+15] ^ s[x+20];
} }
x = 0; while (x < 5) : (x += 1) { x = 0; inline while (x < 5) : (x += 1) {
t[0] = c[M5[x+4]] ^ math.rotl(u64, c[M5[x+1]], usize(1)); t[0] = c[M5[x+4]] ^ math.rotl(u64, c[M5[x+1]], usize(1));
y = 0; while (y < 5) : (y += 1) { y = 0; inline while (y < 5) : (y += 1) {
s[x + y*5] ^= t[0]; s[x + y*5] ^= t[0];
} }
} }
// rho+pi // rho+pi
t[0] = s[1]; t[0] = s[1];
x = 0; while (x < 24) : (x += 1) { x = 0; inline while (x < 24) : (x += 1) {
c[0] = s[PIL[x]]; c[0] = s[PIL[x]];
s[PIL[x]] = math.rotl(u64, t[0], ROTC[x]); s[PIL[x]] = math.rotl(u64, t[0], ROTC[x]);
t[0] = c[0]; t[0] = c[0];
} }
// chi // chi
y = 0; while (y < 5) : (y += 1) { y = 0; inline while (y < 5) : (y += 1) {
x = 0; while (x < 5) : (x += 1) { x = 0; inline while (x < 5) : (x += 1) {
c[x] = s[x + y*5]; c[x] = s[x + y*5];
} }
x = 0; while (x < 5) : (x += 1) { x = 0; inline while (x < 5) : (x += 1) {
s[x + y*5] = c[x] ^ (~c[M5[x+1]] & c[M5[x+2]]); s[x + y*5] = c[x] ^ (~c[M5[x+1]] & c[M5[x+2]]);
} }
} }