mirror of
https://github.com/c3lang/c3c.git
synced 2026-02-27 03:51:18 +00:00
refactor md5.body function to reduce instruction count
- replaced manual unrolling with loop structures and constant arrays - instruction count reduced from 12445 to 4016 - maybe about 1 to 2% performance loss on some benchs but take this number with a grain of salt.
This commit is contained in:
Manuel Barrio Linares
committed by
Christoffer Lerno
Christoffer Lerno
parent
df030ac51c
commit
ec6ba8e7ca
41
benchmarks/stdlib/hash/md5.c3
Normal file
41
benchmarks/stdlib/hash/md5.c3
Normal file
@@ -0,0 +1,41 @@
|
||||
module md5_bench;
|
||||
|
||||
fn void initialize_bench() @init
|
||||
{
|
||||
set_benchmark_warmup_iterations(3);
|
||||
set_benchmark_max_iterations(128);
|
||||
|
||||
input = mem::alloc_array(char, BUFSZ);
|
||||
input[:BUFSZ] = (char[]){ [0..BUFSZ-1] = 0xA5 }[..];
|
||||
input_slice = input[:BUFSZ];
|
||||
}
|
||||
|
||||
fn void teardown_bench() @finalizer
|
||||
{
|
||||
mem::free(input);
|
||||
input = null;
|
||||
}
|
||||
|
||||
char* input;
|
||||
char[] input_slice;
|
||||
const usz BUFSZ = 1024 * 1024;
|
||||
|
||||
module md5_bench @benchmark;
|
||||
|
||||
import std::hash;
|
||||
|
||||
fn void md5_hash()
|
||||
{
|
||||
runtime::@start_benchmark();
|
||||
char[*] myset = md5::hash(input_slice);
|
||||
runtime::@end_benchmark();
|
||||
mem::zero_volatile(myset[..]);
|
||||
}
|
||||
|
||||
fn void compared_with_sha256()
|
||||
{
|
||||
runtime::@start_benchmark();
|
||||
char[*] myset = sha256::hash(input_slice);
|
||||
runtime::@end_benchmark();
|
||||
mem::zero_volatile(myset[..]);
|
||||
}
|
||||
@@ -99,6 +99,32 @@ fn char[HASH_BYTES] Md5.final(&ctx)
|
||||
|
||||
module std::hash::md5 @private;
|
||||
|
||||
const uint[64] MD5_T @private = {
|
||||
0xd76aa478, 0xe8c7b756, 0x242070db, 0xc1bdceee,
|
||||
0xf57c0faf, 0x4787c62a, 0xa8304613, 0xfd469501,
|
||||
0x698098d8, 0x8b44f7af, 0xffff5bb1, 0x895cd7be,
|
||||
0x6b901122, 0xfd987193, 0xa679438e, 0x49b40821,
|
||||
0xf61e2562, 0xc040b340, 0x265e5a51, 0xe9b6c7aa,
|
||||
0xd62f105d, 0x02441453, 0xd8a1e681, 0xe7d3fbc8,
|
||||
0x21e1cde6, 0xc33707d6, 0xf4d50d87, 0x455a14ed,
|
||||
0xa9e3e905, 0xfcefa3f8, 0x676f02d9, 0x8d2a4c8a,
|
||||
0xfffa3942, 0x8771f681, 0x6d9d6122, 0xfde5380c,
|
||||
0xa4beea44, 0x4bdecfa9, 0xf6bb4b60, 0xbebfbc70,
|
||||
0x289b7ec6, 0xeaa127fa, 0xd4ef3085, 0x04881d05,
|
||||
0xd9d4d039, 0xe6db99e5, 0x1fa27cf8, 0xc4ac5665,
|
||||
0xf4292244, 0x432aff97, 0xab9423a7, 0xfc93a039,
|
||||
0x655b59c3, 0x8f0ccc92, 0xffeff47d, 0x85845dd1,
|
||||
0x6fa87e4f, 0xfe2ce6e0, 0xa3014314, 0x4e0811a1,
|
||||
0xf7537e82, 0xbd3af235, 0x2ad7d2bb, 0xeb86d391
|
||||
};
|
||||
|
||||
const int[16] MD5_S @private = {
|
||||
7, 12, 17, 22,
|
||||
5, 9, 14, 20,
|
||||
4, 11, 16, 23,
|
||||
6, 10, 15, 21
|
||||
};
|
||||
|
||||
// Implementation
|
||||
macro @f(x, y, z) => z ^ (x & (y ^ z));
|
||||
macro @g(x, y, z) => y ^ (z & (x ^ y));
|
||||
@@ -116,93 +142,54 @@ macro void @step(#f, a, b, c, d, ptr, n, t, s)
|
||||
|
||||
fn char* body(Md5* ctx, void* data, usz size)
|
||||
{
|
||||
char* ptr;
|
||||
uint a, b, c, d;
|
||||
uint saved_a, saved_b, saved_c, saved_d;
|
||||
ptr = data;
|
||||
a = ctx.a;
|
||||
b = ctx.b;
|
||||
c = ctx.c;
|
||||
d = ctx.d;
|
||||
char* ptr = data;
|
||||
uint a = ctx.a;
|
||||
uint b = ctx.b;
|
||||
uint c = ctx.c;
|
||||
uint d = ctx.d;
|
||||
|
||||
do
|
||||
{
|
||||
saved_a = a;
|
||||
saved_b = b;
|
||||
saved_c = c;
|
||||
saved_d = d;
|
||||
uint saved_a = a;
|
||||
uint saved_b = b;
|
||||
uint saved_c = c;
|
||||
uint saved_d = d;
|
||||
|
||||
/* Round 1 */
|
||||
@step(@f, &a, b, c, d, ptr, 0, 0xd76aa478, 7) ;
|
||||
@step(@f, &d, a, b, c, ptr, 1, 0xe8c7b756, 12) ;
|
||||
@step(@f, &c, d, a, b, ptr, 2, 0x242070db, 17) ;
|
||||
@step(@f, &b, c, d, a, ptr, 3, 0xc1bdceee, 22) ;
|
||||
@step(@f, &a, b, c, d, ptr, 4, 0xf57c0faf, 7) ;
|
||||
@step(@f, &d, a, b, c, ptr, 5, 0x4787c62a, 12) ;
|
||||
@step(@f, &c, d, a, b, ptr, 6, 0xa8304613, 17) ;
|
||||
@step(@f, &b, c, d, a, ptr, 7, 0xfd469501, 22) ;
|
||||
@step(@f, &a, b, c, d, ptr, 8, 0x698098d8, 7) ;
|
||||
@step(@f, &d, a, b, c, ptr, 9, 0x8b44f7af, 12) ;
|
||||
@step(@f, &c, d, a, b, ptr, 10, 0xffff5bb1, 17);
|
||||
@step(@f, &b, c, d, a, ptr, 11, 0x895cd7be, 22);
|
||||
@step(@f, &a, b, c, d, ptr, 12, 0x6b901122, 7) ;
|
||||
@step(@f, &d, a, b, c, ptr, 13, 0xfd987193, 12);
|
||||
@step(@f, &c, d, a, b, ptr, 14, 0xa679438e, 17);
|
||||
@step(@f, &b, c, d, a, ptr, 15, 0x49b40821, 22);
|
||||
/* Round 1 */
|
||||
for (int i = 0; i < 16; i += 4)
|
||||
{
|
||||
@step(@f, &a, b, c, d, ptr, i + 0, MD5_T[i + 0], MD5_S[0]);
|
||||
@step(@f, &d, a, b, c, ptr, i + 1, MD5_T[i + 1], MD5_S[1]);
|
||||
@step(@f, &c, d, a, b, ptr, i + 2, MD5_T[i + 2], MD5_S[2]);
|
||||
@step(@f, &b, c, d, a, ptr, i + 3, MD5_T[i + 3], MD5_S[3]);
|
||||
}
|
||||
|
||||
/* Round 2 */
|
||||
@step(@g, &a, b, c, d, ptr, 1, 0xf61e2562, 5) ;
|
||||
@step(@g, &d, a, b, c, ptr, 6, 0xc040b340, 9) ;
|
||||
@step(@g, &c, d, a, b, ptr, 11, 0x265e5a51, 14);
|
||||
@step(@g, &b, c, d, a, ptr, 0, 0xe9b6c7aa, 20) ;
|
||||
@step(@g, &a, b, c, d, ptr, 5, 0xd62f105d, 5) ;
|
||||
@step(@g, &d, a, b, c, ptr, 10, 0x02441453, 9) ;
|
||||
@step(@g, &c, d, a, b, ptr, 15, 0xd8a1e681, 14);
|
||||
@step(@g, &b, c, d, a, ptr, 4, 0xe7d3fbc8, 20) ;
|
||||
@step(@g, &a, b, c, d, ptr, 9, 0x21e1cde6, 5) ;
|
||||
@step(@g, &d, a, b, c, ptr, 14, 0xc33707d6, 9) ;
|
||||
@step(@g, &c, d, a, b, ptr, 3, 0xf4d50d87, 14) ;
|
||||
@step(@g, &b, c, d, a, ptr, 8, 0x455a14ed, 20) ;
|
||||
@step(@g, &a, b, c, d, ptr, 13, 0xa9e3e905, 5) ;
|
||||
@step(@g, &d, a, b, c, ptr, 2, 0xfcefa3f8, 9) ;
|
||||
@step(@g, &c, d, a, b, ptr, 7, 0x676f02d9, 14) ;
|
||||
@step(@g, &b, c, d, a, ptr, 12, 0x8d2a4c8a, 20);
|
||||
/* Round 2 */
|
||||
for (int i = 0; i < 16; i += 4)
|
||||
{
|
||||
@step(@g, &a, b, c, d, ptr, (1 + 5 * (i + 0)) % 16, MD5_T[16 + i + 0], MD5_S[4]);
|
||||
@step(@g, &d, a, b, c, ptr, (1 + 5 * (i + 1)) % 16, MD5_T[16 + i + 1], MD5_S[5]);
|
||||
@step(@g, &c, d, a, b, ptr, (1 + 5 * (i + 2)) % 16, MD5_T[16 + i + 2], MD5_S[6]);
|
||||
@step(@g, &b, c, d, a, ptr, (1 + 5 * (i + 3)) % 16, MD5_T[16 + i + 3], MD5_S[7]);
|
||||
}
|
||||
|
||||
/* Round 3 */
|
||||
@step(@h, &a, b, c, d, ptr, 5, 0xfffa3942, 4);
|
||||
@step(@h2, &d, a, b, c, ptr, 8, 0x8771f681, 11);
|
||||
@step(@h, &c, d, a, b, ptr, 11, 0x6d9d6122, 16);
|
||||
@step(@h2, &b, c, d, a, ptr, 14, 0xfde5380c, 23);
|
||||
@step(@h, &a, b, c, d, ptr, 1, 0xa4beea44, 4);
|
||||
@step(@h2, &d, a, b, c, ptr, 4, 0x4bdecfa9, 11);
|
||||
@step(@h, &c, d, a, b, ptr, 7, 0xf6bb4b60, 16);
|
||||
@step(@h2, &b, c, d, a, ptr, 10, 0xbebfbc70, 23);
|
||||
@step(@h, &a, b, c, d, ptr, 13, 0x289b7ec6, 4) ;
|
||||
@step(@h2, &d, a, b, c, ptr, 0, 0xeaa127fa, 11) ;
|
||||
@step(@h, &c, d, a, b, ptr, 3, 0xd4ef3085, 16) ;
|
||||
@step(@h2, &b, c, d, a, ptr, 6, 0x04881d05, 23) ;
|
||||
@step(@h, &a, b, c, d, ptr, 9, 0xd9d4d039, 4) ;
|
||||
@step(@h2, &d, a, b, c, ptr, 12, 0xe6db99e5, 11) ;
|
||||
@step(@h, &c, d, a, b, ptr, 15, 0x1fa27cf8, 16) ;
|
||||
@step(@h2, &b, c, d, a, ptr, 2, 0xc4ac5665, 23) ;
|
||||
/* Round 3 */
|
||||
for (int i = 0; i < 16; i += 4)
|
||||
{
|
||||
@step(@h, &a, b, c, d, ptr, (5 + 3 * (i + 0)) % 16, MD5_T[32 + i + 0], MD5_S[8]);
|
||||
@step(@h, &d, a, b, c, ptr, (5 + 3 * (i + 1)) % 16, MD5_T[32 + i + 1], MD5_S[9]);
|
||||
@step(@h, &c, d, a, b, ptr, (5 + 3 * (i + 2)) % 16, MD5_T[32 + i + 2], MD5_S[10]);
|
||||
@step(@h, &b, c, d, a, ptr, (5 + 3 * (i + 3)) % 16, MD5_T[32 + i + 3], MD5_S[11]);
|
||||
}
|
||||
|
||||
/* Round 4 */
|
||||
@step(@i, &a, b, c, d, ptr, 0, 0xf4292244, 6) ;
|
||||
@step(@i, &d, a, b, c, ptr, 7, 0x432aff97, 10) ;
|
||||
@step(@i, &c, d, a, b, ptr, 14, 0xab9423a7, 15) ;
|
||||
@step(@i, &b, c, d, a, ptr, 5, 0xfc93a039, 21) ;
|
||||
@step(@i, &a, b, c, d, ptr, 12, 0x655b59c3, 6) ;
|
||||
@step(@i, &d, a, b, c, ptr, 3, 0x8f0ccc92, 10) ;
|
||||
@step(@i, &c, d, a, b, ptr, 10, 0xffeff47d, 15) ;
|
||||
@step(@i, &b, c, d, a, ptr, 1, 0x85845dd1, 21) ;
|
||||
@step(@i, &a, b, c, d, ptr, 8, 0x6fa87e4f, 6) ;
|
||||
@step(@i, &d, a, b, c, ptr, 15, 0xfe2ce6e0, 10) ;
|
||||
@step(@i, &c, d, a, b, ptr, 6, 0xa3014314, 15) ;
|
||||
@step(@i, &b, c, d, a, ptr, 13, 0x4e0811a1, 21) ;
|
||||
@step(@i, &a, b, c, d, ptr, 4, 0xf7537e82, 6) ;
|
||||
@step(@i, &d, a, b, c, ptr, 11, 0xbd3af235, 10) ;
|
||||
@step(@i, &c, d, a, b, ptr, 2, 0x2ad7d2bb, 15) ;
|
||||
@step(@i, &b, c, d, a, ptr, 9, 0xeb86d391, 21) ;
|
||||
/* Round 4 */
|
||||
for (int i = 0; i < 16; i += 4)
|
||||
{
|
||||
@step(@i, &a, b, c, d, ptr, (7 * (i + 0)) % 16, MD5_T[48 + i + 0], MD5_S[12]);
|
||||
@step(@i, &d, a, b, c, ptr, (7 * (i + 1)) % 16, MD5_T[48 + i + 1], MD5_S[13]);
|
||||
@step(@i, &c, d, a, b, ptr, (7 * (i + 2)) % 16, MD5_T[48 + i + 2], MD5_S[14]);
|
||||
@step(@i, &b, c, d, a, ptr, (7 * (i + 3)) % 16, MD5_T[48 + i + 3], MD5_S[15]);
|
||||
}
|
||||
|
||||
a += saved_a;
|
||||
b += saved_b;
|
||||
|
||||
Reference in New Issue
Block a user