Skip to content

Instantly share code, notes, and snippets.

@leo60228
Created August 11, 2022 14:41
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save leo60228/4a0d74c6c2a11940d4171547098555af to your computer and use it in GitHub Desktop.
Save leo60228/4a0d74c6c2a11940d4171547098555af to your computer and use it in GitHub Desktop.
#include <stdio.h>
#include <stdint.h>
#include <string.h>
#include <arm_neon.h>
#include <arm_acle.h>
const uint32_t K[] = {
0x428A2F98, 0x71374491, 0xB5C0FBCF, 0xE9B5DBA5,
0x3956C25B, 0x59F111F1, 0x923F82A4, 0xAB1C5ED5,
0xD807AA98, 0x12835B01, 0x243185BE, 0x550C7DC3,
0x72BE5D74, 0x80DEB1FE, 0x9BDC06A7, 0xC19BF174,
0xE49B69C1, 0xEFBE4786, 0x0FC19DC6, 0x240CA1CC,
0x2DE92C6F, 0x4A7484AA, 0x5CB0A9DC, 0x76F988DA,
0x983E5152, 0xA831C66D, 0xB00327C8, 0xBF597FC7,
0xC6E00BF3, 0xD5A79147, 0x06CA6351, 0x14292967,
0x27B70A85, 0x2E1B2138, 0x4D2C6DFC, 0x53380D13,
0x650A7354, 0x766A0ABB, 0x81C2C92E, 0x92722C85,
0xA2BFE8A1, 0xA81A664B, 0xC24B8B70, 0xC76C51A3,
0xD192E819, 0xD6990624, 0xF40E3585, 0x106AA070,
0x19A4C116, 0x1E376C08, 0x2748774C, 0x34B0BCB5,
0x391C0CB3, 0x4ED8AA4A, 0x5B9CCA4F, 0x682E6FF3,
0x748F82EE, 0x78A5636F, 0x84C87814, 0x8CC70208,
0x90BEFFFA, 0xA4506CEB, 0xBEF9A3F7, 0xC67178F2,
};
void sha256_16(uint32_t *io) {
uint32_t temp1, temp2, W[64];
uint32_t A[8] = {
0x6A09E667,
0xBB67AE85,
0x3C6EF372,
0xA54FF53A,
0x510E527F,
0x9B05688C,
0x1F83D9AB,
0x5BE0CD19
};
unsigned int i;
// padding and msglen identical/similar to sha1_16
W[0] = io[0];
W[1] = io[1];
W[2] = io[2];
W[3] = io[3];
W[4] = 0x80000000u;
W[5] = 0; W[6] = 0; W[7] = 0;
W[8] = 0; W[9] = 0; W[10] = 0; W[11] = 0;
W[12] = 0; W[13] = 0; W[14] = 0;
W[15] = 0x80u;
uint32x4_t STATE0, STATE1;
uint32x4_t MSG0, MSG1, MSG2, MSG3;
uint32x4_t TMP0, TMP1, TMP2;
STATE0 = vld1q_u32(&A[0]);
STATE1 = vld1q_u32(&A[4]);
MSG0 = vld1q_u32(&W[0]);
MSG1 = vld1q_u32(&W[4]);
MSG2 = vld1q_u32(&W[8]);
MSG3 = vld1q_u32(&W[12]);
TMP0 = vaddq_u32(MSG0, vld1q_u32(&K[0]));
/* Rounds 0-3 */
MSG0 = vsha256su0q_u32(MSG0, MSG1);
TMP2 = STATE0;
TMP1 = vaddq_u32(MSG1, vld1q_u32(&K[0x04]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP0);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP0);
MSG0 = vsha256su1q_u32(MSG0, MSG2, MSG3);
/* Rounds 4-7 */
MSG1 = vsha256su0q_u32(MSG1, MSG2);
TMP2 = STATE0;
TMP0 = vaddq_u32(MSG2, vld1q_u32(&K[0x08]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP1);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP1);
MSG1 = vsha256su1q_u32(MSG1, MSG3, MSG0);
/* Rounds 8-11 */
MSG2 = vsha256su0q_u32(MSG2, MSG3);
TMP2 = STATE0;
TMP1 = vaddq_u32(MSG3, vld1q_u32(&K[0x0c]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP0);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP0);
MSG2 = vsha256su1q_u32(MSG2, MSG0, MSG1);
/* Rounds 12-15 */
MSG3 = vsha256su0q_u32(MSG3, MSG0);
TMP2 = STATE0;
TMP0 = vaddq_u32(MSG0, vld1q_u32(&K[0x10]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP1);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP1);
MSG3 = vsha256su1q_u32(MSG3, MSG1, MSG2);
/* Rounds 16-19 */
MSG0 = vsha256su0q_u32(MSG0, MSG1);
TMP2 = STATE0;
TMP1 = vaddq_u32(MSG1, vld1q_u32(&K[0x14]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP0);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP0);
MSG0 = vsha256su1q_u32(MSG0, MSG2, MSG3);
/* Rounds 20-23 */
MSG1 = vsha256su0q_u32(MSG1, MSG2);
TMP2 = STATE0;
TMP0 = vaddq_u32(MSG2, vld1q_u32(&K[0x18]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP1);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP1);
MSG1 = vsha256su1q_u32(MSG1, MSG3, MSG0);
/* Rounds 24-27 */
MSG2 = vsha256su0q_u32(MSG2, MSG3);
TMP2 = STATE0;
TMP1 = vaddq_u32(MSG3, vld1q_u32(&K[0x1c]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP0);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP0);
MSG2 = vsha256su1q_u32(MSG2, MSG0, MSG1);
/* Rounds 28-31 */
MSG3 = vsha256su0q_u32(MSG3, MSG0);
TMP2 = STATE0;
TMP0 = vaddq_u32(MSG0, vld1q_u32(&K[0x20]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP1);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP1);
MSG3 = vsha256su1q_u32(MSG3, MSG1, MSG2);
/* Rounds 32-35 */
MSG0 = vsha256su0q_u32(MSG0, MSG1);
TMP2 = STATE0;
TMP1 = vaddq_u32(MSG1, vld1q_u32(&K[0x24]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP0);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP0);
MSG0 = vsha256su1q_u32(MSG0, MSG2, MSG3);
/* Rounds 36-39 */
MSG1 = vsha256su0q_u32(MSG1, MSG2);
TMP2 = STATE0;
TMP0 = vaddq_u32(MSG2, vld1q_u32(&K[0x28]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP1);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP1);
MSG1 = vsha256su1q_u32(MSG1, MSG3, MSG0);
/* Rounds 40-43 */
MSG2 = vsha256su0q_u32(MSG2, MSG3);
TMP2 = STATE0;
TMP1 = vaddq_u32(MSG3, vld1q_u32(&K[0x2c]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP0);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP0);
MSG2 = vsha256su1q_u32(MSG2, MSG0, MSG1);
/* Rounds 44-47 */
MSG3 = vsha256su0q_u32(MSG3, MSG0);
TMP2 = STATE0;
TMP0 = vaddq_u32(MSG0, vld1q_u32(&K[0x30]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP1);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP1);
MSG3 = vsha256su1q_u32(MSG3, MSG1, MSG2);
/* Rounds 48-51 */
TMP2 = STATE0;
TMP1 = vaddq_u32(MSG1, vld1q_u32(&K[0x34]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP0);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP0);
/* Rounds 52-55 */
TMP2 = STATE0;
TMP0 = vaddq_u32(MSG2, vld1q_u32(&K[0x38]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP1);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP1);
/* Rounds 56-59 */
TMP2 = STATE0;
TMP1 = vaddq_u32(MSG3, vld1q_u32(&K[0x3c]));
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP0);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP0);
/* Rounds 60-63 */
TMP2 = STATE0;
STATE0 = vsha256hq_u32(STATE0, STATE1, TMP1);
STATE1 = vsha256h2q_u32(STATE1, TMP2, TMP1);
STATE0 = vaddq_u32(STATE0, vld1q_u32(&A[0]));
vst1q_u32(io, STATE0);
}
int main(int argc, char* argv[]) {
unsigned char buf[16];
uint32_t io[4];
if (fread(buf, sizeof(buf), 1, stdin) != 1) {
perror("fread");
return 1;
}
memcpy(io, buf, sizeof(buf));
for (int i = 0; i < 4; i++) {
io[i] = __builtin_bswap32(io[i]);
}
sha256_16(io);
for (int i = 0; i < 4; i++) {
printf("%08x", io[i]);
}
printf("\n");
return 0;
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment