diff options
Diffstat (limited to 'libbb/hash_md5_sha256_x86-64_shaNI.S')
-rw-r--r-- | libbb/hash_md5_sha256_x86-64_shaNI.S | 19 |
1 files changed, 9 insertions, 10 deletions
diff --git a/libbb/hash_md5_sha256_x86-64_shaNI.S b/libbb/hash_md5_sha256_x86-64_shaNI.S index c246762aa..5ed80c2ef 100644 --- a/libbb/hash_md5_sha256_x86-64_shaNI.S +++ b/libbb/hash_md5_sha256_x86-64_shaNI.S | |||
@@ -4,7 +4,7 @@ | |||
4 | // We use shorter insns, even though they are for "wrong" | 4 | // We use shorter insns, even though they are for "wrong" |
5 | // data type (fp, not int). | 5 | // data type (fp, not int). |
6 | // For Intel, there is no penalty for doing it at all | 6 | // For Intel, there is no penalty for doing it at all |
7 | // (CPUs which do have such penalty do not support SHA1 insns). | 7 | // (CPUs which do have such penalty do not support SHA insns). |
8 | // For AMD, the penalty is one extra cycle | 8 | // For AMD, the penalty is one extra cycle |
9 | // (allegedly: I failed to find measurable difference). | 9 | // (allegedly: I failed to find measurable difference). |
10 | 10 | ||
@@ -42,12 +42,13 @@ | |||
42 | .balign 8 # allow decoders to fetch at least 2 first insns | 42 | .balign 8 # allow decoders to fetch at least 2 first insns |
43 | sha256_process_block64_shaNI: | 43 | sha256_process_block64_shaNI: |
44 | 44 | ||
45 | movu128 80+0*16(%rdi), XMMTMP /* DCBA (msb-to-lsb: 3,2,1,0) */ | 45 | movu128 80+0*16(%rdi), XMMTMP /* ABCD (little-endian dword order) */ |
46 | movu128 80+1*16(%rdi), STATE1 /* HGFE */ | 46 | movu128 80+1*16(%rdi), STATE1 /* EFGH */ |
47 | /* shufps takes dwords 0,1 from *2nd* operand, and dwords 2,3 from 1st one */ | 47 | /* shufps takes dwords 0,1 from *2nd* operand, and dwords 2,3 from 1st one */ |
48 | mova128 STATE1, STATE0 | 48 | mova128 STATE1, STATE0 |
49 | shufps SHUF(1,0,1,0), XMMTMP, STATE0 /* ABEF */ | 49 | /* --- -------------- ABCD -- EFGH */ |
50 | shufps SHUF(3,2,3,2), XMMTMP, STATE1 /* CDGH */ | 50 | shufps SHUF(1,0,1,0), XMMTMP, STATE0 /* FEBA */ |
51 | shufps SHUF(3,2,3,2), XMMTMP, STATE1 /* HGDC */ | ||
51 | 52 | ||
52 | /* XMMTMP holds flip mask from here... */ | 53 | /* XMMTMP holds flip mask from here... */ |
53 | mova128 PSHUFFLE_BSWAP32_FLIP_MASK(%rip), XMMTMP | 54 | mova128 PSHUFFLE_BSWAP32_FLIP_MASK(%rip), XMMTMP |
@@ -243,13 +244,11 @@ sha256_process_block64_shaNI: | |||
243 | paddd CDGH_SAVE, STATE1 | 244 | paddd CDGH_SAVE, STATE1 |
244 | 245 | ||
245 | /* Write hash values back in the correct order */ | 246 | /* Write hash values back in the correct order */ |
246 | /* STATE0: ABEF (msb-to-lsb: 3,2,1,0) */ | ||
247 | /* STATE1: CDGH */ | ||
248 | mova128 STATE0, XMMTMP | 247 | mova128 STATE0, XMMTMP |
249 | /* shufps takes dwords 0,1 from *2nd* operand, and dwords 2,3 from 1st one */ | 248 | /* shufps takes dwords 0,1 from *2nd* operand, and dwords 2,3 from 1st one */ |
250 | shufps SHUF(3,2,3,2), STATE1, STATE0 /* DCBA */ | 249 | /* --- -------------- HGDC -- FEBA */ |
251 | shufps SHUF(1,0,1,0), STATE1, XMMTMP /* HGFE */ | 250 | shufps SHUF(3,2,3,2), STATE1, STATE0 /* ABCD */ |
252 | 251 | shufps SHUF(1,0,1,0), STATE1, XMMTMP /* EFGH */ | |
253 | movu128 STATE0, 80+0*16(%rdi) | 252 | movu128 STATE0, 80+0*16(%rdi) |
254 | movu128 XMMTMP, 80+1*16(%rdi) | 253 | movu128 XMMTMP, 80+1*16(%rdi) |
255 | 254 | ||