github_mirrors/vectorscan
3
0
Fork 1
mirror of https://github.com/VectorCamp/vectorscan.git synced 2025-11-19 10:34:25 +03:00
Code Issues Packages Projects Releases Wiki Activity

fix names, use own intrinsic instead of explicit _mm* ones

Browse Source
This commit is contained in:
Konstantinos Margaritis
2020-09-23 11:51:21 +03:00
committed by Konstantinos Margaritis
parent 4fd0723532
commit 8ef26f19fc
15 changed files with 137 additions and 137 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
src/nfa/shufti.c
View File

@@ -159,7 +159,7 @@ const u8 *shuftiExec(m128 mask_lo, m128 mask_hi, const u8 *buf,
}
const m128 zeroes = zeroes128();
const m128 low4bits = _mm_set1_epi8(0xf);
const m128 low4bits = set1_16x8(0xf);
const u8 *rv;
size_t min = (size_t)buf % 16;
@@ -246,7 +246,7 @@ const u8 *rshuftiExec(m128 mask_lo, m128 mask_hi, const u8 *buf,
}
const m128 zeroes = zeroes128();
const m128 low4bits = _mm_set1_epi8(0xf);
const m128 low4bits = set1_16x8(0xf);
const u8 *rv;
assert(buf_end - buf >= 16);
@@ -320,7 +320,7 @@ const u8 *shuftiDoubleExec(m128 mask1_lo, m128 mask1_hi,
m128 mask2_lo, m128 mask2_hi,
const u8 *buf, const u8 *buf_end) {
const m128 ones = ones128();
const m128 low4bits = _mm_set1_epi8(0xf);
const m128 low4bits = set1_16x8(0xf);
const u8 *rv;
size_t min = (size_t)buf % 16;
@@ -455,15 +455,15 @@ const u8 *shuftiExec(m128 mask_lo, m128 mask_hi, const u8 *buf,
buf, buf_end);
}
const m256 low4bits = set32x8(0xf);
const m256 low4bits = set1_32x8(0xf);
if (buf_end - buf <= 32) {
return shuftiFwdShort(mask_lo, mask_hi, buf, buf_end, low4bits);
}
const m256 zeroes = zeroes256();
const m256 wide_mask_lo = set2x128(mask_lo);
const m256 wide_mask_hi = set2x128(mask_hi);
const m256 wide_mask_lo = set1_2x128(mask_lo);
const m256 wide_mask_hi = set1_2x128(mask_hi);
const u8 *rv;
size_t min = (size_t)buf % 32;
@@ -579,15 +579,15 @@ const u8 *rshuftiExec(m128 mask_lo, m128 mask_hi, const u8 *buf,
buf, buf_end);
}
const m256 low4bits = set32x8(0xf);
const m256 low4bits = set1_32x8(0xf);
if (buf_end - buf <= 32) {
return shuftiRevShort(mask_lo, mask_hi, buf, buf_end, low4bits);
}
const m256 zeroes = zeroes256();
const m256 wide_mask_lo = set2x128(mask_lo);
const m256 wide_mask_hi = set2x128(mask_hi);
const m256 wide_mask_lo = set1_2x128(mask_lo);
const m256 wide_mask_hi = set1_2x128(mask_hi);
const u8 *rv;
assert(buf_end - buf >= 32);
@@ -676,7 +676,7 @@ static really_inline
const u8 *shuftiDoubleShort(m128 mask1_lo, m128 mask1_hi, m128 mask2_lo,
m128 mask2_hi, const u8 *buf, const u8 *buf_end) {
DEBUG_PRINTF("buf %p len %zu\n", buf, buf_end - buf);
const m256 low4bits = set32x8(0xf);
const m256 low4bits = set1_32x8(0xf);
// run shufti over two overlapping 16-byte unaligned reads
const m256 mask1 = combine2x128(mask1_hi, mask1_lo);
const m256 mask2 = combine2x128(mask2_hi, mask2_lo);
@@ -708,11 +708,11 @@ const u8 *shuftiDoubleExec(m128 mask1_lo, m128 mask1_hi,
}
const m256 ones = ones256();
const m256 low4bits = set32x8(0xf);
const m256 wide_mask1_lo = set2x128(mask1_lo);
const m256 wide_mask1_hi = set2x128(mask1_hi);
const m256 wide_mask2_lo = set2x128(mask2_lo);
const m256 wide_mask2_hi = set2x128(mask2_hi);
const m256 low4bits = set1_32x8(0xf);
const m256 wide_mask1_lo = set1_2x128(mask1_lo);
const m256 wide_mask1_hi = set1_2x128(mask1_hi);
const m256 wide_mask2_lo = set1_2x128(mask2_lo);
const m256 wide_mask2_hi = set1_2x128(mask2_hi);
const u8 *rv;
size_t min = (size_t)buf % 32;