Change RAID AVX512/GFNI functions to use forward iterations

This commit is contained in:
Andrea Mazzoleni 2026-05-30 10:11:00 +02:00
parent 85b4d65308
commit 2ae305fbc4
3 changed files with 145 additions and 276 deletions

View file

@ -136,16 +136,18 @@ void raid_gen1_sse2(int nd, size_t size, void **vv)
raid_sse_begin();
for (i = 0; i < size; i += 64) {
asm volatile ("movdqa %0,%%xmm0" : : "m" (v[l][i]));
asm volatile ("movdqa %0,%%xmm1" : : "m" (v[l][i + 16]));
asm volatile ("movdqa %0,%%xmm2" : : "m" (v[l][i + 32]));
asm volatile ("movdqa %0,%%xmm3" : : "m" (v[l][i + 48]));
for (d = l - 1; d >= 0; --d) {
asm volatile ("movdqa %0,%%xmm0" : : "m" (v[0][i]));
asm volatile ("movdqa %0,%%xmm1" : : "m" (v[0][i + 16]));
asm volatile ("movdqa %0,%%xmm2" : : "m" (v[0][i + 32]));
asm volatile ("movdqa %0,%%xmm3" : : "m" (v[0][i + 48]));
for (d = 1; d <= l; ++d) {
asm volatile ("pxor %0,%%xmm0" : : "m" (v[d][i]));
asm volatile ("pxor %0,%%xmm1" : : "m" (v[d][i + 16]));
asm volatile ("pxor %0,%%xmm2" : : "m" (v[d][i + 32]));
asm volatile ("pxor %0,%%xmm3" : : "m" (v[d][i + 48]));
}
asm volatile ("movntdq %%xmm0,%0" : "=m" (p[i]));
asm volatile ("movntdq %%xmm1,%0" : "=m" (p[i + 16]));
asm volatile ("movntdq %%xmm2,%0" : "=m" (p[i + 32]));
@ -171,12 +173,14 @@ void raid_gen1_avx2(int nd, size_t size, void **vv)
raid_avx_begin();
for (i = 0; i < size; i += 64) {
asm volatile ("vmovdqa %0,%%ymm0" : : "m" (v[l][i]));
asm volatile ("vmovdqa %0,%%ymm1" : : "m" (v[l][i + 32]));
for (d = l - 1; d >= 0; --d) {
asm volatile ("vmovdqa %0,%%ymm0" : : "m" (v[0][i]));
asm volatile ("vmovdqa %0,%%ymm1" : : "m" (v[0][i + 32]));
for (d = 1; d <= l; ++d) {
asm volatile ("vpxor %0,%%ymm0,%%ymm0" : : "m" (v[d][i]));
asm volatile ("vpxor %0,%%ymm1,%%ymm1" : : "m" (v[d][i + 32]));
}
asm volatile ("vmovntdq %%ymm0,%0" : "=m" (p[i]));
asm volatile ("vmovntdq %%ymm1,%0" : "=m" (p[i + 32]));
}
@ -205,19 +209,19 @@ void raid_gen1_avx512bw(int nd, size_t size, void **vv)
raid_avx_begin();
for (i = 0; i < size; i += 64) {
asm volatile ("vmovdqa64 %0,%%zmm0" : : "m" (v[l][i]));
for (d = l - 1; d >= 1; d -= 2) {
asm volatile ("vmovdqa64 %0,%%zmm0" : : "m" (v[0][i]));
for (d = 1; d <= l - 1; d += 2) {
asm volatile (
"vmovdqa64 %0,%%zmm1\n\t"
"vpternlogq $0x96,%1,%%zmm1,%%zmm0"
:
: "m" (v[d][i]), "m" (v[d-1][i])
: "m" (v[d][i]), "m" (v[d+1][i])
);
}
if (d == 0) {
asm volatile ("vpxorq %0, %%zmm0,%%zmm0" : : "m" (v[0][i]));
if (d == l) {
asm volatile ("vpxorq %0,%%zmm0,%%zmm0" : : "m" (v[l][i]));
}
asm volatile ("vmovntdq %%zmm0,%0" : "=m" (p[i]));
@ -265,6 +269,7 @@ void raid_gen2_sse2(int nd, size_t size, void **vv)
asm volatile ("movdqa %0,%%xmm1" : : "m" (v[l][i + 16]));
asm volatile ("movdqa %xmm0,%xmm2");
asm volatile ("movdqa %xmm1,%xmm3");
for (d = l - 1; d >= 0; --d) {
/* scale by 2 before adding the current disk so the first (last processed) disk with factor 1 avoids doubling */
asm volatile ("pxor %xmm4,%xmm4");
@ -320,6 +325,7 @@ void raid_gen2_avx2(int nd, size_t size, void **vv)
asm volatile ("vmovdqa %0,%%ymm1" : : "m" (v[l][i + 32]));
asm volatile ("vmovdqa %ymm0,%ymm2");
asm volatile ("vmovdqa %ymm1,%ymm3");
for (d = l - 1; d >= 0; --d) {
/* scale by 2 before adding the current disk so the first (last processed) disk with factor 1 avoids doubling */
asm volatile ("vpcmpgtb %ymm2,%ymm6,%ymm4");
@ -338,6 +344,7 @@ void raid_gen2_avx2(int nd, size_t size, void **vv)
asm volatile ("vpxor %ymm4,%ymm2,%ymm2");
asm volatile ("vpxor %ymm5,%ymm3,%ymm3");
}
asm volatile ("vmovntdq %%ymm0,%0" : "=m" (p[i]));
asm volatile ("vmovntdq %%ymm1,%0" : "=m" (p[i + 32]));
asm volatile ("vmovntdq %%ymm2,%0" : "=m" (q[i]));
@ -368,10 +375,10 @@ void raid_gen2_avx512bw(int nd, size_t size, void **vv)
asm volatile ("vpbroadcastb %0,%%zmm31" : : "m" (gfconst16.low4[0]));
for (i = 0; i < size; i += 64) {
asm volatile ("vpxorq %zmm0,%zmm0,%zmm0");
asm volatile ("vpxorq %zmm1,%zmm1,%zmm1");
asm volatile ("vmovdqa64 %0,%%zmm0" : : "m" (v[0][i]));
asm volatile ("vmovdqa64 %zmm0,%zmm1");
for (d = l; d > 0; --d) {
for (d = 1; d <= l; ++d) {
asm volatile ("vmovdqa64 %0,%%zmm10" : : "m" (v[d][i]));
asm volatile ("vpxorq %zmm10,%zmm0,%zmm0");
@ -388,12 +395,6 @@ void raid_gen2_avx512bw(int nd, size_t size, void **vv)
asm volatile ("vpxorq %zmm13,%zmm1,%zmm1");
}
/* first disk with all coefficients at 1 */
asm volatile ("vmovdqa64 %0,%%zmm10" : : "m" (v[0][i]));
asm volatile ("vpxorq %zmm10,%zmm0,%zmm0");
asm volatile ("vpxorq %zmm10,%zmm1,%zmm1");
asm volatile ("vmovntdq %%zmm0,%0" : "=m" (p[i]));
asm volatile ("vmovntdq %%zmm1,%0" : "=m" (q[i]));
}
@ -433,6 +434,7 @@ void raid_gen2_sse2ext(int nd, size_t size, void **vv)
asm volatile ("movdqa %xmm1,%xmm5");
asm volatile ("movdqa %xmm2,%xmm6");
asm volatile ("movdqa %xmm3,%xmm7");
for (d = l - 1; d >= 0; --d) {
/* scale by 2 before adding the current disk so the first (last processed) disk with factor 1 avoids doubling */
asm volatile ("pxor %xmm8,%xmm8");
@ -469,6 +471,7 @@ void raid_gen2_sse2ext(int nd, size_t size, void **vv)
asm volatile ("pxor %xmm10,%xmm6");
asm volatile ("pxor %xmm11,%xmm7");
}
asm volatile ("movntdq %%xmm0,%0" : "=m" (p[i]));
asm volatile ("movntdq %%xmm1,%0" : "=m" (p[i + 16]));
asm volatile ("movntdq %%xmm2,%0" : "=m" (p[i + 32]));
@ -875,11 +878,11 @@ void raid_gen3_avx512bw(int nd, size_t size, void **vv)
asm volatile ("vpbroadcastb %0,%%zmm31" : : "m" (gfconst16.low4[0]));
for (i = 0; i < size; i += 64) {
asm volatile ("vpxorq %zmm0,%zmm0,%zmm0");
asm volatile ("vpxorq %zmm1,%zmm1,%zmm1");
asm volatile ("vpxorq %zmm2,%zmm2,%zmm2");
asm volatile ("vmovdqa64 %0,%%zmm0" : : "m" (v[0][i]));
asm volatile ("vmovdqa64 %zmm0,%zmm1");
asm volatile ("vmovdqa64 %zmm0,%zmm2");
for (d = l; d > 0; --d) {
for (d = 1; d <= l; ++d) {
asm volatile ("vmovdqa64 %0,%%zmm10" : : "m" (v[d][i]));
asm volatile ("vpxorq %zmm10,%zmm0,%zmm0");
@ -902,13 +905,6 @@ void raid_gen3_avx512bw(int nd, size_t size, void **vv)
asm volatile ("vpxorq %zmm15,%zmm2,%zmm2");
}
/* first disk with all coefficients at 1 */
asm volatile ("vmovdqa64 %0,%%zmm10" : : "m" (v[0][i]));
asm volatile ("vpxorq %zmm10,%zmm0,%zmm0");
asm volatile ("vpxorq %zmm10,%zmm1,%zmm1");
asm volatile ("vpxorq %zmm10,%zmm2,%zmm2");
asm volatile ("vmovntdq %%zmm0,%0" : "=m" (p[i]));
asm volatile ("vmovntdq %%zmm1,%0" : "=m" (q[i]));
asm volatile ("vmovntdq %%zmm2,%0" : "=m" (r[i]));
@ -1391,12 +1387,12 @@ void raid_gen4_avx512bw(int nd, size_t size, void **vv)
asm volatile ("vpbroadcastb %0,%%zmm31" : : "m" (gfconst16.low4[0]));
for (i = 0; i < size; i += 64) {
asm volatile ("vpxorq %zmm0,%zmm0,%zmm0");
asm volatile ("vpxorq %zmm1,%zmm1,%zmm1");
asm volatile ("vpxorq %zmm2,%zmm2,%zmm2");
asm volatile ("vpxorq %zmm3,%zmm3,%zmm3");
asm volatile ("vmovdqa64 %0,%%zmm0" : : "m" (v[0][i]));
asm volatile ("vmovdqa64 %zmm0,%zmm1");
asm volatile ("vmovdqa64 %zmm0,%zmm2");
asm volatile ("vmovdqa64 %zmm0,%zmm3");
for (d = l; d > 0; --d) {
for (d = 1; d <= l; ++d) {
asm volatile ("vmovdqa64 %0,%%zmm10" : : "m" (v[d][i]));
asm volatile ("vpxorq %zmm10,%zmm0,%zmm0");
@ -1425,14 +1421,6 @@ void raid_gen4_avx512bw(int nd, size_t size, void **vv)
asm volatile ("vpxorq %zmm17,%zmm3,%zmm3");
}
/* first disk with all coefficients at 1 */
asm volatile ("vmovdqa64 %0,%%zmm10" : : "m" (v[0][i]));
asm volatile ("vpxorq %zmm10,%zmm0,%zmm0");
asm volatile ("vpxorq %zmm10,%zmm1,%zmm1");
asm volatile ("vpxorq %zmm10,%zmm2,%zmm2");
asm volatile ("vpxorq %zmm10,%zmm3,%zmm3");
asm volatile ("vmovntdq %%zmm0,%0" : "=m" (p[i]));
asm volatile ("vmovntdq %%zmm1,%0" : "=m" (q[i]));
asm volatile ("vmovntdq %%zmm2,%0" : "=m" (r[i]));
@ -1868,13 +1856,13 @@ void raid_gen5_avx512bw(int nd, size_t size, void **vv)
asm volatile ("vpbroadcastb %0,%%zmm31" : : "m" (gfconst16.low4[0]));
for (i = 0; i < size; i += 64) {
asm volatile ("vpxorq %zmm0,%zmm0,%zmm0");
asm volatile ("vpxorq %zmm1,%zmm1,%zmm1");
asm volatile ("vpxorq %zmm2,%zmm2,%zmm2");
asm volatile ("vpxorq %zmm3,%zmm3,%zmm3");
asm volatile ("vpxorq %zmm4,%zmm4,%zmm4");
asm volatile ("vmovdqa64 %0,%%zmm0" : : "m" (v[0][i]));
asm volatile ("vmovdqa64 %zmm0,%zmm1");
asm volatile ("vmovdqa64 %zmm0,%zmm2");
asm volatile ("vmovdqa64 %zmm0,%zmm3");
asm volatile ("vmovdqa64 %zmm0,%zmm4");
for (d = l; d > 0; --d) {
for (d = 1; d <= l; ++d) {
asm volatile ("vmovdqa64 %0,%%zmm10" : : "m" (v[d][i]));
asm volatile ("vpxorq %zmm10,%zmm0,%zmm0");
@ -1909,15 +1897,6 @@ void raid_gen5_avx512bw(int nd, size_t size, void **vv)
asm volatile ("vpxorq %zmm19,%zmm4,%zmm4");
}
/* first disk with all coefficients at 1 */
asm volatile ("vmovdqa64 %0,%%zmm10" : : "m" (v[0][i]));
asm volatile ("vpxorq %zmm10,%zmm0,%zmm0");
asm volatile ("vpxorq %zmm10,%zmm1,%zmm1");
asm volatile ("vpxorq %zmm10,%zmm2,%zmm2");
asm volatile ("vpxorq %zmm10,%zmm3,%zmm3");
asm volatile ("vpxorq %zmm10,%zmm4,%zmm4");
asm volatile ("vmovntdq %%zmm0,%0" : "=m" (p[i]));
asm volatile ("vmovntdq %%zmm1,%0" : "=m" (q[i]));
asm volatile ("vmovntdq %%zmm2,%0" : "=m" (r[i]));
@ -2422,15 +2401,14 @@ void raid_gen6_avx512bw(int nd, size_t size, void **vv)
asm volatile ("vpbroadcastb %0,%%zmm31" : : "m" (gfconst16.low4[0]));
for (i = 0; i < size; i += 64) {
asm volatile ("vpxorq %zmm0,%zmm0,%zmm0");
asm volatile ("vpxorq %zmm1,%zmm1,%zmm1");
asm volatile ("vpxorq %zmm2,%zmm2,%zmm2");
asm volatile ("vpxorq %zmm3,%zmm3,%zmm3");
asm volatile ("vpxorq %zmm4,%zmm4,%zmm4");
asm volatile ("vpxorq %zmm5,%zmm5,%zmm5");
asm volatile ("vmovdqa64 %0,%%zmm0" : : "m" (v[0][i]));
asm volatile ("vmovdqa64 %zmm0,%zmm1");
asm volatile ("vmovdqa64 %zmm0,%zmm2");
asm volatile ("vmovdqa64 %zmm0,%zmm3");
asm volatile ("vmovdqa64 %zmm0,%zmm4");
asm volatile ("vmovdqa64 %zmm0,%zmm5");
/* intermediate disks */
for (d = l; d > 0; --d) {
for (d = 1; d <= l; ++d) {
asm volatile ("vmovdqa64 %0,%%zmm10" : : "m" (v[d][i]));
asm volatile ("vpxorq %zmm10,%zmm0,%zmm0");
@ -2471,16 +2449,6 @@ void raid_gen6_avx512bw(int nd, size_t size, void **vv)
asm volatile ("vpxorq %zmm21,%zmm5,%zmm5");
}
/* first disk with all coefficients at 1 */
asm volatile ("vmovdqa64 %0,%%zmm10" : : "m" (v[0][i]));
asm volatile ("vpxorq %zmm10,%zmm0,%zmm0");
asm volatile ("vpxorq %zmm10,%zmm1,%zmm1");
asm volatile ("vpxorq %zmm10,%zmm2,%zmm2");
asm volatile ("vpxorq %zmm10,%zmm3,%zmm3");
asm volatile ("vpxorq %zmm10,%zmm4,%zmm4");
asm volatile ("vpxorq %zmm10,%zmm5,%zmm5");
asm volatile ("vmovntdq %%zmm0,%0" : "=m" (p[i]));
asm volatile ("vmovntdq %%zmm1,%0" : "=m" (q[i]));
asm volatile ("vmovntdq %%zmm2,%0" : "=m" (r[i]));