381 lines
10 KiB
C
381 lines
10 KiB
C
/* x86 */
|
|
#if defined(X86ASM_AVX) && (!defined(SCRYPT_CHOOSE_COMPILETIME) || !defined(SCRYPT_SALSA_INCLUDED)) && !defined(CPU_X86_FORCE_INTRINSICS)
|
|
|
|
#define SCRYPT_SALSA_AVX
|
|
|
|
asm_naked_fn_proto(void, scrypt_ChunkMix_avx)(uint32_t *Bout/*[chunkBytes]*/, uint32_t *Bin/*[chunkBytes]*/, uint32_t *Bxor/*[chunkBytes]*/, uint32_t r)
|
|
asm_naked_fn(scrypt_ChunkMix_avx)
|
|
a1(push ebx)
|
|
a1(push edi)
|
|
a1(push esi)
|
|
a1(push ebp)
|
|
a2(mov ebp,esp)
|
|
a2(mov edi,[ebp+20])
|
|
a2(mov esi,[ebp+24])
|
|
a2(mov eax,[ebp+28])
|
|
a2(mov ebx,[ebp+32])
|
|
a2(sub esp,32)
|
|
a2(and esp,~63)
|
|
a2(lea edx,[ebx*2])
|
|
a2(shl edx,6)
|
|
a2(lea ecx,[edx-64])
|
|
a2(and eax, eax)
|
|
a2(movdqa xmm0,[ecx+esi+0])
|
|
a2(movdqa xmm1,[ecx+esi+16])
|
|
a2(movdqa xmm2,[ecx+esi+32])
|
|
a2(movdqa xmm3,[ecx+esi+48])
|
|
aj(jz scrypt_ChunkMix_avx_no_xor1)
|
|
a3(vpxor xmm0,xmm0,[ecx+eax+0])
|
|
a3(vpxor xmm1,xmm1,[ecx+eax+16])
|
|
a3(vpxor xmm2,xmm2,[ecx+eax+32])
|
|
a3(vpxor xmm3,xmm3,[ecx+eax+48])
|
|
a1(scrypt_ChunkMix_avx_no_xor1:)
|
|
a2(xor ecx,ecx)
|
|
a2(xor ebx,ebx)
|
|
a1(scrypt_ChunkMix_avx_loop:)
|
|
a2(and eax, eax)
|
|
a3(vpxor xmm0,xmm0,[esi+ecx+0])
|
|
a3(vpxor xmm1,xmm1,[esi+ecx+16])
|
|
a3(vpxor xmm2,xmm2,[esi+ecx+32])
|
|
a3(vpxor xmm3,xmm3,[esi+ecx+48])
|
|
aj(jz scrypt_ChunkMix_avx_no_xor2)
|
|
a3(vpxor xmm0,xmm0,[eax+ecx+0])
|
|
a3(vpxor xmm1,xmm1,[eax+ecx+16])
|
|
a3(vpxor xmm2,xmm2,[eax+ecx+32])
|
|
a3(vpxor xmm3,xmm3,[eax+ecx+48])
|
|
a1(scrypt_ChunkMix_avx_no_xor2:)
|
|
a2(vmovdqa [esp+0],xmm0)
|
|
a2(vmovdqa [esp+16],xmm1)
|
|
a2(vmovdqa xmm6,xmm2)
|
|
a2(vmovdqa xmm7,xmm3)
|
|
a2(mov eax,8)
|
|
a1(scrypt_salsa_avx_loop: )
|
|
a3(vpaddd xmm4, xmm1, xmm0)
|
|
a3(vpsrld xmm5, xmm4, 25)
|
|
a3(vpslld xmm4, xmm4, 7)
|
|
a3(vpxor xmm3, xmm3, xmm5)
|
|
a3(vpxor xmm3, xmm3, xmm4)
|
|
a3(vpaddd xmm4, xmm0, xmm3)
|
|
a3(vpsrld xmm5, xmm4, 23)
|
|
a3(vpslld xmm4, xmm4, 9)
|
|
a3(vpxor xmm2, xmm2, xmm5)
|
|
a3(vpxor xmm2, xmm2, xmm4)
|
|
a3(vpaddd xmm4, xmm3, xmm2)
|
|
a3(vpsrld xmm5, xmm4, 19)
|
|
a3(vpslld xmm4, xmm4, 13)
|
|
a3(vpxor xmm1, xmm1, xmm5)
|
|
a3(vpshufd xmm3, xmm3, 0x93)
|
|
a3(vpxor xmm1, xmm1, xmm4)
|
|
a3(vpaddd xmm4, xmm2, xmm1)
|
|
a3(vpsrld xmm5, xmm4, 14)
|
|
a3(vpslld xmm4, xmm4, 18)
|
|
a3(vpxor xmm0, xmm0, xmm5)
|
|
a3(vpshufd xmm2, xmm2, 0x4e)
|
|
a3(vpxor xmm0, xmm0, xmm4)
|
|
a3(vpaddd xmm4, xmm3, xmm0)
|
|
a3(vpshufd xmm1, xmm1, 0x39)
|
|
a3(vpsrld xmm5, xmm4, 25)
|
|
a3(vpslld xmm4, xmm4, 7)
|
|
a3(vpxor xmm1, xmm1, xmm5)
|
|
a3(vpxor xmm1, xmm1, xmm4)
|
|
a3(vpaddd xmm4, xmm0, xmm1)
|
|
a3(vpsrld xmm5, xmm4, 23)
|
|
a3(vpslld xmm4, xmm4, 9)
|
|
a3(vpxor xmm2, xmm2, xmm5)
|
|
a3(vpxor xmm2, xmm2, xmm4)
|
|
a3(vpaddd xmm4, xmm1, xmm2)
|
|
a3(vpsrld xmm5, xmm4, 19)
|
|
a3(vpslld xmm4, xmm4, 13)
|
|
a3(vpxor xmm3, xmm3, xmm5)
|
|
a3(vpshufd xmm1, xmm1, 0x93)
|
|
a3(vpxor xmm3, xmm3, xmm4)
|
|
a3(vpaddd xmm4, xmm2, xmm3)
|
|
a3(vpsrld xmm5, xmm4, 14)
|
|
a3(vpslld xmm4, xmm4, 18)
|
|
a3(vpxor xmm0, xmm0, xmm5)
|
|
a3(vpshufd xmm2, xmm2, 0x4e)
|
|
a3(vpxor xmm0, xmm0, xmm4)
|
|
a3(vpshufd xmm3, xmm3, 0x39)
|
|
a2(sub eax, 2)
|
|
aj(ja scrypt_salsa_avx_loop)
|
|
a3(vpaddd xmm0,xmm0,[esp+0])
|
|
a3(vpaddd xmm1,xmm1,[esp+16])
|
|
a3(vpaddd xmm2,xmm2,xmm6)
|
|
a3(vpaddd xmm3,xmm3,xmm7)
|
|
a2(lea eax,[ebx+ecx])
|
|
a2(xor ebx,edx)
|
|
a2(and eax,~0x7f)
|
|
a2(add ecx,64)
|
|
a2(shr eax,1)
|
|
a2(add eax, edi)
|
|
a2(cmp ecx,edx)
|
|
a2(vmovdqa [eax+0],xmm0)
|
|
a2(vmovdqa [eax+16],xmm1)
|
|
a2(vmovdqa [eax+32],xmm2)
|
|
a2(vmovdqa [eax+48],xmm3)
|
|
a2(mov eax,[ebp+28])
|
|
aj(jne scrypt_ChunkMix_avx_loop)
|
|
a2(mov esp,ebp)
|
|
a1(pop ebp)
|
|
a1(pop esi)
|
|
a1(pop edi)
|
|
a1(pop ebx)
|
|
aret(16)
|
|
asm_naked_fn_end(scrypt_ChunkMix_avx)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
/* x64 */
|
|
#if defined(X86_64ASM_AVX) && (!defined(SCRYPT_CHOOSE_COMPILETIME) || !defined(SCRYPT_SALSA_INCLUDED)) && !defined(CPU_X86_FORCE_INTRINSICS)
|
|
|
|
#define SCRYPT_SALSA_AVX
|
|
|
|
asm_naked_fn_proto(void, scrypt_ChunkMix_avx)(uint32_t *Bout/*[chunkBytes]*/, uint32_t *Bin/*[chunkBytes]*/, uint32_t *Bxor/*[chunkBytes]*/, uint32_t r)
|
|
asm_naked_fn(scrypt_ChunkMix_avx)
|
|
a2(lea rcx,[ecx*2]) /* zero extend uint32_t by using ecx, win64 can leave garbage in the top half */
|
|
a2(shl rcx,6)
|
|
a2(lea r9,[rcx-64])
|
|
a2(lea rax,[rsi+r9])
|
|
a2(lea r9,[rdx+r9])
|
|
a2(and rdx, rdx)
|
|
a2(vmovdqa xmm0,[rax+0])
|
|
a2(vmovdqa xmm1,[rax+16])
|
|
a2(vmovdqa xmm2,[rax+32])
|
|
a2(vmovdqa xmm3,[rax+48])
|
|
aj(jz scrypt_ChunkMix_avx_no_xor1)
|
|
a3(vpxor xmm0,xmm0,[r9+0])
|
|
a3(vpxor xmm1,xmm1,[r9+16])
|
|
a3(vpxor xmm2,xmm2,[r9+32])
|
|
a3(vpxor xmm3,xmm3,[r9+48])
|
|
a1(scrypt_ChunkMix_avx_no_xor1:)
|
|
a2(xor r9,r9)
|
|
a2(xor r8,r8)
|
|
a1(scrypt_ChunkMix_avx_loop:)
|
|
a2(and rdx, rdx)
|
|
a3(vpxor xmm0,xmm0,[rsi+r9+0])
|
|
a3(vpxor xmm1,xmm1,[rsi+r9+16])
|
|
a3(vpxor xmm2,xmm2,[rsi+r9+32])
|
|
a3(vpxor xmm3,xmm3,[rsi+r9+48])
|
|
aj(jz scrypt_ChunkMix_avx_no_xor2)
|
|
a3(vpxor xmm0,xmm0,[rdx+r9+0])
|
|
a3(vpxor xmm1,xmm1,[rdx+r9+16])
|
|
a3(vpxor xmm2,xmm2,[rdx+r9+32])
|
|
a3(vpxor xmm3,xmm3,[rdx+r9+48])
|
|
a1(scrypt_ChunkMix_avx_no_xor2:)
|
|
a2(vmovdqa xmm8,xmm0)
|
|
a2(vmovdqa xmm9,xmm1)
|
|
a2(vmovdqa xmm10,xmm2)
|
|
a2(vmovdqa xmm11,xmm3)
|
|
a2(mov rax,8)
|
|
a1(scrypt_salsa_avx_loop: )
|
|
a3(vpaddd xmm4, xmm1, xmm0)
|
|
a3(vpsrld xmm5, xmm4, 25)
|
|
a3(vpslld xmm4, xmm4, 7)
|
|
a3(vpxor xmm3, xmm3, xmm5)
|
|
a3(vpxor xmm3, xmm3, xmm4)
|
|
a3(vpaddd xmm4, xmm0, xmm3)
|
|
a3(vpsrld xmm5, xmm4, 23)
|
|
a3(vpslld xmm4, xmm4, 9)
|
|
a3(vpxor xmm2, xmm2, xmm5)
|
|
a3(vpxor xmm2, xmm2, xmm4)
|
|
a3(vpaddd xmm4, xmm3, xmm2)
|
|
a3(vpsrld xmm5, xmm4, 19)
|
|
a3(vpslld xmm4, xmm4, 13)
|
|
a3(vpxor xmm1, xmm1, xmm5)
|
|
a3(vpshufd xmm3, xmm3, 0x93)
|
|
a3(vpxor xmm1, xmm1, xmm4)
|
|
a3(vpaddd xmm4, xmm2, xmm1)
|
|
a3(vpsrld xmm5, xmm4, 14)
|
|
a3(vpslld xmm4, xmm4, 18)
|
|
a3(vpxor xmm0, xmm0, xmm5)
|
|
a3(vpshufd xmm2, xmm2, 0x4e)
|
|
a3(vpxor xmm0, xmm0, xmm4)
|
|
a3(vpaddd xmm4, xmm3, xmm0)
|
|
a3(vpshufd xmm1, xmm1, 0x39)
|
|
a3(vpsrld xmm5, xmm4, 25)
|
|
a3(vpslld xmm4, xmm4, 7)
|
|
a3(vpxor xmm1, xmm1, xmm5)
|
|
a3(vpxor xmm1, xmm1, xmm4)
|
|
a3(vpaddd xmm4, xmm0, xmm1)
|
|
a3(vpsrld xmm5, xmm4, 23)
|
|
a3(vpslld xmm4, xmm4, 9)
|
|
a3(vpxor xmm2, xmm2, xmm5)
|
|
a3(vpxor xmm2, xmm2, xmm4)
|
|
a3(vpaddd xmm4, xmm1, xmm2)
|
|
a3(vpsrld xmm5, xmm4, 19)
|
|
a3(vpslld xmm4, xmm4, 13)
|
|
a3(vpxor xmm3, xmm3, xmm5)
|
|
a3(vpshufd xmm1, xmm1, 0x93)
|
|
a3(vpxor xmm3, xmm3, xmm4)
|
|
a3(vpaddd xmm4, xmm2, xmm3)
|
|
a3(vpsrld xmm5, xmm4, 14)
|
|
a3(vpslld xmm4, xmm4, 18)
|
|
a3(vpxor xmm0, xmm0, xmm5)
|
|
a3(vpshufd xmm2, xmm2, 0x4e)
|
|
a3(vpxor xmm0, xmm0, xmm4)
|
|
a3(vpshufd xmm3, xmm3, 0x39)
|
|
a2(sub rax, 2)
|
|
aj(ja scrypt_salsa_avx_loop)
|
|
a3(vpaddd xmm0,xmm0,xmm8)
|
|
a3(vpaddd xmm1,xmm1,xmm9)
|
|
a3(vpaddd xmm2,xmm2,xmm10)
|
|
a3(vpaddd xmm3,xmm3,xmm11)
|
|
a2(lea rax,[r8+r9])
|
|
a2(xor r8,rcx)
|
|
a2(and rax,~0x7f)
|
|
a2(add r9,64)
|
|
a2(shr rax,1)
|
|
a2(add rax, rdi)
|
|
a2(cmp r9,rcx)
|
|
a2(vmovdqa [rax+0],xmm0)
|
|
a2(vmovdqa [rax+16],xmm1)
|
|
a2(vmovdqa [rax+32],xmm2)
|
|
a2(vmovdqa [rax+48],xmm3)
|
|
aj(jne scrypt_ChunkMix_avx_loop)
|
|
a1(ret)
|
|
asm_naked_fn_end(scrypt_ChunkMix_avx)
|
|
|
|
#endif
|
|
|
|
|
|
/* intrinsic */
|
|
#if defined(X86_INTRINSIC_AVX) && (!defined(SCRYPT_CHOOSE_COMPILETIME) || !defined(SCRYPT_SALSA_INCLUDED))
|
|
|
|
#define SCRYPT_SALSA_AVX
|
|
|
|
static void asm_calling_convention NOINLINE
|
|
scrypt_ChunkMix_avx(uint32_t *Bout/*[chunkBytes]*/, uint32_t *Bin/*[chunkBytes]*/, uint32_t *Bxor/*[chunkBytes]*/, uint32_t r) {
|
|
uint32_t i, blocksPerChunk = r * 2, half = 0;
|
|
xmmi *xmmp,x0,x1,x2,x3,x4,x5,t0,t1,t2,t3;
|
|
size_t rounds;
|
|
|
|
/* 1: X = B_{2r - 1} */
|
|
xmmp = (xmmi *)scrypt_block(Bin, blocksPerChunk - 1);
|
|
x0 = xmmp[0];
|
|
x1 = xmmp[1];
|
|
x2 = xmmp[2];
|
|
x3 = xmmp[3];
|
|
|
|
if (Bxor) {
|
|
xmmp = (xmmi *)scrypt_block(Bxor, blocksPerChunk - 1);
|
|
x0 = _mm_xor_si128(x0, xmmp[0]);
|
|
x1 = _mm_xor_si128(x1, xmmp[1]);
|
|
x2 = _mm_xor_si128(x2, xmmp[2]);
|
|
x3 = _mm_xor_si128(x3, xmmp[3]);
|
|
}
|
|
|
|
/* 2: for i = 0 to 2r - 1 do */
|
|
for (i = 0; i < blocksPerChunk; i++, half ^= r) {
|
|
/* 3: X = H(X ^ B_i) */
|
|
xmmp = (xmmi *)scrypt_block(Bin, i);
|
|
x0 = _mm_xor_si128(x0, xmmp[0]);
|
|
x1 = _mm_xor_si128(x1, xmmp[1]);
|
|
x2 = _mm_xor_si128(x2, xmmp[2]);
|
|
x3 = _mm_xor_si128(x3, xmmp[3]);
|
|
|
|
if (Bxor) {
|
|
xmmp = (xmmi *)scrypt_block(Bxor, i);
|
|
x0 = _mm_xor_si128(x0, xmmp[0]);
|
|
x1 = _mm_xor_si128(x1, xmmp[1]);
|
|
x2 = _mm_xor_si128(x2, xmmp[2]);
|
|
x3 = _mm_xor_si128(x3, xmmp[3]);
|
|
}
|
|
|
|
t0 = x0;
|
|
t1 = x1;
|
|
t2 = x2;
|
|
t3 = x3;
|
|
|
|
for (rounds = 8; rounds; rounds -= 2) {
|
|
x4 = x1;
|
|
x4 = _mm_add_epi32(x4, x0);
|
|
x5 = x4;
|
|
x4 = _mm_slli_epi32(x4, 7);
|
|
x5 = _mm_srli_epi32(x5, 25);
|
|
x3 = _mm_xor_si128(x3, x4);
|
|
x4 = x0;
|
|
x3 = _mm_xor_si128(x3, x5);
|
|
x4 = _mm_add_epi32(x4, x3);
|
|
x5 = x4;
|
|
x4 = _mm_slli_epi32(x4, 9);
|
|
x5 = _mm_srli_epi32(x5, 23);
|
|
x2 = _mm_xor_si128(x2, x4);
|
|
x4 = x3;
|
|
x2 = _mm_xor_si128(x2, x5);
|
|
x3 = _mm_shuffle_epi32(x3, 0x93);
|
|
x4 = _mm_add_epi32(x4, x2);
|
|
x5 = x4;
|
|
x4 = _mm_slli_epi32(x4, 13);
|
|
x5 = _mm_srli_epi32(x5, 19);
|
|
x1 = _mm_xor_si128(x1, x4);
|
|
x4 = x2;
|
|
x1 = _mm_xor_si128(x1, x5);
|
|
x2 = _mm_shuffle_epi32(x2, 0x4e);
|
|
x4 = _mm_add_epi32(x4, x1);
|
|
x5 = x4;
|
|
x4 = _mm_slli_epi32(x4, 18);
|
|
x5 = _mm_srli_epi32(x5, 14);
|
|
x0 = _mm_xor_si128(x0, x4);
|
|
x4 = x3;
|
|
x0 = _mm_xor_si128(x0, x5);
|
|
x1 = _mm_shuffle_epi32(x1, 0x39);
|
|
x4 = _mm_add_epi32(x4, x0);
|
|
x5 = x4;
|
|
x4 = _mm_slli_epi32(x4, 7);
|
|
x5 = _mm_srli_epi32(x5, 25);
|
|
x1 = _mm_xor_si128(x1, x4);
|
|
x4 = x0;
|
|
x1 = _mm_xor_si128(x1, x5);
|
|
x4 = _mm_add_epi32(x4, x1);
|
|
x5 = x4;
|
|
x4 = _mm_slli_epi32(x4, 9);
|
|
x5 = _mm_srli_epi32(x5, 23);
|
|
x2 = _mm_xor_si128(x2, x4);
|
|
x4 = x1;
|
|
x2 = _mm_xor_si128(x2, x5);
|
|
x1 = _mm_shuffle_epi32(x1, 0x93);
|
|
x4 = _mm_add_epi32(x4, x2);
|
|
x5 = x4;
|
|
x4 = _mm_slli_epi32(x4, 13);
|
|
x5 = _mm_srli_epi32(x5, 19);
|
|
x3 = _mm_xor_si128(x3, x4);
|
|
x4 = x2;
|
|
x3 = _mm_xor_si128(x3, x5);
|
|
x2 = _mm_shuffle_epi32(x2, 0x4e);
|
|
x4 = _mm_add_epi32(x4, x3);
|
|
x5 = x4;
|
|
x4 = _mm_slli_epi32(x4, 18);
|
|
x5 = _mm_srli_epi32(x5, 14);
|
|
x0 = _mm_xor_si128(x0, x4);
|
|
x3 = _mm_shuffle_epi32(x3, 0x39);
|
|
x0 = _mm_xor_si128(x0, x5);
|
|
}
|
|
|
|
x0 = _mm_add_epi32(x0, t0);
|
|
x1 = _mm_add_epi32(x1, t1);
|
|
x2 = _mm_add_epi32(x2, t2);
|
|
x3 = _mm_add_epi32(x3, t3);
|
|
|
|
/* 4: Y_i = X */
|
|
/* 6: B'[0..r-1] = Y_even */
|
|
/* 6: B'[r..2r-1] = Y_odd */
|
|
xmmp = (xmmi *)scrypt_block(Bout, (i / 2) + half);
|
|
xmmp[0] = x0;
|
|
xmmp[1] = x1;
|
|
xmmp[2] = x2;
|
|
xmmp[3] = x3;
|
|
}
|
|
}
|
|
|
|
#endif
|
|
|
|
#if defined(SCRYPT_SALSA_AVX)
|
|
/* uses salsa_core_tangle_sse2 */
|
|
|
|
#undef SCRYPT_MIX
|
|
#define SCRYPT_MIX "Salsa/8-AVX"
|
|
#undef SCRYPT_SALSA_INCLUDED
|
|
#define SCRYPT_SALSA_INCLUDED
|
|
#endif
|