diff options
Diffstat (limited to 'libbb/hash_md5_sha_x86-64.S')
| -rw-r--r-- | libbb/hash_md5_sha_x86-64.S | 992 |
1 files changed, 558 insertions, 434 deletions
diff --git a/libbb/hash_md5_sha_x86-64.S b/libbb/hash_md5_sha_x86-64.S index 87fb616..069a187 100644 --- a/libbb/hash_md5_sha_x86-64.S +++ b/libbb/hash_md5_sha_x86-64.S @@ -20,16 +20,10 @@ sha1_process_block64: # eax..edx: a..d # ebp: e # esi,edi: temps -# -32+4*n(%rsp),r8...r15: W[0..7,8..15] -# (TODO: actually W[0..7] are used a bit more often, put _them_ into r8..r15?) - movl $3, %eax -1: - movq (%rdi,%rax,8), %rsi - bswapq %rsi - rolq $32, %rsi - movq %rsi, -32(%rsp,%rax,8) - decl %eax - jns 1b +# xmm0..xmm3: W[] +# xmm4,xmm5: temps +# xmm6: current round constant +# -64(%rsp): area for passing RCONST + W[] from vector to integer units movl 80(%rdi), %eax # a = ctx->hash[0] movl 84(%rdi), %ebx # b = ctx->hash[1] @@ -37,587 +31,709 @@ sha1_process_block64: movl 92(%rdi), %edx # d = ctx->hash[3] movl 96(%rdi), %ebp # e = ctx->hash[4] + movaps rconst0x5A827999(%rip), %xmm6 + + # For round 1, steps 0 and 8..15, we pass W[0,8..15] in esi,r8..r15 + # instead of spilling them to stack. + # (We lose parallelized addition of RCONST, but LEA + # can do two additions at once, so...) + movq 4*0(%rdi), %rsi + movq 4*2(%rdi), %r10 + bswapq %rsi + bswapq %r10 + rolq $32, %rsi # rsi = W[1]:W[0] + rolq $32, %r10 + movq %rsi, %xmm0 + movq %r10, %xmm4 + punpcklqdq %xmm4, %xmm0 # xmm0 = r10:rsi = (W[0],W[1],W[2],W[3]) + movaps %xmm0, %xmm4 + paddd %xmm6, %xmm4 + movups %xmm4, -64+4*0(%rsp) + + movq 4*4(%rdi), %r8 + movq 4*6(%rdi), %r10 + bswapq %r8 + bswapq %r10 + rolq $32, %r8 + rolq $32, %r10 + movq %r8, %xmm1 + movq %r10, %xmm4 + punpcklqdq %xmm4, %xmm1 # xmm1 = r10:r8 = (W[4],W[5],W[6],W[7]) + movaps %xmm1, %xmm4 + paddd %xmm6, %xmm4 + movups %xmm4, -64+4*4(%rsp) + movq 4*8(%rdi), %r8 movq 4*10(%rdi), %r10 bswapq %r8 bswapq %r10 + movl %r8d, %r9d # r9d = W[9] + rolq $32, %r8 # r8 = W[9]:W[8] + movl %r10d, %r11d # r11d = W[11] + rolq $32, %r10 # r10 = W[11]:W[10] + movq %r8, %xmm2 + movq %r10, %xmm4 + punpcklqdq %xmm4, %xmm2 # xmm2 = r10:r8 = (W[8],W[9],W[10],W[11]) + movq 4*12(%rdi), %r12 movq 4*14(%rdi), %r14 bswapq %r12 bswapq %r14 - movl %r8d, %r9d - shrq $32, %r8 - movl %r10d, %r11d - shrq $32, %r10 - movl %r12d, %r13d - shrq $32, %r12 - movl %r14d, %r15d - shrq $32, %r14 + movl %r12d, %r13d # r13d = W[13] + rolq $32, %r12 # r12 = W[13]:W[12] + movl %r14d, %r15d # r15d = W[15] + rolq $32, %r14 # r14 = W[15]:W[14] + movq %r12, %xmm3 + movq %r14, %xmm4 + punpcklqdq %xmm4, %xmm3 # xmm3 = r14:r12 = (W[12],W[13],W[14],W[15]) # 0 - # W[0], already in %esi + leal 0x5A827999(%rbp,%rsi), %ebp # e += RCONST + W[n] movl %ecx, %edi # c xorl %edx, %edi # ^d andl %ebx, %edi # &b xorl %edx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rbp,%rsi), %ebp # e += RCONST + W[n] addl %edi, %ebp # e += (((c ^ d) & b) ^ d) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) # 1 - movl -32+4*1(%rsp), %esi # W[n] + addl -64+4*1(%rsp), %edx # e += RCONST + W[n] movl %ebx, %edi # c xorl %ecx, %edi # ^d andl %eax, %edi # &b xorl %ecx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rdx,%rsi), %edx # e += RCONST + W[n] addl %edi, %edx # e += (((c ^ d) & b) ^ d) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) # 2 - movl -32+4*2(%rsp), %esi # W[n] + addl -64+4*2(%rsp), %ecx # e += RCONST + W[n] movl %eax, %edi # c xorl %ebx, %edi # ^d andl %ebp, %edi # &b xorl %ebx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rcx,%rsi), %ecx # e += RCONST + W[n] addl %edi, %ecx # e += (((c ^ d) & b) ^ d) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) # 3 - movl -32+4*3(%rsp), %esi # W[n] + addl -64+4*3(%rsp), %ebx # e += RCONST + W[n] movl %ebp, %edi # c xorl %eax, %edi # ^d andl %edx, %edi # &b xorl %eax, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rbx,%rsi), %ebx # e += RCONST + W[n] addl %edi, %ebx # e += (((c ^ d) & b) ^ d) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) # 4 - movl -32+4*4(%rsp), %esi # W[n] + addl -64+4*4(%rsp), %eax # e += RCONST + W[n] movl %edx, %edi # c xorl %ebp, %edi # ^d andl %ecx, %edi # &b xorl %ebp, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rax,%rsi), %eax # e += RCONST + W[n] addl %edi, %eax # e += (((c ^ d) & b) ^ d) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) # 5 - movl -32+4*5(%rsp), %esi # W[n] + addl -64+4*5(%rsp), %ebp # e += RCONST + W[n] movl %ecx, %edi # c xorl %edx, %edi # ^d andl %ebx, %edi # &b xorl %edx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rbp,%rsi), %ebp # e += RCONST + W[n] addl %edi, %ebp # e += (((c ^ d) & b) ^ d) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) # 6 - movl -32+4*6(%rsp), %esi # W[n] + addl -64+4*6(%rsp), %edx # e += RCONST + W[n] movl %ebx, %edi # c xorl %ecx, %edi # ^d andl %eax, %edi # &b xorl %ecx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rdx,%rsi), %edx # e += RCONST + W[n] addl %edi, %edx # e += (((c ^ d) & b) ^ d) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) # 7 - movl -32+4*7(%rsp), %esi # W[n] + addl -64+4*7(%rsp), %ecx # e += RCONST + W[n] movl %eax, %edi # c xorl %ebx, %edi # ^d andl %ebp, %edi # &b xorl %ebx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rcx,%rsi), %ecx # e += RCONST + W[n] addl %edi, %ecx # e += (((c ^ d) & b) ^ d) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) +# PREP %xmm0 %xmm1 %xmm2 %xmm3 -64+16*0(%rsp) + movaps %xmm3, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm0, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm1, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm2, %xmm0 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm0 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm0, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm0, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm0, %xmm0 # shift left by 1 + psubd %xmm4, %xmm0 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm0 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm0 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm0, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*0(%rsp) # 8 - # W[n], in %r8 + leal 0x5A827999(%rbx,%r8), %ebx # e += RCONST + W[n] movl %ebp, %edi # c xorl %eax, %edi # ^d andl %edx, %edi # &b xorl %eax, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rbx,%r8), %ebx # e += RCONST + W[n] addl %edi, %ebx # e += (((c ^ d) & b) ^ d) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) # 9 - # W[n], in %r9 + leal 0x5A827999(%rax,%r9), %eax # e += RCONST + W[n] movl %edx, %edi # c xorl %ebp, %edi # ^d andl %ecx, %edi # &b xorl %ebp, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rax,%r9), %eax # e += RCONST + W[n] addl %edi, %eax # e += (((c ^ d) & b) ^ d) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) # 10 - # W[n], in %r10 + leal 0x5A827999(%rbp,%r10), %ebp # e += RCONST + W[n] movl %ecx, %edi # c xorl %edx, %edi # ^d andl %ebx, %edi # &b xorl %edx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rbp,%r10), %ebp # e += RCONST + W[n] addl %edi, %ebp # e += (((c ^ d) & b) ^ d) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) # 11 - # W[n], in %r11 + leal 0x5A827999(%rdx,%r11), %edx # e += RCONST + W[n] movl %ebx, %edi # c xorl %ecx, %edi # ^d andl %eax, %edi # &b xorl %ecx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rdx,%r11), %edx # e += RCONST + W[n] addl %edi, %edx # e += (((c ^ d) & b) ^ d) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) + movaps rconst0x6ED9EBA1(%rip), %xmm6 +# PREP %xmm1 %xmm2 %xmm3 %xmm0 -64+16*1(%rsp) + movaps %xmm0, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm1, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm2, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm3, %xmm1 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm1 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm1, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm1, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm1, %xmm1 # shift left by 1 + psubd %xmm4, %xmm1 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm1 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm1 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm1, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*1(%rsp) # 12 - # W[n], in %r12 + leal 0x5A827999(%rcx,%r12), %ecx # e += RCONST + W[n] movl %eax, %edi # c xorl %ebx, %edi # ^d andl %ebp, %edi # &b xorl %ebx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rcx,%r12), %ecx # e += RCONST + W[n] addl %edi, %ecx # e += (((c ^ d) & b) ^ d) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) # 13 - # W[n], in %r13 + leal 0x5A827999(%rbx,%r13), %ebx # e += RCONST + W[n] movl %ebp, %edi # c xorl %eax, %edi # ^d andl %edx, %edi # &b xorl %eax, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rbx,%r13), %ebx # e += RCONST + W[n] addl %edi, %ebx # e += (((c ^ d) & b) ^ d) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) # 14 - # W[n], in %r14 + leal 0x5A827999(%rax,%r14), %eax # e += RCONST + W[n] movl %edx, %edi # c xorl %ebp, %edi # ^d andl %ecx, %edi # &b xorl %ebp, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rax,%r14), %eax # e += RCONST + W[n] addl %edi, %eax # e += (((c ^ d) & b) ^ d) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) # 15 - # W[n], in %r15 + leal 0x5A827999(%rbp,%r15), %ebp # e += RCONST + W[n] movl %ecx, %edi # c xorl %edx, %edi # ^d andl %ebx, %edi # &b xorl %edx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rbp,%r15), %ebp # e += RCONST + W[n] addl %edi, %ebp # e += (((c ^ d) & b) ^ d) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) +# PREP %xmm2 %xmm3 %xmm0 %xmm1 -64+16*2(%rsp) + movaps %xmm1, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm2, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm3, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm0, %xmm2 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm2 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm2, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm2, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm2, %xmm2 # shift left by 1 + psubd %xmm4, %xmm2 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm2 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm2 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm2, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*2(%rsp) # 16 - movl %r13d, %esi # W[(n+13) & 15] - xorl %r8d, %esi # ^W[(n+8) & 15] - xorl -32+4*2(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*0(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*0(%rsp) # store to W[n & 15] movl %ebx, %edi # c xorl %ecx, %edi # ^d andl %eax, %edi # &b xorl %ecx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rdx,%rsi), %edx # e += RCONST + W[n & 15] + addl -64+4*0(%rsp), %edx # e += RCONST + W[n & 15] addl %edi, %edx # e += (((c ^ d) & b) ^ d) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) # 17 - movl %r14d, %esi # W[(n+13) & 15] - xorl %r9d, %esi # ^W[(n+8) & 15] - xorl -32+4*3(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*1(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*1(%rsp) # store to W[n & 15] movl %eax, %edi # c xorl %ebx, %edi # ^d andl %ebp, %edi # &b xorl %ebx, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rcx,%rsi), %ecx # e += RCONST + W[n & 15] + addl -64+4*1(%rsp), %ecx # e += RCONST + W[n & 15] addl %edi, %ecx # e += (((c ^ d) & b) ^ d) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) # 18 - movl %r15d, %esi # W[(n+13) & 15] - xorl %r10d, %esi # ^W[(n+8) & 15] - xorl -32+4*4(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*2(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*2(%rsp) # store to W[n & 15] movl %ebp, %edi # c xorl %eax, %edi # ^d andl %edx, %edi # &b xorl %eax, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rbx,%rsi), %ebx # e += RCONST + W[n & 15] + addl -64+4*2(%rsp), %ebx # e += RCONST + W[n & 15] addl %edi, %ebx # e += (((c ^ d) & b) ^ d) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) # 19 - movl -32+4*0(%rsp), %esi # W[(n+13) & 15] - xorl %r11d, %esi # ^W[(n+8) & 15] - xorl -32+4*5(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*3(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*3(%rsp) # store to W[n & 15] movl %edx, %edi # c xorl %ebp, %edi # ^d andl %ecx, %edi # &b xorl %ebp, %edi # (((c ^ d) & b) ^ d) - leal 0x5A827999(%rax,%rsi), %eax # e += RCONST + W[n & 15] + addl -64+4*3(%rsp), %eax # e += RCONST + W[n & 15] addl %edi, %eax # e += (((c ^ d) & b) ^ d) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) +# PREP %xmm3 %xmm0 %xmm1 %xmm2 -64+16*3(%rsp) + movaps %xmm2, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm3, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm0, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm1, %xmm3 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm3 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm3, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm3, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm3, %xmm3 # shift left by 1 + psubd %xmm4, %xmm3 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm3 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm3 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm3, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*3(%rsp) # 20 - movl -32+4*1(%rsp), %esi # W[(n+13) & 15] - xorl %r12d, %esi # ^W[(n+8) & 15] - xorl -32+4*6(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*4(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*4(%rsp) # store to W[n & 15] movl %ecx, %edi # c xorl %edx, %edi # ^d xorl %ebx, %edi # ^b - leal 0x6ED9EBA1(%rbp,%rsi), %ebp # e += RCONST + W[n & 15] + addl -64+4*4(%rsp), %ebp # e += RCONST + W[n & 15] addl %edi, %ebp # e += (c ^ d ^ b) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) # 21 - movl -32+4*2(%rsp), %esi # W[(n+13) & 15] - xorl %r13d, %esi # ^W[(n+8) & 15] - xorl -32+4*7(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*5(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*5(%rsp) # store to W[n & 15] movl %ebx, %edi # c xorl %ecx, %edi # ^d xorl %eax, %edi # ^b - leal 0x6ED9EBA1(%rdx,%rsi), %edx # e += RCONST + W[n & 15] + addl -64+4*5(%rsp), %edx # e += RCONST + W[n & 15] addl %edi, %edx # e += (c ^ d ^ b) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) # 22 - movl -32+4*3(%rsp), %esi # W[(n+13) & 15] - xorl %r14d, %esi # ^W[(n+8) & 15] - xorl %r8d, %esi # ^W[(n+2) & 15] - xorl -32+4*6(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*6(%rsp) # store to W[n & 15] movl %eax, %edi # c xorl %ebx, %edi # ^d xorl %ebp, %edi # ^b - leal 0x6ED9EBA1(%rcx,%rsi), %ecx # e += RCONST + W[n & 15] + addl -64+4*6(%rsp), %ecx # e += RCONST + W[n & 15] addl %edi, %ecx # e += (c ^ d ^ b) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) # 23 - movl -32+4*4(%rsp), %esi # W[(n+13) & 15] - xorl %r15d, %esi # ^W[(n+8) & 15] - xorl %r9d, %esi # ^W[(n+2) & 15] - xorl -32+4*7(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*7(%rsp) # store to W[n & 15] movl %ebp, %edi # c xorl %eax, %edi # ^d xorl %edx, %edi # ^b - leal 0x6ED9EBA1(%rbx,%rsi), %ebx # e += RCONST + W[n & 15] + addl -64+4*7(%rsp), %ebx # e += RCONST + W[n & 15] addl %edi, %ebx # e += (c ^ d ^ b) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) +# PREP %xmm0 %xmm1 %xmm2 %xmm3 -64+16*0(%rsp) + movaps %xmm3, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm0, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm1, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm2, %xmm0 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm0 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm0, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm0, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm0, %xmm0 # shift left by 1 + psubd %xmm4, %xmm0 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm0 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm0 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm0, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*0(%rsp) # 24 - xorl -32+4*5(%rsp), %r8d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*0(%rsp), %r8d # ^W[(n+8) & 15] - xorl %r10d, %r8d # ^W[(n+2) & 15] - roll %r8d # movl %edx, %edi # c xorl %ebp, %edi # ^d xorl %ecx, %edi # ^b - leal 0x6ED9EBA1(%rax,%r8), %eax # e += RCONST + W[n & 15] + addl -64+4*8(%rsp), %eax # e += RCONST + W[n & 15] addl %edi, %eax # e += (c ^ d ^ b) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) # 25 - xorl -32+4*6(%rsp), %r9d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*1(%rsp), %r9d # ^W[(n+8) & 15] - xorl %r11d, %r9d # ^W[(n+2) & 15] - roll %r9d # movl %ecx, %edi # c xorl %edx, %edi # ^d xorl %ebx, %edi # ^b - leal 0x6ED9EBA1(%rbp,%r9), %ebp # e += RCONST + W[n & 15] + addl -64+4*9(%rsp), %ebp # e += RCONST + W[n & 15] addl %edi, %ebp # e += (c ^ d ^ b) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) # 26 - xorl -32+4*7(%rsp), %r10d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*2(%rsp), %r10d # ^W[(n+8) & 15] - xorl %r12d, %r10d # ^W[(n+2) & 15] - roll %r10d # movl %ebx, %edi # c xorl %ecx, %edi # ^d xorl %eax, %edi # ^b - leal 0x6ED9EBA1(%rdx,%r10), %edx # e += RCONST + W[n & 15] + addl -64+4*10(%rsp), %edx # e += RCONST + W[n & 15] addl %edi, %edx # e += (c ^ d ^ b) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) # 27 - xorl %r8d, %r11d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*3(%rsp), %r11d # ^W[(n+8) & 15] - xorl %r13d, %r11d # ^W[(n+2) & 15] - roll %r11d # movl %eax, %edi # c xorl %ebx, %edi # ^d xorl %ebp, %edi # ^b - leal 0x6ED9EBA1(%rcx,%r11), %ecx # e += RCONST + W[n & 15] + addl -64+4*11(%rsp), %ecx # e += RCONST + W[n & 15] addl %edi, %ecx # e += (c ^ d ^ b) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) +# PREP %xmm1 %xmm2 %xmm3 %xmm0 -64+16*1(%rsp) + movaps %xmm0, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm1, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm2, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm3, %xmm1 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm1 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm1, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm1, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm1, %xmm1 # shift left by 1 + psubd %xmm4, %xmm1 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm1 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm1 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm1, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*1(%rsp) # 28 - xorl %r9d, %r12d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*4(%rsp), %r12d # ^W[(n+8) & 15] - xorl %r14d, %r12d # ^W[(n+2) & 15] - roll %r12d # movl %ebp, %edi # c xorl %eax, %edi # ^d xorl %edx, %edi # ^b - leal 0x6ED9EBA1(%rbx,%r12), %ebx # e += RCONST + W[n & 15] + addl -64+4*12(%rsp), %ebx # e += RCONST + W[n & 15] addl %edi, %ebx # e += (c ^ d ^ b) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) # 29 - xorl %r10d, %r13d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*5(%rsp), %r13d # ^W[(n+8) & 15] - xorl %r15d, %r13d # ^W[(n+2) & 15] - roll %r13d # movl %edx, %edi # c xorl %ebp, %edi # ^d xorl %ecx, %edi # ^b - leal 0x6ED9EBA1(%rax,%r13), %eax # e += RCONST + W[n & 15] + addl -64+4*13(%rsp), %eax # e += RCONST + W[n & 15] addl %edi, %eax # e += (c ^ d ^ b) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) # 30 - xorl %r11d, %r14d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*6(%rsp), %r14d # ^W[(n+8) & 15] - xorl -32+4*0(%rsp), %r14d # ^W[(n+2) & 15] - roll %r14d # movl %ecx, %edi # c xorl %edx, %edi # ^d xorl %ebx, %edi # ^b - leal 0x6ED9EBA1(%rbp,%r14), %ebp # e += RCONST + W[n & 15] + addl -64+4*14(%rsp), %ebp # e += RCONST + W[n & 15] addl %edi, %ebp # e += (c ^ d ^ b) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) # 31 - xorl %r12d, %r15d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*7(%rsp), %r15d # ^W[(n+8) & 15] - xorl -32+4*1(%rsp), %r15d # ^W[(n+2) & 15] - roll %r15d # movl %ebx, %edi # c xorl %ecx, %edi # ^d xorl %eax, %edi # ^b - leal 0x6ED9EBA1(%rdx,%r15), %edx # e += RCONST + W[n & 15] + addl -64+4*15(%rsp), %edx # e += RCONST + W[n & 15] addl %edi, %edx # e += (c ^ d ^ b) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) + movaps rconst0x8F1BBCDC(%rip), %xmm6 +# PREP %xmm2 %xmm3 %xmm0 %xmm1 -64+16*2(%rsp) + movaps %xmm1, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm2, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm3, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm0, %xmm2 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm2 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm2, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm2, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm2, %xmm2 # shift left by 1 + psubd %xmm4, %xmm2 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm2 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm2 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm2, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*2(%rsp) # 32 - movl %r13d, %esi # W[(n+13) & 15] - xorl %r8d, %esi # ^W[(n+8) & 15] - xorl -32+4*2(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*0(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*0(%rsp) # store to W[n & 15] movl %eax, %edi # c xorl %ebx, %edi # ^d xorl %ebp, %edi # ^b - leal 0x6ED9EBA1(%rcx,%rsi), %ecx # e += RCONST + W[n & 15] + addl -64+4*0(%rsp), %ecx # e += RCONST + W[n & 15] addl %edi, %ecx # e += (c ^ d ^ b) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) # 33 - movl %r14d, %esi # W[(n+13) & 15] - xorl %r9d, %esi # ^W[(n+8) & 15] - xorl -32+4*3(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*1(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*1(%rsp) # store to W[n & 15] movl %ebp, %edi # c xorl %eax, %edi # ^d xorl %edx, %edi # ^b - leal 0x6ED9EBA1(%rbx,%rsi), %ebx # e += RCONST + W[n & 15] + addl -64+4*1(%rsp), %ebx # e += RCONST + W[n & 15] addl %edi, %ebx # e += (c ^ d ^ b) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) # 34 - movl %r15d, %esi # W[(n+13) & 15] - xorl %r10d, %esi # ^W[(n+8) & 15] - xorl -32+4*4(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*2(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*2(%rsp) # store to W[n & 15] movl %edx, %edi # c xorl %ebp, %edi # ^d xorl %ecx, %edi # ^b - leal 0x6ED9EBA1(%rax,%rsi), %eax # e += RCONST + W[n & 15] + addl -64+4*2(%rsp), %eax # e += RCONST + W[n & 15] addl %edi, %eax # e += (c ^ d ^ b) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) # 35 - movl -32+4*0(%rsp), %esi # W[(n+13) & 15] - xorl %r11d, %esi # ^W[(n+8) & 15] - xorl -32+4*5(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*3(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*3(%rsp) # store to W[n & 15] movl %ecx, %edi # c xorl %edx, %edi # ^d xorl %ebx, %edi # ^b - leal 0x6ED9EBA1(%rbp,%rsi), %ebp # e += RCONST + W[n & 15] + addl -64+4*3(%rsp), %ebp # e += RCONST + W[n & 15] addl %edi, %ebp # e += (c ^ d ^ b) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) +# PREP %xmm3 %xmm0 %xmm1 %xmm2 -64+16*3(%rsp) + movaps %xmm2, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm3, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm0, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm1, %xmm3 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm3 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm3, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm3, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm3, %xmm3 # shift left by 1 + psubd %xmm4, %xmm3 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm3 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm3 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm3, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*3(%rsp) # 36 - movl -32+4*1(%rsp), %esi # W[(n+13) & 15] - xorl %r12d, %esi # ^W[(n+8) & 15] - xorl -32+4*6(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*4(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*4(%rsp) # store to W[n & 15] movl %ebx, %edi # c xorl %ecx, %edi # ^d xorl %eax, %edi # ^b - leal 0x6ED9EBA1(%rdx,%rsi), %edx # e += RCONST + W[n & 15] + addl -64+4*4(%rsp), %edx # e += RCONST + W[n & 15] addl %edi, %edx # e += (c ^ d ^ b) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) # 37 - movl -32+4*2(%rsp), %esi # W[(n+13) & 15] - xorl %r13d, %esi # ^W[(n+8) & 15] - xorl -32+4*7(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*5(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*5(%rsp) # store to W[n & 15] movl %eax, %edi # c xorl %ebx, %edi # ^d xorl %ebp, %edi # ^b - leal 0x6ED9EBA1(%rcx,%rsi), %ecx # e += RCONST + W[n & 15] + addl -64+4*5(%rsp), %ecx # e += RCONST + W[n & 15] addl %edi, %ecx # e += (c ^ d ^ b) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) # 38 - movl -32+4*3(%rsp), %esi # W[(n+13) & 15] - xorl %r14d, %esi # ^W[(n+8) & 15] - xorl %r8d, %esi # ^W[(n+2) & 15] - xorl -32+4*6(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*6(%rsp) # store to W[n & 15] movl %ebp, %edi # c xorl %eax, %edi # ^d xorl %edx, %edi # ^b - leal 0x6ED9EBA1(%rbx,%rsi), %ebx # e += RCONST + W[n & 15] + addl -64+4*6(%rsp), %ebx # e += RCONST + W[n & 15] addl %edi, %ebx # e += (c ^ d ^ b) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) # 39 - movl -32+4*4(%rsp), %esi # W[(n+13) & 15] - xorl %r15d, %esi # ^W[(n+8) & 15] - xorl %r9d, %esi # ^W[(n+2) & 15] - xorl -32+4*7(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*7(%rsp) # store to W[n & 15] movl %edx, %edi # c xorl %ebp, %edi # ^d xorl %ecx, %edi # ^b - leal 0x6ED9EBA1(%rax,%rsi), %eax # e += RCONST + W[n & 15] + addl -64+4*7(%rsp), %eax # e += RCONST + W[n & 15] addl %edi, %eax # e += (c ^ d ^ b) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) +# PREP %xmm0 %xmm1 %xmm2 %xmm3 -64+16*0(%rsp) + movaps %xmm3, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm0, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm1, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm2, %xmm0 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm0 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm0, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm0, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm0, %xmm0 # shift left by 1 + psubd %xmm4, %xmm0 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm0 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm0 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm0, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*0(%rsp) # 40 movl %ebx, %edi # di: b movl %ebx, %esi # si: b @@ -625,12 +741,8 @@ sha1_process_block64: andl %ecx, %esi # si: b & c andl %edx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl -32+4*5(%rsp), %r8d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*0(%rsp), %r8d # ^W[(n+8) & 15] - xorl %r10d, %r8d # ^W[(n+2) & 15] - roll %r8d # addl %edi, %ebp # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rbp,%r8), %ebp # e += RCONST + W[n & 15] + addl -64+4*8(%rsp), %ebp # e += RCONST + W[n & 15] movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) @@ -642,12 +754,8 @@ sha1_process_block64: andl %ebx, %esi # si: b & c andl %ecx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl -32+4*6(%rsp), %r9d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*1(%rsp), %r9d # ^W[(n+8) & 15] - xorl %r11d, %r9d # ^W[(n+2) & 15] - roll %r9d # addl %edi, %edx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rdx,%r9), %edx # e += RCONST + W[n & 15] + addl -64+4*9(%rsp), %edx # e += RCONST + W[n & 15] movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) @@ -659,12 +767,8 @@ sha1_process_block64: andl %eax, %esi # si: b & c andl %ebx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl -32+4*7(%rsp), %r10d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*2(%rsp), %r10d # ^W[(n+8) & 15] - xorl %r12d, %r10d # ^W[(n+2) & 15] - roll %r10d # addl %edi, %ecx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rcx,%r10), %ecx # e += RCONST + W[n & 15] + addl -64+4*10(%rsp), %ecx # e += RCONST + W[n & 15] movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) @@ -676,16 +780,37 @@ sha1_process_block64: andl %ebp, %esi # si: b & c andl %eax, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl %r8d, %r11d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*3(%rsp), %r11d # ^W[(n+8) & 15] - xorl %r13d, %r11d # ^W[(n+2) & 15] - roll %r11d # addl %edi, %ebx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rbx,%r11), %ebx # e += RCONST + W[n & 15] + addl -64+4*11(%rsp), %ebx # e += RCONST + W[n & 15] movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) +# PREP %xmm1 %xmm2 %xmm3 %xmm0 -64+16*1(%rsp) + movaps %xmm0, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm1, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm2, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm3, %xmm1 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm1 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm1, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm1, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm1, %xmm1 # shift left by 1 + psubd %xmm4, %xmm1 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm1 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm1 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm1, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*1(%rsp) # 44 movl %ecx, %edi # di: b movl %ecx, %esi # si: b @@ -693,12 +818,8 @@ sha1_process_block64: andl %edx, %esi # si: b & c andl %ebp, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl %r9d, %r12d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*4(%rsp), %r12d # ^W[(n+8) & 15] - xorl %r14d, %r12d # ^W[(n+2) & 15] - roll %r12d # addl %edi, %eax # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rax,%r12), %eax # e += RCONST + W[n & 15] + addl -64+4*12(%rsp), %eax # e += RCONST + W[n & 15] movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) @@ -710,12 +831,8 @@ sha1_process_block64: andl %ecx, %esi # si: b & c andl %edx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl %r10d, %r13d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*5(%rsp), %r13d # ^W[(n+8) & 15] - xorl %r15d, %r13d # ^W[(n+2) & 15] - roll %r13d # addl %edi, %ebp # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rbp,%r13), %ebp # e += RCONST + W[n & 15] + addl -64+4*13(%rsp), %ebp # e += RCONST + W[n & 15] movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) @@ -727,12 +844,8 @@ sha1_process_block64: andl %ebx, %esi # si: b & c andl %ecx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl %r11d, %r14d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*6(%rsp), %r14d # ^W[(n+8) & 15] - xorl -32+4*0(%rsp), %r14d # ^W[(n+2) & 15] - roll %r14d # addl %edi, %edx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rdx,%r14), %edx # e += RCONST + W[n & 15] + addl -64+4*14(%rsp), %edx # e += RCONST + W[n & 15] movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) @@ -744,16 +857,37 @@ sha1_process_block64: andl %eax, %esi # si: b & c andl %ebx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl %r12d, %r15d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*7(%rsp), %r15d # ^W[(n+8) & 15] - xorl -32+4*1(%rsp), %r15d # ^W[(n+2) & 15] - roll %r15d # addl %edi, %ecx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rcx,%r15), %ecx # e += RCONST + W[n & 15] + addl -64+4*15(%rsp), %ecx # e += RCONST + W[n & 15] movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) +# PREP %xmm2 %xmm3 %xmm0 %xmm1 -64+16*2(%rsp) + movaps %xmm1, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm2, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm3, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm0, %xmm2 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm2 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm2, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm2, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm2, %xmm2 # shift left by 1 + psubd %xmm4, %xmm2 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm2 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm2 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm2, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*2(%rsp) # 48 movl %edx, %edi # di: b movl %edx, %esi # si: b @@ -761,14 +895,8 @@ sha1_process_block64: andl %ebp, %esi # si: b & c andl %eax, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - movl %r13d, %esi # W[(n+13) & 15] - xorl %r8d, %esi # ^W[(n+8) & 15] - xorl -32+4*2(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*0(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*0(%rsp) # store to W[n & 15] addl %edi, %ebx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rbx,%rsi), %ebx # e += RCONST + W[n & 15] + addl -64+4*0(%rsp), %ebx # e += RCONST + W[n & 15] movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) @@ -780,14 +908,8 @@ sha1_process_block64: andl %edx, %esi # si: b & c andl %ebp, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - movl %r14d, %esi # W[(n+13) & 15] - xorl %r9d, %esi # ^W[(n+8) & 15] - xorl -32+4*3(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*1(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*1(%rsp) # store to W[n & 15] addl %edi, %eax # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rax,%rsi), %eax # e += RCONST + W[n & 15] + addl -64+4*1(%rsp), %eax # e += RCONST + W[n & 15] movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) @@ -799,14 +921,8 @@ sha1_process_block64: andl %ecx, %esi # si: b & c andl %edx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - movl %r15d, %esi # W[(n+13) & 15] - xorl %r10d, %esi # ^W[(n+8) & 15] - xorl -32+4*4(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*2(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*2(%rsp) # store to W[n & 15] addl %edi, %ebp # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rbp,%rsi), %ebp # e += RCONST + W[n & 15] + addl -64+4*2(%rsp), %ebp # e += RCONST + W[n & 15] movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) @@ -818,18 +934,38 @@ sha1_process_block64: andl %ebx, %esi # si: b & c andl %ecx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - movl -32+4*0(%rsp), %esi # W[(n+13) & 15] - xorl %r11d, %esi # ^W[(n+8) & 15] - xorl -32+4*5(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*3(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*3(%rsp) # store to W[n & 15] addl %edi, %edx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rdx,%rsi), %edx # e += RCONST + W[n & 15] + addl -64+4*3(%rsp), %edx # e += RCONST + W[n & 15] movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) + movaps rconst0xCA62C1D6(%rip), %xmm6 +# PREP %xmm3 %xmm0 %xmm1 %xmm2 -64+16*3(%rsp) + movaps %xmm2, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm3, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm0, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm1, %xmm3 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm3 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm3, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm3, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm3, %xmm3 # shift left by 1 + psubd %xmm4, %xmm3 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm3 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm3 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm3, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*3(%rsp) # 52 movl %ebp, %edi # di: b movl %ebp, %esi # si: b @@ -837,14 +973,8 @@ sha1_process_block64: andl %eax, %esi # si: b & c andl %ebx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - movl -32+4*1(%rsp), %esi # W[(n+13) & 15] - xorl %r12d, %esi # ^W[(n+8) & 15] - xorl -32+4*6(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*4(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*4(%rsp) # store to W[n & 15] addl %edi, %ecx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rcx,%rsi), %ecx # e += RCONST + W[n & 15] + addl -64+4*4(%rsp), %ecx # e += RCONST + W[n & 15] movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) @@ -856,14 +986,8 @@ sha1_process_block64: andl %ebp, %esi # si: b & c andl %eax, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - movl -32+4*2(%rsp), %esi # W[(n+13) & 15] - xorl %r13d, %esi # ^W[(n+8) & 15] - xorl -32+4*7(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*5(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*5(%rsp) # store to W[n & 15] addl %edi, %ebx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rbx,%rsi), %ebx # e += RCONST + W[n & 15] + addl -64+4*5(%rsp), %ebx # e += RCONST + W[n & 15] movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) @@ -875,14 +999,8 @@ sha1_process_block64: andl %edx, %esi # si: b & c andl %ebp, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - movl -32+4*3(%rsp), %esi # W[(n+13) & 15] - xorl %r14d, %esi # ^W[(n+8) & 15] - xorl %r8d, %esi # ^W[(n+2) & 15] - xorl -32+4*6(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*6(%rsp) # store to W[n & 15] addl %edi, %eax # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rax,%rsi), %eax # e += RCONST + W[n & 15] + addl -64+4*6(%rsp), %eax # e += RCONST + W[n & 15] movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) @@ -894,18 +1012,37 @@ sha1_process_block64: andl %ecx, %esi # si: b & c andl %edx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - movl -32+4*4(%rsp), %esi # W[(n+13) & 15] - xorl %r15d, %esi # ^W[(n+8) & 15] - xorl %r9d, %esi # ^W[(n+2) & 15] - xorl -32+4*7(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*7(%rsp) # store to W[n & 15] addl %edi, %ebp # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rbp,%rsi), %ebp # e += RCONST + W[n & 15] + addl -64+4*7(%rsp), %ebp # e += RCONST + W[n & 15] movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) +# PREP %xmm0 %xmm1 %xmm2 %xmm3 -64+16*0(%rsp) + movaps %xmm3, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm0, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm1, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm2, %xmm0 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm0 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm0, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm0, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm0, %xmm0 # shift left by 1 + psubd %xmm4, %xmm0 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm0 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm0 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm0, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*0(%rsp) # 56 movl %eax, %edi # di: b movl %eax, %esi # si: b @@ -913,12 +1050,8 @@ sha1_process_block64: andl %ebx, %esi # si: b & c andl %ecx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl -32+4*5(%rsp), %r8d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*0(%rsp), %r8d # ^W[(n+8) & 15] - xorl %r10d, %r8d # ^W[(n+2) & 15] - roll %r8d # addl %edi, %edx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rdx,%r8), %edx # e += RCONST + W[n & 15] + addl -64+4*8(%rsp), %edx # e += RCONST + W[n & 15] movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) @@ -930,12 +1063,8 @@ sha1_process_block64: andl %eax, %esi # si: b & c andl %ebx, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl -32+4*6(%rsp), %r9d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*1(%rsp), %r9d # ^W[(n+8) & 15] - xorl %r11d, %r9d # ^W[(n+2) & 15] - roll %r9d # addl %edi, %ecx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rcx,%r9), %ecx # e += RCONST + W[n & 15] + addl -64+4*9(%rsp), %ecx # e += RCONST + W[n & 15] movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) @@ -947,12 +1076,8 @@ sha1_process_block64: andl %ebp, %esi # si: b & c andl %eax, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl -32+4*7(%rsp), %r10d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*2(%rsp), %r10d # ^W[(n+8) & 15] - xorl %r12d, %r10d # ^W[(n+2) & 15] - roll %r10d # addl %edi, %ebx # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rbx,%r10), %ebx # e += RCONST + W[n & 15] + addl -64+4*10(%rsp), %ebx # e += RCONST + W[n & 15] movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) @@ -964,307 +1089,282 @@ sha1_process_block64: andl %edx, %esi # si: b & c andl %ebp, %edi # di: (b | c) & d orl %esi, %edi # ((b | c) & d) | (b & c) - xorl %r8d, %r11d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*3(%rsp), %r11d # ^W[(n+8) & 15] - xorl %r13d, %r11d # ^W[(n+2) & 15] - roll %r11d # addl %edi, %eax # += ((b | c) & d) | (b & c) - leal -0x70E44324(%rax,%r11), %eax # e += RCONST + W[n & 15] + addl -64+4*11(%rsp), %eax # e += RCONST + W[n & 15] movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) +# PREP %xmm1 %xmm2 %xmm3 %xmm0 -64+16*1(%rsp) + movaps %xmm0, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm1, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm2, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm3, %xmm1 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm1 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm1, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm1, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm1, %xmm1 # shift left by 1 + psubd %xmm4, %xmm1 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm1 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm1 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm1, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*1(%rsp) # 60 - xorl %r9d, %r12d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*4(%rsp), %r12d # ^W[(n+8) & 15] - xorl %r14d, %r12d # ^W[(n+2) & 15] - roll %r12d # movl %ecx, %edi # c xorl %edx, %edi # ^d xorl %ebx, %edi # ^b - leal -0x359D3E2A(%rbp,%r12), %ebp # e += RCONST + W[n & 15] + addl -64+4*12(%rsp), %ebp # e += RCONST + W[n & 15] addl %edi, %ebp # e += (c ^ d ^ b) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) # 61 - xorl %r10d, %r13d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*5(%rsp), %r13d # ^W[(n+8) & 15] - xorl %r15d, %r13d # ^W[(n+2) & 15] - roll %r13d # movl %ebx, %edi # c xorl %ecx, %edi # ^d xorl %eax, %edi # ^b - leal -0x359D3E2A(%rdx,%r13), %edx # e += RCONST + W[n & 15] + addl -64+4*13(%rsp), %edx # e += RCONST + W[n & 15] addl %edi, %edx # e += (c ^ d ^ b) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) # 62 - xorl %r11d, %r14d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*6(%rsp), %r14d # ^W[(n+8) & 15] - xorl -32+4*0(%rsp), %r14d # ^W[(n+2) & 15] - roll %r14d # movl %eax, %edi # c xorl %ebx, %edi # ^d xorl %ebp, %edi # ^b - leal -0x359D3E2A(%rcx,%r14), %ecx # e += RCONST + W[n & 15] + addl -64+4*14(%rsp), %ecx # e += RCONST + W[n & 15] addl %edi, %ecx # e += (c ^ d ^ b) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) # 63 - xorl %r12d, %r15d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*7(%rsp), %r15d # ^W[(n+8) & 15] - xorl -32+4*1(%rsp), %r15d # ^W[(n+2) & 15] - roll %r15d # movl %ebp, %edi # c xorl %eax, %edi # ^d xorl %edx, %edi # ^b - leal -0x359D3E2A(%rbx,%r15), %ebx # e += RCONST + W[n & 15] + addl -64+4*15(%rsp), %ebx # e += RCONST + W[n & 15] addl %edi, %ebx # e += (c ^ d ^ b) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) +# PREP %xmm2 %xmm3 %xmm0 %xmm1 -64+16*2(%rsp) + movaps %xmm1, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm2, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm3, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm0, %xmm2 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm2 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm2, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm2, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm2, %xmm2 # shift left by 1 + psubd %xmm4, %xmm2 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm2 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm2 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm2, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*2(%rsp) # 64 - movl %r13d, %esi # W[(n+13) & 15] - xorl %r8d, %esi # ^W[(n+8) & 15] - xorl -32+4*2(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*0(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*0(%rsp) # store to W[n & 15] movl %edx, %edi # c xorl %ebp, %edi # ^d xorl %ecx, %edi # ^b - leal -0x359D3E2A(%rax,%rsi), %eax # e += RCONST + W[n & 15] + addl -64+4*0(%rsp), %eax # e += RCONST + W[n & 15] addl %edi, %eax # e += (c ^ d ^ b) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) # 65 - movl %r14d, %esi # W[(n+13) & 15] - xorl %r9d, %esi # ^W[(n+8) & 15] - xorl -32+4*3(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*1(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*1(%rsp) # store to W[n & 15] movl %ecx, %edi # c xorl %edx, %edi # ^d xorl %ebx, %edi # ^b - leal -0x359D3E2A(%rbp,%rsi), %ebp # e += RCONST + W[n & 15] + addl -64+4*1(%rsp), %ebp # e += RCONST + W[n & 15] addl %edi, %ebp # e += (c ^ d ^ b) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) # 66 - movl %r15d, %esi # W[(n+13) & 15] - xorl %r10d, %esi # ^W[(n+8) & 15] - xorl -32+4*4(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*2(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*2(%rsp) # store to W[n & 15] movl %ebx, %edi # c xorl %ecx, %edi # ^d xorl %eax, %edi # ^b - leal -0x359D3E2A(%rdx,%rsi), %edx # e += RCONST + W[n & 15] + addl -64+4*2(%rsp), %edx # e += RCONST + W[n & 15] addl %edi, %edx # e += (c ^ d ^ b) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) # 67 - movl -32+4*0(%rsp), %esi # W[(n+13) & 15] - xorl %r11d, %esi # ^W[(n+8) & 15] - xorl -32+4*5(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*3(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*3(%rsp) # store to W[n & 15] movl %eax, %edi # c xorl %ebx, %edi # ^d xorl %ebp, %edi # ^b - leal -0x359D3E2A(%rcx,%rsi), %ecx # e += RCONST + W[n & 15] + addl -64+4*3(%rsp), %ecx # e += RCONST + W[n & 15] addl %edi, %ecx # e += (c ^ d ^ b) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) +# PREP %xmm3 %xmm0 %xmm1 %xmm2 -64+16*3(%rsp) + movaps %xmm2, %xmm4 + psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0) + pshufd $0x4e, %xmm3, %xmm5 # 01001110=2,3,0,1 shuffle, ([2],[3],x,x) + punpcklqdq %xmm0, %xmm5 # T2 = W4[0..63]:T2[0..63] = ([2],[3],[4],[5]) + xorps %xmm1, %xmm3 # ([8],[9],[10],[11]) ^ ([0],[1],[2],[3]) + xorps %xmm4, %xmm5 # ([13],[14],[15],0) ^ ([2],[3],[4],[5]) + xorps %xmm5, %xmm3 # ^ + # W0 = unrotated (W[0]..W[3]), still needs W[3] fixup + movaps %xmm3, %xmm5 + xorps %xmm4, %xmm4 # rol(W0,1): + pcmpgtd %xmm3, %xmm4 # ffffffff for elements <0 (ones with msb bit 1) + paddd %xmm3, %xmm3 # shift left by 1 + psubd %xmm4, %xmm3 # add 1 to those who had msb bit 1 + # W0 = rotated (W[0]..W[3]), still needs W[3] fixup + pslldq $12, %xmm5 # lshift by 12 bytes: T2 = (0,0,0,unrotW[0]) + movaps %xmm5, %xmm4 + pslld $2, %xmm5 + psrld $30, %xmm4 +# xorps %xmm4, %xmm5 # rol((0,0,0,unrotW[0]),2) + xorps %xmm4, %xmm3 # same result, but does not depend on/does not modify T2 + xorps %xmm5, %xmm3 # W0 = rol(W[0]..W[3],1) ^ (0,0,0,rol(unrotW[0],2)) + movaps %xmm3, %xmm5 + paddd %xmm6, %xmm5 + movups %xmm5, -64+16*3(%rsp) # 68 - movl -32+4*1(%rsp), %esi # W[(n+13) & 15] - xorl %r12d, %esi # ^W[(n+8) & 15] - xorl -32+4*6(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*4(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*4(%rsp) # store to W[n & 15] movl %ebp, %edi # c xorl %eax, %edi # ^d xorl %edx, %edi # ^b - leal -0x359D3E2A(%rbx,%rsi), %ebx # e += RCONST + W[n & 15] + addl -64+4*4(%rsp), %ebx # e += RCONST + W[n & 15] addl %edi, %ebx # e += (c ^ d ^ b) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) # 69 - movl -32+4*2(%rsp), %esi # W[(n+13) & 15] - xorl %r13d, %esi # ^W[(n+8) & 15] - xorl -32+4*7(%rsp), %esi # ^W[(n+2) & 15] - xorl -32+4*5(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*5(%rsp) # store to W[n & 15] movl %edx, %edi # c xorl %ebp, %edi # ^d xorl %ecx, %edi # ^b - leal -0x359D3E2A(%rax,%rsi), %eax # e += RCONST + W[n & 15] + addl -64+4*5(%rsp), %eax # e += RCONST + W[n & 15] addl %edi, %eax # e += (c ^ d ^ b) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) # 70 - movl -32+4*3(%rsp), %esi # W[(n+13) & 15] - xorl %r14d, %esi # ^W[(n+8) & 15] - xorl %r8d, %esi # ^W[(n+2) & 15] - xorl -32+4*6(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*6(%rsp) # store to W[n & 15] movl %ecx, %edi # c xorl %edx, %edi # ^d xorl %ebx, %edi # ^b - leal -0x359D3E2A(%rbp,%rsi), %ebp # e += RCONST + W[n & 15] + addl -64+4*6(%rsp), %ebp # e += RCONST + W[n & 15] addl %edi, %ebp # e += (c ^ d ^ b) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) # 71 - movl -32+4*4(%rsp), %esi # W[(n+13) & 15] - xorl %r15d, %esi # ^W[(n+8) & 15] - xorl %r9d, %esi # ^W[(n+2) & 15] - xorl -32+4*7(%rsp), %esi # ^W[n & 15] - roll %esi # - movl %esi, -32+4*7(%rsp) # store to W[n & 15] movl %ebx, %edi # c xorl %ecx, %edi # ^d xorl %eax, %edi # ^b - leal -0x359D3E2A(%rdx,%rsi), %edx # e += RCONST + W[n & 15] + addl -64+4*7(%rsp), %edx # e += RCONST + W[n & 15] addl %edi, %edx # e += (c ^ d ^ b) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) # 72 - xorl -32+4*5(%rsp), %r8d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*0(%rsp), %r8d # ^W[(n+8) & 15] - xorl %r10d, %r8d # ^W[(n+2) & 15] - roll %r8d # movl %eax, %edi # c xorl %ebx, %edi # ^d xorl %ebp, %edi # ^b - leal -0x359D3E2A(%rcx,%r8), %ecx # e += RCONST + W[n & 15] + addl -64+4*8(%rsp), %ecx # e += RCONST + W[n & 15] addl %edi, %ecx # e += (c ^ d ^ b) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) # 73 - xorl -32+4*6(%rsp), %r9d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*1(%rsp), %r9d # ^W[(n+8) & 15] - xorl %r11d, %r9d # ^W[(n+2) & 15] - roll %r9d # movl %ebp, %edi # c xorl %eax, %edi # ^d xorl %edx, %edi # ^b - leal -0x359D3E2A(%rbx,%r9), %ebx # e += RCONST + W[n & 15] + addl -64+4*9(%rsp), %ebx # e += RCONST + W[n & 15] addl %edi, %ebx # e += (c ^ d ^ b) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) # 74 - xorl -32+4*7(%rsp), %r10d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*2(%rsp), %r10d # ^W[(n+8) & 15] - xorl %r12d, %r10d # ^W[(n+2) & 15] - roll %r10d # movl %edx, %edi # c xorl %ebp, %edi # ^d xorl %ecx, %edi # ^b - leal -0x359D3E2A(%rax,%r10), %eax # e += RCONST + W[n & 15] + addl -64+4*10(%rsp), %eax # e += RCONST + W[n & 15] addl %edi, %eax # e += (c ^ d ^ b) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %eax # e += rotl32(a,5) rorl $2, %ecx # b = rotl32(b,30) # 75 - xorl %r8d, %r11d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*3(%rsp), %r11d # ^W[(n+8) & 15] - xorl %r13d, %r11d # ^W[(n+2) & 15] - roll %r11d # movl %ecx, %edi # c xorl %edx, %edi # ^d xorl %ebx, %edi # ^b - leal -0x359D3E2A(%rbp,%r11), %ebp # e += RCONST + W[n & 15] + addl -64+4*11(%rsp), %ebp # e += RCONST + W[n & 15] addl %edi, %ebp # e += (c ^ d ^ b) movl %eax, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebp # e += rotl32(a,5) rorl $2, %ebx # b = rotl32(b,30) # 76 - xorl %r9d, %r12d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*4(%rsp), %r12d # ^W[(n+8) & 15] - xorl %r14d, %r12d # ^W[(n+2) & 15] - roll %r12d # movl %ebx, %edi # c xorl %ecx, %edi # ^d xorl %eax, %edi # ^b - leal -0x359D3E2A(%rdx,%r12), %edx # e += RCONST + W[n & 15] + addl -64+4*12(%rsp), %edx # e += RCONST + W[n & 15] addl %edi, %edx # e += (c ^ d ^ b) movl %ebp, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %edx # e += rotl32(a,5) rorl $2, %eax # b = rotl32(b,30) # 77 - xorl %r10d, %r13d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*5(%rsp), %r13d # ^W[(n+8) & 15] - xorl %r15d, %r13d # ^W[(n+2) & 15] - roll %r13d # movl %eax, %edi # c xorl %ebx, %edi # ^d xorl %ebp, %edi # ^b - leal -0x359D3E2A(%rcx,%r13), %ecx # e += RCONST + W[n & 15] + addl -64+4*13(%rsp), %ecx # e += RCONST + W[n & 15] addl %edi, %ecx # e += (c ^ d ^ b) movl %edx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ecx # e += rotl32(a,5) rorl $2, %ebp # b = rotl32(b,30) # 78 - xorl %r11d, %r14d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*6(%rsp), %r14d # ^W[(n+8) & 15] - xorl -32+4*0(%rsp), %r14d # ^W[(n+2) & 15] - roll %r14d # movl %ebp, %edi # c xorl %eax, %edi # ^d xorl %edx, %edi # ^b - leal -0x359D3E2A(%rbx,%r14), %ebx # e += RCONST + W[n & 15] + addl -64+4*14(%rsp), %ebx # e += RCONST + W[n & 15] addl %edi, %ebx # e += (c ^ d ^ b) movl %ecx, %esi # roll $5, %esi # rotl32(a,5) addl %esi, %ebx # e += rotl32(a,5) rorl $2, %edx # b = rotl32(b,30) # 79 - xorl %r12d, %r15d # W[n & 15] ^= W[(n+13) & 15] - xorl -32+4*7(%rsp), %r15d # ^W[(n+8) & 15] - xorl -32+4*1(%rsp), %r15d # ^W[(n+2) & 15] - roll %r15d # movl %edx, %edi # c xorl %ebp, %edi # ^d xorl %ecx, %edi # ^b - leal -0x359D3E2A(%rax,%r15), %eax # e += RCONST + W[n & 15] + addl -64+4*15(%rsp), %eax # e += RCONST + W[n & 15] addl %edi, %eax # e += (c ^ d ^ b) movl %ebx, %esi # roll $5, %esi # rotl32(a,5) @@ -1286,4 +1386,28 @@ sha1_process_block64: ret .size sha1_process_block64, .-sha1_process_block64 + + .section .rodata.cst16.sha1const, "aM", @progbits, 16 + .align 16 +rconst0x5A827999: + .long 0x5A827999 + .long 0x5A827999 + .long 0x5A827999 + .long 0x5A827999 +rconst0x6ED9EBA1: + .long 0x6ED9EBA1 + .long 0x6ED9EBA1 + .long 0x6ED9EBA1 + .long 0x6ED9EBA1 +rconst0x8F1BBCDC: + .long 0x8F1BBCDC + .long 0x8F1BBCDC + .long 0x8F1BBCDC + .long 0x8F1BBCDC +rconst0xCA62C1D6: + .long 0xCA62C1D6 + .long 0xCA62C1D6 + .long 0xCA62C1D6 + .long 0xCA62C1D6 + #endif |