Integration of cnv2 and asm optimizations for cnv1&cnv2 thx to @SChernykh (#185)

src/crypto/asm/cnv2_main_loop_ivybridge.inc

@@ -0,0 +1,186 @@
+	mov	 QWORD PTR [rsp+24], rbx
+	push	 rbp
+	push	 rsi
+	push	 rdi
+	push	 r12
+	push	 r13
+	push	 r14
+	push	 r15
+	sub	 rsp, 80
+
+	stmxcsr DWORD PTR [rsp]
+	mov DWORD PTR [rsp+4], 24448
+	ldmxcsr DWORD PTR [rsp+4]
+
+	mov	 rax, QWORD PTR [rcx+48]
+	mov	 r9, rcx
+	xor	 rax, QWORD PTR [rcx+16]
+	mov	 esi, 524288
+	mov	 r8, QWORD PTR [rcx+32]
+	mov	 r13d, -2147483647
+	xor	 r8, QWORD PTR [rcx]
+	mov	 r11, QWORD PTR [rcx+40]
+	mov	 r10, r8
+	mov	 rdx, QWORD PTR [rcx+56]
+	movq	 xmm4, rax
+	xor	 rdx, QWORD PTR [rcx+24]
+	xor	 r11, QWORD PTR [rcx+8]
+	mov	 rbx, QWORD PTR [rcx+224]
+	mov	 rax, QWORD PTR [r9+80]
+	xor	 rax, QWORD PTR [r9+64]
+	movq	 xmm0, rdx
+	mov	 rcx, QWORD PTR [rcx+88]
+	xor	 rcx, QWORD PTR [r9+72]
+	movq	 xmm3, QWORD PTR [r9+104]
+	movaps	 XMMWORD PTR [rsp+64], xmm6
+	movaps	 XMMWORD PTR [rsp+48], xmm7
+	movaps	 XMMWORD PTR [rsp+32], xmm8
+	and	 r10d, 2097136
+	movq	 xmm5, rax
+
+	mov ax, 1023
+	shl rax, 52
+	movq xmm8, rax
+	mov r15, QWORD PTR [r9+96]
+	punpcklqdq xmm4, xmm0
+	movq	 xmm0, rcx
+	punpcklqdq xmm5, xmm0
+	movdqu	 xmm6, XMMWORD PTR [r10+rbx]
+
+        #ifdef __APPLE__
+        ALIGN 16
+        #else
+        ALIGN 64
+        #endif
+$main_loop_ivybridge:
+	lea	 rdx, QWORD PTR [r10+rbx]
+	mov	 ecx, r10d
+	mov	 eax, r10d
+	mov rdi, r15
+	xor	 ecx, 16
+	xor	 eax, 32
+	xor	 r10d, 48
+	movq	 xmm0, r11
+	movq	 xmm7, r8
+	punpcklqdq xmm7, xmm0
+	aesenc	 xmm6, xmm7
+	movq	 rbp, xmm6
+	mov	 r9, rbp
+	and	 r9d, 2097136
+	movdqu	 xmm2, XMMWORD PTR [rcx+rbx]
+	movdqu	 xmm1, XMMWORD PTR [rax+rbx]
+	movdqu	 xmm0, XMMWORD PTR [r10+rbx]
+	paddq	 xmm1, xmm7
+	paddq	 xmm0, xmm5
+	paddq	 xmm2, xmm4
+	movdqu	 XMMWORD PTR [rcx+rbx], xmm0
+	movdqu	 XMMWORD PTR [rax+rbx], xmm2
+	movdqu	 XMMWORD PTR [r10+rbx], xmm1
+	mov r10, r9
+	xor r10d, 32
+	movq	 rcx, xmm3
+	mov	 rax, rcx
+	shl	 rax, 32
+	xor	 rdi, rax
+	movdqa	 xmm0, xmm6
+	pxor	 xmm0, xmm4
+	movdqu	 XMMWORD PTR [rdx], xmm0
+	xor	 rdi, QWORD PTR [r9+rbx]
+	lea	 r14, QWORD PTR [r9+rbx]
+	mov	 r12, QWORD PTR [r14+8]
+	xor	 edx, edx
+	lea	 r9d, DWORD PTR [ecx+ecx]
+	add	 r9d, ebp
+	movdqa	 xmm0, xmm6
+	psrldq	 xmm0, 8
+	or	 r9d, r13d
+	movq	 rax, xmm0
+	div	 r9
+	xorps xmm3, xmm3
+	mov	 eax, eax
+	shl	 rdx, 32
+	add	 rdx, rax
+	lea	 r9, QWORD PTR [rdx+rbp]
+	mov r15, rdx
+	mov	 rax, r9
+	shr	 rax, 12
+	movq	 xmm0, rax
+	paddq	 xmm0, xmm8
+	sqrtsd	 xmm3, xmm0
+	movq	 rdx, xmm3
+	test	 edx, 524287
+	je	 $sqrt_fixup_ivybridge
+	psrlq	 xmm3, 19
+$sqrt_fixup_ivybridge_ret:
+
+	mov	 ecx, r10d
+	mov	 rax, rdi
+	mul	 rbp
+	movq xmm2, rdx
+	xor rdx, [rcx+rbx]
+	add	 r8, rdx
+	mov	 QWORD PTR [r14], r8
+	xor	 r8, rdi
+	mov edi, r8d
+	and edi, 2097136
+	movq xmm0, rax
+	xor rax, [rcx+rbx+8]
+	add	 r11, rax
+	mov	 QWORD PTR [r14+8], r11
+	punpcklqdq xmm2, xmm0
+
+	mov	 r9d, r10d
+	xor	 r9d, 48
+	xor	 r10d, 16
+	pxor	 xmm2, XMMWORD PTR [r9+rbx]
+	movdqu	 xmm0, XMMWORD PTR [r10+rbx]
+	paddq	 xmm0, xmm5
+	movdqu	 xmm1, XMMWORD PTR [rcx+rbx]
+	paddq	 xmm2, xmm4
+	paddq	 xmm1, xmm7
+	movdqa	 xmm5, xmm4
+	movdqu	 XMMWORD PTR [r9+rbx], xmm0
+	movdqa	 xmm4, xmm6
+	movdqu	 XMMWORD PTR [rcx+rbx], xmm2
+	movdqu	 XMMWORD PTR [r10+rbx], xmm1
+	movdqu xmm6, [rdi+rbx]
+	mov	 r10d, edi
+	xor	 r11, r12
+	dec rsi
+	jne	 $main_loop_ivybridge
+
+	ldmxcsr DWORD PTR [rsp]
+	mov	 rbx, QWORD PTR [rsp+160]
+	movaps	 xmm6, XMMWORD PTR [rsp+64]
+	movaps	 xmm7, XMMWORD PTR [rsp+48]
+	movaps	 xmm8, XMMWORD PTR [rsp+32]
+	add	 rsp, 80
+	pop	 r15
+	pop	 r14
+	pop	 r13
+	pop	 r12
+	pop	 rdi
+	pop	 rsi
+	pop	 rbp
+	jmp $cnv2_main_loop_ivybridge_endp
+
+$sqrt_fixup_ivybridge:
+	dec	 rdx
+	mov r13d, -1022
+	shl r13, 32
+	mov	 rax, rdx
+	shr	 rdx, 19
+	shr	 rax, 20
+	mov	 rcx, rdx
+	sub	 rcx, rax
+	add	 rax, r13
+	not r13
+	sub	 rcx, r13
+	mov	 r13d, -2147483647
+	imul	 rcx, rax
+	sub	 rcx, r9
+	adc	 rdx, 0
+	movq	 xmm3, rdx
+	jmp	 $sqrt_fixup_ivybridge_ret
+
+$cnv2_main_loop_ivybridge_endp: