Land BoringSSL roll on master

third_party/boringssl/mac-x86_64/crypto/bn/x86_64-mont.S

Index: third_party/boringssl/mac-x86_64/crypto/bn/x86_64-mont.S
diff --git a/third_party/boringssl/mac-x86_64/crypto/bn/x86_64-mont.S b/third_party/boringssl/mac-x86_64/crypto/bn/x86_64-mont.S
index 6b9bc05bf4e2f96c68a3e97315d4d6596428b4f9..51e5d19931f0545ef6d41e790a0403ba68058a20 100644
--- a/third_party/boringssl/mac-x86_64/crypto/bn/x86_64-mont.S
+++ b/third_party/boringssl/mac-x86_64/crypto/bn/x86_64-mont.S
@@ -634,20 +634,20 @@ L$sqr8x_enter:
 
 
 
-	leaq	-64(%rsp,%r9,4),%r11
+	leaq	-64(%rsp,%r9,2),%r11
 	movq	(%r8),%r8
 	subq	%rsi,%r11
 	andq	$4095,%r11
 	cmpq	%r11,%r10
 	jb	L$sqr8x_sp_alt
 	subq	%r11,%rsp
-	leaq	-64(%rsp,%r9,4),%rsp
+	leaq	-64(%rsp,%r9,2),%rsp
 	jmp	L$sqr8x_sp_done
 
 .p2align	5
 L$sqr8x_sp_alt:
-	leaq	4096-64(,%r9,4),%r10
-	leaq	-64(%rsp,%r9,4),%rsp
+	leaq	4096-64(,%r9,2),%r10
+	leaq	-64(%rsp,%r9,2),%rsp
 	subq	%r10,%r11
 	movq	$0,%r10
 	cmovcq	%r10,%r11
@@ -657,58 +657,80 @@ L$sqr8x_sp_done:
 	movq	%r9,%r10
 	negq	%r9
 
-	leaq	64(%rsp,%r9,2),%r11
 	movq	%r8,32(%rsp)
 	movq	%rax,40(%rsp)
 L$sqr8x_body:
 
-	movq	%r9,%rbp
-.byte	102,73,15,110,211
-	shrq	$3+2,%rbp
-	movl	_OPENSSL_ia32cap_P+8(%rip),%eax
-	jmp	L$sqr8x_copy_n
-
-.p2align	5
-L$sqr8x_copy_n:
-	movq	0(%rcx),%xmm0
-	movq	8(%rcx),%xmm1
-	movq	16(%rcx),%xmm3
-	movq	24(%rcx),%xmm4
-	leaq	32(%rcx),%rcx
-	movdqa	%xmm0,0(%r11)
-	movdqa	%xmm1,16(%r11)
-	movdqa	%xmm3,32(%r11)
-	movdqa	%xmm4,48(%r11)
-	leaq	64(%r11),%r11
-	decq	%rbp
-	jnz	L$sqr8x_copy_n
-
+.byte	102,72,15,110,209
 	pxor	%xmm0,%xmm0
 .byte	102,72,15,110,207
 .byte	102,73,15,110,218
 	call	_bn_sqr8x_internal
 
+
+
+
+	leaq	(%rdi,%r9,1),%rbx
+	movq	%r9,%rcx
+	movq	%r9,%rdx
+.byte	102,72,15,126,207
+	sarq	$3+2,%rcx
+	jmp	L$sqr8x_sub
+
+.p2align	5
+L$sqr8x_sub:
+	movq	0(%rbx),%r12
+	movq	8(%rbx),%r13
+	movq	16(%rbx),%r14
+	movq	24(%rbx),%r15
+	leaq	32(%rbx),%rbx
+	sbbq	0(%rbp),%r12
+	sbbq	8(%rbp),%r13
+	sbbq	16(%rbp),%r14
+	sbbq	24(%rbp),%r15
+	leaq	32(%rbp),%rbp
+	movq	%r12,0(%rdi)
+	movq	%r13,8(%rdi)
+	movq	%r14,16(%rdi)
+	movq	%r15,24(%rdi)
+	leaq	32(%rdi),%rdi
+	incq	%rcx
+	jnz	L$sqr8x_sub
+
+	sbbq	$0,%rax
+	leaq	(%rbx,%r9,1),%rbx
+	leaq	(%rdi,%r9,1),%rdi
+
+.byte	102,72,15,110,200
 	pxor	%xmm0,%xmm0
-	leaq	48(%rsp),%rax
-	leaq	64(%rsp,%r9,2),%rdx
-	shrq	$3+2,%r9
+	pshufd	$0,%xmm1,%xmm1
 	movq	40(%rsp),%rsi
-	jmp	L$sqr8x_zero
+	jmp	L$sqr8x_cond_copy
 
 .p2align	5
-L$sqr8x_zero:
-	movdqa	%xmm0,0(%rax)
-	movdqa	%xmm0,16(%rax)
-	movdqa	%xmm0,32(%rax)
-	movdqa	%xmm0,48(%rax)
-	leaq	64(%rax),%rax
-	movdqa	%xmm0,0(%rdx)
-	movdqa	%xmm0,16(%rdx)
-	movdqa	%xmm0,32(%rdx)
-	movdqa	%xmm0,48(%rdx)
-	leaq	64(%rdx),%rdx
-	decq	%r9
-	jnz	L$sqr8x_zero
+L$sqr8x_cond_copy:
+	movdqa	0(%rbx),%xmm2
+	movdqa	16(%rbx),%xmm3
+	leaq	32(%rbx),%rbx
+	movdqu	0(%rdi),%xmm4
+	movdqu	16(%rdi),%xmm5
+	leaq	32(%rdi),%rdi
+	movdqa	%xmm0,-32(%rbx)
+	movdqa	%xmm0,-16(%rbx)
+	movdqa	%xmm0,-32(%rbx,%rdx,1)
+	movdqa	%xmm0,-16(%rbx,%rdx,1)
+	pcmpeqd	%xmm1,%xmm0
+	pand	%xmm1,%xmm2
+	pand	%xmm1,%xmm3
+	pand	%xmm0,%xmm4
+	pand	%xmm0,%xmm5
+	pxor	%xmm0,%xmm0
+	por	%xmm2,%xmm4
+	por	%xmm3,%xmm5
+	movdqu	%xmm4,-32(%rdi)
+	movdqu	%xmm5,-16(%rdi)
+	addq	$32,%r9
+	jnz	L$sqr8x_cond_copy
 
 	movq	$1,%rax
 	movq	-48(%rsi),%r15