Update x86-64 mpn routines from GMP 5.0.1.

1 files changed, 92 insertions, 23 deletions

diff --git a/sysdeps/x86_64/addmul_1.S b/sysdeps/x86_64/addmul_1.S
index bdb5226..e997896 100644
--- a/sysdeps/x86_64/addmul_1.S
+++ b/sysdeps/x86_64/addmul_1.S
@@ -1,6 +1,6 @@
-/* AMD64 __mpn_addmul_1 -- Multiply a limb vector with a limb and add
+/* x86-64 __mpn_addmul_1 -- Multiply a limb vector with a limb and add
    the result to a second limb vector.
-   Copyright (C) 2004 Free Software Foundation, Inc.
+   Copyright (C) 2003,2004,2005,2007,2008,2009 Free Software Foundation, Inc.
    This file is part of the GNU MP Library.
 
    The GNU MP Library is free software; you can redistribute it and/or modify
@@ -21,26 +21,95 @@
 #include "sysdep.h"
 #include "asm-syntax.h"
 
+#define rp	%rdi
+#define up	%rsi
+#define n	%rdx
+#define v0	%rcx
+
+#ifndef func
+# define func __mpn_addmul_1
+# define ADDSUB add
+#endif
+
 	.text
-ENTRY (__mpn_addmul_1)
-	movq	%rdx, %r11
-	leaq	(%rsi,%rdx,8), %rsi
-	leaq	(%rdi,%rdx,8), %rdi
-	negq	%r11
-	xorl	%r8d, %r8d
-	xorl	%r10d, %r10d
-	.p2align 2
-L(loop):
-	movq	(%rsi,%r11,8), %rax
-	mulq	%rcx
-	addq	(%rdi,%r11,8), %rax
-	adcq	%r10, %rdx
-	addq	%r8, %rax
-	movq	%r10, %r8
-	movq	%rax, (%rdi,%r11,8)
-	adcq	%rdx, %r8
-	incq	%r11
-	jne	L(loop)
-	movq	%r8, %rax
+ENTRY (func)
+	push	%rbx
+	push	%rbp
+	lea	(%rdx), %rbx
+	neg	%rbx
+
+	mov	(up), %rax
+	mov	(rp), %r10
+
+	lea	-16(rp,%rdx,8), rp
+	lea	(up,%rdx,8), up
+	mul	%rcx
+
+	bt	$0, %ebx
+	jc	L(odd)
+
+	lea	(%rax), %r11
+	mov	8(up,%rbx,8), %rax
+	lea	(%rdx), %rbp
+	mul	%rcx
+	add	$2, %rbx
+	jns	L(n2)
+
+	lea	(%rax), %r8
+	mov	(up,%rbx,8), %rax
+	lea	(%rdx), %r9
+	jmp	L(mid)
+
+L(odd):	add	$1, %rbx
+	jns	L(n1)
+
+	lea	(%rax), %r8
+	mov	(up,%rbx,8), %rax
+	lea	(%rdx), %r9
+	mul	%rcx
+	lea	(%rax), %r11
+	mov	8(up,%rbx,8), %rax
+	lea	(%rdx), %rbp
+	jmp	L(e)
+
+	.p2align 4
+L(top):	mul	%rcx
+	ADDSUB	%r8, %r10
+	lea	(%rax), %r8
+	mov	(up,%rbx,8), %rax
+	adc	%r9, %r11
+	mov	%r10, -8(rp,%rbx,8)
+	mov	(rp,%rbx,8), %r10
+	lea	(%rdx), %r9
+	adc	$0, %rbp
+L(mid):	mul	%rcx
+	ADDSUB	%r11, %r10
+	lea	(%rax), %r11
+	mov	8(up,%rbx,8), %rax
+	adc	%rbp, %r8
+	mov	%r10, (rp,%rbx,8)
+	mov	8(rp,%rbx,8), %r10
+	lea	(%rdx), %rbp
+	adc	$0, %r9
+L(e):	add	$2, %rbx
+	js	L(top)
+
+	mul	%rcx
+	ADDSUB	%r8, %r10
+	adc	%r9, %r11
+	mov	%r10, -8(rp)
+	adc	$0, %rbp
+L(n2):	mov	(rp), %r10
+	ADDSUB	%r11, %r10
+	adc	%rbp, %rax
+	mov	%r10, (rp)
+	adc	$0, %rdx
+L(n1):	mov	8(rp), %r10
+	ADDSUB	%rax, %r10
+	mov	%r10, 8(rp)
+	mov	%ebx, %eax	/* zero rax */
+	adc	%rdx, %rax
+	pop	%rbp
+	pop	%rbx
 	ret
-END (__mpn_addmul_1)
+END (func)