Home | History | Annotate | Line # | Download | only in x86_64 
      1 dnl  AMD64 mpn_sqr_diag_addlsh1
      2 
      3 dnl  Contributed to the GNU project by Torbjrn Granlund.
      4 
      5 dnl  Copyright 2011-2013 Free Software Foundation, Inc.
      6 
      7 dnl  This file is part of the GNU MP Library.
      8 dnl
      9 dnl  The GNU MP Library is free software; you can redistribute it and/or modify
     10 dnl  it under the terms of either:
     11 dnl
     12 dnl    * the GNU Lesser General Public License as published by the Free
     13 dnl      Software Foundation; either version 3 of the License, or (at your
     14 dnl      option) any later version.
     15 dnl
     16 dnl  or
     17 dnl
     18 dnl    * the GNU General Public License as published by the Free Software
     19 dnl      Foundation; either version 2 of the License, or (at your option) any
     20 dnl      later version.
     21 dnl
     22 dnl  or both in parallel, as here.
     23 dnl
     24 dnl  The GNU MP Library is distributed in the hope that it will be useful, but
     25 dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
     26 dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
     27 dnl  for more details.
     28 dnl
     29 dnl  You should have received copies of the GNU General Public License and the
     30 dnl  GNU Lesser General Public License along with the GNU MP Library.  If not,
     31 dnl  see https://www.gnu.org/licenses/.
     32 
     33 include(`../config.m4')
     34 
     35 C	     cycles/limb
     36 C AMD K8,K9	 2.5
     37 C AMD K10	 2.5
     38 C AMD bull	 3.6
     39 C AMD pile	 3.6
     40 C AMD steam	 ?
     41 C AMD bobcat	 4
     42 C AMD jaguar	 ?
     43 C Intel P4	 11.5
     44 C Intel core	 4
     45 C Intel NHM	 3.6
     46 C Intel SBR	 3.15
     47 C Intel IBR	 3.0
     48 C Intel HWL	 2.6
     49 C Intel BWL	 ?
     50 C Intel atom	14
     51 C VIA nano	 3.5
     52 
     53 C When playing with pointers, set this to $2 to fall back to conservative
     54 C indexing in wind-down code.
     55 define(`I',`$1')
     56 
     57 define(`rp',     `%rdi')
     58 define(`tp',     `%rsi')
     59 define(`up_arg', `%rdx')
     60 define(`n',      `%rcx')
     61 
     62 define(`up',     `%r11')
     63 
     64 ABI_SUPPORT(DOS64)
     65 ABI_SUPPORT(STD64)
     66 
     67 ASM_START()
     68 	TEXT
     69 	ALIGN(32)
     70 PROLOGUE(mpn_sqr_diag_addlsh1)
     71 	FUNC_ENTRY(4)
     72 	push	%rbx
     73 
     74 	dec	n
     75 	shl	n
     76 
     77 	mov	(up_arg), %rax
     78 
     79 	lea	(rp,n,8), rp
     80 	lea	(tp,n,8), tp
     81 	lea	(up_arg,n,4), up
     82 	neg	n
     83 
     84 	mul	%rax
     85 	mov	%rax, (rp,n,8)
     86 
     87 	xor	R32(%rbx), R32(%rbx)
     88 	jmp	L(mid)
     89 
     90 	ALIGN(16)
     91 L(top):	add	%r10, %r8
     92 	adc	%rax, %r9
     93 	mov	%r8, -8(rp,n,8)
     94 	mov	%r9, (rp,n,8)
     95 L(mid):	mov	8(up,n,4), %rax
     96 	mov	(tp,n,8), %r8
     97 	mov	8(tp,n,8), %r9
     98 	adc	%r8, %r8
     99 	adc	%r9, %r9
    100 	lea	(%rdx,%rbx), %r10
    101 	setc	R8(%rbx)
    102 	mul	%rax
    103 	add	$2, n
    104 	js	L(top)
    105 
    106 L(end):	add	%r10, %r8
    107 	adc	%rax, %r9
    108 	mov	%r8, I(-8(rp),-8(rp,n,8))
    109 	mov	%r9, I((rp),(rp,n,8))
    110 	adc	%rbx, %rdx
    111 	mov	%rdx, I(8(rp),8(rp,n,8))
    112 
    113 	pop	%rbx
    114 	FUNC_EXIT()
    115 	ret
    116 EPILOGUE()
    117