mpir/mpn/x86/nehalem/mul_basecase.asm
2009-09-27 14:54:29 +00:00

165 lines
3.5 KiB
NASM

dnl Intel Pentium-4 mpn_mul_basecase -- mpn by mpn multiplication.
dnl Copyright 2001, 2002 Free Software Foundation, Inc.
dnl
dnl This file is part of the GNU MP Library.
dnl
dnl The GNU MP Library is free software; you can redistribute it and/or
dnl modify it under the terms of the GNU Lesser General Public License as
dnl published by the Free Software Foundation; either version 2.1 of the
dnl License, or (at your option) any later version.
dnl
dnl The GNU MP Library is distributed in the hope that it will be useful,
dnl but WITHOUT ANY WARRANTY; without even the implied warranty of
dnl MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
dnl Lesser General Public License for more details.
dnl
dnl You should have received a copy of the GNU Lesser General Public
dnl License along with the GNU MP Library; see the file COPYING.LIB. If
dnl not, write to the Free Software Foundation, Inc., 51 Franklin Street,
dnl Fifth Floor, Boston, MA 02110-1301, USA.
include(`../config.m4')
C P4: 6.0 cycles/crossproduct (approx)
C void mpn_mul_basecase (mp_ptr wp,
C mp_srcptr xp, mp_size_t xsize,
C mp_srcptr yp, mp_size_t ysize);
C
C Nothing special here, basically just mpn/generic/mul_basecase.c done with
C mpn_mul_1 and mpn_addmul_1 inline. As per mpn_addmul_1, the dependent
C chain in the inner loop is 4 c/l, but measures about 6.
C
C Enhancements:
C
C Perhaps some sort of vertical method would suit, though there'd be branch
C mispredictions on the end sections. But it's not clear how to get less
C than 4 instructions per crossproduct, and unless that can be done then a
C basic addmul_1 style may as well be used (assuming it can be brought up to
C its proper 4 c/l).
defframe(PARAM_YSIZE, 20)
defframe(PARAM_YP, 16)
defframe(PARAM_XSIZE, 12)
defframe(PARAM_XP, 8)
defframe(PARAM_WP, 4)
define(SAVE_EBX,`PARAM_XP')
define(SAVE_ESI,`PARAM_YP')
define(SAVE_EDI,`PARAM_YSIZE')
define(SAVE_EBP,`PARAM_WP')
TEXT
ALIGN(8)
PROLOGUE(mpn_mul_basecase)
deflit(`FRAME',0)
movl PARAM_XP, %eax
movl %ebx, SAVE_EBX
pxor %mm0, %mm0 C initial carry
movl PARAM_YP, %edx
movl %esi, SAVE_ESI
movl PARAM_WP, %ebx
movl %ebp, SAVE_EBP
movl %eax, %esi C xp
movd (%edx), %mm7 C yp[0]
movl PARAM_XSIZE, %ecx
movl PARAM_YSIZE, %ebp
movl %edi, SAVE_EDI
movl %ebx, %edi C wp
L(mul1):
C eax xp, incrementing
C ebx wp, incrementing
C ecx xsize, decrementing
C edx yp
C esi xp
C edi wp
C ebp ysize
C
C mm0 carry limb
C mm7 multiplier
movd (%eax), %mm1
addl $4, %eax
pmuludq %mm7, %mm1
paddq %mm1, %mm0
movd %mm0, (%ebx)
addl $4, %ebx
psrlq $32, %mm0
subl $1, %ecx
jnz L(mul1)
movd %mm0, (%ebx)
subl $1, %ebp
jz L(done)
L(outer):
C eax
C ebx
C ecx
C edx yp, incrementing
C esi xp
C edi wp, incrementing
C ebp ysize, decrementing
movl %esi, %eax C xp
leal 4(%edi), %ebx C next wp
addl $4, %edi
movd 4(%edx), %mm7 C next yp limb
addl $4, %edx
pxor %mm0, %mm0 C initial carry
movl PARAM_XSIZE, %ecx
L(inner):
C eax xp, incrementing
C ebx wp, incrementing
C ecx xsize, decrementing
C edx outer yp
C esi outer xp
C edi outer wp
C ebp outer ysize
movd (%eax), %mm1
leal 4(%eax), %eax
movd (%ebx),%mm2
pmuludq %mm7, %mm1
paddq %mm2, %mm1
paddq %mm1, %mm0
subl $1, %ecx
movd %mm0, (%ebx)
psrlq $32, %mm0
leal 4(%ebx), %ebx
jnz L(inner)
movd %mm0, (%ebx)
subl $1, %ebp
jnz L(outer)
L(done):
movl SAVE_EBX, %ebx
movl SAVE_ESI, %esi
movl SAVE_EDI, %edi
movl SAVE_EBP, %ebp
emms
ret
EPILOGUE()