/*********************************************************************/
/* Copyright 2009, 2010 The University of Texas at Austin. */
/* All rights reserved. */
/* */
/* Redistribution and use in source and binary forms, with or */
/* without modification, are permitted provided that the following */
/* conditions are met: */
/* */
/* 1. Redistributions of source code must retain the above */
/* copyright notice, this list of conditions and the following */
/* disclaimer. */
/* */
/* 2. Redistributions in binary form must reproduce the above */
/* copyright notice, this list of conditions and the following */
/* disclaimer in the documentation and/or other materials */
/* provided with the distribution. */
/* */
/* THIS SOFTWARE IS PROVIDED BY THE UNIVERSITY OF TEXAS AT */
/* AUSTIN ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, */
/* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */
/* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */
/* DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY OF TEXAS AT */
/* AUSTIN OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, */
/* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES */
/* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE */
/* GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR */
/* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF */
/* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT */
/* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT */
/* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */
/* POSSIBILITY OF SUCH DAMAGE. */
/* */
/* The views and conclusions contained in the software and */
/* documentation are those of the authors and should not be */
/* interpreted as representing official policies, either expressed */
/* or implied, of The University of Texas at Austin. */
/*********************************************************************/
#define ASSEMBLER
#include "common.h"
#define STACK 12
#define ARGS 0
#define STACK_N 4 + STACK + ARGS(%esp)
#define STACK_X 8 + STACK + ARGS(%esp)
#define STACK_INCX 12 + STACK + ARGS(%esp)
#define STACK_Y 16 + STACK + ARGS(%esp)
#define STACK_INCY 20 + STACK + ARGS(%esp)
#define N %ecx
#define X %esi
#define INCX %ebx
#define Y %edi
#define INCY %edx
#include "l1param.h"
PROLOGUE
PROFCODE
pushl %edi
pushl %esi
pushl %ebx
movl STACK_N, N
movl STACK_X, X
movl STACK_INCX, INCX
movl STACK_Y, Y
movl STACK_INCY, INCY
#ifdef F_INTERFACE
movl (N), N # N
movl (INCX),INCX # INCX
movl (INCY),INCY # INCY
#endif
leal (, INCX, SIZE), INCX
leal (, INCY, SIZE), INCY
xorps %xmm0, %xmm0
xorps %xmm1, %xmm1
xorps %xmm2, %xmm2
xorps %xmm3, %xmm3
cmpl $0, N
jle .L999
cmpl $SIZE, INCX
jne .L50
cmpl $SIZE, INCY
jne .L50
subl $-32 * SIZE, X
subl $-32 * SIZE, Y
cmpl $3, N
jle .L17
testl $SIZE, Y
je .L05
movss -32 * SIZE(X), %xmm0
mulss -32 * SIZE(Y), %xmm0
addl $1 * SIZE, X
addl $1 * SIZE, Y
decl N
ALIGN_2
.L05:
testl $2 * SIZE, Y
je .L10
#ifdef movsd
xorps %xmm4, %xmm4
#endif
movsd -32 * SIZE(X), %xmm4
#ifdef movsd
xorps %xmm1, %xmm1
#endif
movsd -32 * SIZE(Y), %xmm1
mulps %xmm4, %xmm1
addl $2 * SIZE, X
addl $2 * SIZE, Y
subl $2, N
jle .L999
ALIGN_2
.L10:
#ifdef ALIGNED_ACCESS
testl $2 * SIZE, X
jne .L30
testl $SIZE, X
jne .L20
#else
testl $3 * SIZE, X
jne .L20
#endif
movl N, %eax
sarl $5, %eax
jle .L14
movaps -32 * SIZE(X), %xmm4
movaps -28 * SIZE(X), %xmm5
movaps -24 * SIZE(X), %xmm6
movaps -20 * SIZE(X), %xmm7
decl %eax
jle .L12
ALIGN_3
.L11:
#ifdef PREFETCH
PREFETCH (PREFETCHSIZE + 0) - PREOFFSET(X)
#endif
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -16 * SIZE(X), %xmm4
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps -12 * SIZE(X), %xmm5
#ifdef PREFETCH
PREFETCH (PREFETCHSIZE + 0) - PREOFFSET(Y)
#endif
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps -8 * SIZE(X), %xmm6
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps -4 * SIZE(X), %xmm7
#if defined(PREFETCH) && !defined(FETCH128)
PREFETCH (PREFETCHSIZE + 64) - PREOFFSET(X)
#endif
mulps -16 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps 0 * SIZE(X), %xmm4
mulps -12 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps 4 * SIZE(X), %xmm5
#if defined(PREFETCH) && !defined(FETCH128)
PREFETCH (PREFETCHSIZE + 64) - PREOFFSET(Y)
#endif
mulps -8 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps 8 * SIZE(X), %xmm6
mulps -4 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps 12 * SIZE(X), %xmm7
subl $-32 * SIZE, X
subl $-32 * SIZE, Y
decl %eax
jg .L11
ALIGN_3
.L12:
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -16 * SIZE(X), %xmm4
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps -12 * SIZE(X), %xmm5
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps -8 * SIZE(X), %xmm6
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps -4 * SIZE(X), %xmm7
mulps -16 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
mulps -12 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
mulps -8 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
mulps -4 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
subl $-32 * SIZE, X
subl $-32 * SIZE, Y
ALIGN_3
.L14:
testl $31, N
jle .L999
testl $16, N
jle .L15
movaps -32 * SIZE(X), %xmm4
movaps -28 * SIZE(X), %xmm5
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps -24 * SIZE(X), %xmm6
movaps -20 * SIZE(X), %xmm7
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
addl $16 * SIZE, X
addl $16 * SIZE, Y
ALIGN_3
.L15:
testl $8, N
jle .L16
movaps -32 * SIZE(X), %xmm4
movaps -28 * SIZE(X), %xmm5
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
addl $8 * SIZE, X
addl $8 * SIZE, Y
ALIGN_3
.L16:
testl $4, N
jle .L17
movaps -32 * SIZE(X), %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm2
addl $4 * SIZE, X
addl $4 * SIZE, Y
ALIGN_3
.L17:
testl $2, N
jle .L18
#ifdef movsd
xorps %xmm4, %xmm4
#endif
movsd -32 * SIZE(X), %xmm4
#ifdef movsd
xorps %xmm6, %xmm6
#endif
movsd -32 * SIZE(Y), %xmm6
mulps %xmm6, %xmm4
addps %xmm4, %xmm3
addl $2 * SIZE, X
addl $2 * SIZE, Y
ALIGN_3
.L18:
testl $1, N
jle .L999
movss -32 * SIZE(X), %xmm4
mulss -32 * SIZE(Y), %xmm4
addss %xmm4, %xmm0
jmp .L999
ALIGN_3
#ifdef ALIGNED_ACCESS
.L20:
movaps -33 * SIZE(X), %xmm4
addl $3 * SIZE, X
movl N, %eax
sarl $5, %eax
jle .L24
movaps -32 * SIZE(X), %xmm5
movaps -28 * SIZE(X), %xmm6
movaps -24 * SIZE(X), %xmm7
decl %eax
jle .L22
ALIGN_3
.L21:
#ifdef PREFETCH
PREFETCH (PREFETCHSIZE + 0) - PREOFFSET(X)
#endif
movss %xmm5, %xmm4
PSHUFD1($0x39, %xmm4)
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -20 * SIZE(X), %xmm4
movss %xmm6, %xmm5
PSHUFD1($0x39, %xmm5)
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps -16 * SIZE(X), %xmm5
#ifdef PREFETCH
PREFETCH (PREFETCHSIZE + 0) - PREOFFSET(Y)
#endif
movss %xmm7, %xmm6
PSHUFD1($0x39, %xmm6)
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps -12 * SIZE(X), %xmm6
movss %xmm4, %xmm7
PSHUFD1($0x39, %xmm7)
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps -8 * SIZE(X), %xmm7
#if defined(PREFETCH) && !defined(FETCH128)
PREFETCH (PREFETCHSIZE + 64) - PREOFFSET(X)
#endif
movss %xmm5, %xmm4
PSHUFD1($0x39, %xmm4)
mulps -16 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -4 * SIZE(X), %xmm4
movss %xmm6, %xmm5
PSHUFD1($0x39, %xmm5)
mulps -12 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps 0 * SIZE(X), %xmm5
#if defined(PREFETCH) && !defined(FETCH128)
PREFETCH (PREFETCHSIZE + 64) - PREOFFSET(Y)
#endif
movss %xmm7, %xmm6
PSHUFD1($0x39, %xmm6)
mulps -8 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps 4 * SIZE(X), %xmm6
movss %xmm4, %xmm7
PSHUFD1($0x39, %xmm7)
mulps -4 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps 8 * SIZE(X), %xmm7
subl $-32 * SIZE, X
subl $-32 * SIZE, Y
decl %eax
jg .L21
ALIGN_3
.L22:
movss %xmm5, %xmm4
PSHUFD1($0x39, %xmm4)
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -20 * SIZE(X), %xmm4
movss %xmm6, %xmm5
PSHUFD1($0x39, %xmm5)
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps -16 * SIZE(X), %xmm5
movss %xmm7, %xmm6
PSHUFD1($0x39, %xmm6)
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps -12 * SIZE(X), %xmm6
movss %xmm4, %xmm7
PSHUFD1($0x39, %xmm7)
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps -8 * SIZE(X), %xmm7
movss %xmm5, %xmm4
PSHUFD1($0x39, %xmm4)
mulps -16 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -4 * SIZE(X), %xmm4
movss %xmm6, %xmm5
PSHUFD1($0x39, %xmm5)
mulps -12 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movss %xmm7, %xmm6
PSHUFD1($0x39, %xmm6)
mulps -8 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movss %xmm4, %xmm7
PSHUFD1($0x39, %xmm7)
mulps -4 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
subl $-32 * SIZE, X
subl $-32 * SIZE, Y
ALIGN_3
.L24:
testl $31, N
jle .L999
testl $16, N
jle .L25
movaps -32 * SIZE(X), %xmm5
movaps -28 * SIZE(X), %xmm6
movaps -24 * SIZE(X), %xmm7
movss %xmm5, %xmm4
PSHUFD1($0x39, %xmm4)
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -20 * SIZE(X), %xmm4
movss %xmm6, %xmm5
PSHUFD1($0x39, %xmm5)
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movss %xmm7, %xmm6
PSHUFD1($0x39, %xmm6)
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movss %xmm4, %xmm7
PSHUFD1($0x39, %xmm7)
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
addl $16 * SIZE, X
addl $16 * SIZE, Y
ALIGN_3
.L25:
testl $8, N
jle .L26
movaps -32 * SIZE(X), %xmm5
movaps -28 * SIZE(X), %xmm6
movss %xmm5, %xmm4
PSHUFD1($0x39, %xmm4)
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movss %xmm6, %xmm5
PSHUFD1($0x39, %xmm5)
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps %xmm6, %xmm4
addl $8 * SIZE, X
addl $8 * SIZE, Y
ALIGN_3
.L26:
testl $4, N
jle .L27
movaps -32 * SIZE(X), %xmm5
movss %xmm5, %xmm4
PSHUFD1($0x39, %xmm4)
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm2
movaps %xmm5, %xmm4
addl $4 * SIZE, X
addl $4 * SIZE, Y
ALIGN_3
.L27:
testl $2, N
jle .L28
#ifdef movsd
xorps %xmm6, %xmm6
#endif
movsd -32 * SIZE(Y), %xmm6
PSHUFD2($0x39, %xmm4, %xmm5)
mulps %xmm6, %xmm5
addps %xmm5, %xmm3
movhlps %xmm4, %xmm4
addl $2 * SIZE, X
addl $2 * SIZE, Y
ALIGN_3
.L28:
testl $1, N
jle .L999
PSHUFD1($0x39, %xmm4)
mulss -32 * SIZE(Y), %xmm4
addss %xmm4, %xmm0
jmp .L999
ALIGN_3
.L30:
testl $SIZE, X
jne .L40
movhps -32 * SIZE(X), %xmm4
addl $2 * SIZE, X
movl N, %eax
sarl $5, %eax
jle .L34
movaps -32 * SIZE(X), %xmm5
movaps -28 * SIZE(X), %xmm6
movaps -24 * SIZE(X), %xmm7
decl %eax
jle .L32
ALIGN_3
.L31:
#ifdef PREFETCH
PREFETCH (PREFETCHSIZE + 0) - PREOFFSET(X)
#endif
SHUFPD_1 %xmm5, %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -20 * SIZE(X), %xmm4
SHUFPD_1 %xmm6, %xmm5
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps -16 * SIZE(X), %xmm5
#ifdef PREFETCH
PREFETCH (PREFETCHSIZE + 0) - PREOFFSET(Y)
#endif
SHUFPD_1 %xmm7, %xmm6
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps -12 * SIZE(X), %xmm6
SHUFPD_1 %xmm4, %xmm7
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps -8 * SIZE(X), %xmm7
#if defined(PREFETCH) && !defined(FETCH128)
PREFETCH (PREFETCHSIZE + 64) - PREOFFSET(X)
#endif
SHUFPD_1 %xmm5, %xmm4
mulps -16 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -4 * SIZE(X), %xmm4
SHUFPD_1 %xmm6, %xmm5
mulps -12 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps 0 * SIZE(X), %xmm5
#if defined(PREFETCH) && !defined(FETCH128)
PREFETCH (PREFETCHSIZE + 64) - PREOFFSET(Y)
#endif
SHUFPD_1 %xmm7, %xmm6
mulps -8 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps 4 * SIZE(X), %xmm6
SHUFPD_1 %xmm4, %xmm7
mulps -4 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps 8 * SIZE(X), %xmm7
subl $-32 * SIZE, X
subl $-32 * SIZE, Y
decl %eax
jg .L31
ALIGN_3
.L32:
SHUFPD_1 %xmm5, %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -20 * SIZE(X), %xmm4
SHUFPD_1 %xmm6, %xmm5
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps -16 * SIZE(X), %xmm5
SHUFPD_1 %xmm7, %xmm6
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps -12 * SIZE(X), %xmm6
SHUFPD_1 %xmm4, %xmm7
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps -8 * SIZE(X), %xmm7
SHUFPD_1 %xmm5, %xmm4
mulps -16 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -4 * SIZE(X), %xmm4
SHUFPD_1 %xmm6, %xmm5
mulps -12 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
SHUFPD_1 %xmm7, %xmm6
mulps -8 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
SHUFPD_1 %xmm4, %xmm7
mulps -4 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
subl $-32 * SIZE, X
subl $-32 * SIZE, Y
ALIGN_3
.L34:
testl $31, N
jle .L999
testl $16, N
jle .L35
movaps -32 * SIZE(X), %xmm5
movaps -28 * SIZE(X), %xmm6
movaps -24 * SIZE(X), %xmm7
SHUFPD_1 %xmm5, %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -20 * SIZE(X), %xmm4
SHUFPD_1 %xmm6, %xmm5
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
SHUFPD_1 %xmm7, %xmm6
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
SHUFPD_1 %xmm4, %xmm7
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
addl $16 * SIZE, X
addl $16 * SIZE, Y
ALIGN_3
.L35:
testl $8, N
jle .L36
movaps -32 * SIZE(X), %xmm5
movaps -28 * SIZE(X), %xmm6
SHUFPD_1 %xmm5, %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
SHUFPD_1 %xmm6, %xmm5
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps %xmm6, %xmm4
addl $8 * SIZE, X
addl $8 * SIZE, Y
ALIGN_3
.L36:
testl $4, N
jle .L37
movaps -32 * SIZE(X), %xmm5
SHUFPD_1 %xmm5, %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps %xmm5, %xmm4
addl $4 * SIZE, X
addl $4 * SIZE, Y
ALIGN_3
.L37:
testl $2, N
jle .L38
xorps %xmm5, %xmm5
movhlps %xmm4, %xmm5
mulps -32 * SIZE(Y), %xmm5
addps %xmm5, %xmm0
addl $2 * SIZE, X
addl $2 * SIZE, Y
ALIGN_3
.L38:
testl $1, N
jle .L999
movss -34 * SIZE(X), %xmm4
mulss -32 * SIZE(Y), %xmm4
addss %xmm4, %xmm0
jmp .L999
ALIGN_3
.L40:
movaps -35 * SIZE(X), %xmm4
addl $SIZE, X
movl N, %eax
sarl $5, %eax
jle .L44
movaps -32 * SIZE(X), %xmm5
movaps -28 * SIZE(X), %xmm6
movaps -24 * SIZE(X), %xmm7
decl %eax
jle .L42
ALIGN_3
.L41:
#ifdef PREFETCH
PREFETCH (PREFETCHSIZE + 0) - PREOFFSET(X)
#endif
movss %xmm5, %xmm4
shufps $0x93, %xmm5, %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -20 * SIZE(X), %xmm4
movss %xmm6, %xmm5
shufps $0x93, %xmm6, %xmm5
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps -16 * SIZE(X), %xmm5
#ifdef PREFETCH
PREFETCH (PREFETCHSIZE + 0) - PREOFFSET(Y)
#endif
movss %xmm7, %xmm6
shufps $0x93, %xmm7, %xmm6
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps -12 * SIZE(X), %xmm6
movss %xmm4, %xmm7
shufps $0x93, %xmm4, %xmm7
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps -8 * SIZE(X), %xmm7
#if defined(PREFETCH) && !defined(FETCH128)
PREFETCH (PREFETCHSIZE + 64) - PREOFFSET(X)
#endif
movss %xmm5, %xmm4
shufps $0x93, %xmm5, %xmm4
mulps -16 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -4 * SIZE(X), %xmm4
movss %xmm6, %xmm5
shufps $0x93, %xmm6, %xmm5
mulps -12 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps 0 * SIZE(X), %xmm5
#if defined(PREFETCH) && !defined(FETCH128)
PREFETCH (PREFETCHSIZE + 64) - PREOFFSET(Y)
#endif
movss %xmm7, %xmm6
shufps $0x93, %xmm7, %xmm6
mulps -8 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps 4 * SIZE(X), %xmm6
movss %xmm4, %xmm7
shufps $0x93, %xmm4, %xmm7
mulps -4 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps 8 * SIZE(X), %xmm7
subl $-32 * SIZE, X
subl $-32 * SIZE, Y
decl %eax
jg .L41
ALIGN_3
.L42:
movss %xmm5, %xmm4
shufps $0x93, %xmm5, %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -20 * SIZE(X), %xmm4
movss %xmm6, %xmm5
shufps $0x93, %xmm6, %xmm5
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps -16 * SIZE(X), %xmm5
movss %xmm7, %xmm6
shufps $0x93, %xmm7, %xmm6
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movaps -12 * SIZE(X), %xmm6
movss %xmm4, %xmm7
shufps $0x93, %xmm4, %xmm7
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movaps -8 * SIZE(X), %xmm7
movss %xmm5, %xmm4
shufps $0x93, %xmm5, %xmm4
mulps -16 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -4 * SIZE(X), %xmm4
movss %xmm6, %xmm5
shufps $0x93, %xmm6, %xmm5
mulps -12 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movss %xmm7, %xmm6
shufps $0x93, %xmm7, %xmm6
mulps -8 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movss %xmm4, %xmm7
shufps $0x93, %xmm4, %xmm7
mulps -4 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
subl $-32 * SIZE, X
subl $-32 * SIZE, Y
ALIGN_3
.L44:
testl $31, N
jle .L999
testl $16, N
jle .L45
movaps -32 * SIZE(X), %xmm5
movaps -28 * SIZE(X), %xmm6
movaps -24 * SIZE(X), %xmm7
movss %xmm5, %xmm4
shufps $0x93, %xmm5, %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movaps -20 * SIZE(X), %xmm4
movss %xmm6, %xmm5
shufps $0x93, %xmm6, %xmm5
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movss %xmm7, %xmm6
shufps $0x93, %xmm7, %xmm6
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movss %xmm4, %xmm7
shufps $0x93, %xmm4, %xmm7
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
addl $16 * SIZE, X
addl $16 * SIZE, Y
ALIGN_3
.L45:
testl $8, N
jle .L46
movaps -32 * SIZE(X), %xmm5
movaps -28 * SIZE(X), %xmm6
movss %xmm5, %xmm4
shufps $0x93, %xmm5, %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movss %xmm6, %xmm5
shufps $0x93, %xmm6, %xmm5
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movaps %xmm6, %xmm4
addl $8 * SIZE, X
addl $8 * SIZE, Y
ALIGN_3
.L46:
testl $4, N
jle .L47
movaps -32 * SIZE(X), %xmm5
movss %xmm5, %xmm4
shufps $0x93, %xmm5, %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm2
movaps %xmm5, %xmm4
addl $4 * SIZE, X
addl $4 * SIZE, Y
ALIGN_3
.L47:
testl $2, N
jle .L48
movaps -32 * SIZE(X), %xmm5
#ifdef movsd
xorps %xmm7, %xmm7
#endif
movsd -32 * SIZE(Y), %xmm7
movss %xmm5, %xmm4
shufps $0x93, %xmm5, %xmm4
mulps %xmm7, %xmm4
addps %xmm4, %xmm3
movlhps %xmm5, %xmm4
addl $2 * SIZE, X
addl $2 * SIZE, Y
ALIGN_3
.L48:
testl $1, N
jle .L999
PSHUFD1($0x93, %xmm4)
mulss -32 * SIZE(Y), %xmm4
addss %xmm4, %xmm0
jmp .L999
ALIGN_4
#else
.L20:
movl N, %eax
sarl $5, %eax
jle .L24
movlps -32 * SIZE(X), %xmm4
movhps -30 * SIZE(X), %xmm4
movlps -28 * SIZE(X), %xmm5
movhps -26 * SIZE(X), %xmm5
movlps -24 * SIZE(X), %xmm6
movhps -22 * SIZE(X), %xmm6
movlps -20 * SIZE(X), %xmm7
movhps -18 * SIZE(X), %xmm7
decl %eax
jle .L22
ALIGN_3
.L21:
#ifdef PREFETCH
PREFETCH (PREFETCHSIZE + 0) - PREOFFSET(X)
#endif
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movlps -16 * SIZE(X), %xmm4
movhps -14 * SIZE(X), %xmm4
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movlps -12 * SIZE(X), %xmm5
movhps -10 * SIZE(X), %xmm5
#ifdef PREFETCH
PREFETCH (PREFETCHSIZE + 0) - PREOFFSET(Y)
#endif
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movlps -8 * SIZE(X), %xmm6
movhps -6 * SIZE(X), %xmm6
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movlps -4 * SIZE(X), %xmm7
movhps -2 * SIZE(X), %xmm7
#if defined(PREFETCH) && !defined(FETCH128)
PREFETCH (PREFETCHSIZE + 64) - PREOFFSET(X)
#endif
mulps -16 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movlps 0 * SIZE(X), %xmm4
movhps 2 * SIZE(X), %xmm4
mulps -12 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movlps 4 * SIZE(X), %xmm5
movhps 6 * SIZE(X), %xmm5
#if defined(PREFETCH) && !defined(FETCH128)
PREFETCH (PREFETCHSIZE + 64) - PREOFFSET(Y)
#endif
mulps -8 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movlps 8 * SIZE(X), %xmm6
movhps 10 * SIZE(X), %xmm6
mulps -4 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movlps 12 * SIZE(X), %xmm7
movhps 14 * SIZE(X), %xmm7
subl $-32 * SIZE, X
subl $-32 * SIZE, Y
decl %eax
jg .L21
ALIGN_3
.L22:
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
movlps -16 * SIZE(X), %xmm4
movhps -14 * SIZE(X), %xmm4
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
movlps -12 * SIZE(X), %xmm5
movhps -10 * SIZE(X), %xmm5
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
movlps -8 * SIZE(X), %xmm6
movhps -6 * SIZE(X), %xmm6
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
movlps -4 * SIZE(X), %xmm7
movhps -2 * SIZE(X), %xmm7
mulps -16 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
mulps -12 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
mulps -8 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
mulps -4 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
subl $-32 * SIZE, X
subl $-32 * SIZE, Y
ALIGN_3
.L24:
testl $31, N
jle .L999
testl $16, N
jle .L25
movlps -32 * SIZE(X), %xmm4
movhps -30 * SIZE(X), %xmm4
movlps -28 * SIZE(X), %xmm5
movhps -26 * SIZE(X), %xmm5
movlps -24 * SIZE(X), %xmm6
movhps -22 * SIZE(X), %xmm6
movlps -20 * SIZE(X), %xmm7
movhps -18 * SIZE(X), %xmm7
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
mulps -24 * SIZE(Y), %xmm6
addps %xmm6, %xmm2
mulps -20 * SIZE(Y), %xmm7
addps %xmm7, %xmm3
addl $16 * SIZE, X
addl $16 * SIZE, Y
ALIGN_3
.L25:
testl $8, N
jle .L26
movlps -32 * SIZE(X), %xmm4
movhps -30 * SIZE(X), %xmm4
movlps -28 * SIZE(X), %xmm5
movhps -26 * SIZE(X), %xmm5
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm0
mulps -28 * SIZE(Y), %xmm5
addps %xmm5, %xmm1
addl $8 * SIZE, X
addl $8 * SIZE, Y
ALIGN_3
.L26:
testl $4, N
jle .L27
movlps -32 * SIZE(X), %xmm4
movhps -30 * SIZE(X), %xmm4
mulps -32 * SIZE(Y), %xmm4
addps %xmm4, %xmm2
addl $4 * SIZE, X
addl $4 * SIZE, Y
ALIGN_3
.L27:
testl $2, N
jle .L28
#ifdef movsd
xorps %xmm4, %xmm4
#endif
movsd -32 * SIZE(X), %xmm4
#ifdef movsd
xorps %xmm6, %xmm6
#endif
movsd -32 * SIZE(Y), %xmm6
mulps %xmm6, %xmm4
addps %xmm4, %xmm3
addl $2 * SIZE, X
addl $2 * SIZE, Y
ALIGN_3
.L28:
testl $1, N
jle .L999
movss -32 * SIZE(X), %xmm4
mulss -32 * SIZE(Y), %xmm4
addss %xmm4, %xmm0
jmp .L999
ALIGN_3
#endif
.L50:
movl N, %eax
sarl $2, %eax
jle .L55
ALIGN_3
.L53:
movss 0 * SIZE(X), %xmm4
addl INCX, X
mulss 0 * SIZE(Y), %xmm4
addl INCY, Y
movss 0 * SIZE(X), %xmm5
addl INCX, X
mulss 0 * SIZE(Y), %xmm5
addl INCY, Y
movss 0 * SIZE(X), %xmm6
addl INCX, X
mulss 0 * SIZE(Y), %xmm6
addl INCY, Y
movss 0 * SIZE(X), %xmm7
addl INCX, X
mulss 0 * SIZE(Y), %xmm7
addl INCY, Y
addss %xmm4, %xmm0
addss %xmm5, %xmm1
addss %xmm6, %xmm2
addss %xmm7, %xmm3
decl %eax
jg .L53
ALIGN_3
.L55:
movl N, %eax
andl $3, %eax
jle .L999
ALIGN_3
.L56:
movss 0 * SIZE(X), %xmm4
addl INCX, X
mulss 0 * SIZE(Y), %xmm4
addl INCY, Y
addss %xmm4, %xmm0
decl %eax
jg .L56
ALIGN_3
.L999:
addps %xmm1, %xmm0
addps %xmm3, %xmm2
addps %xmm2, %xmm0
#if defined(HAVE_SSE3) && !defined(__INTERIX)
haddps %xmm0, %xmm0
haddps %xmm0, %xmm0
#elif defined(HAVE_SSE2)
movhlps %xmm0, %xmm1
addps %xmm1, %xmm0
PSHUFD2($1, %xmm0, %xmm1)
addss %xmm1, %xmm0
#else
movhlps %xmm0, %xmm1
addps %xmm1, %xmm0
movaps %xmm0, %xmm1
shufps $1, %xmm0, %xmm0
addss %xmm1, %xmm0
#endif
movss %xmm0, STACK_N
flds STACK_N
popl %ebx
popl %esi
popl %edi
ret
EPILOGUE