|
- /*********************************************************************/
- /* 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_M 4 + STACK + ARGS(%esp)
- #define STACK_ALPHA 16 + STACK + ARGS(%esp)
- #define STACK_X 20 + STACK + ARGS(%esp)
- #define STACK_INCX 24 + STACK + ARGS(%esp)
-
- #define M %ebx
- #define X %ecx
- #define INCX %edx
- #define I %esi
- #define XX %edi
-
- #include "l1param.h"
-
- PROLOGUE
-
- pushl %edi
- pushl %esi
- pushl %ebx
-
- PROFCODE
-
- movl STACK_M, M
- movl STACK_X, X
- movl STACK_INCX, INCX
-
- lea (, INCX, SIZE), INCX
-
- movss STACK_ALPHA, %xmm0
-
- testl M, M
- jle .L999
-
- xorps %xmm1, %xmm1
- comiss %xmm0, %xmm1
- shufps $0, %xmm0, %xmm0
-
- jne .L100
-
- /* Alpha == ZERO */
- cmpl $SIZE, INCX
- jne .L50
-
- /* INCX == 1 */
- cmpl $3, M
- jle .L14
-
- testl $4, X # aligned for double word?
- je .L05
-
- movss %xmm1, 0 * SIZE(X)
- addl $SIZE, X
- decl M
- jle .L999
- ALIGN_3
-
- .L05:
- testl $8, X # aligned for quad word?
- je .L06
-
- movsd %xmm1, 0 * SIZE(X)
- addl $2 * SIZE, X
- subl $2, M
- jle .L999
- ALIGN_3
-
- .L06:
- movl M, I
- sarl $4, I
- jle .L12
- ALIGN_4
-
- .L11:
- #ifdef PREFETCHW
- PREFETCHW (PREFETCHSIZE + 0) - PREOFFSET(X)
- #endif
-
- movaps %xmm1, 0 * SIZE(X)
- movaps %xmm1, 4 * SIZE(X)
- movaps %xmm1, 8 * SIZE(X)
- movaps %xmm1, 12 * SIZE(X)
- addl $16 * SIZE, X
- decl I
- jg .L11
- ALIGN_4
-
- .L12:
- testl $15, M
- je .L999
- testl $8, M
- je .L13
-
- movaps %xmm1, 0 * SIZE(X)
- movaps %xmm1, 4 * SIZE(X)
- addl $8 * SIZE, X
- ALIGN_3
-
- .L13:
- testl $4, M
- je .L14
-
- movaps %xmm1, 0 * SIZE(X)
- addl $4 * SIZE, X
- ALIGN_3
-
- .L14:
- testl $2, M
- je .L15
-
- movsd %xmm1, 0 * SIZE(X)
- addl $2 * SIZE, X
- ALIGN_3
-
- .L15:
- testl $1, M
- je .L999
-
- movss %xmm1, 0 * SIZE(X)
- jmp .L999
- ALIGN_4
-
- /* incx != 1 */
- .L50:
- movl M, I # rcx = n
- sarl $3, I # (n >> 3)
- jle .L52
- ALIGN_4
-
- .L51:
- #ifdef PREFETCHW
- PREFETCHW (PREFETCHSIZE + 0) - PREOFFSET(X)
- #endif
-
- movss %xmm1, (X)
- addl INCX, X
- movss %xmm1, (X)
- addl INCX, X
- movss %xmm1, (X)
- addl INCX, X
- movss %xmm1, (X)
- addl INCX, X
- movss %xmm1, (X)
- addl INCX, X
- movss %xmm1, (X)
- addl INCX, X
- movss %xmm1, (X)
- addl INCX, X
- movss %xmm1, (X)
- addl INCX, X
-
- decl I
- jg .L51
- ALIGN_4
-
- .L52:
- testl $7, M
- je .L999
-
- testl $4, M
- je .L53
-
- movss %xmm1, (X)
- addl INCX, X
- movss %xmm1, (X)
- addl INCX, X
- movss %xmm1, (X)
- addl INCX, X
- movss %xmm1, (X)
- addl INCX, X
- ALIGN_3
-
- .L53:
- testl $2, M
- je .L54
-
- movss %xmm1, (X)
- addl INCX, X
- movss %xmm1, (X)
- addl INCX, X
- ALIGN_3
-
- .L54:
- testl $1, M
- je .L999
-
- movss %xmm1, (X)
- jmp .L999
- ALIGN_4
-
- /* Alpha != ZERO */
-
- .L100:
- cmpl $SIZE, INCX
- jne .L150
-
- subl $-32 * SIZE, X
-
- cmpl $3, M
- jle .L116
-
- testl $SIZE, X
- je .L105
-
- movss -32 * SIZE(X), %xmm1
- mulss %xmm0, %xmm1
- movss %xmm1, -32 * SIZE(X)
- addl $SIZE, X
- decl M
- jle .L999
- ALIGN_3
-
- .L105:
- testl $2 * SIZE, X
- je .L110
-
- movsd -32 * SIZE(X), %xmm1
- mulps %xmm0, %xmm1
- movsd %xmm1, -32 * SIZE(X)
- addl $2 * SIZE, X
- subl $2, M
- jle .L999
- ALIGN_3
-
- .L110:
- movl M, I
- sarl $5, I
- jle .L113
-
- #if defined(BARCELONA) || defined(BULLDOZER)
-
- movaps %xmm0, %xmm1
- mulps -32 * SIZE(X), %xmm1
- movaps %xmm0, %xmm2
- mulps -28 * SIZE(X), %xmm2
- movaps %xmm0, %xmm3
- mulps -24 * SIZE(X), %xmm3
- movaps %xmm0, %xmm4
- mulps -20 * SIZE(X), %xmm4
-
- decl I
- jle .L112
- ALIGN_4
-
- .L111:
- #ifdef PREFETCHW
- PREFETCHW (PREFETCHSIZE + 0) - PREOFFSET(X)
- #endif
-
- movaps %xmm1, -32 * SIZE(X)
- movaps %xmm0, %xmm1
- mulps -16 * SIZE(X), %xmm1
-
- movaps %xmm2, -28 * SIZE(X)
- movaps %xmm0, %xmm2
- mulps -12 * SIZE(X), %xmm2
-
- movaps %xmm3, -24 * SIZE(X)
- movaps %xmm0, %xmm3
- mulps -8 * SIZE(X), %xmm3
-
- movaps %xmm4, -20 * SIZE(X)
- movaps %xmm0, %xmm4
- mulps -4 * SIZE(X), %xmm4
-
- #ifdef PREFETCHW
- PREFETCHW (PREFETCHSIZE + 64) - PREOFFSET(X)
- #endif
-
- movaps %xmm1, -16 * SIZE(X)
- movaps %xmm0, %xmm1
- mulps 0 * SIZE(X), %xmm1
-
- movaps %xmm2, -12 * SIZE(X)
- movaps %xmm0, %xmm2
- mulps 4 * SIZE(X), %xmm2
-
- movaps %xmm3, -8 * SIZE(X)
- movaps %xmm0, %xmm3
- mulps 8 * SIZE(X), %xmm3
-
- movaps %xmm4, -4 * SIZE(X)
- movaps %xmm0, %xmm4
- mulps 12 * SIZE(X), %xmm4
-
- subl $-32 * SIZE, X
- decl I
- jg .L111
- ALIGN_4
-
- .L112:
- movaps %xmm1, -32 * SIZE(X)
- movaps %xmm0, %xmm1
- mulps -16 * SIZE(X), %xmm1
-
- movaps %xmm2, -28 * SIZE(X)
- movaps %xmm0, %xmm2
- mulps -12 * SIZE(X), %xmm2
-
- movaps %xmm3, -24 * SIZE(X)
- movaps %xmm0, %xmm3
- mulps -8 * SIZE(X), %xmm3
-
- movaps %xmm4, -20 * SIZE(X)
- movaps %xmm0, %xmm4
- mulps -4 * SIZE(X), %xmm4
-
- movaps %xmm1, -16 * SIZE(X)
- movaps %xmm2, -12 * SIZE(X)
- movaps %xmm3, -8 * SIZE(X)
- movaps %xmm4, -4 * SIZE(X)
-
- #else
-
- movaps -32 * SIZE(X), %xmm1
- movaps -28 * SIZE(X), %xmm2
- movaps -24 * SIZE(X), %xmm3
- movaps -20 * SIZE(X), %xmm4
-
- decl I
- jle .L112
- ALIGN_4
-
- .L111:
- #ifdef PREFETCHW
- PREFETCHW (PREFETCHSIZE + 0) - PREOFFSET(X)
- #endif
-
- mulps %xmm0, %xmm1
- movaps %xmm1, -32 * SIZE(X)
- movaps -16 * SIZE(X), %xmm1
-
- mulps %xmm0, %xmm2
- movaps %xmm2, -28 * SIZE(X)
- movaps -12 * SIZE(X), %xmm2
-
- mulps %xmm0, %xmm3
- movaps %xmm3, -24 * SIZE(X)
- movaps -8 * SIZE(X), %xmm3
-
- mulps %xmm0, %xmm4
- movaps %xmm4, -20 * SIZE(X)
- movaps -4 * SIZE(X), %xmm4
-
- #ifdef PREFETCHW
- PREFETCHW (PREFETCHSIZE + 64) - PREOFFSET(X)
- #endif
-
- mulps %xmm0, %xmm1
- movaps %xmm1, -16 * SIZE(X)
- movaps 0 * SIZE(X), %xmm1
-
- mulps %xmm0, %xmm2
- movaps %xmm2, -12 * SIZE(X)
- movaps 4 * SIZE(X), %xmm2
-
- mulps %xmm0, %xmm3
- movaps %xmm3, -8 * SIZE(X)
- movaps 8 * SIZE(X), %xmm3
-
- mulps %xmm0, %xmm4
- movaps %xmm4, -4 * SIZE(X)
- movaps 12 * SIZE(X), %xmm4
-
- subl $-32 * SIZE, X
- decl I
- jg .L111
- ALIGN_4
-
- .L112:
- mulps %xmm0, %xmm1
- movaps %xmm1, -32 * SIZE(X)
- movaps -16 * SIZE(X), %xmm1
-
- mulps %xmm0, %xmm2
- movaps %xmm2, -28 * SIZE(X)
- movaps -12 * SIZE(X), %xmm2
-
- mulps %xmm0, %xmm3
- movaps %xmm3, -24 * SIZE(X)
- movaps -8 * SIZE(X), %xmm3
-
- mulps %xmm0, %xmm4
- movaps %xmm4, -20 * SIZE(X)
- movaps -4 * SIZE(X), %xmm4
-
- mulps %xmm0, %xmm1
- movaps %xmm1, -16 * SIZE(X)
- mulps %xmm0, %xmm2
- movaps %xmm2, -12 * SIZE(X)
- mulps %xmm0, %xmm3
- movaps %xmm3, -8 * SIZE(X)
- mulps %xmm0, %xmm4
- movaps %xmm4, -4 * SIZE(X)
-
- #endif
-
- subl $-32 * SIZE, X
- ALIGN_3
-
- .L113:
- testl $31, M
- je .L999
-
- testl $16, M
- je .L114
-
- movaps -32 * SIZE(X), %xmm1
- movaps -28 * SIZE(X), %xmm3
- movaps -24 * SIZE(X), %xmm5
- movaps -20 * SIZE(X), %xmm7
-
- mulps %xmm0, %xmm1
- movaps %xmm1, -32 * SIZE(X)
- mulps %xmm0, %xmm3
- movaps %xmm3, -28 * SIZE(X)
- mulps %xmm0, %xmm5
- movaps %xmm5, -24 * SIZE(X)
- mulps %xmm0, %xmm7
- movaps %xmm7, -20 * SIZE(X)
-
- addl $16 * SIZE, X
- ALIGN_3
-
- .L114:
- testl $8, M
- je .L115
-
- movaps -32 * SIZE(X), %xmm1
- movaps -28 * SIZE(X), %xmm3
-
- mulps %xmm0, %xmm1
- movaps %xmm1, -32 * SIZE(X)
- mulps %xmm0, %xmm3
- movaps %xmm3, -28 * SIZE(X)
- addl $8 * SIZE, X
- ALIGN_3
-
- .L115:
- testl $4, M
- je .L116
-
- movaps -32 * SIZE(X), %xmm1
- mulps %xmm0, %xmm1
- movaps %xmm1, -32 * SIZE(X)
- addl $4 * SIZE, X
- ALIGN_3
-
- .L116:
- testl $2, M
- je .L117
-
- movsd -32 * SIZE(X), %xmm1
- mulps %xmm0, %xmm1
- movsd %xmm1, -32 * SIZE(X)
- addl $2 * SIZE, X
- ALIGN_3
-
- .L117:
- testl $1, M
- je .L999
-
- movss -32 * SIZE(X), %xmm1
- mulss %xmm0, %xmm1
- movss %xmm1, -32 * SIZE(X)
- jmp .L999
- ALIGN_3
-
- /* incx != 1 */
-
- .L150:
- movl X, XX
- movl M, I # rcx = n
- sarl $3, I # (n >> 3)
- jle .L152
- ALIGN_4
-
- .L151:
- movss (X), %xmm1
- addl INCX, X
- movss (X), %xmm2
- addl INCX, X
- movss (X), %xmm3
- addl INCX, X
- movss (X), %xmm4
- addl INCX, X
-
- mulss %xmm0, %xmm1
- mulss %xmm0, %xmm2
- mulss %xmm0, %xmm3
- mulss %xmm0, %xmm4
-
- movss %xmm1, (XX)
- addl INCX, XX
- movss %xmm2, (XX)
- addl INCX, XX
- movss %xmm3, (XX)
- addl INCX, XX
- movss %xmm4, (XX)
- addl INCX, XX
-
- movss (X), %xmm1
- addl INCX, X
- movss (X), %xmm2
- addl INCX, X
- movss (X), %xmm3
- addl INCX, X
- movss (X), %xmm4
- addl INCX, X
-
- mulss %xmm0, %xmm1
- mulss %xmm0, %xmm2
- mulss %xmm0, %xmm3
- mulss %xmm0, %xmm4
-
- movss %xmm1, (XX)
- addl INCX, XX
- movss %xmm2, (XX)
- addl INCX, XX
- movss %xmm3, (XX)
- addl INCX, XX
- movss %xmm4, (XX)
- addl INCX, XX
-
- decl I
- jg .L151
- ALIGN_4
-
- .L152:
- testl $7, M
- je .L999
-
- testl $4, M
- je .L153
-
- movss (X), %xmm1
- addl INCX, X
- movss (X), %xmm2
- addl INCX, X
- movss (X), %xmm3
- addl INCX, X
- movss (X), %xmm4
- addl INCX, X
-
- mulss %xmm0, %xmm1
- mulss %xmm0, %xmm2
- mulss %xmm0, %xmm3
- mulss %xmm0, %xmm4
-
- movss %xmm1, (XX)
- addl INCX, XX
- movss %xmm2, (XX)
- addl INCX, XX
- movss %xmm3, (XX)
- addl INCX, XX
- movss %xmm4, (XX)
- addl INCX, XX
- ALIGN_3
-
- .L153:
- testl $2, M
- je .L154
-
- movss (X), %xmm1
- addl INCX, X
- movss (X), %xmm2
- addl INCX, X
-
- mulss %xmm0, %xmm1
- mulss %xmm0, %xmm2
-
- movss %xmm1, (XX)
- addl INCX, XX
- movss %xmm2, (XX)
- addl INCX, XX
- ALIGN_3
-
- .L154:
- testl $1, M
- je .L999
-
- movss (X), %xmm1
- mulss %xmm0, %xmm1
- movss %xmm1, (X)
- ALIGN_4
-
- .L999:
- xorl %eax, %eax
-
- popl %ebx
- popl %esi
- popl %edi
- ret
-
- EPILOGUE
|