927 lines
21 KiB
ArmAsm
927 lines
21 KiB
ArmAsm
#define ASSEMBLER
|
|
|
|
#include "common.h"
|
|
#define N $r4
|
|
#define X $r5
|
|
#define INCX $r6
|
|
#define Y $r7
|
|
#define INCY $r8
|
|
#define C $f0
|
|
#define S $f1
|
|
|
|
#define I $r12
|
|
#define TEMP $r13
|
|
#define t1 $r14
|
|
#define t2 $r16
|
|
#define t3 $r15
|
|
#define t4 $r17
|
|
#define XX $r18
|
|
#define YY $r19
|
|
#define a1 $f12
|
|
#define VX0 $xr8
|
|
#define VX1 $xr20
|
|
#define VX2 $xr21
|
|
#define VX3 $xr22
|
|
#define VT0 $xr10
|
|
#define VT1 $xr18
|
|
#define VXC $xr23
|
|
#define VXS $xr9
|
|
#define VXZ $xr19
|
|
|
|
PROLOGUE
|
|
|
|
bge $r0, N, .L999
|
|
li.d TEMP, 1
|
|
movgr2fr.d a1, $r0
|
|
ffint.d.l a1, a1
|
|
slli.d TEMP, TEMP, BASE_SHIFT
|
|
slli.d INCX, INCX, BASE_SHIFT
|
|
slli.d INCY, INCY, BASE_SHIFT
|
|
movfr2gr.d t1, C
|
|
xvreplgr2vr.d VXC, t1
|
|
movfr2gr.d t2, S
|
|
xvreplgr2vr.d VXS, t2
|
|
movfr2gr.d t3, a1
|
|
xvreplgr2vr.d VXZ, t3
|
|
srai.d I, N, 3
|
|
bne INCX, TEMP, .L20
|
|
bne INCY, TEMP, .L12 // INCX==1 and INCY!=1
|
|
b .L11 // INCX==1 and INCY==1
|
|
.L20:
|
|
bne INCY, TEMP, .L22 // INCX!=1 and INCY!=1
|
|
b .L21 // INCX!=1 and INCY==1
|
|
|
|
.L11:
|
|
bge $r0, I, .L997
|
|
fcmp.ceq.d $fcc0, C, a1
|
|
bcnez $fcc0, .L110
|
|
fcmp.ceq.d $fcc0, S, a1
|
|
bcnez $fcc0, .L112 // C!=0 S==0
|
|
b .L111 // C!=0 S!=0
|
|
.align 3
|
|
|
|
.L110:
|
|
fcmp.ceq.d $fcc0, S, a1
|
|
bcnez $fcc0, .L114 // C==0 S==0
|
|
b .L113 // C==0 S!=0
|
|
.align 3
|
|
|
|
.L111: // C!=0 S!=0
|
|
xvld VX0, X, 0 * SIZE
|
|
xvld VX2, Y, 0 * SIZE
|
|
xvld VX1, X, 4 * SIZE
|
|
xvld VX3, Y, 4 * SIZE
|
|
xvfmul.d VT0, VX0, VXC
|
|
xvfmadd.d VT0, VX2, VXS, VT0
|
|
xvfmul.d VT1, VX0, VXS
|
|
xvfmsub.d VT1, VX2, VXC, VT1
|
|
xvst VT0, X, 0 * SIZE
|
|
xvst VT1, Y, 0 * SIZE
|
|
xvfmul.d VT0, VX1, VXC
|
|
xvfmadd.d VT0, VX3, VXS, VT0
|
|
xvfmul.d VT1, VX1, VXS
|
|
xvfmsub.d VT1, VX3, VXC, VT1
|
|
xvst VT0, X, 4 * SIZE
|
|
xvst VT1, Y, 4 * SIZE
|
|
addi.d X, X, 8 * SIZE
|
|
addi.d Y, Y, 8 * SIZE
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L111
|
|
b .L997
|
|
.align 3
|
|
|
|
.L112: // C!=0 S==0
|
|
xvld VX0, X, 0 * SIZE
|
|
xvld VX2, Y, 0 * SIZE
|
|
xvld VX1, X, 4 * SIZE
|
|
xvld VX3, Y, 4 * SIZE
|
|
xvfmul.d VT0, VX0, VXC
|
|
xvfmul.d VT1, VX2, VXC
|
|
xvst VT0, X, 0 * SIZE
|
|
xvst VT1, Y, 0 * SIZE
|
|
xvfmul.d VT0, VX1, VXC
|
|
xvfmul.d VT1, VX3, VXC
|
|
xvst VT0, X, 4 * SIZE
|
|
xvst VT1, Y, 4 * SIZE
|
|
addi.d X, X, 8 * SIZE
|
|
addi.d Y, Y, 8 * SIZE
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L112
|
|
b .L997
|
|
.align 3
|
|
|
|
.L113: // C==0 S!=0
|
|
xvld VX0, X, 0 * SIZE
|
|
xvld VX2, Y, 0 * SIZE
|
|
xvld VX1, X, 4 * SIZE
|
|
xvld VX3, Y, 4 * SIZE
|
|
xvfmul.d VT0, VX2, VXS
|
|
xvfmul.d VT1, VX0, VXS
|
|
xvfsub.d VT1, VXZ, VT1
|
|
xvst VT0, X, 0 * SIZE
|
|
xvst VT1, Y, 0 * SIZE
|
|
xvfmul.d VT0, VX3, VXS
|
|
xvfmul.d VT1, VX1, VXS
|
|
xvfsub.d VT1, VXZ, VT1
|
|
xvst VT0, X, 4 * SIZE
|
|
xvst VT1, Y, 4 * SIZE
|
|
addi.d X, X, 8 * SIZE
|
|
addi.d Y, Y, 8 * SIZE
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L113
|
|
b .L997
|
|
.align 3
|
|
|
|
.L114: // C==0 S==0
|
|
xvst VXZ, X, 0 * SIZE
|
|
xvst VXZ, Y, 0 * SIZE
|
|
xvst VXZ, X, 4 * SIZE
|
|
xvst VXZ, Y, 4 * SIZE
|
|
addi.d X, X, 8 * SIZE
|
|
addi.d Y, Y, 8 * SIZE
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L114
|
|
b .L997
|
|
.align 3
|
|
|
|
.L12: // INCX==1 and INCY!=1
|
|
bge $r0, I, .L997
|
|
move YY, Y
|
|
move XX, X
|
|
fcmp.ceq.d $fcc0, C, a1
|
|
bcnez $fcc0, .L120
|
|
fcmp.ceq.d $fcc0, S, a1
|
|
bcnez $fcc0, .L122 // C!=0 S==0
|
|
b .L121 // C!=0 S!=0
|
|
.align 3
|
|
|
|
.L120:
|
|
fcmp.ceq.d $fcc0, S, a1
|
|
bcnez $fcc0, .L124 // C==0 S==0
|
|
b .L123 // C==0 S!=0
|
|
.align 3
|
|
|
|
.L121: // C!=0 S!=0
|
|
xvld VX0, X, 0 * SIZE
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX2, t1, 0
|
|
xvinsgr2vr.d VX2, t2, 1
|
|
xvinsgr2vr.d VX2, t3, 2
|
|
xvinsgr2vr.d VX2, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX0, VXC
|
|
xvfmadd.d VT0, VX2, VXS, VT0
|
|
xvfmul.d VT1, VX0, VXS
|
|
xvfmsub.d VT1, VX2, VXC, VT1
|
|
xvld VX1, X, 4 * SIZE
|
|
xvst VT0, X, 0 * SIZE
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX3, t1, 0
|
|
xvinsgr2vr.d VX3, t2, 1
|
|
xvinsgr2vr.d VX3, t3, 2
|
|
xvinsgr2vr.d VX3, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX1, VXC
|
|
xvfmadd.d VT0, VX3, VXS, VT0
|
|
xvfmul.d VT1, VX1, VXS
|
|
xvfmsub.d VT1, VX3, VXC, VT1
|
|
addi.d I, I, -1
|
|
xvst VT0, X, 4 * SIZE
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
addi.d X, X, 8 * SIZE
|
|
blt $r0, I, .L121
|
|
b .L997
|
|
.align 3
|
|
|
|
.L122: // C!=0 S==0
|
|
xvld VX0, X, 0 * SIZE
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX2, t1, 0
|
|
xvinsgr2vr.d VX2, t2, 1
|
|
xvinsgr2vr.d VX2, t3, 2
|
|
xvinsgr2vr.d VX2, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX0, VXC
|
|
xvfmul.d VT1, VX2, VXC
|
|
xvld VX1, X, 4 * SIZE
|
|
xvst VT0, X, 0 * SIZE
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX3, t1, 0
|
|
xvinsgr2vr.d VX3, t2, 1
|
|
xvinsgr2vr.d VX3, t3, 2
|
|
xvinsgr2vr.d VX3, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX1, VXC
|
|
xvfmul.d VT1, VX3, VXC
|
|
addi.d I, I, -1
|
|
xvst VT0, X, 4 * SIZE
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
addi.d X, X, 8 * SIZE
|
|
blt $r0, I, .L122
|
|
b .L997
|
|
.align 3
|
|
|
|
.L123: // C==0 S!=0
|
|
xvld VX0, X, 0 * SIZE
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX2, t1, 0
|
|
xvinsgr2vr.d VX2, t2, 1
|
|
xvinsgr2vr.d VX2, t3, 2
|
|
xvinsgr2vr.d VX2, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX2, VXS
|
|
xvfmul.d VT1, VX0, VXS
|
|
xvfsub.d VT1, VXZ, VT1
|
|
xvst VT0, X, 0 * SIZE
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
xvld VX1, X, 4 * SIZE
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX3, t1, 0
|
|
xvinsgr2vr.d VX3, t2, 1
|
|
xvinsgr2vr.d VX3, t3, 2
|
|
xvinsgr2vr.d VX3, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX3, VXS
|
|
xvfmul.d VT1, VX1, VXS
|
|
xvfsub.d VT1, VXZ, VT1
|
|
addi.d I, I, -1
|
|
xvst VT0, X, 4 * SIZE
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
addi.d X, X, 8 * SIZE
|
|
blt $r0, I, .L123
|
|
b .L997
|
|
.align 3
|
|
|
|
.L124: // C==0 S==0
|
|
xvst VXZ, X, 0 * SIZE
|
|
xvst VXZ, X, 4 * SIZE
|
|
xvstelm.d VXZ, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L124
|
|
b .L997
|
|
.align 3
|
|
|
|
.L21:// INCX!=1 and INCY==1
|
|
bge $r0, I, .L997
|
|
move XX, X
|
|
fcmp.ceq.d $fcc0, C, a1
|
|
bcnez $fcc0, .L210
|
|
fcmp.ceq.d $fcc0, S, a1
|
|
bcnez $fcc0, .L212 // C!=0 S==0
|
|
b .L211 // C!=0 S!=0
|
|
.align 3
|
|
|
|
.L210:
|
|
fcmp.ceq.d $fcc0, S, a1
|
|
bcnez $fcc0, .L214 // C==0 S==0
|
|
b .L213 // C==0 S!=0
|
|
.align 3
|
|
|
|
.L211: // C!=0 S!=0
|
|
xvld VX2, Y, 0 * SIZE
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
xvinsgr2vr.d VX0, t1, 0
|
|
xvinsgr2vr.d VX0, t2, 1
|
|
xvinsgr2vr.d VX0, t3, 2
|
|
xvinsgr2vr.d VX0, t4, 3
|
|
add.d X, X, INCX
|
|
xvfmul.d VT0, VXC, VX0
|
|
xvfmadd.d VT0, VX2, VXS, VT0
|
|
xvfmul.d VT1, VXS, VX0
|
|
xvfmsub.d VT1, VX2, VXC, VT1
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvst VT1, Y, 0 * SIZE
|
|
xvld VX3, Y, 4 * SIZE
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
xvinsgr2vr.d VX1, t1, 0
|
|
xvinsgr2vr.d VX1, t2, 1
|
|
xvinsgr2vr.d VX1, t3, 2
|
|
xvinsgr2vr.d VX1, t4, 3
|
|
add.d X, X, INCX
|
|
xvfmul.d VT0, VX1, VXC
|
|
xvfmadd.d VT0, VX3, VXS, VT0
|
|
xvfmul.d VT1, VX1, VXS
|
|
xvfmsub.d VT1, VX3, VXC, VT1
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvst VT1, Y, 4 * SIZE
|
|
addi.d Y, Y, 8 * SIZE
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L211
|
|
b .L997
|
|
.align 3
|
|
|
|
.L212: // C!=0 S==0
|
|
xvld VX2, Y, 0 * SIZE
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
xvinsgr2vr.d VX0, t1, 0
|
|
xvinsgr2vr.d VX0, t2, 1
|
|
xvinsgr2vr.d VX0, t3, 2
|
|
xvinsgr2vr.d VX0, t4, 3
|
|
add.d X, X, INCX
|
|
xvfmul.d VT0, VXC, VX0
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvfmul.d VT1, VX2, VXC
|
|
xvst VT1, Y, 0 * SIZE
|
|
xvld VX3, Y, 4 * SIZE
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
xvinsgr2vr.d VX1, t1, 0
|
|
xvinsgr2vr.d VX1, t2, 1
|
|
xvinsgr2vr.d VX1, t3, 2
|
|
xvinsgr2vr.d VX1, t4, 3
|
|
add.d X, X, INCX
|
|
xvfmul.d VT0, VX1, VXC
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvfmul.d VT1, VX3, VXS
|
|
xvst VT1, Y, 4 * SIZE
|
|
addi.d Y, Y, 8 * SIZE
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L212
|
|
b .L997
|
|
.align 3
|
|
|
|
.L213: // C==0 S!=0
|
|
xvld VX2, Y, 0 * SIZE
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
xvinsgr2vr.d VX0, t1, 0
|
|
xvinsgr2vr.d VX0, t2, 1
|
|
xvinsgr2vr.d VX0, t3, 2
|
|
xvinsgr2vr.d VX0, t4, 3
|
|
add.d X, X, INCX
|
|
xvfmul.d VT0, VXS, VX2
|
|
xvfmul.d VT1, VXS, VX0
|
|
xvfsub.d VT1, VXZ, VT1
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvst VT1, Y, 0 * SIZE
|
|
xvld VX3, Y, 4 * SIZE
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
xvinsgr2vr.d VX1, t1, 0
|
|
xvinsgr2vr.d VX1, t2, 1
|
|
xvinsgr2vr.d VX1, t3, 2
|
|
xvinsgr2vr.d VX1, t4, 3
|
|
add.d X, X, INCX
|
|
xvfmul.d VT0, VX3, VXS
|
|
xvfmul.d VT1, VX1, VXS
|
|
xvfsub.d VT1, VXZ, VT1
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvst VT1, Y, 4 * SIZE
|
|
addi.d Y, Y, 8 * SIZE
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L213
|
|
b .L997
|
|
.align 3
|
|
|
|
.L214: // C==0 S==0
|
|
xvstelm.d VXZ, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvst VT1, Y, 0 * SIZE
|
|
xvstelm.d VXZ, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvst VT1, Y, 4 * SIZE
|
|
addi.d Y, Y, 8 * SIZE
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L211
|
|
b .L997
|
|
.align 3
|
|
|
|
.L22:
|
|
bge $r0, I, .L997
|
|
move YY, Y
|
|
move XX, X
|
|
fcmp.ceq.d $fcc0, C, a1
|
|
bcnez $fcc0, .L220
|
|
fcmp.ceq.d $fcc0, S, a1
|
|
bcnez $fcc0, .L222 // C!=0 S==0
|
|
b .L221 // C!=0 S!=0
|
|
.align 3
|
|
|
|
.L220:
|
|
fcmp.ceq.d $fcc0, S, a1
|
|
bcnez $fcc0, .L224 // C==0 S==0
|
|
b .L223 // C==0 S!=0
|
|
.align 3
|
|
|
|
.L221: // C!=0 S!=0
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
xvinsgr2vr.d VX0, t1, 0
|
|
xvinsgr2vr.d VX0, t2, 1
|
|
xvinsgr2vr.d VX0, t3, 2
|
|
xvinsgr2vr.d VX0, t4, 3
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX2, t1, 0
|
|
xvinsgr2vr.d VX2, t2, 1
|
|
xvinsgr2vr.d VX2, t3, 2
|
|
xvinsgr2vr.d VX2, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX0, VXC
|
|
xvfmadd.d VT0, VX2, VXS, VT0
|
|
xvfmul.d VT1, VX0, VXS
|
|
xvfmsub.d VT1, VX2, VXC, VT1
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
xvinsgr2vr.d VX1, t1, 0
|
|
xvinsgr2vr.d VX1, t2, 1
|
|
xvinsgr2vr.d VX1, t3, 2
|
|
xvinsgr2vr.d VX1, t4, 3
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX3, t1, 0
|
|
xvinsgr2vr.d VX3, t2, 1
|
|
xvinsgr2vr.d VX3, t3, 2
|
|
xvinsgr2vr.d VX3, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX1, VXC
|
|
xvfmadd.d VT0, VX3, VXS, VT0
|
|
xvfmul.d VT1, VX0, VXS
|
|
xvfmsub.d VT1, VX3, VXC, VT1
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L221
|
|
b .L997
|
|
.align 3
|
|
|
|
.L222: // C!=0 S==0
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
xvinsgr2vr.d VX0, t1, 0
|
|
xvinsgr2vr.d VX0, t2, 1
|
|
xvinsgr2vr.d VX0, t3, 2
|
|
xvinsgr2vr.d VX0, t4, 3
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX2, t1, 0
|
|
xvinsgr2vr.d VX2, t2, 1
|
|
xvinsgr2vr.d VX2, t3, 2
|
|
xvinsgr2vr.d VX2, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX0, VXC
|
|
xvfmul.d VT1, VX2, VXC
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
xvinsgr2vr.d VX1, t1, 0
|
|
xvinsgr2vr.d VX1, t2, 1
|
|
xvinsgr2vr.d VX1, t3, 2
|
|
xvinsgr2vr.d VX1, t4, 3
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX3, t1, 0
|
|
xvinsgr2vr.d VX3, t2, 1
|
|
xvinsgr2vr.d VX3, t3, 2
|
|
xvinsgr2vr.d VX3, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX1, VXC
|
|
xvfmul.d VT1, VX3, VXC
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L222
|
|
b .L997
|
|
.align 3
|
|
|
|
.L223: // C==0 S!=0
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
xvinsgr2vr.d VX0, t1, 0
|
|
xvinsgr2vr.d VX0, t2, 1
|
|
xvinsgr2vr.d VX0, t3, 2
|
|
xvinsgr2vr.d VX0, t4, 3
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX2, t1, 0
|
|
xvinsgr2vr.d VX2, t2, 1
|
|
xvinsgr2vr.d VX2, t3, 2
|
|
xvinsgr2vr.d VX2, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX2, VXS
|
|
xvfmul.d VT1, VX0, VXS
|
|
xvfsub.d VT1, VXZ, VT1
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
ld.d t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.d t4, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
xvinsgr2vr.d VX1, t1, 0
|
|
xvinsgr2vr.d VX1, t2, 1
|
|
xvinsgr2vr.d VX1, t3, 2
|
|
xvinsgr2vr.d VX1, t4, 3
|
|
ld.d t1, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t2, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t3, Y, 0 * SIZE
|
|
add.d Y, Y, INCY
|
|
ld.d t4, Y, 0 * SIZE
|
|
xvinsgr2vr.d VX3, t1, 0
|
|
xvinsgr2vr.d VX3, t2, 1
|
|
xvinsgr2vr.d VX3, t3, 2
|
|
xvinsgr2vr.d VX3, t4, 3
|
|
add.d Y, Y, INCY
|
|
xvfmul.d VT0, VX3, VXS
|
|
xvfmul.d VT1, VX0, VXS
|
|
xvfsub.d VT1, VXZ, VT1
|
|
xvstelm.d VT0, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT0, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VT1, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VT1, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L223
|
|
b .L997
|
|
.align 3
|
|
|
|
.L224: // C==0 S==0
|
|
xvstelm.d VXZ, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, XX, 0, 0
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 1
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 2
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, XX, 0, 3
|
|
add.d XX, XX, INCX
|
|
xvstelm.d VXZ, YY, 0, 0
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 1
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 2
|
|
add.d YY, YY, INCY
|
|
xvstelm.d VXZ, YY, 0, 3
|
|
add.d YY, YY, INCY
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L224
|
|
b .L997
|
|
.align 3
|
|
|
|
.L997:
|
|
andi I, N, 7
|
|
bge $r0, I, .L999
|
|
.align 3
|
|
|
|
.L998:
|
|
fld.d $f12, X, 0 * SIZE
|
|
fld.d $f13, Y, 0 * SIZE
|
|
fmul.d $f10, $f12, C
|
|
fmadd.d $f10, $f13, S, $f10
|
|
fst.d $f10, X, 0 * SIZE
|
|
addi.d I, I, -1
|
|
fmul.d $f20, $f12, S
|
|
fmsub.d $f20, $f13, C, $f20
|
|
fst.d $f20, Y, 0 * SIZE
|
|
add.d X, X, INCX
|
|
add.d Y, Y, INCY
|
|
blt $r0, I, .L998
|
|
.align 3
|
|
|
|
.L999:
|
|
move $r4, $r12
|
|
jirl $r0, $r1, 0x0
|
|
.align 3
|
|
|
|
EPILOGUE |