157 lines
3.2 KiB
ArmAsm
157 lines
3.2 KiB
ArmAsm
#define ASSEMBLER
|
|
#include "common.h"
|
|
#define N $r4
|
|
#define X $r5
|
|
#define INCX $r6
|
|
#define I $r17
|
|
#define TEMP $r18
|
|
#define t1 $r15
|
|
#define t2 $r12
|
|
#define t3 $r13
|
|
#define t4 $r14
|
|
#define VX0 $xr12
|
|
#define VX1 $xr13
|
|
#define VX2 $xr14
|
|
#define VX3 $xr15
|
|
#define VT0 $xr23
|
|
#define VT1 $xr22
|
|
#define res1 $xr16
|
|
#define res2 $xr17
|
|
#define res0 $xr18
|
|
#define neg1 $xr19
|
|
|
|
PROLOGUE
|
|
xvxor.v res1, res1, res1
|
|
xvxor.v res2, res2, res2
|
|
xvxor.v res0, res0, res0
|
|
bge $r0, N, .L999
|
|
bge $r0, INCX, .L999
|
|
li.w t1, -1
|
|
xvreplgr2vr.w neg1, t1
|
|
xvffint.s.w neg1, neg1
|
|
li.d TEMP, SIZE
|
|
slli.d INCX, INCX, BASE_SHIFT
|
|
srai.d I, N, 3
|
|
bne INCX, TEMP, .L20
|
|
bge $r0, I, .L13
|
|
.align 3
|
|
|
|
.L11:
|
|
xvld VX0, X, 0 * SIZE
|
|
xvfmul.s VX2, neg1, VX0
|
|
xvfcmp.clt.s VT0, VX0, res0
|
|
xvbitsel.v VX0, VX0, VX2, VT0
|
|
xvfadd.s res1, VX0, res1
|
|
addi.d X, X, 8 * SIZE
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L11
|
|
.align 3
|
|
|
|
.L12:
|
|
xvfadd.s res2, res1, res2
|
|
xvpickve.w VX1, res1, 1
|
|
xvpickve.w VX2, res1, 2
|
|
xvpickve.w VX3, res1, 3
|
|
xvfadd.s res1, VX1, res1
|
|
xvfadd.s res1, VX2, res1
|
|
xvfadd.s res1, VX3, res1
|
|
xvpickve.w VX0, res2, 4
|
|
xvpickve.w VX1, res2, 5
|
|
xvpickve.w VX2, res2, 6
|
|
xvpickve.w VX3, res2, 7
|
|
xvfadd.s res1, VX0, res1
|
|
xvfadd.s res1, VX1, res1
|
|
xvfadd.s res1, VX2, res1
|
|
xvfadd.s res1, VX2, res1
|
|
.align 3
|
|
|
|
.L13:
|
|
andi I, N, 7
|
|
bge $r0, I, .L999
|
|
.align 3
|
|
|
|
.L14:
|
|
fld.s $f12, X, 0 * SIZE
|
|
fabs.s $f12, $f12
|
|
fadd.s $f16, $f12, $f16
|
|
addi.d I, I, -1
|
|
addi.d X, X, SIZE
|
|
blt $r0, I, .L14
|
|
b .L999
|
|
.align 3
|
|
|
|
.L20:
|
|
bge $r0, I, .L23
|
|
.align 3
|
|
|
|
.L21:
|
|
ld.w t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.w t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.w t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.w t4, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
xvinsgr2vr.w VX0, t1, 0
|
|
xvinsgr2vr.w VX0, t2, 1
|
|
xvinsgr2vr.w VX0, t3, 2
|
|
xvinsgr2vr.w VX0, t4, 3
|
|
ld.w t1, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.w t2, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.w t3, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
ld.w t4, X, 0 * SIZE
|
|
add.d X, X, INCX
|
|
xvinsgr2vr.w VX0, t1, 4
|
|
xvinsgr2vr.w VX0, t2, 5
|
|
xvinsgr2vr.w VX0, t3, 6
|
|
xvinsgr2vr.w VX0, t4, 7
|
|
xvfmul.s VX2, neg1, VX0
|
|
xvfcmp.clt.s VT0, VX0, res0
|
|
xvbitsel.v VX0, VX0, VX2, VT0
|
|
xvfadd.s res1, VX0, res1
|
|
addi.d I, I, -1
|
|
blt $r0, I, .L21
|
|
.align 3
|
|
|
|
.L22:
|
|
xvfadd.s res2, res1, res2
|
|
xvpickve.w VX1, res1, 1
|
|
xvpickve.w VX2, res1, 2
|
|
xvpickve.w VX3, res1, 3
|
|
xvfadd.s res1, VX1, res1
|
|
xvfadd.s res1, VX2, res1
|
|
xvfadd.s res1, VX3, res1
|
|
xvpickve.w VX0, res2, 4
|
|
xvpickve.w VX1, res2, 5
|
|
xvpickve.w VX2, res2, 6
|
|
xvpickve.w VX3, res2, 7
|
|
xvfadd.s res1, VX0, res1
|
|
xvfadd.s res1, VX1, res1
|
|
xvfadd.s res1, VX2, res1
|
|
xvfadd.s res1, VX2, res1
|
|
.align 3
|
|
|
|
.L23:
|
|
andi I, N, 7
|
|
bge $r0, I, .L999
|
|
.align 3
|
|
|
|
.L24:
|
|
fld.s $f12, X, 0 * SIZE
|
|
fabs.s $f12, $f12
|
|
fadd.s $f16, $f12, $f16
|
|
addi.d I, I, -1
|
|
add.d X, X, INCX
|
|
blt $r0, I, .L24
|
|
.align 3
|
|
|
|
.L999:
|
|
fmov.s $f0, $f16
|
|
jirl $r0, $r1, 0x0
|
|
.align 3
|
|
|
|
EPILOGUE |