OpenBLAS/kernel/loongarch64/dgemm_tcopy_6.S

556 lines
15 KiB
ArmAsm

/*******************************************************************************
Copyright (c) 2024, The OpenBLAS Project
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.
3. Neither the name of the OpenBLAS project nor the names of
its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "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 OPENBLAS PROJECT 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.
*******************************************************************************/
#define ASSEMBLER
#include "common.h"
/* Function parameters */
#define M $r4 // param 1: m
#define N $r5 // param 2: n
#define SRC $r6 // param 3: src
#define LDA $r7 // param 4: lda
#define DST $r8 // param 5: dst
#define I $r9
#define J $r10
#define S0 $r11
#define S1 $r12
#define S2 $r13
#define S3 $r14
#define S4 $r15
#define S5 $r16
#define S6 $r17
#define S7 $r18
#define S8 $r19
#define P0 $r20
#define P1 $r23
#define P2 $r24
#define P3 $r25
#define P4 $r26
#define T0 $r27
#define T1 $r28
#define T2 $r29
#define TL $r7
#define ZERO $r0
#define F0 $f0
#define F1 $f1
#define F2 $f2
#define F3 $f3
#define F4 $f4
#define F5 $f5
#define F6 $f6
#define F7 $f7
/* LSX vectors */
#define V0 $vr0
#define V1 $vr1
#define V2 $vr2
#define V3 $vr3
/* LASX vectors */
#define U0 $xr4
#define U1 $xr5
#define U2 $xr6
#define U3 $xr7
#define U4 $xr8
#define U5 $xr9
#define U6 $xr10
#define U7 $xr11
PROLOGUE
addi.d $sp, $sp, -56
SDARG $r23, $sp, 0
SDARG $r24, $sp, 8
SDARG $r25, $sp, 16
SDARG $r26, $sp, 24
SDARG $r27, $sp, 32
SDARG $r28, $sp, 40
SDARG $r29, $sp, 48
move S0, SRC
move P0, DST
addi.d I, ZERO, 0x06
div.d T0, N, I // 1
mul.d T1, I, T0 // 6
sub.d N, N, T1 // 1
srai.d T2, N, 0x02
slli.d T2, T2, 0x02
add.d T2, T1, T2
mul.d P2, M, T1
mul.d P3, M, T2
slli.d P2, P2, 0x03
slli.d P3, P3, 0x03
srai.d T2, N, 0x01
srai.d J, M, 0x03
slli.d T2, T2, 0x01
add.d T2, T1, T2
add.d P2, DST, P2
mul.d P4, M, T2
add.d P3, DST, P3
slli.d P4, P4, 0x03
slli.d TL, LDA, 0x03
add.d P4, DST, P4
slli.d T2, TL, 0x01
slli.d T1, M, 0x03
mul.d T1, T1, I
beq ZERO, J, .L_M7
.L_J1: /* J-- */
move S1, S0
add.d S2, S0, TL
add.d S3, S1, T2
add.d S4, S2, T2
add.d S5, S3, T2
add.d S6, S4, T2
add.d S7, S5, T2
add.d S8, S6, T2
add.d S0, S7, T2
move P1, P0
addi.d P0, P0, 0x180
move I, T0
addi.d J, J, -1
beq ZERO, I, .L_N7
.L_I1: /* I-- */
xvld U0, S1, 0x00
vld V0, S1, 0x20
xvld U1, S2, 0x00
vld V1, S2, 0x20
xvld U2, S3, 0x00
vld V2, S3, 0x20
xvld U3, S4, 0x00
vld V3, S4, 0x20
xvst U0, P1, 0x00
vst V0, P1, 0x20
xvst U1, P1, 0x30
vst V1, P1, 0x50
xvst U2, P1, 0x60
vst V2, P1, 0x80
xvst U3, P1, 0x90
vst V3, P1, 0xB0
xvld U0, S5, 0x00
vld V0, S5, 0x20
xvld U1, S6, 0x00
vld V1, S6, 0x20
xvld U2, S7, 0x00
vld V2, S7, 0x20
xvld U3, S8, 0x00
vld V3, S8, 0x20
xvst U0, P1, 0xC0
vst V0, P1, 0xE0
xvst U1, P1, 0xF0
vst V1, P1, 0x110
xvst U2, P1, 0x120
vst V2, P1, 0x140
xvst U3, P1, 0x150
vst V3, P1, 0x170
addi.d S1, S1, 0x30
addi.d S2, S2, 0x30
addi.d S3, S3, 0x30
addi.d S4, S4, 0x30
addi.d S5, S5, 0x30
addi.d S6, S6, 0x30
addi.d S7, S7, 0x30
addi.d S8, S8, 0x30
addi.d I, I, -1
add.d P1, P1, T1
blt ZERO, I, .L_I1
.L_N7:
andi I, N, 0x04
beq ZERO, I, .L_N3
xvld U0, S1, 0x00
xvld U1, S2, 0x00
xvld U2, S3, 0x00
xvld U3, S4, 0x00
xvld U4, S5, 0x00
xvld U5, S6, 0x00
xvld U6, S7, 0x00
xvld U7, S8, 0x00
xvst U0, P2, 0x00
xvst U1, P2, 0x20
xvst U2, P2, 0x40
xvst U3, P2, 0x60
xvst U4, P2, 0x80
xvst U5, P2, 0xA0
xvst U6, P2, 0xC0
xvst U7, P2, 0xE0
addi.d S1, S1, 0x20
addi.d S2, S2, 0x20
addi.d S3, S3, 0x20
addi.d S4, S4, 0x20
addi.d S5, S5, 0x20
addi.d S6, S6, 0x20
addi.d S7, S7, 0x20
addi.d S8, S8, 0x20
addi.d P2, P2, 0x100
.L_N3:
andi I, N, 0x02
beq ZERO, I, .L_N1
xvld U0, S1, 0x00
xvld U1, S2, 0x00
xvld U2, S3, 0x00
xvld U3, S4, 0x00
xvld U4, S5, 0x00
xvld U5, S6, 0x00
xvld U6, S7, 0x00
xvld U7, S8, 0x00
xvpermi.q U0, U1, 0x02
xvpermi.q U2, U3, 0x02
xvpermi.q U4, U5, 0x02
xvpermi.q U6, U7, 0x02
xvst U0, P3, 0x00
xvst U2, P3, 0x20
xvst U4, P3, 0x40
xvst U6, P3, 0x60
addi.d S1, S1, 0x10
addi.d S2, S2, 0x10
addi.d S3, S3, 0x10
addi.d S4, S4, 0x10
addi.d S5, S5, 0x10
addi.d S6, S6, 0x10
addi.d S7, S7, 0x10
addi.d S8, S8, 0x10
addi.d P3, P3, 0x80
.L_N1:
andi I, N, 0x01
beq ZERO, I, .L_N0
fld.d F0, S1, 0x00
fld.d F1, S2, 0x00
fld.d F2, S3, 0x00
fld.d F3, S4, 0x00
fld.d F4, S5, 0x00
fld.d F5, S6, 0x00
fld.d F6, S7, 0x00
fld.d F7, S8, 0x00
fst.d F0, P4, 0x00
fst.d F1, P4, 0x08
fst.d F2, P4, 0x10
fst.d F3, P4, 0x18
fst.d F4, P4, 0x20
fst.d F5, P4, 0x28
fst.d F6, P4, 0x30
fst.d F7, P4, 0x38
addi.d S1, S1, 0x08
addi.d S2, S2, 0x08
addi.d S3, S3, 0x08
addi.d S4, S4, 0x08
addi.d S5, S5, 0x08
addi.d S6, S6, 0x08
addi.d S7, S7, 0x08
addi.d S8, S8, 0x08
addi.d P4, P4, 0x40
.L_N0:
blt ZERO, J, .L_J1
.L_M7:
andi J, M, 0x04
beq ZERO, J, .L_M3
move S1, S0
add.d S2, S0, TL
add.d S3, S1, T2
add.d S4, S2, T2
add.d S0, S3, T2
move P1, P0
addi.d P0, P0, 0xC0
move I, T0
beq ZERO, I, .L_4N7
.L_4I1: /* I-- */
xvld U0, S1, 0x00
vld V0, S1, 0x20
xvld U1, S2, 0x00
vld V1, S2, 0x20
xvld U2, S3, 0x00
vld V2, S3, 0x20
xvld U3, S4, 0x00
vld V3, S4, 0x20
xvst U0, P1, 0x00
vst V0, P1, 0x20
xvst U1, P1, 0x30
vst V1, P1, 0x50
xvst U2, P1, 0x60
vst V2, P1, 0x80
xvst U3, P1, 0x90
vst V3, P1, 0xB0
addi.d S1, S1, 0x30
addi.d S2, S2, 0x30
addi.d S3, S3, 0x30
addi.d S4, S4, 0x30
addi.d I, I, -1
add.d P1, P1, T1
blt ZERO, I, .L_4I1
.L_4N7:
andi I, N, 0x04
beq ZERO, I, .L_4N3
xvld U0, S1, 0x00
xvld U1, S2, 0x00
xvld U2, S3, 0x00
xvld U3, S4, 0x00
xvst U0, P2, 0x00
xvst U1, P2, 0x20
xvst U2, P2, 0x40
xvst U3, P2, 0x60
addi.d S1, S1, 0x20
addi.d S2, S2, 0x20
addi.d S3, S3, 0x20
addi.d S4, S4, 0x20
addi.d P2, P2, 0x80
.L_4N3:
andi I, N, 0x02
beq ZERO, I, .L_4N1
xvld U0, S1, 0x00
xvld U1, S2, 0x00
xvld U2, S3, 0x00
xvld U3, S4, 0x00
xvpermi.q U0, U1, 0x02
xvpermi.q U2, U3, 0x02
xvst U0, P3, 0x00
xvst U2, P3, 0x20
addi.d S1, S1, 0x10
addi.d S2, S2, 0x10
addi.d S3, S3, 0x10
addi.d S4, S4, 0x10
addi.d P3, P3, 0x40
.L_4N1:
andi I, N, 0x01
beq ZERO, I, .L_M3
fld.d F0, S1, 0x00
fld.d F1, S2, 0x00
fld.d F2, S3, 0x00
fld.d F3, S4, 0x00
fst.d F0, P4, 0x00
fst.d F1, P4, 0x08
fst.d F2, P4, 0x10
fst.d F3, P4, 0x18
addi.d S1, S1, 0x08
addi.d S2, S2, 0x08
addi.d S3, S3, 0x08
addi.d S4, S4, 0x08
addi.d P4, P4, 0x20
.L_M3:
andi J, M, 0x02
beq ZERO, J, .L_M1
move S1, S0
add.d S2, S0, TL
add.d S0, S0, T2
move P1, P0
addi.d P0, P0, 0x60
move I, T0
beq ZERO, I, .L_2N7
.L_2I1: /* I-- */
xvld U0, S1, 0x00
vld V0, S1, 0x20
xvld U1, S2, 0x00
vld V1, S2, 0x20
xvst U0, P1, 0x00
vst V0, P1, 0x20
xvst U1, P1, 0x30
vst V1, P1, 0x50
addi.d S1, S1, 0x30
addi.d S2, S2, 0x30
addi.d I, I, -1
add.d P1, P1, T1
blt ZERO, I, .L_2I1
.L_2N7:
andi I, N, 0x04
beq ZERO, I, .L_2N3
xvld U0, S1, 0x00
xvld U1, S2, 0x00
xvst U0, P2, 0x00
xvst U1, P2, 0x20
addi.d S1, S1, 0x20
addi.d S2, S2, 0x20
addi.d P2, P2, 0x40
.L_2N3:
andi I, N, 0x02
beq ZERO, I, .L_2N1
xvld U0, S1, 0x00
xvld U1, S2, 0x00
xvpermi.q U0, U1, 0x02
xvst U0, P3, 0x00
addi.d S1, S1, 0x10
addi.d S2, S2, 0x10
addi.d P3, P3, 0x20
.L_2N1:
andi I, N, 0x01
beq ZERO, I, .L_M1
fld.d F0, S1, 0x00
fld.d F1, S2, 0x00
fst.d F0, P4, 0x00
fst.d F1, P4, 0x08
addi.d S1, S1, 0x08
addi.d S2, S2, 0x08
addi.d P4, P4, 0x10
.L_M1:
andi J, M, 0x01
beq ZERO, J, .L_M0
move S1, S0
add.d S2, S0, TL
move P1, P0
addi.d P0, P0, 0x30
move I, T0
beq ZERO, I, .L_1N7
.L_1I1: /* I-- */
xvld U0, S1, 0x00
vld V0, S1, 0x20
xvst U0, P1, 0x00
vst V0, P1, 0x20
addi.d S1, S1, 0x30
addi.d I, I, -1
add.d P1, P1, T1
blt ZERO, I, .L_1I1
.L_1N7:
andi I, N, 0x04
beq ZERO, I, .L_1N3
xvld U0, S1, 0x00
xvst U0, P2, 0x00
addi.d S1, S1, 0x20
addi.d P2, P2, 0x20
.L_1N3:
andi I, N, 0x02
beq ZERO, I, .L_1N1
fld.d F0, S1, 0x00
fld.d F1, S1, 0x08
fst.d F0, P3, 0x00
fst.d F1, P3, 0x08
addi.d S1, S1, 0x10
addi.d P3, P3, 0x10
.L_1N1:
andi I, N, 0x01
beq ZERO, I, .L_M0
fld.d F0, S1, 0x00
fst.d F0, P4, 0x00
addi.d S1, S1, 0x08
addi.d P4, P4, 0x08
.L_M0:
LDARG $r23, $sp, 0
LDARG $r24, $sp, 8
LDARG $r25, $sp, 16
LDARG $r26, $sp, 24
LDARG $r27, $sp, 32
LDARG $r28, $sp, 40
LDARG $r29, $sp, 48
addi.d $sp, $sp, 56
jirl $r0, $r1, 0x00
EPILOGUE