loongarch64: Refine copy,swap,nrm2,sum optimization.

Author: yinshiyou

kernel/loongarch64/KERNEL.LOONGSON2K1000

@@ -31,20 +31,20 @@ IDAMAXKERNEL = idamax_lsx.S
 ISAMINKERNEL = isamin_lsx.S
 IDAMINKERNEL = idamin_lsx.S
 
-SCOPYKERNEL =  scopy_lsx.S
-DCOPYKERNEL =  dcopy_lsx.S
+SCOPYKERNEL =  copy_lsx.S
+DCOPYKERNEL =  copy_lsx.S
 
-SSWAPKERNEL =  sswap_lsx.S
-DSWAPKERNEL =  dswap_lsx.S
+SSWAPKERNEL =  swap_lsx.S
+DSWAPKERNEL =  swap_lsx.S
 
 SAXPYKERNEL =  saxpy_lsx.S
 DAXPYKERNEL =  daxpy_lsx.S
 
 SAXPBYKERNEL = saxpby_lsx.S
 DAXPBYKERNEL = daxpby_lsx.S
 
-SSUMKERNEL  =  ssum_lsx.S
-DSUMKERNEL  =  dsum_lsx.S
+SSUMKERNEL  =  sum_lsx.S
+DSUMKERNEL  =  sum_lsx.S
 
 SASUMKERNEL =  sasum_lsx.S
 DASUMKERNEL =  dasum_lsx.S

kernel/loongarch64/KERNEL.LOONGSON3R5

@@ -31,20 +31,20 @@ IDAMAXKERNEL = idamax_lasx.S
 ISAMINKERNEL = isamin_lasx.S
 IDAMINKERNEL = idamin_lasx.S
 
-SCOPYKERNEL =  scopy_lasx.S
-DCOPYKERNEL =  dcopy_lasx.S
+SCOPYKERNEL =  copy_lasx.S
+DCOPYKERNEL =  copy_lasx.S
 
-SSWAPKERNEL =  sswap_lasx.S
-DSWAPKERNEL =  dswap_lasx.S
+SSWAPKERNEL =  swap_lasx.S
+DSWAPKERNEL =  swap_lasx.S
 
 SAXPYKERNEL =  saxpy_lasx.S
 DAXPYKERNEL =  daxpy_lasx.S
 
 SAXPBYKERNEL = saxpby_lasx.S
 DAXPBYKERNEL = daxpby_lasx.S
 
-SSUMKERNEL  =  ssum_lasx.S
-DSUMKERNEL  =  dsum_lasx.S
+SSUMKERNEL  =  sum_lasx.S
+DSUMKERNEL  =  sum_lasx.S
 
 SASUMKERNEL =  sasum_lasx.S
 DASUMKERNEL =  dasum_lasx.S

kernel/loongarch64/copy_lasx.S

@@ -0,0 +1,306 @@
+/*****************************************************************************
+Copyright (c) 2023, 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 COPYRIGHT OWNER 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"
+#define N      $r4
+#define X      $r5
+#define INCX   $r6
+#define Y      $r7
+#define INCY   $r8
+#define I      $r17
+#define TEMP   $r18
+#define t1     $r14
+#define t2     $r15
+#define t3     $r16
+#define t4     $r19
+#define a1     $f12
+#define a2     $f13
+#define a3     $f14
+#define a4     $f15
+#define VX0    $xr12
+#define VX1    $xr13
+
+    PROLOGUE
+    bge $r0, N, .L999
+    li.d TEMP, 1
+    slli.d  TEMP, TEMP, BASE_SHIFT
+    slli.d  INCX, INCX, BASE_SHIFT
+    slli.d  INCY, INCY, BASE_SHIFT
+    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
+
+/* INCX==1 and INCY==1 */
+.L11:
+    bge $r0, I, .L112
+    .align 3
+
+.L111:
+    xvld VX0, X, 0
+    addi.d  I, I, -1
+    xvst VX0, Y, 0
+#ifdef DOUBLE
+    xvld VX0, X, 32
+    xvst VX0, Y, 32
+#endif
+    addi.d X, X, 8 * SIZE
+    addi.d Y, Y, 8 * SIZE
+    blt $r0, I, .L111
+    .align 3
+
+.L112:
+    andi I, N, 7
+    bge $r0, I, .L999
+    .align 3
+
+.L113:
+    LD    $f12, X, 0
+    addi.d I, I, -1
+    addi.d  X, X, SIZE
+    ST    $f12, Y, 0
+    addi.d  Y, Y, SIZE
+    blt $r0, I, .L113
+    b .L999
+    .align 3
+
+/* INCX==1 and INCY!=1 */
+.L12:
+    bge $r0, I, .L122
+    .align 3
+
+.L121:
+#ifdef DOUBLE
+    xvld VX0, X, 0
+    xvld VX1, X, 32
+    xvstelm.d VX0, Y, 0, 0
+    add.d Y, Y, INCY
+    xvstelm.d VX0, Y, 0, 1
+    add.d Y, Y, INCY
+    xvstelm.d VX0, Y, 0, 2
+    add.d Y, Y, INCY
+    xvstelm.d VX0, Y, 0, 3
+    add.d Y, Y, INCY
+    xvstelm.d VX1, Y, 0, 0
+    add.d Y, Y, INCY
+    xvstelm.d VX1, Y, 0, 1
+    add.d Y, Y, INCY
+    xvstelm.d VX1, Y, 0, 2
+    add.d Y, Y, INCY
+    xvstelm.d VX1, Y, 0, 3
+    add.d Y, Y, INCY
+#else
+    xvld VX0, X, 0
+    xvstelm.w VX0, Y, 0, 0
+    add.d Y, Y, INCY
+    xvstelm.w VX0, Y, 0, 1
+    add.d Y, Y, INCY
+    xvstelm.w VX0, Y, 0, 2
+    add.d Y, Y, INCY
+    xvstelm.w VX0, Y, 0, 3
+    add.d Y, Y, INCY
+    xvstelm.w VX0, Y, 0, 4
+    add.d Y, Y, INCY
+    xvstelm.w VX0, Y, 0, 5
+    add.d Y, Y, INCY
+    xvstelm.w VX0, Y, 0, 6
+    add.d Y, Y, INCY
+    xvstelm.w VX0, Y, 0, 7
+    add.d Y, Y, INCY
+#endif
+    addi.d X, X, 8 * SIZE
+    addi.d  I, I, -1
+    blt $r0, I, .L121
+    .align 3
+
+.L122:
+    andi I, N, 7
+    bge $r0, I, .L999
+    .align 3
+
+.L123:
+    LD    $f12, X, 0
+    addi.d I, I, -1
+    addi.d  X, X, SIZE
+    ST    $f12, Y, 0
+    add.d  Y, Y, INCY
+    blt $r0, I, .L123
+    b .L999
+    .align 3
+
+/* INCX!=1 and INCY==1 */
+.L21:
+    bge $r0, I, .L212
+    .align 3
+
+.L211:
+#ifdef DOUBLE
+    ld.d t1, X, 0
+    add.d X, X, INCX
+    ld.d t2, X, 0
+    add.d X, X, INCX
+    ld.d t3, X, 0
+    add.d X, X, INCX
+    ld.d t4, X, 0
+    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
+    xvst VX0, Y, 0
+    ld.d t1, X, 0
+    add.d X, X, INCX
+    ld.d t2, X, 0
+    add.d X, X, INCX
+    ld.d t3, X, 0
+    add.d X, X, INCX
+    ld.d t4, X, 0
+    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
+    xvst VX1, Y, 32
+#else
+    ld.w t1, X, 0
+    add.d X, X, INCX
+    ld.w t2, X, 0
+    add.d X, X, INCX
+    ld.w t3, X, 0
+    add.d X, X, INCX
+    ld.w t4, X, 0
+    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
+    add.d X, X, INCX
+    ld.w t2, X, 0
+    add.d X, X, INCX
+    ld.w t3, X, 0
+    add.d X, X, INCX
+    ld.w t4, X, 0
+    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
+    xvst VX0, Y, 0
+#endif
+    addi.d Y, Y, 8 * SIZE
+    addi.d  I, I, -1
+    blt $r0, I, .L211
+    .align 3
+
+.L212:
+    andi I, N, 7
+    bge $r0, I, .L999
+    .align 3
+
+.L213:
+    LD    $f12, X, 0
+    addi.d I, I, -1
+    ST    $f12, Y, 0
+    add.d  X, X, INCX
+    addi.d  Y, Y, SIZE
+    blt $r0, I, .L213
+    b .L999
+    .align 3
+
+/* INCX!=1 and INCY!=1 */
+.L22:
+    bge $r0, I, .L223
+    .align 3
+
+.L222:
+    LD    a1, X, 0
+    add.d X, X, INCX
+    LD    a2, X, 0
+    add.d X, X, INCX
+    LD    a3, X, 0
+    add.d X, X, INCX
+    LD    a4, X, 0
+    add.d X, X, INCX
+    ST    a1, Y, 0
+    add.d Y, Y, INCY
+    ST    a2, Y, 0
+    add.d Y, Y, INCY
+    ST    a3, X, 0
+    add.d Y, Y, INCY
+    ST    a4, X, 0
+    add.d Y, Y, INCY
+    LD    a1, X, 0
+    add.d X, X, INCX
+    LD    a2, X, 0
+    add.d X, X, INCX
+    LD    a3, X, 0
+    add.d X, X, INCX
+    LD    a4, X, 0
+    add.d X, X, INCX
+    ST    a1, Y, 0
+    add.d Y, Y, INCY
+    ST    a2, Y, 0
+    add.d Y, Y, INCY
+    ST    a3, X, 0
+    add.d Y, Y, INCY
+    ST    a4, X, 0
+    add.d Y, Y, INCY
+    addi.d  I, I, -1
+    blt $r0, I, .L222
+    .align 3
+
+.L223:
+    andi I, N, 7
+    bge $r0, I, .L999
+    .align 3
+
+.L224:
+    LD    $f12, X, 0
+    addi.d I, I, -1
+    ST    $f12, Y, 0
+    add.d  X, X, INCX
+    add.d  Y, Y, INCY
+    blt $r0, I, .L224
+    .align 3
+
+.L999:
+    move $r4, $r12
+    jirl $r0, $r1, 0x0
+    .align 3
+
+    EPILOGUE