loongarch64: Add optimizations for copy.

Author: yancheng

kernel/loongarch64/KERNEL.LOONGSON2K1000

@@ -31,4 +31,7 @@ IDAMAXKERNEL = idamax_lsx.S
 ISAMINKERNEL = isamin_lsx.S
 IDAMINKERNEL = idamin_lsx.S
 
+SCOPYKERNEL =  scopy_lsx.S
+DCOPYKERNEL =  dcopy_lsx.S
+
 endif

kernel/loongarch64/KERNEL.LOONGSON3R5

@@ -31,6 +31,9 @@ IDAMAXKERNEL = idamax_lasx.S
 ISAMINKERNEL = isamin_lasx.S
 IDAMINKERNEL = idamin_lasx.S
 
+SCOPYKERNEL =  scopy_lasx.S
+DCOPYKERNEL =  dcopy_lasx.S
+
 DGEMMKERNEL    = dgemm_kernel_16x4.S
 DGEMMINCOPY    = dgemm_ncopy_16.S
 DGEMMITCOPY    = dgemm_tcopy_16.S

kernel/loongarch64/dcopy_lasx.S

@@ -0,0 +1,224 @@
+#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
+
+.L11:
+    bge $r0, I, .L112
+    .align 3
+
+.L111:
+    xvld VX0, X, 0 * SIZE
+    xvld VX1, X, 4 * SIZE
+    xvst VX0, Y, 0 * SIZE
+    xvst VX1, Y, 4 * SIZE
+    addi.d X, X, 8 * SIZE
+    addi.d Y, Y, 8 * SIZE
+    addi.d  I, I, -1
+    blt $r0, I, .L111
+    .align 3
+
+.L112:
+    andi I, N, 7
+    bge $r0, I, .L999
+    .align 3
+
+.L113:
+    fld.d $f12, X, 0 * SIZE
+    addi.d I, I, -1
+    addi.d  X, X, SIZE
+    fst.d $f12, Y, 0 * SIZE
+    addi.d  Y, Y, SIZE
+    blt $r0, I, .L113
+    b .L999
+    .align 3
+
+.L12:
+    bge $r0, I, .L122
+    .align 3
+
+.L121:
+    xvld VX0, X, 0 * SIZE
+    xvld VX1, X, 4 * SIZE
+    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
+    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:
+    fld.d $f12, X, 0 * SIZE
+    addi.d I, I, -1
+    addi.d  X, X, SIZE
+    fst.d $f12, Y, 0 * SIZE
+    add.d  Y, Y, INCY
+    blt $r0, I, .L123
+    b .L999
+    .align 3
+
+.L21:
+    bge $r0, I, .L212
+    .align 3
+
+.L211:
+    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
+    xvst VX0, 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
+    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, 4 * SIZE
+    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:
+    fld.d $f12, X, 0 * SIZE
+    addi.d I, I, -1
+    fst.d $f12, Y, 0 * SIZE
+    add.d  X, X, INCX
+    addi.d  Y, Y, SIZE
+    blt $r0, I, .L213
+    b .L999
+    .align 3
+
+.L22:
+    bgez INCX, .L220
+    .align 3
+
+.L220:
+    bge $r0, I, .L223
+    .align 3
+
+.L222:
+    fld.d a1, X, 0 * SIZE
+    add.d X, X, INCX
+    fld.d a2, X, 0 * SIZE
+    add.d X, X, INCX
+    fld.d a3, X, 0 * SIZE
+    add.d X, X, INCX
+    fld.d a4, X, 0 * SIZE
+    add.d X, X, INCX
+    fst.d a1, Y, 0 * SIZE
+    add.d Y, Y, INCY
+    fst.d a2, Y, 0 * SIZE
+    add.d Y, Y, INCY
+    fst.d a3, X, 0 * SIZE
+    add.d Y, Y, INCY
+    fst.d a4, X, 0 * SIZE
+    add.d Y, Y, INCY
+    fld.d a1, X, 0 * SIZE
+    add.d X, X, INCX
+    fld.d a2, X, 0 * SIZE
+    add.d X, X, INCX
+    fld.d a3, X, 0 * SIZE
+    add.d X, X, INCX
+    fld.d a4, X, 0 * SIZE
+    add.d X, X, INCX
+    fst.d a1, Y, 0 * SIZE
+    add.d Y, Y, INCY
+    fst.d a2, Y, 0 * SIZE
+    add.d Y, Y, INCY
+    fst.d a3, X, 0 * SIZE
+    add.d Y, Y, INCY
+    fst.d a4, X, 0 * SIZE
+    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:
+    fld.d $f12, X, 0 * SIZE
+    addi.d I, I, -1
+    fst.d $f12, Y, 0 * SIZE
+    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