diff --git a/kernel/loongarch64/KERNEL.LOONGSON3R5 b/kernel/loongarch64/KERNEL.LOONGSON3R5
index cce4093e3..bb0441ab2 100644
--- a/kernel/loongarch64/KERNEL.LOONGSON3R5
+++ b/kernel/loongarch64/KERNEL.LOONGSON3R5
@@ -1 +1,14 @@
-#TODO: Add loongarch64 SIMD optimizations
+DGEMMKERNEL    = dgemm_kernel_16x4.S
+DGEMMINCOPY    = dgemm_ncopy_16.S
+DGEMMITCOPY    = dgemm_tcopy_16.S
+DGEMMONCOPY    = dgemm_ncopy_4.S
+DGEMMOTCOPY    = dgemm_tcopy_4.S
+DGEMMINCOPYOBJ = dgemm_incopy.o
+DGEMMITCOPYOBJ = dgemm_itcopy.o
+DGEMMONCOPYOBJ = dgemm_oncopy.o
+DGEMMOTCOPYOBJ = dgemm_otcopy.o
+
+DTRSMKERNEL_LN  = ../generic/trsm_kernel_LN.c
+DTRSMKERNEL_LT  = ../generic/trsm_kernel_LT.c
+DTRSMKERNEL_RN  = ../generic/trsm_kernel_RN.c
+DTRSMKERNEL_RT  = ../generic/trsm_kernel_RT.c
diff --git a/kernel/loongarch64/dgemm_kernel_16x4.S b/kernel/loongarch64/dgemm_kernel_16x4.S
new file mode 100644
index 000000000..13faa977e
--- /dev/null
+++ b/kernel/loongarch64/dgemm_kernel_16x4.S
@@ -0,0 +1,4250 @@
+/*******************************************************************************
+Copyright (c) 2021, 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: bm
+#define N      $r5   // param 2: bn
+#define K      $r6   // param 3: bk
+#define ALPHA  $f0   // param 4: alpha
+#define A      $r7   // param 5: ba
+#define B      $r8   // param 6: bb
+#define C      $r9   // param 7: bc
+#define LDC    $r10  // param 8: ldc
+
+#ifdef TRMMKERNEL
+#define OFFSET $r11  // param 9: offset
+#endif
+#define OFF    $r12
+
+/* Cycle control parameters */
+#define I      $r13
+#define J      $r14
+#define L      $r15
+#define TL     $r16
+/* Matrix address */
+#define A0     $r17
+#define B0     $r18
+#define C0     $r19
+#define C1     $r20
+#define C2     $r23
+#define C3     $r24
+#define T0     $r25 /* !! DO NOT USE $r21 and $r22 !! */
+#define T1     $r26
+#define T2     $r27
+#define ZERO   $r0
+
+/* LASX vectors */
+#define U0     $xr0
+#define U1     $xr1
+#define U2     $xr2
+#define U3     $xr3
+#define U4     $xr4
+#define U5     $xr5
+#define U6     $xr6
+#define D0     $xr7
+#define D1     $xr8
+#define D2     $xr9
+#define D3     $xr10
+#define D4     $xr11
+#define D5     $xr12
+#define D6     $xr13
+#define D7     $xr14
+#define D8     $xr15
+#define D9     $xr16
+#define D10    $xr17
+#define D11    $xr18
+#define D12    $xr19
+#define D13    $xr20
+#define D14    $xr21
+#define D15    $xr22
+#define VALPHA $xr23
+
+/* Prefetch interval */
+#define A_PRE  0x200
+#define B_PRE  0x100
+
+    PROLOGUE
+
+    addi.d   $sp,   $sp,   -56
+    /* Store regs */
+    SDARG    $r23,  $sp,   0
+    SDARG    $r24,  $sp,   8
+    SDARG    $r25,  $sp,   16
+    SDARG    $r26,  $sp,   24
+    SDARG    $r27,  $sp,   32
+    ST       $f23,  $sp,   40
+    ST       ALPHA, $sp,   48
+
+    /* VALPHA = {ALPHA, ALPHA, ALPHA, ALPHA} */
+    xvld         VALPHA, $sp,  48
+    xvreplve0.d  VALPHA, VALPHA
+
+#if defined (TRMMKERNEL) && !defined(LEFT)
+    sub.d   OFF,   ZERO,  OFFSET
+#else
+    xor     OFF,   OFF,   OFF
+#endif
+
+    /* if (!(N >> 2)) goto L_N3 */
+    srai.d   J,     N,     2     /* J = bn >> 2 */
+    andi     N,     N,     0x03
+    beq      ZERO,  J,     .L_N3
+
+.L_J1: /* J-- && This loop include Condition 1 */
+
+/************************* Condition 1 if((N >> 2) && (M >> 4)) START !!! *************************
+*                                                   dgemm_core_16x4                                */
+    move     C0,    C
+    move     A0,    A
+    slli.d   T0,    LDC,   3
+    add.d    C1,    C0,    T0
+    addi.d   J,     J,     -1   /* J-- */
+    add.d    C2,    C1,    T0
+    add.d    C3,    C2,    T0
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+    move     OFF,   OFFSET
+#endif
+
+    /* if (!(M >> 4)) goto L_M8 */
+    srai.d   I,     M,     4     /* I = bm >> 4 */
+    beq      ZERO,  I,     .L_M8
+
+.L_I1: /* I-- */
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x07
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x05
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  16
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  4
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+    /* Calculate the first set of D0~D15,
+     * avoidig set 0 operation
+     * Load 16 * 64 from A0
+     * U0 = {a3,  a2,  a1,  a0}
+     * U1 = {a7,  a6,  a5,  a4}
+     * U2 = {a11, a10, a9,  a8}
+     * U3 = {a15, a14, a13, a12}
+     */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d     U4, B0, 0x00
+    preld          0,  C0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+    xvfmul.d  D1,  U1, U4
+    preld     0,   C0, 0x40
+    xvfmul.d  D2,  U2, U4
+    xvfmul.d  D3,  U3, U4
+
+    xvldrepl.d     U4, B0, 0x08
+    preld          0,   C1,    0x00
+    /* line 2 */
+    xvfmul.d  D4,  U0, U4
+    xvfmul.d  D5,  U1, U4
+    preld     0,   C1,    0x40
+    xvfmul.d  D6,  U2, U4
+    xvfmul.d  D7,  U3, U4
+
+    xvldrepl.d     U4, B0, 0x10
+    preld          0,   C2,    0x00
+    /* line 3 */
+    xvfmul.d  D8,  U0, U4
+    xvfmul.d  D9,  U1, U4
+    preld     0,   C2,    0x40
+    xvfmul.d  D10, U2, U4
+    xvfmul.d  D11, U3, U4
+
+    xvldrepl.d     U4, B0, 0x18
+    preld          0,   C3,    0x00
+    /* line 4 */
+    xvfmul.d  D12, U0, U4
+    xvfmul.d  D13, U1, U4
+    preld     0,   C3,    0x40
+    xvfmul.d  D14, U2, U4
+    xvfmul.d  D15, U3, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x80
+    addi.d    B0,  B0, 0x20
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_L7 */
+    beq       ZERO,TL, .L_L7
+
+    /* Calculate 8 sets of D0~D15 */
+.L_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    /* Cumulative D0~D15 */
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+    preld      0,   B0, B_PRE
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+    preld      0,   A0, A_PRE
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+    xvfmadd.d  D10, U2, U4, D10
+    xvfmadd.d  D11, U3, U4, D11
+    preld      0,   A0, A_PRE + 0x40
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+    xvfmadd.d  D14, U2, U4, D14
+    xvfmadd.d  D15, U3, U4, D15
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x20
+
+           /***8-2***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    /* Cumulative D0~D15 */
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+    preld      0,   B0, B_PRE
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+    preld      0,   A0, A_PRE
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+    xvfmadd.d  D10, U2, U4, D10
+    xvfmadd.d  D11, U3, U4, D11
+    preld      0,   A0, A_PRE + 0x40
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+    xvfmadd.d  D14, U2, U4, D14
+    xvfmadd.d  D15, U3, U4, D15
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x20
+
+           /***8-3***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    /* Cumulative D0~D15 */
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+    preld      0,   B0, B_PRE
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+    preld      0,   A0, A_PRE
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+    xvfmadd.d  D10, U2, U4, D10
+    xvfmadd.d  D11, U3, U4, D11
+    preld      0,   A0, A_PRE + 0x40
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+    xvfmadd.d  D14, U2, U4, D14
+    xvfmadd.d  D15, U3, U4, D15
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x20
+
+           /***8-4***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    /* Cumulative D0~D15 */
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+    preld      0,   B0, B_PRE
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+    preld      0,   A0, A_PRE
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+    xvfmadd.d  D10, U2, U4, D10
+    xvfmadd.d  D11, U3, U4, D11
+    preld      0,   A0, A_PRE + 0x40
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+    xvfmadd.d  D14, U2, U4, D14
+    xvfmadd.d  D15, U3, U4, D15
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x20
+
+           /***8-5***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    /* Cumulative D0~D15 */
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+    preld      0,   B0, B_PRE
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+    preld      0,   A0, A_PRE
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+    xvfmadd.d  D10, U2, U4, D10
+    xvfmadd.d  D11, U3, U4, D11
+    preld      0,   A0, A_PRE + 0x40
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+    xvfmadd.d  D14, U2, U4, D14
+    xvfmadd.d  D15, U3, U4, D15
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x20
+
+           /***8-6***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    /* Cumulative D0~D15 */
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+    preld      0,   B0, B_PRE
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+    preld      0,   A0, A_PRE
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+    xvfmadd.d  D10, U2, U4, D10
+    xvfmadd.d  D11, U3, U4, D11
+    preld      0,   A0, A_PRE + 0x40
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+    xvfmadd.d  D14, U2, U4, D14
+    xvfmadd.d  D15, U3, U4, D15
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x20
+
+           /***8-7***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    /* Cumulative D0~D15 */
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+    preld      0,   B0, B_PRE
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+    preld      0,   A0, A_PRE
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+    xvfmadd.d  D10, U2, U4, D10
+    xvfmadd.d  D11, U3, U4, D11
+    preld      0,   A0, A_PRE + 0x40
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+    xvfmadd.d  D14, U2, U4, D14
+    xvfmadd.d  D15, U3, U4, D15
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x20
+
+           /***8-8***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    /* Cumulative D0~D15 */
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+    preld      0,   B0, B_PRE
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+    preld      0,   A0, A_PRE
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+    xvfmadd.d  D10, U2, U4, D10
+    xvfmadd.d  D11, U3, U4, D11
+    preld      0,   A0, A_PRE + 0x40
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+    xvfmadd.d  D14, U2, U4, D14
+    xvfmadd.d  D15, U3, U4, D15
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x20
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_TL1
+
+   /* Maybe we need calculate the last
+    * 7 sets of D0~D15?
+    */
+.L_L7:
+    /* if (!(L & 7)) goto L_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_L0
+
+.L_L71:
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    /* Cumulative D0~D15 */
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+    xvfmadd.d  D10, U2, U4, D10
+    xvfmadd.d  D11, U3, U4, D11
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+    xvfmadd.d  D14, U2, U4, D14
+    xvfmadd.d  D15, U3, U4, D15
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x20
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_L71
+
+.L_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D1,   D1,  VALPHA
+    xvfmul.d  D2,   D2,  VALPHA
+    xvfmul.d  D3,   D3,  VALPHA
+    xvfmul.d  D4,   D4,  VALPHA
+    xvfmul.d  D5,   D5,  VALPHA
+    xvfmul.d  D6,   D6,  VALPHA
+    xvfmul.d  D7,   D7,  VALPHA
+    xvfmul.d  D8,   D8,  VALPHA
+    xvfmul.d  D9,   D9,  VALPHA
+    xvfmul.d  D10,  D10, VALPHA
+    xvfmul.d  D11,  D11, VALPHA
+    xvfmul.d  D12,  D12, VALPHA
+    xvfmul.d  D13,  D13, VALPHA
+    xvfmul.d  D14,  D14, VALPHA
+    xvfmul.d  D15,  D15, VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvld      U1,  C0,  0x20
+    xvld      U2,  C0,  0x40
+    xvld      U3,  C0,  0x60
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+    xvfmadd.d D1,  D1,  VALPHA,  U1
+    xvfmadd.d D2,  D2,  VALPHA,  U2
+    xvfmadd.d D3,  D3,  VALPHA,  U3
+
+    /* Load C1  */
+    xvld      U0,  C1,  0x00
+    xvld      U1,  C1,  0x20
+    xvld      U2,  C1,  0x40
+    xvld      U3,  C1,  0x60
+    xvfmadd.d D4,  D4,  VALPHA,  U0
+    xvfmadd.d D5,  D5,  VALPHA,  U1
+    xvfmadd.d D6,  D6,  VALPHA,  U2
+    xvfmadd.d D7,  D7,  VALPHA,  U3
+
+    /* Load C2  */
+    xvld      U0,  C2,  0x00
+    xvld      U1,  C2,  0x20
+    xvld      U2,  C2,  0x40
+    xvld      U3,  C2,  0x60
+    xvfmadd.d D8,  D8,  VALPHA,  U0
+    xvfmadd.d D9,  D9,  VALPHA,  U1
+    xvfmadd.d D10, D10, VALPHA,  U2
+    xvfmadd.d D11, D11, VALPHA,  U3
+
+    /* Load C3  */
+    xvld      U0,  C3,  0x00
+    xvld      U1,  C3,  0x20
+    xvld      U2,  C3,  0x40
+    xvld      U3,  C3,  0x60
+    xvfmadd.d D12, D12, VALPHA,  U0
+    xvfmadd.d D13, D13, VALPHA,  U1
+    xvfmadd.d D14, D14, VALPHA,  U2
+    xvfmadd.d D15, D15, VALPHA,  U3
+#endif // #if defined(TRMMKERNEL)
+
+    /* Store C0 */
+    xvst      D0,  C0,  0x00
+    xvst      D1,  C0,  0x20
+    xvst      D2,  C0,  0x40
+    xvst      D3,  C0,  0x60
+    /* Store C1 */
+    xvst      D4,  C1,  0x00
+    xvst      D5,  C1,  0x20
+    xvst      D6,  C1,  0x40
+    xvst      D7,  C1,  0x60
+    /* Store C2 */
+    xvst      D8,  C2,  0x00
+    xvst      D9,  C2,  0x20
+    xvst      D10, C2,  0x40
+    xvst      D11, C2,  0x60
+    /* Store C3 */
+    xvst      D12, C3,  0x00
+    xvst      D13, C3,  0x20
+    xvst      D14, C3,  0x40
+    xvst      D15, C3,  0x60
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x80
+    addi.d    C1,  C1,  0x80
+    addi.d    C2,  C2,  0x80
+    addi.d    C3,  C3,  0x80
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    /* number of values  in A */
+    addi.d    L,   L,   -16
+#else
+    /* number of values in B */
+    addi.d    L,   L,   -4
+#endif
+    slli.d    T0,  L,  0x07
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x05
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    addi.d    OFF, OFF, 0x10
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+    addi.d    I,   I,   -1  /* I-- */
+    blt       ZERO,I,   .L_I1
+
+.L_M8:
+    /* We have done M & 16, considering M=8/4/2/1 */
+    andi      I,   M,   15
+    beq       ZERO,I,   .L_M0
+
+    andi      I,   M,   8
+    beq       ZERO,I,   .L_M4
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x06
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x05
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  8
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  4
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif  // #if defined(TRMMKERNEL)
+
+    /* Load 8 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+    xvfmul.d  D1,  U1, U4
+
+    xvldrepl.d     U4, B0, 0x08
+    /* line 2 */
+    xvfmul.d  D4,  U0, U4
+    xvfmul.d  D5,  U1, U4
+
+    xvldrepl.d     U4, B0, 0x10
+    /* line 3 */
+    xvfmul.d  D8,  U0, U4
+    xvfmul.d  D9,  U1, U4
+
+    xvldrepl.d     U4, B0, 0x18
+    /* line 4 */
+    xvfmul.d  D12, U0, U4
+    xvfmul.d  D13, U1, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x40
+    addi.d    B0,  B0, 0x20
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_M8_L7 */
+    beq       ZERO,TL, .L_M8_L7
+
+.L_M8_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x20
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x20
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x20
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x20
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x20
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x20
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x20
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x20
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_M8_TL1
+
+.L_M8_L7:
+    /* if (!(L & 7)) goto L_M8_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_M8_L0
+
+.L_M8_L71:
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+    xvfmadd.d  D9,  U1, U4, D9
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+    xvfmadd.d  D13, U1, U4, D13
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x20
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_M8_L71
+
+.L_M8_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D1,   D1,  VALPHA
+    xvfmul.d  D4,   D4,  VALPHA
+    xvfmul.d  D5,   D5,  VALPHA
+    xvfmul.d  D8,   D8,  VALPHA
+    xvfmul.d  D9,   D9,  VALPHA
+    xvfmul.d  D12,  D12, VALPHA
+    xvfmul.d  D13,  D13, VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvld      U1,  C0,  0x20
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+    xvfmadd.d D1,  D1,  VALPHA,  U1
+
+    /* Load C1  */
+    xvld      U0,  C1,  0x00
+    xvld      U1,  C1,  0x20
+    xvfmadd.d D4,  D4,  VALPHA,  U0
+    xvfmadd.d D5,  D5,  VALPHA,  U1
+
+    /* Load C2  */
+    xvld      U0,  C2,  0x00
+    xvld      U1,  C2,  0x20
+    xvfmadd.d D8,  D8,  VALPHA,  U0
+    xvfmadd.d D9,  D9,  VALPHA,  U1
+
+    /* Load C3  */
+    xvld      U0,  C3,  0x00
+    xvld      U1,  C3,  0x20
+    xvfmadd.d D12, D12, VALPHA,  U0
+    xvfmadd.d D13, D13, VALPHA,  U1
+#endif   // #if defined(TRMMKERNEL)
+
+    /* Store C0 */
+    xvst      D0,  C0,  0x00
+    xvst      D1,  C0,  0x20
+    /* Store C1 */
+    xvst      D4,  C1,  0x00
+    xvst      D5,  C1,  0x20
+    /* Store C2 */
+    xvst      D8,  C2,  0x00
+    xvst      D9,  C2,  0x20
+    /* Store C3 */
+    xvst      D12, C3,  0x00
+    xvst      D13, C3,  0x20
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x40
+    addi.d    C1,  C1,  0x40
+    addi.d    C2,  C2,  0x40
+    addi.d    C3,  C3,  0x40
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    /* number of values in A */
+    addi.d    L,   L,   -8
+#else
+    /* number of values in B */
+    addi.d    L,   L,   -4
+#endif
+    slli.d    T0,  L,  0x06
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x05
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    /* number of values in A */
+    addi.d    OFF,   OFF,  0x08
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N >> 2 ) && (M & 8)) End************/
+
+.L_M4:
+    andi      I,   M,   4
+    beq       ZERO,I,   .L_M2
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x05
+    add.d    A0,    A0,   T0
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  4
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  4
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 4 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x08
+    /* line 2 */
+    xvfmul.d  D4,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x10
+    /* line 3 */
+    xvfmul.d  D8,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x18
+    /* line 4 */
+    xvfmul.d  D12, U0, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x20
+    addi.d    B0,  B0, 0x20
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_M4_L7 */
+    beq       ZERO,TL, .L_M4_L7
+
+.L_M4_TL1: /* TL-- */
+           /***8-1***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x20
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x20
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x20
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x20
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x20
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x20
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x20
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x20
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_M4_TL1
+
+.L_M4_L7:
+    /* if (!(L & 7)) goto L_M4_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_M4_L0
+
+.L_M4_L71:
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x20
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_M4_L71
+
+.L_M4_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D4,   D4,  VALPHA
+    xvfmul.d  D8,   D8,  VALPHA
+    xvfmul.d  D12,  D12, VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+
+    /* Load C1  */
+    xvld      U0,  C1,  0x00
+    xvfmadd.d D4,  D4,  VALPHA,  U0
+
+    /* Load C2  */
+    xvld      U0,  C2,  0x00
+    xvfmadd.d D8,  D8,  VALPHA,  U0
+
+    /* Load C3  */
+    xvld      U0,  C3,  0x00
+    xvfmadd.d D12, D12, VALPHA,  U0
+#endif   // #if defined(TRMMKERNEL)
+
+    /* Store C0 */
+    xvst      D0,  C0,  0x00
+    /* Store C1 */
+    xvst      D4,  C1,  0x00
+    /* Store C2 */
+    xvst      D8,  C2,  0x00
+    /* Store C3 */
+    xvst      D12, C3,  0x00
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x20
+    addi.d    C1,  C1,  0x20
+    addi.d    C2,  C2,  0x20
+    addi.d    C3,  C3,  0x20
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    /* number of values in A */
+    addi.d    L,   L,   -4
+#else
+    /* number of values in B */
+    addi.d    L,   L,   -4
+#endif
+    slli.d    T0,  L,  0x05
+    add.d     A0,  A0, T0
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    /* number of values in A */
+    addi.d    OFF,   OFF,  0x04
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N >> 2 ) && (M & 4) ) End************/
+
+.L_M2:
+    andi      I,   M,   2
+    beq       ZERO,I,   .L_M1
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x04
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x05
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  2
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  4
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 2 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x08
+    /* line 2 */
+    xvfmul.d  D4,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x10
+    /* line 3 */
+    xvfmul.d  D8,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x18
+    /* line 4 */
+    xvfmul.d  D12, U0, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x10
+    addi.d    B0,  B0, 0x20
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_M2_L7 */
+    beq       ZERO,TL, .L_M2_L7
+
+.L_M2_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 2 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x20
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x20
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x20
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x20
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x20
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x20
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x20
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x20
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_M2_TL1
+
+.L_M2_L7:
+    /* if (!(L & 7)) goto L_M2_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_M2_L0
+
+.L_M2_L71:
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x20
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_M2_L71
+
+.L_M2_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D4,   D4,  VALPHA
+    xvfmul.d  D8,   D8,  VALPHA
+    xvfmul.d  D12,  D12, VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+
+    /* Load C1  */
+    xvld      U0,  C1,  0x00
+    xvfmadd.d D4,  D4,  VALPHA,  U0
+
+    /* Load C2  */
+    xvld      U0,  C2,  0x00
+    xvfmadd.d D8,  D8,  VALPHA,  U0
+
+    /* Load C3  */
+    xvld      U0,  C3,  0x00
+    xvfmadd.d D12, D12, VALPHA,  U0
+#endif   // #if defined(TRMMKERNEL)
+
+    xvstelm.d D0,  C0,  0x00,    0x00
+    xvstelm.d D4,  C1,  0x00,    0x00
+    xvstelm.d D8,  C2,  0x00,    0x00
+    xvstelm.d D12, C3,  0x00,    0x00
+    xvstelm.d D0,  C0,  0x08,    0x01
+    xvstelm.d D4,  C1,  0x08,    0x01
+    xvstelm.d D8,  C2,  0x08,    0x01
+    xvstelm.d D12, C3,  0x08,    0x01
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x10
+    addi.d    C1,  C1,  0x10
+    addi.d    C2,  C2,  0x10
+    addi.d    C3,  C3,  0x10
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    /* number of values in A */
+    addi.d    L,   L,   -2
+#else
+    /* number of values in B */
+    addi.d    L,   L,   -4
+#endif
+    slli.d    T0,  L,  0x04
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x05
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    /* number of values in A */
+    addi.d    OFF,   OFF,  0x02
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N >> 2 ) && (M & 2) ) End************/
+
+.L_M1:
+    andi      I,   M,   1
+    beq       ZERO,I,   .L_M0
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x03
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x05
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  1
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  4
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 1 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x08
+    /* line 2 */
+    xvfmul.d  D4,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x10
+    /* line 3 */
+    xvfmul.d  D8,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x18
+    /* line 4 */
+    xvfmul.d  D12, U0, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x08
+    addi.d    B0,  B0, 0x20
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_M1_L7 */
+    beq       ZERO,TL, .L_M1_L7
+
+.L_M1_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 1 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x20
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x20
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x20
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x20
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x20
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x20
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x20
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x20
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_M1_TL1
+
+.L_M1_L7:
+    /* if (!(L & 7)) goto L_M1_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_M1_L0
+
+.L_M1_L71:
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    xvldrepl.d U4,  B0, 0x10
+    xvfmadd.d  D8,  U0, U4, D8
+
+    xvldrepl.d U4,  B0, 0x18
+    xvfmadd.d  D12, U0, U4, D12
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x20
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_M1_L71
+
+.L_M1_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D4,   D4,  VALPHA
+    xvfmul.d  D8,   D8,  VALPHA
+    xvfmul.d  D12,  D12, VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+
+    /* Load C1  */
+    xvld      U0,  C1,  0x00
+    xvfmadd.d D4,  D4,  VALPHA,  U0
+
+    /* Load C2  */
+    xvld      U0,  C2,  0x00
+    xvfmadd.d D8,  D8,  VALPHA,  U0
+
+    /* Load C3  */
+    xvld      U0,  C3,  0x00
+    xvfmadd.d D12, D12, VALPHA,  U0
+#endif   // #if defined(TRMMKERNEL)
+
+    xvstelm.d D0,  C0,  0x00,    0x00
+    xvstelm.d D4,  C1,  0x00,    0x00
+    xvstelm.d D8,  C2,  0x00,    0x00
+    xvstelm.d D12, C3,  0x00,    0x00
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x08
+    addi.d    C1,  C1,  0x08
+    addi.d    C2,  C2,  0x08
+    addi.d    C3,  C3,  0x08
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    /* number of values in A */
+    addi.d    L,   L,   -1
+#else
+    /* number of values in B */
+    addi.d    L,   L,   -4
+#endif
+    slli.d    T0,  L,  0x03
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x05
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    /* number of values in A */
+    addi.d    OFF,   OFF,  0x01
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N >> 2 ) && (M & 1) ) End************/
+
+.L_M0:
+    /* Add stride for B and C
+     * B += (K * 32)
+     * C += (LDC * 32)
+     */
+    /* since the array type is double,
+     * so we must mul 32
+     */
+    slli.d    T0,   K,   5
+    slli.d    T1,   LDC, 5
+    add.d     B,    B,   T0
+    add.d     C,    C,   T1
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+    addi.d    OFF,  OFF, 0x04
+#endif
+
+    blt      ZERO,  J,   .L_J1
+
+//////////////// go back to L_J1 /////////////////
+/////////////////////////////////////////////////
+/************************ Condition 1 if((N >> 2) && (M >> 4)) END !!! ************************/
+
+.L_N3:
+    andi     J,    N,   2
+    beq      ZERO, J,   .L_N1
+
+/************************* Condition 2 if((N & 2) && (M >> 4)) START !!! *************************
+*                                                   dgemm_core_16x2                                */
+
+    move     C0,    C
+    move     A0,    A
+    slli.d   T0,    LDC,   3
+    add.d    C1,    C0,    T0
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+    move     OFF,   OFFSET
+#endif
+
+    /* if (!(M >> 4)) goto L_N3_M8 */
+    srai.d   I,     M,     4     /* I = bm >> 4 */
+    beq      ZERO,  I,     .L_N3_M8
+
+.L_N3_I1:
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x07
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x04
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  16
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  2
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 16 * 64 from A0
+     * U0 = {a3,  a2,  a1,  a0}
+     * U1 = {a7,  a6,  a5,  a4}
+     * U2 = {a11, a10, a9,  a8}
+     * U3 = {a15, a14, a13, a12}
+     */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+    xvfmul.d  D1,  U1, U4
+    xvfmul.d  D2,  U2, U4
+    xvfmul.d  D3,  U3, U4
+
+    xvldrepl.d     U4, B0, 0x08
+    /* line 2 */
+    xvfmul.d  D4,  U0, U4
+    xvfmul.d  D5,  U1, U4
+    xvfmul.d  D6,  U2, U4
+    xvfmul.d  D7,  U3, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x80
+    addi.d    B0,  B0, 0x10
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_N3_L7 */
+    beq       ZERO,TL, .L_N3_L7
+
+.L_N3_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x10
+
+           /***8-2***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x10
+
+           /***8-3***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x10
+
+           /***8-4***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x10
+
+           /***8-5***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x10
+
+           /***8-6***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x10
+
+           /***8-7***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x10
+
+           /***8-8***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x10
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_N3_TL1
+
+.L_N3_L7:
+    /* if (!(L & 7)) goto L_N3_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_N3_L0
+
+.L_N3_L71:
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+    xvfmadd.d  D6,  U2, U4, D6
+    xvfmadd.d  D7,  U3, U4, D7
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x10
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_N3_L71
+
+.L_N3_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D1,   D1,  VALPHA
+    xvfmul.d  D2,   D2,  VALPHA
+    xvfmul.d  D3,   D3,  VALPHA
+    xvfmul.d  D4,   D4,  VALPHA
+    xvfmul.d  D5,   D5,  VALPHA
+    xvfmul.d  D6,   D6,  VALPHA
+    xvfmul.d  D7,   D7,  VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvld      U1,  C0,  0x20
+    xvld      U2,  C0,  0x40
+    xvld      U3,  C0,  0x60
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+    xvfmadd.d D1,  D1,  VALPHA,  U1
+    xvfmadd.d D2,  D2,  VALPHA,  U2
+    xvfmadd.d D3,  D3,  VALPHA,  U3
+
+    /* Load C1  */
+    xvld      U0,  C1,  0x00
+    xvld      U1,  C1,  0x20
+    xvld      U2,  C1,  0x40
+    xvld      U3,  C1,  0x60
+    xvfmadd.d D4,  D4,  VALPHA,  U0
+    xvfmadd.d D5,  D5,  VALPHA,  U1
+    xvfmadd.d D6,  D6,  VALPHA,  U2
+    xvfmadd.d D7,  D7,  VALPHA,  U3
+#endif // #if defined(TRMMKERNEL)
+
+    /* Store C0 */
+    xvst      D0,  C0,  0x00
+    xvst      D1,  C0,  0x20
+    xvst      D2,  C0,  0x40
+    xvst      D3,  C0,  0x60
+    /* Store C1 */
+    xvst      D4,  C1,  0x00
+    xvst      D5,  C1,  0x20
+    xvst      D6,  C1,  0x40
+    xvst      D7,  C1,  0x60
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x80
+    addi.d    C1,  C1,  0x80
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    addi.d    L,   L,   -16
+#else
+    addi.d    L,   L,   -2
+#endif
+    slli.d    T0,  L,  0x07
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x04
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    addi.d    OFF,   OFF,  0x10
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+    addi.d    I,   I,   -1  /* I-- */
+    blt       ZERO,I,   .L_N3_I1
+
+.L_N3_M8:
+    /* We have done M & 16, considering M=8/4/2/1 */
+    andi      I,   M,   15
+    beq       ZERO,I,   .L_N3_M0
+
+    andi      I,   M,   8
+    beq       ZERO,I,   .L_N3_M4
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x06
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x04
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  8
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  2
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 8 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+    xvfmul.d  D1,  U1, U4
+
+    xvldrepl.d     U4, B0, 0x08
+    /* line 2 */
+    xvfmul.d  D4,  U0, U4
+    xvfmul.d  D5,  U1, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x40
+    addi.d    B0,  B0, 0x10
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_N3_M8_L7 */
+    beq       ZERO,TL, .L_N3_M8_L7
+
+.L_N3_M8_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x10
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x10
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x10
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x10
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    /* Cumulative D0~D15 */
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x10
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x10
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x10
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x10
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_N3_M8_TL1
+
+.L_N3_M8_L7:
+    /* if (!(L & 7)) goto L_N3_M8_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_N3_M8_L0
+
+.L_N3_M8_L71:
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+    xvfmadd.d  D5,  U1, U4, D5
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x10
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_N3_M8_L71
+
+.L_N3_M8_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D1,   D1,  VALPHA
+    xvfmul.d  D4,   D4,  VALPHA
+    xvfmul.d  D5,   D5,  VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvld      U1,  C0,  0x20
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+    xvfmadd.d D1,  D1,  VALPHA,  U1
+
+    /* Load C1  */
+    xvld      U0,  C1,  0x00
+    xvld      U1,  C1,  0x20
+    xvfmadd.d D4,  D4,  VALPHA,  U0
+    xvfmadd.d D5,  D5,  VALPHA,  U1
+#endif // #if defined(TRMMKERNEL)
+
+    /* Store C0 */
+    xvst      D0,  C0,  0x00
+    xvst      D1,  C0,  0x20
+    /* Store C1 */
+    xvst      D4,  C1,  0x00
+    xvst      D5,  C1,  0x20
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x40
+    addi.d    C1,  C1,  0x40
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    addi.d    L,   L,   -8
+#else
+    addi.d    L,   L,   -2
+#endif
+    slli.d    T0,  L,  0x06
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x04
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    addi.d    OFF,   OFF,  0x08
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N & 2) && (M & 8) ) End************/
+
+.L_N3_M4:
+    andi      I,   M,   4
+    beq       ZERO,I,   .L_N3_M2
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x05
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x04
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  4
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  2
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 4 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x08
+    /* line 2 */
+    xvfmul.d  D4,  U0, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x20
+    addi.d    B0,  B0, 0x10
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_N3_M4_L7 */
+    beq       ZERO,TL, .L_N3_M4_L7
+
+.L_N3_M4_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 8 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x10
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x10
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x10
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x10
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+
+    /* Cumulative D0~D15 */
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x10
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x10
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x10
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x10
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_N3_M4_TL1
+
+.L_N3_M4_L7:
+    /* if (!(L & 7)) goto L_N3_M4_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_N3_M4_L0
+
+.L_N3_M4_L71:
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x10
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_N3_M4_L71
+
+.L_N3_M4_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D4,   D4,  VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+
+    /* Load C1  */
+    xvld      U0,  C1,  0x00
+    xvfmadd.d D4,  D4,  VALPHA,  U0
+#endif // #if defined(TRMMKERNEL)
+
+    /* Store C0 */
+    xvst      D0,  C0,  0x00
+    /* Store C1 */
+    xvst      D4,  C1,  0x00
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x20
+    addi.d    C1,  C1,  0x20
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    addi.d    L,   L,   -4
+#else
+    addi.d    L,   L,   -2
+#endif
+    slli.d    T0,  L,  0x05
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x04
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    addi.d    OFF,   OFF,  0x04
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N & 2 ) && (M & 4) ) End************/
+
+.L_N3_M2:
+    andi      I,   M,   2
+    beq       ZERO,I,   .L_N3_M1
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x04
+    add.d    A0,    A0,   T0
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  2
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  2
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 2 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x08
+    /* line 2 */
+    xvfmul.d  D4,  U0, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x10
+    addi.d    B0,  B0, 0x10
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_N3_M2_L7 */
+    beq       ZERO,TL, .L_N3_M2_L7
+
+.L_N3_M2_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 2 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x10
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x10
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x10
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x10
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x10
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x10
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x10
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x10
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_N3_M2_TL1
+
+.L_N3_M2_L7:
+    /* if (!(L & 7)) goto L_N3_M2_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_N3_M2_L0
+
+.L_N3_M2_L71:
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x10
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_N3_M2_L71
+
+.L_N3_M2_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D4,   D4,  VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+
+    /* Load C1  */
+    xvld      U0,  C1,  0x00
+    xvfmadd.d D4,  D4,  VALPHA,  U0
+#endif // #if defined(TRMMKERNEL)
+
+    xvstelm.d D0,  C0,  0x00,    0x00
+    xvstelm.d D4,  C1,  0x00,    0x00
+    xvstelm.d D0,  C0,  0x08,    0x01
+    xvstelm.d D4,  C1,  0x08,    0x01
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x10
+    addi.d    C1,  C1,  0x10
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    addi.d    L,   L,   -2
+#else
+    addi.d    L,   L,   -2
+#endif
+    slli.d    T0,  L,  0x04
+    add.d     A0,  A0, T0
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    addi.d    OFF,   OFF,  0x02
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N & 2 ) && (M & 2) ) End************/
+
+.L_N3_M1:
+    andi      I,   M,   1
+    beq       ZERO,I,   .L_N3_M0
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x03
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x04
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  1
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  2
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 1 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+
+    xvldrepl.d     U4, B0, 0x08
+    /* line 2 */
+    xvfmul.d  D4,  U0, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x08
+    addi.d    B0,  B0, 0x10
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_N3_M1_L7 */
+    beq       ZERO,TL, .L_N3_M1_L7
+
+.L_N3_M1_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 1 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x10
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x10
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x10
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x10
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x10
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x10
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x10
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x10
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_N3_M1_TL1
+
+.L_N3_M1_L7:
+    /* if (!(L & 7)) goto L_N3_M1_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_N3_M1_L0
+
+.L_N3_M1_L71:
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    xvldrepl.d U4,  B0, 0x08
+    xvfmadd.d  D4,  U0, U4, D4
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x10
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_N3_M1_L71
+
+.L_N3_M1_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D4,   D4,  VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+
+    /* Load C1  */
+    xvld      U0,  C1,  0x00
+    xvfmadd.d D4,  D4,  VALPHA,  U0
+#endif // #if defined(TRMMKERNEL)
+
+    xvstelm.d D0,  C0,  0x00,    0x00
+    xvstelm.d D4,  C1,  0x00,    0x00
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x08
+    addi.d    C1,  C1,  0x08
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    addi.d    L,   L,   -1
+#else
+    addi.d    L,   L,   -2
+#endif
+    slli.d    T0,  L,  0x03
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x04
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    addi.d    OFF,   OFF,  0x01
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N & 2 ) && (M & 1) ) End************/
+
+.L_N3_M0:
+    /* Add stride for B and C
+     * B += (K * 16)
+     * C += (LDC * 16)
+     */
+    /* since the array type is double,
+     * so we must mul 16
+     */
+    slli.d    T0,   K,   4
+    slli.d    T1,   LDC, 4
+    add.d     B,    B,   T0
+    add.d     C,    C,   T1
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+    addi.d    OFF,  OFF, 0x02
+#endif
+
+    /* We must reinit I */
+    srai.d   I,     M,   4     /* I = bm >> 4 */
+
+/************************* Condition 2 if((N & 2) && (M >> 4)) End !!! *************************
+*                                                   dgemm_core_16x2                                */
+
+.L_N1:
+    andi     J,    N,   1
+    beq      ZERO, J,   .L_N0
+
+/************************* Condition 3 if((N & 1) && (M >> 4)) START !!! *************************
+*                                                   dgemm_core_16x1                                */
+
+    move     C0,    C
+    move     A0,    A
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+    move     OFF,   OFFSET
+#endif
+
+    /* if (!(M >> 4)) goto L_N1_M8 */
+    srai.d   I,     M,     4     /* I = bm >> 4 */
+    beq      ZERO,  I,     .L_N1_M8
+
+.L_N1_I1:
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x07
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x03
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  16
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  1
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 16 * 64 from A0
+     * U0 = {a3,  a2,  a1,  a0}
+     * U1 = {a7,  a6,  a5,  a4}
+     * U2 = {a11, a10, a9,  a8}
+     * U3 = {a15, a14, a13, a12}
+     */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+    xvfmul.d  D1,  U1, U4
+    xvfmul.d  D2,  U2, U4
+    xvfmul.d  D3,  U3, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x80
+    addi.d    B0,  B0, 0x08
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_N1_L7 */
+    beq       ZERO,TL, .L_N1_L7
+
+.L_N1_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x08
+
+           /***8-2***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x08
+
+           /***8-3***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x08
+
+           /***8-4***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x08
+
+           /***8-5***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x08
+
+           /***8-6***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x08
+
+           /***8-7***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x08
+
+           /***8-8***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x08
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_N1_TL1
+
+.L_N1_L7:
+    /* if (!(L & 7)) goto L_N1_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_N1_L0
+
+.L_N1_L71:
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+    xvld     U2,   A0,    0x40
+    xvld     U3,   A0,    0x60
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+    xvfmadd.d  D2,  U2, U4, D2
+    xvfmadd.d  D3,  U3, U4, D3
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x80
+    addi.d     B0,  B0, 0x08
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_N1_L71
+
+.L_N1_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D1,   D1,  VALPHA
+    xvfmul.d  D2,   D2,  VALPHA
+    xvfmul.d  D3,   D3,  VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvld      U1,  C0,  0x20
+    xvld      U2,  C0,  0x40
+    xvld      U3,  C0,  0x60
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+    xvfmadd.d D1,  D1,  VALPHA,  U1
+    xvfmadd.d D2,  D2,  VALPHA,  U2
+    xvfmadd.d D3,  D3,  VALPHA,  U3
+#endif // #if defined(TRMMKERNEL)
+
+    /* Store C0 */
+    xvst      D0,  C0,  0x00
+    xvst      D1,  C0,  0x20
+    xvst      D2,  C0,  0x40
+    xvst      D3,  C0,  0x60
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x80
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    addi.d    L,   L,   -16
+#else
+    addi.d    L,   L,   -1
+#endif
+    slli.d    T0,  L,  0x07
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x03
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    addi.d    OFF,   OFF,  0x10
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+    addi.d    I,   I,   -1  /* I-- */
+    blt       ZERO,I,   .L_N1_I1
+
+.L_N1_M8:
+    /* We have done M & 16, considering M=8/4/2/1 */
+    andi      I,   M,   15
+    beq       ZERO,I,   .L_N1_M0
+
+    andi      I,   M,   8
+    beq       ZERO,I,   .L_N1_M4
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x06
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x03
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  8
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  1
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 8 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+    xvfmul.d  D1,  U1, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x40
+    addi.d    B0,  B0, 0x08
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_N1_M8_L7 */
+    beq       ZERO,TL, .L_N1_M8_L7
+
+.L_N1_M8_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 16 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x08
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x08
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x08
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x08
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x08
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x08
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x08
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x08
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_N1_M8_TL1
+
+.L_N1_M8_L7:
+    /* if (!(L & 7)) goto L_N1_M8_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_N1_M8_L0
+
+.L_N1_M8_L71:
+    xvld     U0,   A0,    0x00
+    xvld     U1,   A0,    0x20
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+    xvfmadd.d  D1,  U1, U4, D1
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x40
+    addi.d     B0,  B0, 0x08
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_N1_M8_L71
+
+.L_N1_M8_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+    xvfmul.d  D1,   D1,  VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvld      U1,  C0,  0x20
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+    xvfmadd.d D1,  D1,  VALPHA,  U1
+#endif // #if defined(TRMMKERNEL)
+
+    /* Store C0 */
+    xvst      D0,  C0,  0x00
+    xvst      D1,  C0,  0x20
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x40
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    addi.d    L,   L,   -8
+#else
+    addi.d    L,   L,   -1
+#endif
+    slli.d    T0,  L,  0x06
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x03
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    addi.d    OFF,   OFF,  0x08
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N & 1) && (M & 8) ) End************/
+
+.L_N1_M4:
+    andi      I,   M,   4
+    beq       ZERO,I,   .L_N1_M2
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x05
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x03
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  4
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  1
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 4 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x20
+    addi.d    B0,  B0, 0x08
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_N1_M4_L7 */
+    beq       ZERO,TL, .L_N1_M4_L7
+
+.L_N1_M4_TL1: /* TL-- */
+           /***8-1***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x08
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x08
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x08
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x08
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x08
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x08
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x08
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x08
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_N1_M4_TL1
+
+.L_N1_M4_L7:
+    /* if (!(L & 7)) goto L_N1_M4_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_N1_M4_L0
+
+.L_N1_M4_L71:
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x20
+    addi.d     B0,  B0, 0x08
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_N1_M4_L71
+
+.L_N1_M4_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+#endif // #if defined(TRMMKERNEL)
+
+    /* Store C0 */
+    xvst      D0,  C0,  0x00
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x20
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    addi.d    L,   L,   -4
+#else
+    addi.d    L,   L,   -1
+#endif
+    slli.d    T0,  L,  0x05
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x03
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    addi.d    OFF,   OFF,  0x04
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N & 1) && (M & 4) ) End************/
+
+.L_N1_M2:
+    andi      I,   M,   2
+    beq       ZERO,I,   .L_N1_M1
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x04
+    add.d    A0,    A0,   T0
+    slli.d   T0,    OFF,  0x03
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  2
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  1
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 2 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x10
+    addi.d    B0,  B0, 0x08
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_N1_M2_L7 */
+    beq       ZERO,TL, .L_N1_M2_L7
+
+.L_N1_M2_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 2 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x08
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x08
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x08
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x08
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x08
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x08
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x08
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x08
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_N1_M2_TL1
+
+.L_N1_M2_L7:
+    /* if (!(L & 7)) goto L_N1_M2_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_N1_M2_L0
+
+.L_N1_M2_L71:
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x10
+    addi.d     B0,  B0, 0x08
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_N1_M2_L71
+
+.L_N1_M2_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+#endif // #if defined(TRMMKERNEL)
+
+    xvstelm.d D0,  C0,  0x00,    0x00
+    xvstelm.d D0,  C0,  0x08,    0x01
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x10
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    addi.d    L,   L,   -2
+#else
+    addi.d    L,   L,   -1
+#endif
+    slli.d    T0,  L,  0x04
+    add.d     A0,  A0, T0
+    slli.d    T0,  L,  0x03
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    addi.d    OFF,   OFF,  0x02
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N & 1 ) && (M & 2) ) End************/
+
+.L_N1_M1:
+    andi      I,   M,   1
+    beq       ZERO,I,   .L_N1_M0
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    move     B0,    B
+#else
+    slli.d   T0,    OFF,  0x03
+    add.d    A0,    A0,   T0
+    add.d    B0,    B,    T0
+#endif
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+    sub.d    L,     K,    OFF
+#elif defined(LEFT)
+    /* number of values in A */
+    addi.d   L,     OFF,  1
+#else
+    /* number of values in B */
+    addi.d   L,     OFF,  1
+#endif
+#else   // #if !defined(TRMMKERNEL)
+    move     B0,    B
+    move     L,     K /* L = bk */
+#endif
+
+    /* Load 1 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d     U4, B0, 0x00
+    /* line 1 */
+    xvfmul.d  D0,  U0, U4
+
+    /* Add stride for A0 and B0 */
+    addi.d    A0,  A0, 0x08
+    addi.d    B0,  B0, 0x08
+    /* Reduce L */
+    addi.d    L,   L,  -1
+    srai.d    TL,  L,  3  /* TL = (L-1) >> 3 */
+    /* if (TL < 1) goto L_N1_M1_L7 */
+    beq       ZERO,TL, .L_N1_M1_L7
+
+.L_N1_M1_TL1: /* TL-- */
+           /***8-1***/
+    /* Load 1 * 64 from A0 */
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x08
+
+           /***8-2***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x08
+
+           /***8-3***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x08
+
+           /***8-4***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x08
+
+           /***8-5***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x08
+
+           /***8-6***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x08
+
+           /***8-7***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x08
+
+           /***8-8***/
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x08
+
+    addi.d    TL,  TL, -1 /* TL-- */
+    blt       ZERO,TL, .L_N1_M1_TL1
+
+.L_N1_M1_L7:
+    /* if (!(L & 7)) goto L_N1_M1_L0 */
+    andi      TL,  L,   7
+    beq       TL,  ZERO,.L_N1_M1_L0
+
+.L_N1_M1_L71:
+    xvld     U0,   A0,    0x00
+
+    xvldrepl.d U4,  B0, 0x00
+    xvfmadd.d  D0,  U0, U4, D0
+
+    /* Add stride for A0, B0 */
+    addi.d     A0,  A0, 0x08
+    addi.d     B0,  B0, 0x08
+
+    addi.d     TL,  TL, -1
+    blt        ZERO,TL, .L_N1_M1_L71
+
+.L_N1_M1_L0:
+#if defined(TRMMKERNEL)
+    xvfmul.d  D0,   D0,  VALPHA
+#else
+    /* Load C0  */
+    xvld      U0,  C0,  0x00
+    xvfmadd.d D0,  D0,  VALPHA,  U0 /* D0 = U0 + (D0 * VALPHA) */
+#endif // #if defined(TRMMKERNEL)
+
+    xvstelm.d D0,  C0,  0x00,    0x00
+
+    /* Add stride for C */
+    addi.d    C0,  C0,  0x08
+
+#if defined(TRMMKERNEL)
+#if (defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+    sub.d     L,   K,   OFF
+#ifdef LEFT
+    addi.d    L,   L,   -1
+#else
+    addi.d    L,   L,   -1
+#endif
+    slli.d    T0,  L,  0x03
+    add.d     A0,  A0, T0
+    add.d     B0,  B0, T0
+#endif
+
+#ifdef LEFT
+    addi.d    OFF,   OFF,  0x01
+#endif
+#endif   // #if defined(TRMMKERNEL)
+
+/********LOOP (if(N & 1 ) && (M & 1) ) End************/
+
+.L_N1_M0:
+
+/************************* Condition 3 if((N & 1) && (M >> 4)) End !!! *************************
+*                                                   dgemm_core_16x1                                */
+
+.L_N0:
+    /* Restore regs */
+    LDARG    $r23,  $sp,   0
+    LDARG    $r24,  $sp,   8
+    LDARG    $r25,  $sp,   16
+    LDARG    $r26,  $sp,   24
+    LDARG    $r27,  $sp,   32
+    LD       $f23,  $sp,   40
+    addi.d   $sp,   $sp,   56
+
+    jirl    $r0, $r1, 0x0
+
+    EPILOGUE
diff --git a/kernel/loongarch64/dgemm_ncopy_16.S b/kernel/loongarch64/dgemm_ncopy_16.S
new file mode 100644
index 000000000..95c879031
--- /dev/null
+++ b/kernel/loongarch64/dgemm_ncopy_16.S
@@ -0,0 +1,691 @@
+/*******************************************************************************
+Copyright (c) 2021, 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 S1     $r12
+#define S2     $r13
+#define S3     $r14
+#define S4     $r15
+#define S5     $r16
+#define S6     $r17
+#define S7     $r18
+#define S8     $r19
+#define S9     $r20
+#define S10    $r23
+#define S11    $r24
+#define S12    $r25
+#define S13    $r26
+#define S14    $r27
+#define S15    $r28
+#define S16    $r29
+#define TD     $r30
+#define TS     $r31
+#define TL     $r7
+#define T0     $r6
+#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
+/* LASX vectors */
+#define U0     $xr0
+#define U1     $xr1
+#define U2     $xr2
+#define U3     $xr3
+#define U4     $xr4
+#define U5     $xr5
+#define U6     $xr6
+#define U7     $xr7
+#define U8     $xr8
+#define U9     $xr9
+#define U10    $xr10
+#define U11    $xr11
+#define U12    $xr12
+#define U13    $xr13
+#define U14    $xr14
+#define U15    $xr15
+#define D0     $xr16
+#define D1     $xr17
+#define D2     $xr18
+#define D3     $xr19
+#define D4     $xr20
+#define D5     $xr21
+#define D6     $xr22
+#define D7     $xr23
+#define D8     $xr24
+#define D9     $xr25
+#define D10    $xr26
+#define D11    $xr27
+#define D12    $xr28
+#define D13    $xr29
+#define D14    $xr30
+#define D15    $xr31
+
+    PROLOGUE
+
+    addi.d     $sp,  $sp,  -0x90
+    SDARG      $r23, $sp,  0x00
+    SDARG      $r24, $sp,  0x08
+    SDARG      $r25, $sp,  0x10
+    SDARG      $r26, $sp,  0x18
+    SDARG      $r27, $sp,  0x20
+    SDARG      $r28, $sp,  0x28
+    SDARG      $r29, $sp,  0x30
+    SDARG      $r30, $sp,  0x38
+    SDARG      $r31, $sp,  0x40
+    ST         $f23, $sp,  0x48
+    ST         $f24, $sp,  0x50
+    ST         $f25, $sp,  0x58
+    ST         $f26, $sp,  0x60
+    ST         $f27, $sp,  0x68
+    ST         $f28, $sp,  0x70
+    ST         $f29, $sp,  0x78
+    ST         $f30, $sp,  0x80
+    ST         $f31, $sp,  0x88
+
+    move       TD,   DST
+    move       TS,   SRC
+    slli.d     TL,   LDA,  0x03
+    slli.d     T0,   TL,   0x01
+    srai.d     J,    N,    0x04
+    beq        J,    ZERO, .L_N8
+
+.L_J1: /* J-- */
+    move       S1,   TS
+    add.d      S2,   TS,   TL
+    srai.d     I,    M,    0x03
+    add.d      S3,   S2,   TL
+    addi.d     J,    J,    -1
+    add.d      S4,   S3,   TL
+    add.d      S5,   S3,   T0
+    add.d      S6,   S4,   T0
+    add.d      S7,   S5,   T0
+    add.d      S8,   S6,   T0
+    add.d      S9,   S7,   T0
+    add.d      S10,  S8,   T0
+    add.d      S11,  S9,   T0
+    add.d      S12,  S10,  T0
+    add.d      S13,  S11,  T0
+    add.d      S14,  S12,  T0
+    add.d      S15,  S13,  T0
+    add.d      S16,  S14,  T0
+    add.d      TS,   S15,  T0
+    beq        I,    ZERO, .L_I7
+
+.L_I1: /* I-- */
+    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
+    xvld       U8,   S9,   0x00
+    xvld       U9,   S10,  0x00
+    xvld       U10,  S11,  0x00
+    xvld       U11,  S12,  0x00
+    xvld       U12,  S13,  0x00
+    xvld       U13,  S14,  0x00
+    xvld       U14,  S15,  0x00
+    xvld       U15,  S16,  0x00
+
+    xvpackev.d D0,   U1,   U0
+    xvpackod.d D1,   U1,   U0
+    xvpackev.d D2,   U3,   U2
+    xvpackod.d D3,   U3,   U2
+    xvpackev.d D4,   U5,   U4
+    xvpackod.d D5,   U5,   U4
+    xvpackev.d D6,   U7,   U6
+    xvpackod.d D7,   U7,   U6
+
+    xvpackev.d D8,   U9,   U8
+    xvpackod.d D9,   U9,   U8
+    xvpackev.d D10,  U11,  U10
+    xvpackod.d D11,  U11,  U10
+    xvpackev.d D12,  U13,  U12
+    xvpackod.d D13,  U13,  U12
+    xvpackev.d D14,  U15,  U14
+    xvpackod.d D15,  U15,  U14
+
+    xvand.v    U0,   D0,   D0
+    xvpermi.q  D0,   D2,   0x02  // 0
+    xvand.v    U4,   D4,   D4
+    xvpermi.q  D4,   D6,   0x02  // 1
+    xvand.v    U1,   D1,   D1
+    xvpermi.q  D1,   D3,   0x02  // 4
+    xvand.v    U5,   D5,   D5
+    xvpermi.q  D5,   D7,   0x02  // 5
+    xvpermi.q  D2,   U0,   0x31  // 8
+    xvpermi.q  D6,   U4,   0x31  // 9
+    xvpermi.q  D3,   U1,   0x31  // 12
+    xvpermi.q  D7,   U5,   0x31  // 13
+
+    xvand.v    U8,   D8,   D8
+    xvpermi.q  D8,   D10,  0x02  // 2
+    xvand.v    U12,  D12,  D12
+    xvpermi.q  D12,  D14,  0x02  // 3
+    xvand.v    U9,   D9,   D9
+    xvpermi.q  D9,   D11,  0x02  // 6
+    xvand.v    U13,  D13,  D13
+    xvpermi.q  D13,  D15,  0x02  // 7
+    xvpermi.q  D10,  U8,   0x31  // 10
+    xvpermi.q  D14,  U12,  0x31  // 11
+    xvpermi.q  D11,  U9,   0x31  // 14
+    xvpermi.q  D15,  U13,  0x31  // 15
+
+    xvst       D0,   TD,   0x00  // 0
+    xvst       D4,   TD,   0x20  // 1
+    xvst       D8,   TD,   0x40  // 2
+    xvst       D12,  TD,   0x60  // 3
+    xvst       D1,   TD,   0x80  // 4
+    xvst       D5,   TD,   0xA0  // 5
+    xvst       D9,   TD,   0xC0  // 6
+    xvst       D13,  TD,   0xE0  // 7
+    addi.d     TD,   TD,   0x100
+    xvst       D2,   TD,   0x00  // 8
+    xvst       D6,   TD,   0x20  // 9
+    xvst       D10,  TD,   0x40  // 10
+    xvst       D14,  TD,   0x60  // 11
+    xvst       D3,   TD,   0x80  // 12
+    xvst       D7,   TD,   0xA0  // 13
+    xvst       D11,  TD,   0xC0  // 14
+    xvst       D15,  TD,   0xE0  // 15
+    addi.d     TD,   TD,   0x100
+
+    xvld       U0,   S1,   0x20
+    xvld       U1,   S2,   0x20
+    xvld       U2,   S3,   0x20
+    xvld       U3,   S4,   0x20
+    xvld       U4,   S5,   0x20
+    xvld       U5,   S6,   0x20
+    xvld       U6,   S7,   0x20
+    xvld       U7,   S8,   0x20
+    xvld       U8,   S9,   0x20
+    xvld       U9,   S10,  0x20
+    xvld       U10,  S11,  0x20
+    xvld       U11,  S12,  0x20
+    xvld       U12,  S13,  0x20
+    xvld       U13,  S14,  0x20
+    xvld       U14,  S15,  0x20
+    xvld       U15,  S16,  0x20
+
+    xvpackev.d D0,   U1,   U0
+    xvpackod.d D1,   U1,   U0
+    xvpackev.d D2,   U3,   U2
+    xvpackod.d D3,   U3,   U2
+    xvpackev.d D4,   U5,   U4
+    xvpackod.d D5,   U5,   U4
+    xvpackev.d D6,   U7,   U6
+    xvpackod.d D7,   U7,   U6
+
+    xvpackev.d D8,   U9,   U8
+    xvpackod.d D9,   U9,   U8
+    xvpackev.d D10,  U11,  U10
+    xvpackod.d D11,  U11,  U10
+    xvpackev.d D12,  U13,  U12
+    xvpackod.d D13,  U13,  U12
+    xvpackev.d D14,  U15,  U14
+    xvpackod.d D15,  U15,  U14
+
+    xvand.v    U0,   D0,   D0
+    xvpermi.q  D0,   D2,   0x02  // 0
+    xvand.v    U4,   D4,   D4
+    xvpermi.q  D4,   D6,   0x02  // 1
+    xvand.v    U1,   D1,   D1
+    xvpermi.q  D1,   D3,   0x02  // 4
+    xvand.v    U5,   D5,   D5
+    xvpermi.q  D5,   D7,   0x02  // 5
+    xvpermi.q  D2,   U0,   0x31  // 8
+    xvpermi.q  D6,   U4,   0x31  // 9
+    xvpermi.q  D3,   U1,   0x31  // 12
+    xvpermi.q  D7,   U5,   0x31  // 13
+
+    xvand.v    U8,   D8,   D8
+    xvpermi.q  D8,   D10,  0x02  // 2
+    xvand.v    U12,  D12,  D12
+    xvpermi.q  D12,  D14,  0x02  // 3
+    xvand.v    U9,   D9,   D9
+    xvpermi.q  D9,   D11,  0x02  // 6
+    xvand.v    U13,  D13,  D13
+    xvpermi.q  D13,  D15,  0x02  // 7
+    xvpermi.q  D10,  U8,   0x31  // 10
+    xvpermi.q  D14,  U12,  0x31  // 11
+    xvpermi.q  D11,  U9,   0x31  // 14
+    xvpermi.q  D15,  U13,  0x31  // 15
+
+    xvst       D0,   TD,   0x00  // 0
+    xvst       D4,   TD,   0x20  // 1
+    xvst       D8,   TD,   0x40  // 2
+    xvst       D12,  TD,   0x60  // 3
+    xvst       D1,   TD,   0x80  // 4
+    xvst       D5,   TD,   0xA0  // 5
+    xvst       D9,   TD,   0xC0  // 6
+    xvst       D13,  TD,   0xE0  // 7
+    addi.d     TD,   TD,   0x100
+    xvst       D2,   TD,   0x00  // 8
+    xvst       D6,   TD,   0x20  // 9
+    xvst       D10,  TD,   0x40  // 10
+    xvst       D14,  TD,   0x60  // 11
+    xvst       D3,   TD,   0x80  // 12
+    xvst       D7,   TD,   0xA0  // 13
+    xvst       D11,  TD,   0xC0  // 14
+    xvst       D15,  TD,   0xE0  // 15
+    addi.d     TD,   TD,   0x100
+
+
+    addi.d     S1,   S1,   0x40
+    addi.d     S2,   S2,   0x40
+    addi.d     S3,   S3,   0x40
+    addi.d     S4,   S4,   0x40
+    addi.d     S5,   S5,   0x40
+    addi.d     S6,   S6,   0x40
+    addi.d     S7,   S7,   0x40
+    addi.d     S8,   S8,   0x40
+    addi.d     S9,   S9,   0x40
+    addi.d     S10,  S10,  0x40
+    addi.d     S11,  S11,  0x40
+    addi.d     S12,  S12,  0x40
+    addi.d     S13,  S13,  0x40
+    addi.d     S14,  S14,  0x40
+    addi.d     S15,  S15,  0x40
+    addi.d     S16,  S16,  0x40
+
+    addi.d     I,    I,    -1
+    blt        ZERO, I,    .L_I1
+
+.L_I7:
+    andi      I,     M,    0x07
+    beq       I,     ZERO, .L_I0
+
+.L_II1: /* I-- */
+    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,    TD,  0x00
+    addi.d    S1,    S1,  0x08
+    fst.d     F1,    TD,  0x08
+    addi.d    S2,    S2,  0x08
+    fst.d     F2,    TD,  0x10
+    addi.d    S3,    S3,  0x08
+    fst.d     F3,    TD,  0x18
+    addi.d    S4,    S4,  0x08
+    fst.d     F4,    TD,  0x20
+    addi.d    S5,    S5,  0x08
+    fst.d     F5,    TD,  0x28
+    addi.d    S6,    S6,  0x08
+    fst.d     F6,    TD,  0x30
+    addi.d    S7,    S7,  0x08
+    fst.d     F7,    TD,  0x38
+    addi.d    S8,    S8,  0x08
+    addi.d    TD,    TD,  0x40
+
+    fld.d     F0,    S9,  0x00
+    fld.d     F1,    S10, 0x00
+    fld.d     F2,    S11, 0x00
+    fld.d     F3,    S12, 0x00
+    fld.d     F4,    S13, 0x00
+    fld.d     F5,    S14, 0x00
+    fld.d     F6,    S15, 0x00
+    fld.d     F7,    S16, 0x00
+
+    fst.d     F0,    TD,  0x00
+    addi.d    S9,    S9,  0x08
+    fst.d     F1,    TD,  0x08
+    addi.d    S10,   S10, 0x08
+    fst.d     F2,    TD,  0x10
+    addi.d    S11,   S11, 0x08
+    fst.d     F3,    TD,  0x18
+    addi.d    S12,   S12, 0x08
+    fst.d     F4,    TD,  0x20
+    addi.d    S13,   S13, 0x08
+    fst.d     F5,    TD,  0x28
+    addi.d    S14,   S14, 0x08
+    fst.d     F6,    TD,  0x30
+    addi.d    S15,   S15, 0x08
+    fst.d     F7,    TD,  0x38
+    addi.d    S16,   S16, 0x08
+    addi.d    TD,    TD,  0x40
+
+    addi.d    I,     I,   -1
+    blt       ZERO,  I,   .L_II1
+
+.L_I0:
+    blt       ZERO,  J,   .L_J1
+
+.L_N8:
+    andi      J,     N,   0x08
+    beq       ZERO,  J,   .L_N4
+
+    move       S1,   TS
+    add.d      S2,   TS,   TL
+    srai.d     I,    M,    0x03
+    add.d      S3,   S2,   TL
+    add.d      S4,   S2,   T0
+    add.d      S5,   S3,   T0
+    add.d      S6,   S4,   T0
+    add.d      S7,   S5,   T0
+    add.d      S8,   S6,   T0
+    add.d      TS,   S7,   T0
+    beq        I,    ZERO, .L_8I3
+
+.L_8I1:  /* I-- */
+    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
+
+    xvpackev.d D0,   U1,   U0
+    xvpackod.d D1,   U1,   U0
+    xvpackev.d D2,   U3,   U2
+    xvpackod.d D3,   U3,   U2
+    xvpackev.d D4,   U5,   U4
+    xvpackod.d D5,   U5,   U4
+    xvpackev.d D6,   U7,   U6
+    xvpackod.d D7,   U7,   U6
+
+    xvand.v    U0,   D0,   D0
+    xvpermi.q  D0,   D2,   0x02  // 0
+    xvand.v    U4,   D4,   D4
+    xvpermi.q  D4,   D6,   0x02  // 1
+    xvand.v    U1,   D1,   D1
+    xvpermi.q  D1,   D3,   0x02  // 2
+    xvand.v    U5,   D5,   D5
+    xvpermi.q  D5,   D7,   0x02  // 3
+    xvpermi.q  D2,   U0,   0x31  // 4
+    xvpermi.q  D6,   U4,   0x31  // 5
+    xvpermi.q  D3,   U1,   0x31  // 6
+    xvpermi.q  D7,   U5,   0x31  // 7
+
+    xvst       D0,   TD,   0x00
+    xvst       D4,   TD,   0x20
+    xvst       D1,   TD,   0x40
+    xvst       D5,   TD,   0x60
+    xvst       D2,   TD,   0x80
+    xvst       D6,   TD,   0xA0
+    xvst       D3,   TD,   0xC0
+    xvst       D7,   TD,   0xE0
+    addi.d     TD,   TD,   0x100
+
+    xvld       U0,   S1,   0x20
+    xvld       U1,   S2,   0x20
+    xvld       U2,   S3,   0x20
+    xvld       U3,   S4,   0x20
+    xvld       U4,   S5,   0x20
+    xvld       U5,   S6,   0x20
+    xvld       U6,   S7,   0x20
+    xvld       U7,   S8,   0x20
+
+    xvpackev.d D0,   U1,   U0
+    xvpackod.d D1,   U1,   U0
+    xvpackev.d D2,   U3,   U2
+    xvpackod.d D3,   U3,   U2
+    xvpackev.d D4,   U5,   U4
+    xvpackod.d D5,   U5,   U4
+    xvpackev.d D6,   U7,   U6
+    xvpackod.d D7,   U7,   U6
+
+    xvand.v    U0,   D0,   D0
+    xvpermi.q  D0,   D2,   0x02  // 0
+    xvand.v    U4,   D4,   D4
+    xvpermi.q  D4,   D6,   0x02  // 1
+    xvand.v    U1,   D1,   D1
+    xvpermi.q  D1,   D3,   0x02  // 2
+    xvand.v    U5,   D5,   D5
+    xvpermi.q  D5,   D7,   0x02  // 3
+    xvpermi.q  D2,   U0,   0x31  // 4
+    xvpermi.q  D6,   U4,   0x31  // 5
+    xvpermi.q  D3,   U1,   0x31  // 6
+    xvpermi.q  D7,   U5,   0x31  // 7
+
+    xvst       D0,   TD,   0x00
+    xvst       D4,   TD,   0x20
+    xvst       D1,   TD,   0x40
+    xvst       D5,   TD,   0x60
+    xvst       D2,   TD,   0x80
+    xvst       D6,   TD,   0xA0
+    xvst       D3,   TD,   0xC0
+    xvst       D7,   TD,   0xE0
+    addi.d     TD,   TD,   0x100
+
+    addi.d     S1,   S1,   0x40
+    addi.d     S2,   S2,   0x40
+    addi.d     S3,   S3,   0x40
+    addi.d     S4,   S4,   0x40
+    addi.d     S5,   S5,   0x40
+    addi.d     S6,   S6,   0x40
+    addi.d     S7,   S7,   0x40
+    addi.d     S8,   S8,   0x40
+
+    addi.d     I,    I,    -1
+    blt        ZERO, I,    .L_8I1
+
+.L_8I3:
+    andi      I,     M,    0x07
+    beq       I,     ZERO, .L_N4
+
+.L_8I11:
+    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,    TD,  0x00
+    addi.d    S1,    S1,  0x08
+    fst.d     F1,    TD,  0x08
+    addi.d    S2,    S2,  0x08
+    fst.d     F2,    TD,  0x10
+    addi.d    S3,    S3,  0x08
+    fst.d     F3,    TD,  0x18
+    addi.d    S4,    S4,  0x08
+    fst.d     F4,    TD,  0x20
+    addi.d    S5,    S5,  0x08
+    fst.d     F5,    TD,  0x28
+    addi.d    S6,    S6,  0x08
+    fst.d     F6,    TD,  0x30
+    addi.d    S7,    S7,  0x08
+    fst.d     F7,    TD,  0x38
+    addi.d    S8,    S8,  0x08
+
+    addi.d    TD,    TD,  0x40
+    addi.d    I,     I,   -1
+    blt       ZERO,  I,   .L_8I11
+
+.L_N4:
+    andi      J,     N,   0x04
+    beq       ZERO,  J,   .L_N2
+
+    move       S1,   TS
+    add.d      S2,   TS,   TL
+    srai.d     I,    M,    0x02
+    add.d      S3,   S2,   TL
+    add.d      S4,   S2,   T0
+    add.d      TS,   S3,   T0
+    beq        I,    ZERO, .L_I3
+
+.L_4I1: /* I-- */
+    xvld       U0,   S1,   0x00
+    xvld       U1,   S2,   0x00
+    xvld       U2,   S3,   0x00
+    xvld       U3,   S4,   0x00
+
+    xvpackev.d D0,   U1,   U0
+    xvpackod.d D1,   U1,   U0
+    xvpackev.d D2,   U3,   U2
+    xvpackod.d D3,   U3,   U2
+
+    xvand.v    U0,   D0,   D0
+    xvpermi.q  D0,   D2,   0x02  // 0
+    xvand.v    U1,   D1,   D1
+    xvpermi.q  D1,   D3,   0x02  // 1
+    xvpermi.q  D2,   U0,   0x31  // 2
+    xvpermi.q  D3,   U1,   0x31  // 3
+
+    xvst       D0,   TD,   0x00
+    xvst       D1,   TD,   0x20
+    xvst       D2,   TD,   0x40
+    xvst       D3,   TD,   0x60
+
+    addi.d     S1,   S1,   0x20
+    addi.d     S2,   S2,   0x20
+    addi.d     S3,   S3,   0x20
+    addi.d     S4,   S4,   0x20
+    addi.d     TD,   TD,   0x80
+
+    addi.d     I,    I,    -1
+    blt        ZERO, I,    .L_4I1
+
+.L_I3:
+    andi      I,     M,    0x03
+    beq       I,     ZERO, .L_N2
+
+.L_4II1:
+    fld.d     F0,    S1,  0x00
+    fld.d     F1,    S2,  0x00
+    fld.d     F2,    S3,  0x00
+    fld.d     F3,    S4,  0x00
+
+    fst.d     F0,    TD,  0x00
+    addi.d    S1,    S1,  0x08
+    fst.d     F1,    TD,  0x08
+    addi.d    S2,    S2,  0x08
+    fst.d     F2,    TD,  0x10
+    addi.d    S3,    S3,  0x08
+    fst.d     F3,    TD,  0x18
+    addi.d    S4,    S4,  0x08
+
+    addi.d    TD,    TD,  0x20
+    addi.d    I,     I,   -1
+    blt       ZERO,  I,   .L_4II1
+
+.L_N2:
+    andi      J,     N,   0x02
+    beq       ZERO,  J,   .L_N1
+
+    move       S1,   TS
+    add.d      S2,   TS,   TL
+    srai.d     I,    M,    0x01
+    add.d      TS,   S2,   TL
+    beq        I,    ZERO, .L_NI1
+
+.L_2I1: /* I-- */
+    xvld       U0,   S1,   0x00
+    xvld       U1,   S2,   0x00
+
+    xvpackev.d D0,   U1,   U0
+    xvpackod.d D1,   U1,   U0
+
+    xvpermi.q  D0,   D1,   0x02  // 0
+
+    xvst       D0,   TD,   0x00
+
+    addi.d     S1,   S1,   0x10
+    addi.d     S2,   S2,   0x10
+    addi.d     TD,   TD,   0x20
+
+    addi.d     I,    I,    -1
+    blt        ZERO, I,    .L_2I1
+
+.L_NI1:
+    andi      I,     M,    0x01
+    beq       I,     ZERO, .L_N1
+
+
+    fld.d     F0,    S1,  0x00
+    fld.d     F1,    S2,  0x00
+
+    fst.d     F0,    TD,  0x00
+    addi.d    S1,    S1,  0x08
+    fst.d     F1,    TD,  0x08
+    addi.d    S2,    S2,  0x08
+    addi.d    TD,    TD,  0x10
+
+.L_N1:
+    move      S1,    TS
+    beq       ZERO,  M,   .L_N0
+
+.L_M1:
+    fld.d     F0,    S1,  0x00
+    addi.d    S1,    S1,  0x08
+    fst.d     F0,    TD,  0x00
+    addi.d    TD,    TD,  0x08
+    addi.d    M,     M,   -1
+    blt       ZERO,  M,   .L_M1
+
+.L_N0:
+    LDARG      $r23, $sp,  0x00
+    LDARG      $r24, $sp,  0x08
+    LDARG      $r25, $sp,  0x10
+    LDARG      $r26, $sp,  0x18
+    LDARG      $r27, $sp,  0x20
+    LDARG      $r28, $sp,  0x28
+    LDARG      $r29, $sp,  0x30
+    LDARG      $r30, $sp,  0x38
+    LDARG      $r31, $sp,  0x40
+    LD         $f23, $sp,  0x48
+    LD         $f24, $sp,  0x50
+    LD         $f25, $sp,  0x58
+    LD         $f26, $sp,  0x60
+    LD         $f27, $sp,  0x68
+    LD         $f28, $sp,  0x70
+    LD         $f29, $sp,  0x78
+    LD         $f30, $sp,  0x80
+    LD         $f31, $sp,  0x88
+    addi.d     $sp,  $sp,  0x90
+    jirl       $r0,  $r1,  0x00
+
+    EPILOGUE
diff --git a/kernel/loongarch64/dgemm_ncopy_4.S b/kernel/loongarch64/dgemm_ncopy_4.S
new file mode 100644
index 000000000..b1f322a06
--- /dev/null
+++ b/kernel/loongarch64/dgemm_ncopy_4.S
@@ -0,0 +1,237 @@
+/*******************************************************************************
+Copyright (c) 2021, 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 S1     $r12
+#define S2     $r13
+#define S3     $r14
+#define S4     $r15
+#define S5     $r16
+#define S6     $r17
+#define S7     $r18
+#define S8     $r19
+#define TD     $r20
+#define TS     $r11
+#define TL     $r7
+#define T0     $r23
+#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
+/* LASX vectors */
+#define U0     $xr0
+#define U1     $xr1
+#define U2     $xr2
+#define U3     $xr3
+#define U4     $xr4
+#define U5     $xr5
+#define U6     $xr6
+#define U7     $xr7
+#define D0     $xr14
+#define D1     $xr8
+#define D2     $xr9
+#define D3     $xr10
+#define D4     $xr11
+#define D5     $xr12
+#define D6     $xr13
+#define D7     $xr15
+
+    PROLOGUE
+
+    addi.d     $sp,  $sp,  -8
+    SDARG      $r23, $sp,  0
+
+    move       TD,   DST
+    move       TS,   SRC
+    slli.d     TL,   LDA,  0x03
+    slli.d     T0,   TL,   0x01
+    srai.d     J,    N,    0x02
+    beq        J,    ZERO, .L_N2
+
+.L_J1: /* J-- */
+    move       S1,   TS
+    add.d      S2,   TS,   TL
+    srai.d     I,    M,    0x02
+    add.d      S3,   S2,   TL
+    add.d      S4,   S2,   T0
+    add.d      TS,   S3,   T0
+    addi.d     J,    J,    -1
+    beq        I,    ZERO, .L_I3
+
+.L_I1: /* I-- */
+    xvld       U0,   S1,   0x00
+    xvld       U1,   S2,   0x00
+    xvld       U2,   S3,   0x00
+    xvld       U3,   S4,   0x00
+
+    xvpackev.d D0,   U1,   U0
+    xvpackod.d D1,   U1,   U0
+    xvpackev.d D2,   U3,   U2
+    xvpackod.d D3,   U3,   U2
+
+    xvand.v    U0,   D0,   D0
+    xvpermi.q  D0,   D2,   0x02  // 0
+    xvand.v    U1,   D1,   D1
+    xvpermi.q  D1,   D3,   0x02  // 1
+    xvpermi.q  D2,   U0,   0x31  // 2
+    xvpermi.q  D3,   U1,   0x31  // 3
+
+    xvst       D0,   TD,   0x00
+    xvst       D1,   TD,   0x20
+    xvst       D2,   TD,   0x40
+    xvst       D3,   TD,   0x60
+
+    addi.d     S1,   S1,   0x20
+    addi.d     S2,   S2,   0x20
+    addi.d     S3,   S3,   0x20
+    addi.d     S4,   S4,   0x20
+    addi.d     TD,   TD,   0x80
+
+    addi.d     I,    I,    -1
+    blt        ZERO, I,    .L_I1
+
+.L_I3:
+    andi      I,     M,    0x03
+    beq       I,     ZERO, .L_I0
+
+.L_II1:
+    fld.d     F0,    S1,  0x00
+    fld.d     F1,    S2,  0x00
+    fld.d     F2,    S3,  0x00
+    fld.d     F3,    S4,  0x00
+
+    fst.d     F0,    TD,  0x00
+    addi.d    S1,    S1,  0x08
+    fst.d     F1,    TD,  0x08
+    addi.d    S2,    S2,  0x08
+    fst.d     F2,    TD,  0x10
+    addi.d    S3,    S3,  0x08
+    fst.d     F3,    TD,  0x18
+    addi.d    S4,    S4,  0x08
+
+    addi.d    TD,    TD,  0x20
+    addi.d    I,     I,   -1
+    blt       ZERO,  I,   .L_II1
+
+.L_I0:
+    blt       ZERO,  J,   .L_J1
+
+.L_N2:
+    andi      J,     N,   0x02
+    beq       ZERO,  J,   .L_N1
+
+    move       S1,   TS
+    add.d      S2,   TS,   TL
+    srai.d     I,    M,    0x02
+    add.d      TS,   S2,   TL
+    beq        I,    ZERO, .L_2I3
+
+.L_2I1: /* I-- */
+    xvld       U0,   S1,   0x00
+    xvld       U1,   S2,   0x00
+
+    xvpackev.d D0,   U1,   U0
+    xvpackod.d D1,   U1,   U0
+
+    xvand.v    U0,   D0,   D0
+    xvpermi.q  D0,   D1,   0x02  // 0
+    xvpermi.q  D1,   U0,   0x31  // 1
+
+    xvst       D0,   TD,   0x00
+    xvst       D1,   TD,   0x20
+    addi.d     S1,   S1,   0x20
+    addi.d     S2,   S2,   0x20
+    addi.d     TD,   TD,   0x40
+    addi.d     I,    I,    -1
+    blt        ZERO, I,    .L_2I1
+
+.L_2I3:
+    andi       I,    M,    0x03
+    beq        ZERO, I,    .L_N1
+
+.L_2II1: /* I-- */
+    fld.d      F0,   S1,   0x00
+    fld.d      F1,   S2,   0x00
+    fst.d      F0,   TD,   0x00
+    addi.d     I,    I,    -1
+    fst.d      F1,   TD,   0x08
+    addi.d     S1,   S1,   0x08
+    addi.d     S2,   S2,   0x08
+    addi.d     TD,   TD,   0x10
+    blt        ZERO, I,    .L_2II1
+
+.L_N1:
+    andi       J,    N,    0x01
+    beq        ZERO, J,    .L_N0
+
+    move       S1,   TS
+    srai.d     I,    M,    0x02
+    beq        ZERO, I,    .L_1I3
+
+.L_1I1:
+    xvld       U0,   S1,   0x00
+    addi.d     S1,   S1,   0x20
+    xvst       U0,   TD,   0x00
+    addi.d     I,    I,    -1
+    addi.d     TD,   TD,   0x20
+    blt        ZERO, I,    .L_1I1
+
+.L_1I3:
+    andi       I,    M,    0x03
+    beq        ZERO, I,    .L_N0
+
+.L_1II1:
+    fld.d      F0,   S1,   0x00
+    addi.d     S1,   S1,   0x08
+    fst.d      F0,   TD,   0x00
+    addi.d     I,    I,    -1
+    addi.d     TD,   TD,   0x08
+    blt        ZERO, I,    .L_1II1
+
+.L_N0:
+    LDARG     $r23,  $sp, 0
+    addi.d    $sp,   $sp, 8
+    jirl      $r0,   $r1, 0x00
+
+    EPILOGUE
diff --git a/kernel/loongarch64/dgemm_tcopy_16.S b/kernel/loongarch64/dgemm_tcopy_16.S
new file mode 100644
index 000000000..afafe5b37
--- /dev/null
+++ b/kernel/loongarch64/dgemm_tcopy_16.S
@@ -0,0 +1,710 @@
+/*******************************************************************************
+Copyright (c) 2021, 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 P5     $r27
+#define T0     $r28
+#define T1     $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
+/* LASX vectors */
+#define U0     $xr0
+#define U1     $xr1
+#define U2     $xr2
+#define U3     $xr3
+#define U4     $xr4
+#define U5     $xr5
+#define U6     $xr6
+#define U7     $xr7
+
+    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
+
+    srai.d     T0,     N,     0x04
+    srai.d     T1,     N,     0x03
+    slli.d     T0,     T0,    0x04
+    slli.d     T1,     T1,    0x03
+    mul.d      P2,     M,     T0
+    mul.d      P3,     M,     T1
+    slli.d     P2,     P2,    0x03
+    slli.d     P3,     P3,    0x03
+    add.d      P2,     DST,   P2
+    add.d      P3,     DST,   P3
+
+    srai.d     T0,     N,     0x02
+    srai.d     T1,     N,     0x01
+    slli.d     T0,     T0,    0x02
+    slli.d     T1,     T1,    0x01
+    mul.d      P4,     M,     T0
+    mul.d      P5,     M,     T1
+    slli.d     P4,     P4,    0x03
+    slli.d     P5,     P5,    0x03
+    add.d      P4,     DST,   P4
+    add.d      P5,     DST,   P5
+
+    slli.d     TL,     LDA,   0x03
+    srai.d     J,      M,     0x03
+    slli.d     T0,     TL,    0x01
+    slli.d     T1,     M,     0x07
+    beq        ZERO,   J,     .L_M7
+
+.L_J1: /* J-- */
+    move       S1,     S0
+    add.d      S2,     S0,    TL
+    add.d      S3,     S1,    T0
+    add.d      S4,     S2,    T0
+    add.d      S5,     S3,    T0
+    add.d      S6,     S4,    T0
+    add.d      S7,     S5,    T0
+    add.d      S8,     S6,    T0
+    add.d      S0,     S7,    T0
+
+    move       P1,     P0
+    addi.d     P0,     P0,    0x400
+
+    srai.d     I,      N,     0x04
+    addi.d     J,      J,     -1
+    beq        ZERO,   I,     .L_N15
+
+.L_I1: /* I-- */
+    xvld       U0,     S1,    0x00
+    xvld       U1,     S1,    0x20
+    xvld       U2,     S1,    0x40
+    xvld       U3,     S1,    0x60
+    xvld       U4,     S2,    0x00
+    xvld       U5,     S2,    0x20
+    xvld       U6,     S2,    0x40
+    xvld       U7,     S2,    0x60
+
+    xvst       U0,     P1,    0x00
+    xvst       U1,     P1,    0x20
+    xvst       U2,     P1,    0x40
+    xvst       U3,     P1,    0x60
+    xvst       U4,     P1,    0x80
+    xvst       U5,     P1,    0xA0
+    xvst       U6,     P1,    0xC0
+    xvst       U7,     P1,    0xE0
+
+    xvld       U0,     S3,    0x00
+    xvld       U1,     S3,    0x20
+    xvld       U2,     S3,    0x40
+    xvld       U3,     S3,    0x60
+    xvld       U4,     S4,    0x00
+    xvld       U5,     S4,    0x20
+    xvld       U6,     S4,    0x40
+    xvld       U7,     S4,    0x60
+
+    xvst       U0,     P1,    0x100
+    xvst       U1,     P1,    0x120
+    xvst       U2,     P1,    0x140
+    xvst       U3,     P1,    0x160
+    xvst       U4,     P1,    0x180
+    xvst       U5,     P1,    0x1A0
+    xvst       U6,     P1,    0x1C0
+    xvst       U7,     P1,    0x1E0
+
+    xvld       U0,     S5,    0x00
+    xvld       U1,     S5,    0x20
+    xvld       U2,     S5,    0x40
+    xvld       U3,     S5,    0x60
+    xvld       U4,     S6,    0x00
+    xvld       U5,     S6,    0x20
+    xvld       U6,     S6,    0x40
+    xvld       U7,     S6,    0x60
+
+    xvst       U0,     P1,    0x200
+    xvst       U1,     P1,    0x220
+    xvst       U2,     P1,    0x240
+    xvst       U3,     P1,    0x260
+    xvst       U4,     P1,    0x280
+    xvst       U5,     P1,    0x2A0
+    xvst       U6,     P1,    0x2C0
+    xvst       U7,     P1,    0x2E0
+
+    xvld       U0,     S7,    0x00
+    xvld       U1,     S7,    0x20
+    xvld       U2,     S7,    0x40
+    xvld       U3,     S7,    0x60
+    xvld       U4,     S8,    0x00
+    xvld       U5,     S8,    0x20
+    xvld       U6,     S8,    0x40
+    xvld       U7,     S8,    0x60
+
+    xvst       U0,     P1,    0x300
+    xvst       U1,     P1,    0x320
+    xvst       U2,     P1,    0x340
+    xvst       U3,     P1,    0x360
+    xvst       U4,     P1,    0x380
+    xvst       U5,     P1,    0x3A0
+    xvst       U6,     P1,    0x3C0
+    xvst       U7,     P1,    0x3E0
+
+    addi.d     S1,     S1,    0x80
+    addi.d     S2,     S2,    0x80
+    addi.d     S3,     S3,    0x80
+    addi.d     S4,     S4,    0x80
+    addi.d     S5,     S5,    0x80
+    addi.d     S6,     S6,    0x80
+    addi.d     S7,     S7,    0x80
+    addi.d     S8,     S8,    0x80
+    addi.d     I,      I,     -1
+    add.d      P1,     P1,    T1
+    blt        ZERO,   I,     .L_I1
+
+.L_N15:
+    andi       I,      N,     0x08
+    beq        ZERO,   I,     .L_N7
+
+    xvld       U0,     S1,    0x00
+    xvld       U1,     S1,    0x20
+    xvld       U2,     S2,    0x00
+    xvld       U3,     S2,    0x20
+    xvld       U4,     S3,    0x00
+    xvld       U5,     S3,    0x20
+    xvld       U6,     S4,    0x00
+    xvld       U7,     S4,    0x20
+
+    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
+
+    xvld       U0,     S5,    0x00
+    xvld       U1,     S5,    0x20
+    xvld       U2,     S6,    0x00
+    xvld       U3,     S6,    0x20
+    xvld       U4,     S7,    0x00
+    xvld       U5,     S7,    0x20
+    xvld       U6,     S8,    0x00
+    xvld       U7,     S8,    0x20
+
+    xvst       U0,     P2,    0x100
+    xvst       U1,     P2,    0x120
+    xvst       U2,     P2,    0x140
+    xvst       U3,     P2,    0x160
+    xvst       U4,     P2,    0x180
+    xvst       U5,     P2,    0x1A0
+    xvst       U6,     P2,    0x1C0
+    xvst       U7,     P2,    0x1E0
+
+    addi.d     S1,     S1,    0x40
+    addi.d     S2,     S2,    0x40
+    addi.d     S3,     S3,    0x40
+    addi.d     S4,     S4,    0x40
+    addi.d     S5,     S5,    0x40
+    addi.d     S6,     S6,    0x40
+    addi.d     S7,     S7,    0x40
+    addi.d     S8,     S8,    0x40
+    addi.d     P2,     P2,    0x200
+
+.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,     P3,    0x00
+    xvst       U1,     P3,    0x20
+    xvst       U2,     P3,    0x40
+    xvst       U3,     P3,    0x60
+    xvst       U4,     P3,    0x80
+    xvst       U5,     P3,    0xA0
+    xvst       U6,     P3,    0xC0
+    xvst       U7,     P3,    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     P3,     P3,    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,     P4,    0x00
+    xvst       U2,     P4,    0x20
+    xvst       U4,     P4,    0x40
+    xvst       U6,     P4,    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     P4,     P4,    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,     P5,    0x00
+    fst.d      F1,     P5,    0x08
+    fst.d      F2,     P5,    0x10
+    fst.d      F3,     P5,    0x18
+    fst.d      F4,     P5,    0x20
+    fst.d      F5,     P5,    0x28
+    fst.d      F6,     P5,    0x30
+    fst.d      F7,     P5,    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     P5,     P5,    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,    T0
+    add.d      S4,     S2,    T0
+    add.d      S0,     S3,    T0
+
+    move       P1,     P0
+    addi.d     P0,     P0,    0x200
+
+    srai.d     I,      N,     0x04
+    beq        ZERO,   I,     .L_4N15
+
+.L_4I1: /* I-- */
+    xvld       U0,     S1,    0x00
+    xvld       U1,     S1,    0x20
+    xvld       U2,     S1,    0x40
+    xvld       U3,     S1,    0x60
+    xvld       U4,     S2,    0x00
+    xvld       U5,     S2,    0x20
+    xvld       U6,     S2,    0x40
+    xvld       U7,     S2,    0x60
+
+    xvst       U0,     P1,    0x00
+    xvst       U1,     P1,    0x20
+    xvst       U2,     P1,    0x40
+    xvst       U3,     P1,    0x60
+    xvst       U4,     P1,    0x80
+    xvst       U5,     P1,    0xA0
+    xvst       U6,     P1,    0xC0
+    xvst       U7,     P1,    0xE0
+
+    xvld       U0,     S3,    0x00
+    xvld       U1,     S3,    0x20
+    xvld       U2,     S3,    0x40
+    xvld       U3,     S3,    0x60
+    xvld       U4,     S4,    0x00
+    xvld       U5,     S4,    0x20
+    xvld       U6,     S4,    0x40
+    xvld       U7,     S4,    0x60
+
+    xvst       U0,     P1,    0x100
+    xvst       U1,     P1,    0x120
+    xvst       U2,     P1,    0x140
+    xvst       U3,     P1,    0x160
+    xvst       U4,     P1,    0x180
+    xvst       U5,     P1,    0x1A0
+    xvst       U6,     P1,    0x1C0
+    xvst       U7,     P1,    0x1E0
+
+    addi.d     S1,     S1,    0x80
+    addi.d     S2,     S2,    0x80
+    addi.d     S3,     S3,    0x80
+    addi.d     S4,     S4,    0x80
+    addi.d     I,      I,     -1
+    add.d      P1,     P1,    T1
+    blt        ZERO,   I,     .L_4I1
+
+.L_4N15:
+    andi       I,      N,     0x08
+    beq        ZERO,   I,     .L_4N7
+
+    xvld       U0,     S1,    0x00
+    xvld       U1,     S1,    0x20
+    xvld       U2,     S2,    0x00
+    xvld       U3,     S2,    0x20
+    xvld       U4,     S3,    0x00
+    xvld       U5,     S3,    0x20
+    xvld       U6,     S4,    0x00
+    xvld       U7,     S4,    0x20
+
+    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,    0x40
+    addi.d     S2,     S2,    0x40
+    addi.d     S3,     S3,    0x40
+    addi.d     S4,     S4,    0x40
+    addi.d     P2,     P2,    0x100
+
+.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,     P3,    0x00
+    xvst       U1,     P3,    0x20
+    xvst       U2,     P3,    0x40
+    xvst       U3,     P3,    0x60
+
+    addi.d     S1,     S1,    0x20
+    addi.d     S2,     S2,    0x20
+    addi.d     S3,     S3,    0x20
+    addi.d     S4,     S4,    0x20
+    addi.d     P3,     P3,    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,     P4,    0x00
+    xvst       U2,     P4,    0x20
+
+    addi.d     S1,     S1,    0x10
+    addi.d     S2,     S2,    0x10
+    addi.d     S3,     S3,    0x10
+    addi.d     S4,     S4,    0x10
+    addi.d     P4,     P4,    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,     P5,    0x00
+    fst.d      F1,     P5,    0x08
+    fst.d      F2,     P5,    0x10
+    fst.d      F3,     P5,    0x18
+
+    addi.d     S1,     S1,    0x08
+    addi.d     S2,     S2,    0x08
+    addi.d     S3,     S3,    0x08
+    addi.d     S4,     S4,    0x08
+    addi.d     P5,     P5,    0x20
+
+.L_M3:
+    andi       J,      M,     0x02
+    beq        ZERO,   J,     .L_M1
+
+    move       S1,     S0
+    add.d      S2,     S0,    TL
+    add.d      S0,     S0,    T0
+
+    move       P1,     P0
+    addi.d     P0,     P0,    0x100
+
+    srai.d     I,      N,     0x04
+    beq        ZERO,   I,     .L_2N15
+
+.L_2I1: /* I-- */
+    xvld       U0,     S1,    0x00
+    xvld       U1,     S1,    0x20
+    xvld       U2,     S1,    0x40
+    xvld       U3,     S1,    0x60
+    xvld       U4,     S2,    0x00
+    xvld       U5,     S2,    0x20
+    xvld       U6,     S2,    0x40
+    xvld       U7,     S2,    0x60
+
+    xvst       U0,     P1,    0x00
+    xvst       U1,     P1,    0x20
+    xvst       U2,     P1,    0x40
+    xvst       U3,     P1,    0x60
+    xvst       U4,     P1,    0x80
+    xvst       U5,     P1,    0xA0
+    xvst       U6,     P1,    0xC0
+    xvst       U7,     P1,    0xE0
+
+    addi.d     S1,     S1,    0x80
+    addi.d     S2,     S2,    0x80
+    addi.d     I,      I,     -1
+    add.d      P1,     P1,    T1
+    blt        ZERO,   I,     .L_2I1
+
+.L_2N15:
+    andi       I,      N,     0x08
+    beq        ZERO,   I,     .L_2N7
+
+    xvld       U0,     S1,    0x00
+    xvld       U1,     S1,    0x20
+    xvld       U2,     S2,    0x00
+    xvld       U3,     S2,    0x20
+
+    xvst       U0,     P2,    0x00
+    xvst       U1,     P2,    0x20
+    xvst       U2,     P2,    0x40
+    xvst       U3,     P2,    0x60
+
+    addi.d     S1,     S1,    0x40
+    addi.d     S2,     S2,    0x40
+    addi.d     P2,     P2,    0x80
+
+.L_2N7:
+    andi       I,      N,     0x04
+    beq        ZERO,   I,     .L_2N3
+
+    xvld       U0,     S1,    0x00
+    xvld       U1,     S2,    0x00
+
+    xvst       U0,     P3,    0x00
+    xvst       U1,     P3,    0x20
+
+    addi.d     S1,     S1,    0x20
+    addi.d     S2,     S2,    0x20
+    addi.d     P3,     P3,    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,     P4,    0x00
+
+    addi.d     S1,     S1,    0x10
+    addi.d     S2,     S2,    0x10
+    addi.d     P4,     P4,    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,     P5,    0x00
+    fst.d      F1,     P5,    0x08
+
+    addi.d     S1,     S1,    0x08
+    addi.d     S2,     S2,    0x08
+    addi.d     P5,     P5,    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,    0x80
+
+    srai.d     I,      N,     0x04
+    beq        ZERO,   I,     .L_1N15
+
+.L_1I1: /* I-- */
+    xvld       U0,     S1,    0x00
+    xvld       U1,     S1,    0x20
+    xvld       U2,     S1,    0x40
+    xvld       U3,     S1,    0x60
+
+    xvst       U0,     P1,    0x00
+    xvst       U1,     P1,    0x20
+    xvst       U2,     P1,    0x40
+    xvst       U3,     P1,    0x60
+
+    addi.d     S1,     S1,    0x80
+    addi.d     I,      I,     -1
+    add.d      P1,     P1,    T1
+    blt        ZERO,   I,     .L_1I1
+
+.L_1N15:
+    andi       I,      N,     0x08
+    beq        ZERO,   I,     .L_1N7
+
+    xvld       U0,     S1,    0x00
+    xvld       U1,     S1,    0x20
+
+    xvst       U0,     P2,    0x00
+    xvst       U1,     P2,    0x20
+
+    addi.d     S1,     S1,    0x40
+    addi.d     P2,     P2,    0x40
+
+.L_1N7:
+    andi       I,      N,     0x04
+    beq        ZERO,   I,     .L_1N3
+
+    xvld       U0,     S1,    0x00
+
+    xvst       U0,     P3,    0x00
+
+    addi.d     S1,     S1,    0x20
+    addi.d     P3,     P3,    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,     P4,    0x00
+    fst.d      F1,     P4,    0x08
+
+    addi.d     S1,     S1,    0x10
+    addi.d     P4,     P4,    0x10
+
+.L_1N1:
+    andi       I,      N,     0x01
+    beq        ZERO,   I,     .L_M0
+
+    fld.d      F0,     S1,    0x00
+
+    fst.d      F0,     P5,    0x00
+
+    addi.d     S1,     S1,    0x08
+    addi.d     P5,     P5,    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
diff --git a/kernel/loongarch64/dgemm_tcopy_4.S b/kernel/loongarch64/dgemm_tcopy_4.S
new file mode 100644
index 000000000..700989ca1
--- /dev/null
+++ b/kernel/loongarch64/dgemm_tcopy_4.S
@@ -0,0 +1,270 @@
+/*******************************************************************************
+Copyright (c) 2021, 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 P0     $r16
+#define P1     $r17
+#define P2     $r18
+#define P3     $r19
+#define T0     $r20
+#define T1     $r23
+#define TL     $r7
+#define ZERO   $r0
+
+#define F0     $f0
+#define F1     $f1
+#define F2     $f2
+#define F3     $f3
+/* LASX vectors */
+#define U0     $xr0
+#define U1     $xr1
+#define U2     $xr2
+#define U3     $xr3
+
+    PROLOGUE
+
+    addi.d     $sp,    $sp,   -8
+    SDARG      $r23,   $sp,   0
+
+    move       S0,     SRC
+    move       P0,     DST
+
+    srai.d     T0,     N,     0x02
+    slli.d     T0,     T0,    0x02
+    srai.d     T1,     N,     0x01
+    slli.d     T1,     T1,    0x01
+    mul.d      T0,     M,     T0
+    mul.d      T1,     M,     T1
+    slli.d     T0,     T0,    0x03
+    slli.d     T1,     T1,    0x03
+    add.d      P2,     DST,   T0
+    add.d      P3,     DST,   T1
+
+    slli.d     TL,     LDA,   0x03
+    srai.d     J,      M,     0x02
+    slli.d     T0,     TL,    0x01
+    slli.d     T1,     M,     0x05
+    beq        ZERO,   J,     .L_M3
+
+.L_J1: /* J-- */
+    move       S1,     S0
+    add.d      S2,     S0,    TL
+    add.d      S3,     S1,    T0
+    add.d      S4,     S2,    T0
+    add.d      S0,     S3,    T0
+
+    move       P1,     P0
+    addi.d     P0,     P0,    0x80
+
+    srai.d     I,      N,     0x02
+    addi.d     J,      J,     -1
+    beq        ZERO,   I,     .L_N3
+
+.L_I1: /* I-- */
+    xvld       U0,     S1,    0x00
+    xvld       U1,     S2,    0x00
+    xvld       U2,     S3,    0x00
+    xvld       U3,     S4,    0x00
+
+    xvst       U0,     P1,    0x00
+    xvst       U1,     P1,    0x20
+    xvst       U2,     P1,    0x40
+    xvst       U3,     P1,    0x60
+
+    addi.d     S1,     S1,    0x20
+    addi.d     S2,     S2,    0x20
+    addi.d     S3,     S3,    0x20
+    addi.d     S4,     S4,    0x20
+    add.d      P1,     P1,    T1
+
+    addi.d     I,      I,    -1
+    blt        ZERO,   I,    .L_I1
+
+.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
+
+    xvpermi.q  U0,     U1,    0x02
+    xvpermi.q  U2,     U3,    0x02
+
+    xvst       U0,     P2,    0x00
+    xvst       U2,     P2,    0x20
+
+    addi.d     S1,     S1,    0x10
+    addi.d     S2,     S2,    0x10
+    addi.d     S3,     S3,    0x10
+    addi.d     S4,     S4,    0x10
+    addi.d     P2,     P2,    0x40
+
+.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
+
+    fst.d      F0,     P3,    0x00
+    fst.d      F1,     P3,    0x08
+    fst.d      F2,     P3,    0x10
+    fst.d      F3,     P3,    0x18
+
+    addi.d     S1,     S1,    0x08
+    addi.d     S2,     S2,    0x08
+    addi.d     S3,     S3,    0x08
+    addi.d     S4,     S4,    0x08
+    addi.d     P3,     P3,    0x20
+
+.L_N0:
+    blt        ZERO,   J,     .L_J1
+
+.L_M3:
+    andi       J,      M,      0x02
+    beq        ZERO,   J,      .L_M1
+
+    move       S1,     S0
+    add.d      S2,     S0,     TL
+    add.d      S0,     S0,     T0
+
+    move       P1,     P0
+    addi.d     P0,     P0,     0x40
+
+    srai.d     I,      N,      0x02
+    beq        ZERO,   I,      .L_2N3
+
+.L_2I1:   /* I-- */
+    xvld       U0,     S1,     0x00
+    xvld       U1,     S2,     0x00
+
+    xvst       U0,     P1,     0x00
+    xvst       U1,     P1,     0x20
+
+    addi.d     S1,     S1,     0x20
+    addi.d     S2,     S2,     0x20
+    addi.d     I,      I,      -1
+    add.d      P1,     P1,     T1
+
+    blt        ZERO,   I,     .L_2I1
+
+.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,     P2,     0x00
+
+    addi.d     S1,     S1,     0x10
+    addi.d     S2,     S2,     0x10
+    addi.d     P2,     P2,     0x20
+
+.L_2N1:
+    addi.d     I,      N,      0x01
+    beq        ZERO,   I,      .L_M1
+
+    fld.d      F0,     S1,     0x00
+    fld.d      F1,     S2,     0x00
+
+    fst.d      F0,     P3,     0x00
+    fst.d      F1,     P3,     0x08
+
+    addi.d     S1,     S1,     0x08
+    addi.d     S2,     S2,     0x08
+    addi.d     P3,     P3,     0x10
+
+.L_M1:
+    andi       J,      M,      0x01
+    beq        ZERO,   J,      .L_M0
+
+    move       S1,     S0
+    move       P1,     P0
+
+    srai.d     I,      N,      0x02
+    beq        ZERO,   I,      .L_1N3
+
+.L_1I1:
+    xvld       U0,    S1,      0x00
+
+    xvst       U0,    P1,      0x00
+
+    addi.d     S1,    S1,      0x20
+    addi.d     I,     I,       -1
+    add.d      P1,    P1,      T1
+
+    blt        ZERO,  I,       .L_1I1
+
+.L_1N3:
+    andi       I,     N,       0x02
+    beq        I,     ZERO,    .L_1N1
+
+    fld.d      F0,    S1,      0x00
+    fld.d      F1,    S1,      0x08
+
+    fst.d      F0,    P2,      0x00
+    fst.d      F1,    P2,      0x08
+
+    addi.d     S1,    S1,      0x10
+    addi.d     P2,    P2,      0x10
+
+.L_1N1:
+    andi       I,     N,       0x01
+    beq        I,     ZERO,    .L_M0
+
+    fld.d      F0,    S1,      0x00
+
+    fst.d      F0,    P3,      0x00
+
+.L_M0:
+    LDARG      $r23,   $sp,   0
+    addi.d     $sp,    $sp,   8
+    jirl       $r0,    $r1,   0x00
+
+    EPILOGUE
diff --git a/param.h b/param.h
index 8dd2a7461..2dffaae3c 100644
--- a/param.h
+++ b/param.h
@@ -2852,35 +2852,35 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define GEMM_DEFAULT_ALIGN 0x0ffffUL
 
 #define SGEMM_DEFAULT_UNROLL_N 8
-#define DGEMM_DEFAULT_UNROLL_N 8
+#define DGEMM_DEFAULT_UNROLL_N 4
 #define QGEMM_DEFAULT_UNROLL_N 2
 #define CGEMM_DEFAULT_UNROLL_N 4
 #define ZGEMM_DEFAULT_UNROLL_N 4
 #define XGEMM_DEFAULT_UNROLL_N 1
 
 #define SGEMM_DEFAULT_UNROLL_M 2
-#define DGEMM_DEFAULT_UNROLL_M 2
+#define DGEMM_DEFAULT_UNROLL_M 16
 #define QGEMM_DEFAULT_UNROLL_M 2
 #define CGEMM_DEFAULT_UNROLL_M 1
 #define ZGEMM_DEFAULT_UNROLL_M 1
 #define XGEMM_DEFAULT_UNROLL_M 1
 
 #define SGEMM_DEFAULT_P sgemm_p
-#define DGEMM_DEFAULT_P dgemm_p
+#define DGEMM_DEFAULT_P 32
 #define QGEMM_DEFAULT_P qgemm_p
 #define CGEMM_DEFAULT_P cgemm_p
 #define ZGEMM_DEFAULT_P zgemm_p
 #define XGEMM_DEFAULT_P xgemm_p
 
 #define SGEMM_DEFAULT_R sgemm_r
-#define DGEMM_DEFAULT_R dgemm_r
+#define DGEMM_DEFAULT_R 858
 #define QGEMM_DEFAULT_R qgemm_r
 #define CGEMM_DEFAULT_R cgemm_r
 #define ZGEMM_DEFAULT_R zgemm_r
 #define XGEMM_DEFAULT_R xgemm_r
 
 #define SGEMM_DEFAULT_Q 128
-#define DGEMM_DEFAULT_Q 128
+#define DGEMM_DEFAULT_Q 152
 #define QGEMM_DEFAULT_Q 128
 #define CGEMM_DEFAULT_Q 128
 #define ZGEMM_DEFAULT_Q 128