diff --git a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32Opt.S b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32Opt.S
index 7dda0cfa8e4..999ae943bbb 100644
--- a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32Opt.S
+++ b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32Opt.S
@@ -1345,239 +1345,162 @@ LoopRow4:
             st1 {v30.4s}, [x11], #16
             b WriteEnd
         C4Write5:
-            add x19, x11, #16
-            st1 {v8.4s}, [x11]
-            add x11, x11, #20
-            str s9, [x19]
-            add x19, x19, #20
+            add x19, x11, x8
+            st1 {v8.4s}, [x11], #16
+            str s9, [x19], #4
             cmp x6, #1
             beq WriteEnd
-
-            st1 {v10.4s}, [x11]
-            add x11, x11, #20
-            str s11, [x19]
-            add x19, x19, #20
+            st1 {v10.4s}, [x11], #16
+            str s11, [x19], #4
             cmp x6, #2
             beq WriteEnd
-
-            st1 {v12.4s}, [x11]
-            add x11, x11, #20
-            str s13, [x19]
-            add x19, x19, #20
+            st1 {v12.4s}, [x11], #16
+            str s13, [x19], #4
             cmp x6, #3
             beq WriteEnd
-
-            st1 {v14.4s}, [x11]
-            add x11, x11, #20
-            str s15, [x19]
-            add x19, x19, #20
+            st1 {v14.4s}, [x11], #16
+            str s15, [x19], #4
             cmp x6, #4
             beq WriteEnd
-
-            st1 {v16.4s}, [x11]
-            add x11, x11, #20
-            str s17, [x19]
-            add x19, x19, #20
+            st1 {v16.4s}, [x11], #16
+            str s17, [x19], #4
             cmp x6, #5
             beq WriteEnd
-
-            st1 {v18.4s}, [x11]
-            add x11, x11, #20
-            str s19, [x19]
-            add x19, x19, #20
+            st1 {v18.4s}, [x11], #16
+            str s19, [x19], #4
             cmp x6, #6
             beq WriteEnd
-
-            st1 {v20.4s}, [x11]
-            add x11, x11, #20
-            str s21, [x19]
-            add x19, x19, #20
+            st1 {v20.4s}, [x11], #16
+            str s21, [x19], #4
             cmp x6, #7
             beq WriteEnd
-
-            st1 {v22.4s}, [x11]
-            add x11, x11, #20
-            str s23, [x19]
-            add x19, x19, #20
+            st1 {v22.4s}, [x11], #16
+            str s23, [x19], #4
             cmp x6, #8
             beq WriteEnd
-
-            st1 {v24.4s}, [x11]
-            add x11, x11, #20
-            str s25, [x19]
-            add x19, x19, #20
+            st1 {v24.4s}, [x11], #16
+            str s25, [x19], #4
             cmp x6, #9
             beq WriteEnd
-
-            st1 {v26.4s}, [x11]
-            add x11, x11, #20
-            str s27, [x19]
-            add x19, x19, #20
+            st1 {v26.4s}, [x11], #16
+            str s27, [x19], #4
             cmp x6, #10
             beq WriteEnd
-
-            st1 {v28.4s}, [x11]
-            add x11, x11, #20
-            str s29, [x19]
-            add x19, x19, #20
+            st1 {v28.4s}, [x11], #16
+            str s29, [x19], #4
             cmp x6, #11
             beq WriteEnd
-
-            st1 {v30.4s}, [x11]
-            str s31, [x19]
+            st1 {v30.4s}, [x11], #16
+            str s31, [x19], #4
             b WriteEnd
         C4Write6:
-            add x19, x11, #16
-            st1 {v8.4s}, [x11]
-            add x11, x11, #24
-            st1 {v9.2s}, [x19]
-            add x19, x19, #24
+            add x19, x11, x8
+            st1 {v8.4s}, [x11], #16
+            st1 {v9.2s}, [x19], #8
             cmp x6, #1
             beq WriteEnd
-
-            st1 {v10.4s}, [x11]
-            add x11, x11, #24
-            st1 {v11.2s}, [x19]
-            add x19, x19, #24
+            st1 {v10.4s}, [x11], #16
+            st1 {v11.2s}, [x19], #8
             cmp x6, #2
             beq WriteEnd
-
-            st1 {v12.4s}, [x11]
-            add x11, x11, #24
-            st1 {v13.2s}, [x19]
-            add x19, x19, #24
+            st1 {v12.4s}, [x11], #16
+            st1 {v13.2s}, [x19], #8
             cmp x6, #3
             beq WriteEnd
-
-            st1 {v14.4s}, [x11]
-            add x11, x11, #24
-            st1 {v15.2s}, [x19]
-            add x19, x19, #24
+            st1 {v14.4s}, [x11], #16
+            st1 {v15.2s}, [x19], #8
             cmp x6, #4
             beq WriteEnd
-
-            st1 {v16.4s}, [x11]
-            add x11, x11, #24
-            st1 {v17.2s}, [x19]
-            add x19, x19, #24
+            st1 {v16.4s}, [x11], #16
+            st1 {v17.2s}, [x19], #8
             cmp x6, #5
             beq WriteEnd
-
-            st1 {v18.4s}, [x11]
-            add x11, x11, #24
-            st1 {v19.2s}, [x19]
-            add x19, x19, #24
+            st1 {v18.4s}, [x11], #16
+            st1 {v19.2s}, [x19], #8
             cmp x6, #6
             beq WriteEnd
-
-            st1 {v20.4s}, [x11]
-            add x11, x11, #24
-            st1 {v21.2s}, [x19]
-            add x19, x19, #24
+            st1 {v20.4s}, [x11], #16
+            st1 {v21.2s}, [x19], #8
             cmp x6, #7
             beq WriteEnd
-
-            st1 {v22.4s}, [x11]
-            add x11, x11, #24
-            st1 {v23.2s}, [x19]
-            add x19, x19, #24
+            st1 {v22.4s}, [x11], #16
+            st1 {v23.2s}, [x19], #8
             cmp x6, #8
             beq WriteEnd
-
-            st1 {v24.4s}, [x11]
-            add x11, x11, #24
-            st1 {v25.2s}, [x19]
-            add x19, x19, #24
+            st1 {v24.4s}, [x11], #16
+            st1 {v25.2s}, [x19], #8
             cmp x6, #9
             beq WriteEnd
-
-            st1 {v26.4s}, [x11]
-            add x11, x11, #24
-            st1 {v27.2s}, [x19]
-            add x19, x19, #24
+            st1 {v26.4s}, [x11], #16
+            st1 {v27.2s}, [x19], #8
             cmp x6, #10
             beq WriteEnd
-
-            st1 {v28.4s}, [x11]
-            add x11, x11, #24
-            st1 {v29.2s}, [x19]
-            add x19, x19, #24
+            st1 {v28.4s}, [x11], #16
+            st1 {v29.2s}, [x19], #8
             cmp x6, #11
             beq WriteEnd
-
-            st1 {v30.4s}, [x11]
-            st1 {v31.2s}, [x19]
+            st1 {v30.4s}, [x11], #16
+            st1 {v31.2s}, [x19], #8
             b WriteEnd
         C4Write7:
-            add x19, x11, #16
-            add x16, x11, #24
-            mov x10, #28
-            st1 {v8.4s}, [x11], x10
-            st1 {v9.2s}, [x19], x10
-            st1 {v9.s}[2], [x16], x10
+            add x19, x11, x8
+            add x16, x19, #8
+            mov x15, #12
+            st1 {v8.4s}, [x11], #16
+            st1 {v9.2s}, [x19], x15
+            st1 {v9.s}[2], [x16], x15
             cmp x6, #1
             beq WriteEnd
-
-            st1 {v10.4s}, [x11], x10
-            st1 {v11.2s}, [x19], x10
-            st1 {v11.s}[2], [x16], x10
+            st1 {v10.4s}, [x11], #16
+            st1 {v11.2s}, [x19], x15
+            st1 {v11.s}[2], [x16], x15
             cmp x6, #2
             beq WriteEnd
-
-            st1 {v12.4s}, [x11], x10
-            st1 {v13.2s}, [x19], x10
-            st1 {v13.s}[2], [x16], x10
+            st1 {v12.4s}, [x11], #16
+            st1 {v13.2s}, [x19], x15
+            st1 {v13.s}[2], [x16], x15
             cmp x6, #3
             beq WriteEnd
-
-            st1 {v14.4s}, [x11], x10
-            st1 {v15.2s}, [x19], x10
-            st1 {v15.s}[2], [x16], x10
+            st1 {v14.4s}, [x11], #16
+            st1 {v15.2s}, [x19], x15
+            st1 {v15.s}[2], [x16], x15
             cmp x6, #4
             beq WriteEnd
-
-            st1 {v16.4s}, [x11], x10
-            st1 {v17.2s}, [x19], x10
-            st1 {v17.s}[2], [x16], x10
+            st1 {v16.4s}, [x11], #16
+            st1 {v17.2s}, [x19], x15
+            st1 {v17.s}[2], [x16], x15
             cmp x6, #5
             beq WriteEnd
-
-            st1 {v18.4s}, [x11], x10
-            st1 {v19.2s}, [x19], x10
-            st1 {v19.s}[2], [x16], x10
+            st1 {v18.4s}, [x11], #16
+            st1 {v19.2s}, [x19], x15
+            st1 {v19.s}[2], [x16], x15
             cmp x6, #6
             beq WriteEnd
-
-            st1 {v20.4s}, [x11], x10
-            st1 {v21.2s}, [x19], x10
-            st1 {v21.s}[2], [x16], x10
+            st1 {v20.4s}, [x11], #16
+            st1 {v21.2s}, [x19], x15
+            st1 {v21.s}[2], [x16], x15
             cmp x6, #7
             beq WriteEnd
-
-            st1 {v22.4s}, [x11], x10
-            st1 {v23.2s}, [x19], x10
-            st1 {v23.s}[2], [x16], x10
+            st1 {v22.4s}, [x11], #16
+            st1 {v23.2s}, [x19], x15
+            st1 {v23.s}[2], [x16], x15
             cmp x6, #8
             beq WriteEnd
-
-            st1 {v24.4s}, [x11], x10
-            st1 {v25.2s}, [x19], x10
-            st1 {v25.s}[2], [x16], x10
+            st1 {v24.4s}, [x11], #16
+            st1 {v25.2s}, [x19], x15
+            st1 {v25.s}[2], [x16], x15
             cmp x6, #9
             beq WriteEnd
-
-            st1 {v26.4s}, [x11], x10
-            st1 {v27.2s}, [x19], x10
-            st1 {v27.s}[2], [x16], x10
+            st1 {v26.4s}, [x11], #16
+            st1 {v27.2s}, [x19], x15
+            st1 {v27.s}[2], [x16], x15
             cmp x6, #10
             beq WriteEnd
-
-            st1 {v28.4s}, [x11], x10
-            st1 {v29.2s}, [x19], x10
-            st1 {v29.s}[2], [x16], x10
+            st1 {v28.4s}, [x11], #16
+            st1 {v29.2s}, [x19], x15
+            st1 {v29.s}[2], [x16], x15
             cmp x6, #11
             beq WriteEnd
-
             st1 {v30.4s}, [x11]
             st1 {v31.2s}, [x19]
             st1 {v31.s}[2], [x16]
diff --git a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow12.S b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow12.S
index eae7a436fea..15d9b14ecb3 100644
--- a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow12.S
+++ b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow12.S
@@ -398,7 +398,7 @@ LoopRow:
             str s26, [x11]
             cmp x6, #10
             beq WriteEnd
-add x11, x11, x8
+            add x11, x11, x8
             str s28, [x11]
             cmp x6, #11
             beq WriteEnd
@@ -972,204 +972,160 @@ add x11, x11, x8
             st1 {v30.4s}, [x11], #16
             b WriteEnd
         C4Write5:
-            add x19, x11, #16
-            st1 {v8.4s}, [x11]
-            add x11, x11, #20
-            str s9, [x19]
-            add x19, x19, #20
+            add x19, x11, x8
+            st1 {v8.4s}, [x11], #16
+            str s9, [x19], #4
             cmp x6, #1
             beq WriteEnd
-            st1 {v10.4s}, [x11]
-            add x11, x11, #20
-            str s11, [x19]
-            add x19, x19, #20
+            st1 {v10.4s}, [x11], #16
+            str s11, [x19], #4
             cmp x6, #2
             beq WriteEnd
-            st1 {v12.4s}, [x11]
-            add x11, x11, #20
-            str s13, [x19]
-            add x19, x19, #20
+            st1 {v12.4s}, [x11], #16
+            str s13, [x19], #4
             cmp x6, #3
             beq WriteEnd
-            st1 {v14.4s}, [x11]
-            add x11, x11, #20
-            str s15, [x19]
-            add x19, x19, #20
+            st1 {v14.4s}, [x11], #16
+            str s15, [x19], #4
             cmp x6, #4
             beq WriteEnd
-            st1 {v16.4s}, [x11]
-            add x11, x11, #20
-            str s17, [x19]
-            add x19, x19, #20
+            st1 {v16.4s}, [x11], #16
+            str s17, [x19], #4
             cmp x6, #5
             beq WriteEnd
-            st1 {v18.4s}, [x11]
-            add x11, x11, #20
-            str s19, [x19]
-            add x19, x19, #20
+            st1 {v18.4s}, [x11], #16
+            str s19, [x19], #4
             cmp x6, #6
             beq WriteEnd
-            st1 {v20.4s}, [x11]
-            add x11, x11, #20
-            str s21, [x19]
-            add x19, x19, #20
+            st1 {v20.4s}, [x11], #16
+            str s21, [x19], #4
             cmp x6, #7
             beq WriteEnd
-            st1 {v22.4s}, [x11]
-            add x11, x11, #20
-            str s23, [x19]
-            add x19, x19, #20
+            st1 {v22.4s}, [x11], #16
+            str s23, [x19], #4
             cmp x6, #8
             beq WriteEnd
-            st1 {v24.4s}, [x11]
-            add x11, x11, #20
-            str s25, [x19]
-            add x19, x19, #20
+            st1 {v24.4s}, [x11], #16
+            str s25, [x19], #4
             cmp x6, #9
             beq WriteEnd
-            st1 {v26.4s}, [x11]
-            add x11, x11, #20
-            str s27, [x19]
-            add x19, x19, #20
+            st1 {v26.4s}, [x11], #16
+            str s27, [x19], #4
             cmp x6, #10
             beq WriteEnd
-            st1 {v28.4s}, [x11]
-            add x11, x11, #20
-            str s29, [x19]
-            add x19, x19, #20
+            st1 {v28.4s}, [x11], #16
+            str s29, [x19], #4
             cmp x6, #11
             beq WriteEnd
-            st1 {v30.4s}, [x11]
-            str s31, [x19]
+            st1 {v30.4s}, [x11], #16
+            str s31, [x19], #4
             b WriteEnd
         C4Write6:
-            add x19, x11, #16
-            st1 {v8.4s}, [x11]
-            add x11, x11, #24
-            st1 {v9.2s}, [x19]
-            add x19, x19, #24
+            add x19, x11, x8
+            st1 {v8.4s}, [x11], #16
+            st1 {v9.2s}, [x19], #8
             cmp x6, #1
             beq WriteEnd
-            st1 {v10.4s}, [x11]
-            add x11, x11, #24
-            st1 {v11.2s}, [x19]
-            add x19, x19, #24
+            st1 {v10.4s}, [x11], #16
+            st1 {v11.2s}, [x19], #8
             cmp x6, #2
             beq WriteEnd
-            st1 {v12.4s}, [x11]
-            add x11, x11, #24
-            st1 {v13.2s}, [x19]
-            add x19, x19, #24
+            st1 {v12.4s}, [x11], #16
+            st1 {v13.2s}, [x19], #8
             cmp x6, #3
             beq WriteEnd
-            st1 {v14.4s}, [x11]
-            add x11, x11, #24
-            st1 {v15.2s}, [x19]
-            add x19, x19, #24
+            st1 {v14.4s}, [x11], #16
+            st1 {v15.2s}, [x19], #8
             cmp x6, #4
             beq WriteEnd
-            st1 {v16.4s}, [x11]
-            add x11, x11, #24
-            st1 {v17.2s}, [x19]
-            add x19, x19, #24
+            st1 {v16.4s}, [x11], #16
+            st1 {v17.2s}, [x19], #8
             cmp x6, #5
             beq WriteEnd
-            st1 {v18.4s}, [x11]
-            add x11, x11, #24
-            st1 {v19.2s}, [x19]
-            add x19, x19, #24
+            st1 {v18.4s}, [x11], #16
+            st1 {v19.2s}, [x19], #8
             cmp x6, #6
             beq WriteEnd
-            st1 {v20.4s}, [x11]
-            add x11, x11, #24
-            st1 {v21.2s}, [x19]
-            add x19, x19, #24
+            st1 {v20.4s}, [x11], #16
+            st1 {v21.2s}, [x19], #8
             cmp x6, #7
             beq WriteEnd
-            st1 {v22.4s}, [x11]
-            add x11, x11, #24
-            st1 {v23.2s}, [x19]
-            add x19, x19, #24
+            st1 {v22.4s}, [x11], #16
+            st1 {v23.2s}, [x19], #8
             cmp x6, #8
             beq WriteEnd
-            st1 {v24.4s}, [x11]
-            add x11, x11, #24
-            st1 {v25.2s}, [x19]
-            add x19, x19, #24
+            st1 {v24.4s}, [x11], #16
+            st1 {v25.2s}, [x19], #8
             cmp x6, #9
             beq WriteEnd
-            st1 {v26.4s}, [x11]
-            add x11, x11, #24
-            st1 {v27.2s}, [x19]
-            add x19, x19, #24
+            st1 {v26.4s}, [x11], #16
+            st1 {v27.2s}, [x19], #8
             cmp x6, #10
             beq WriteEnd
-            st1 {v28.4s}, [x11]
-            add x11, x11, #24
-            st1 {v29.2s}, [x19]
-            add x19, x19, #24
+            st1 {v28.4s}, [x11], #16
+            st1 {v29.2s}, [x19], #8
             cmp x6, #11
             beq WriteEnd
-            st1 {v30.4s}, [x11]
-            st1 {v31.2s}, [x19]
+            st1 {v30.4s}, [x11], #16
+            st1 {v31.2s}, [x19], #8
             b WriteEnd
         C4Write7:
-            add x19, x11, #16
-            add x16, x11, #24
-            mov x10, #28
-            st1 {v8.4s}, [x11], x10
-            st1 {v9.2s}, [x19], x10
-            st1 {v9.s}[2], [x16], x10
+            add x19, x11, x8
+            add x16, x19, #8
+            mov x15, #12
+            st1 {v8.4s}, [x11], #16
+            st1 {v9.2s}, [x19], x15
+            st1 {v9.s}[2], [x16], x15
             cmp x6, #1
             beq WriteEnd
-            st1 {v10.4s}, [x11], x10
-            st1 {v11.2s}, [x19], x10
-            st1 {v11.s}[2], [x16], x10
+            st1 {v10.4s}, [x11], #16
+            st1 {v11.2s}, [x19], x15
+            st1 {v11.s}[2], [x16], x15
             cmp x6, #2
             beq WriteEnd
-            st1 {v12.4s}, [x11], x10
-            st1 {v13.2s}, [x19], x10
-            st1 {v13.s}[2], [x16], x10
+            st1 {v12.4s}, [x11], #16
+            st1 {v13.2s}, [x19], x15
+            st1 {v13.s}[2], [x16], x15
             cmp x6, #3
             beq WriteEnd
-            st1 {v14.4s}, [x11], x10
-            st1 {v15.2s}, [x19], x10
-            st1 {v15.s}[2], [x16], x10
+            st1 {v14.4s}, [x11], #16
+            st1 {v15.2s}, [x19], x15
+            st1 {v15.s}[2], [x16], x15
             cmp x6, #4
             beq WriteEnd
-            st1 {v16.4s}, [x11], x10
-            st1 {v17.2s}, [x19], x10
-            st1 {v17.s}[2], [x16], x10
+            st1 {v16.4s}, [x11], #16
+            st1 {v17.2s}, [x19], x15
+            st1 {v17.s}[2], [x16], x15
             cmp x6, #5
             beq WriteEnd
-            st1 {v18.4s}, [x11], x10
-            st1 {v19.2s}, [x19], x10
-            st1 {v19.s}[2], [x16], x10
+            st1 {v18.4s}, [x11], #16
+            st1 {v19.2s}, [x19], x15
+            st1 {v19.s}[2], [x16], x15
             cmp x6, #6
             beq WriteEnd
-            st1 {v20.4s}, [x11], x10
-            st1 {v21.2s}, [x19], x10
-            st1 {v21.s}[2], [x16], x10
+            st1 {v20.4s}, [x11], #16
+            st1 {v21.2s}, [x19], x15
+            st1 {v21.s}[2], [x16], x15
             cmp x6, #7
             beq WriteEnd
-            st1 {v22.4s}, [x11], x10
-            st1 {v23.2s}, [x19], x10
-            st1 {v23.s}[2], [x16], x10
+            st1 {v22.4s}, [x11], #16
+            st1 {v23.2s}, [x19], x15
+            st1 {v23.s}[2], [x16], x15
             cmp x6, #8
             beq WriteEnd
-            st1 {v24.4s}, [x11], x10
-            st1 {v25.2s}, [x19], x10
-            st1 {v25.s}[2], [x16], x10
+            st1 {v24.4s}, [x11], #16
+            st1 {v25.2s}, [x19], x15
+            st1 {v25.s}[2], [x16], x15
             cmp x6, #9
             beq WriteEnd
-            st1 {v26.4s}, [x11], x10
-            st1 {v27.2s}, [x19], x10
-            st1 {v27.s}[2], [x16], x10
+            st1 {v26.4s}, [x11], #16
+            st1 {v27.2s}, [x19], x15
+            st1 {v27.s}[2], [x16], x15
             cmp x6, #10
             beq WriteEnd
-            st1 {v28.4s}, [x11], x10
-            st1 {v29.2s}, [x19], x10
-            st1 {v29.s}[2], [x16], x10
+            st1 {v28.4s}, [x11], #16
+            st1 {v29.2s}, [x19], x15
+            st1 {v29.s}[2], [x16], x15
             cmp x6, #11
             beq WriteEnd
             st1 {v30.4s}, [x11]
diff --git a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow4.S b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow4.S
index eaa9e47db50..78242eb717d 100644
--- a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow4.S
+++ b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow4.S
@@ -475,73 +475,61 @@ LoopRow4:
             st1 {v14.4s}, [x11], #16
             b WriteEnd
         C4Write5:
-            add x19, x11, #16
-            st1 {v8.4s}, [x11]
-            add x11, x11, #20
-            str s9, [x19]
-            add x19, x19, #20
+            add x19, x11, x8
+            st1 {v8.4s}, [x11], #16
+            str s9, [x19], #4
             cmp x6, #1
             beq WriteEnd
-            st1 {v10.4s}, [x11]
-            add x11, x11, #20
-            str s11, [x19]
-            add x19, x19, #20
+            st1 {v10.4s}, [x11], #16
+            str s11, [x19], #4
             cmp x6, #2
             beq WriteEnd
-            st1 {v12.4s}, [x11]
-            add x11, x11, #20
-            str s13, [x19]
-            add x19, x19, #20
+            st1 {v12.4s}, [x11], #16
+            str s13, [x19], #4
             cmp x6, #3
             beq WriteEnd
-            st1 {v14.4s}, [x11]
-            str s15, [x19]
+            st1 {v14.4s}, [x11], #16
+            str s15, [x19], #4
             b WriteEnd
         C4Write6:
-            add x19, x11, #16
-            st1 {v8.4s}, [x11]
-            add x11, x11, #24
-            st1 {v9.2s}, [x19]
-            add x19, x19, #24
+            add x19, x11, x8
+            st1 {v8.4s}, [x11], #16
+            st1 {v9.2s}, [x19], #8
             cmp x6, #1
             beq WriteEnd
-            st1 {v10.4s}, [x11]
-            add x11, x11, #24
-            st1 {v11.2s}, [x19]
-            add x19, x19, #24
+            st1 {v10.4s}, [x11], #16
+            st1 {v11.2s}, [x19], #8
             cmp x6, #2
             beq WriteEnd
-            st1 {v12.4s}, [x11]
-            add x11, x11, #24
-            st1 {v13.2s}, [x19]
-            add x19, x19, #24
+            st1 {v12.4s}, [x11], #16
+            st1 {v13.2s}, [x19], #8
             cmp x6, #3
             beq WriteEnd
-            st1 {v14.4s}, [x11]
-            st1 {v15.2s}, [x19]
+            st1 {v14.4s}, [x11], #16
+            st1 {v15.2s}, [x19], #8
             b WriteEnd
         C4Write7:
-            add x19, x11, #16
-            add x16, x11, #24
-            mov x10, #28
-            st1 {v8.4s}, [x11], x10
-            st1 {v9.2s}, [x19], x10
-            st1 {v9.s}[2], [x16], x10
+            add x19, x11, x8
+            add x16, x19, #8
+            mov x15, #12
+            st1 {v8.4s}, [x11], #16
+            st1 {v9.2s}, [x19], x15
+            st1 {v9.s}[2], [x16], x15
             cmp x6, #1
             beq WriteEnd
-            st1 {v10.4s}, [x11], x10
-            st1 {v11.2s}, [x19], x10
-            st1 {v11.s}[2], [x16], x10
+            st1 {v10.4s}, [x11], #16
+            st1 {v11.2s}, [x19], x15
+            st1 {v11.s}[2], [x16], x15
             cmp x6, #2
             beq WriteEnd
-            st1 {v12.4s}, [x11], x10
-            st1 {v13.2s}, [x19], x10
-            st1 {v13.s}[2], [x16], x10
+            st1 {v12.4s}, [x11], #16
+            st1 {v13.2s}, [x19], x15
+            st1 {v13.s}[2], [x16], x15
             cmp x6, #3
             beq WriteEnd
-            st1 {v14.4s}, [x11], x10
-            st1 {v15.2s}, [x19], x10
-            st1 {v15.s}[2], [x16], x10
+            st1 {v14.4s}, [x11], #16
+            st1 {v15.2s}, [x19], x15
+            st1 {v15.s}[2], [x16], x15
             b WriteEnd
         C4Write8:
             add x19, x11, x8
diff --git a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow8.S b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow8.S
index c6dc3191259..b262455782d 100644
--- a/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow8.S
+++ b/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/assembly/arm64/MatmulFp32OptRow8.S
@@ -722,141 +722,113 @@ LoopRow8:
             st1 {v22.4s}, [x11], #16
             b WriteEnd
         C4Write5:
-            add x19, x11, #16
-            st1 {v8.4s}, [x11]
-            add x11, x11, #20
-            str s9, [x19]
-            add x19, x19, #20
+            add x19, x11, x8
+            st1 {v8.4s}, [x11], #16
+            str s9, [x19], #4
             cmp x6, #1
             beq WriteEnd
-            st1 {v10.4s}, [x11]
-            add x11, x11, #20
-            str s11, [x19]
-            add x19, x19, #20
+            st1 {v10.4s}, [x11], #16
+            str s11, [x19], #4
             cmp x6, #2
             beq WriteEnd
-            st1 {v12.4s}, [x11]
-            add x11, x11, #20
-            str s13, [x19]
-            add x19, x19, #20
+            st1 {v12.4s}, [x11], #16
+            str s13, [x19], #4
             cmp x6, #3
             beq WriteEnd
-            st1 {v14.4s}, [x11]
-            add x11, x11, #20
-            str s15, [x19]
-            add x19, x19, #20
+            st1 {v14.4s}, [x11], #16
+            str s15, [x19], #4
             cmp x6, #4
             beq WriteEnd
-            st1 {v16.4s}, [x11]
-            add x11, x11, #20
-            str s17, [x19]
-            add x19, x19, #20
+            st1 {v16.4s}, [x11], #16
+            str s17, [x19], #4
             cmp x6, #5
             beq WriteEnd
-            st1 {v18.4s}, [x11]
-            add x11, x11, #20
-            str s19, [x19]
-            add x19, x19, #20
+            st1 {v18.4s}, [x11], #16
+            str s19, [x19], #4
             cmp x6, #6
             beq WriteEnd
-            st1 {v20.4s}, [x11]
-            add x11, x11, #20
-            str s21, [x19]
-            add x19, x19, #20
+            st1 {v20.4s}, [x11], #16
+            str s21, [x19], #4
             cmp x6, #7
             beq WriteEnd
-            st1 {v22.4s}, [x11]
-            str s23, [x19]
+            st1 {v22.4s}, [x11], #16
+            str s23, [x19], #4
             b WriteEnd
         C4Write6:
-            add x19, x11, #16
-            st1 {v8.4s}, [x11]
-            add x11, x11, #24
-            st1 {v9.2s}, [x19]
-            add x19, x19, #24
+            add x19, x11, x8
+            st1 {v8.4s}, [x11], #16
+            st1 {v9.2s}, [x19], #8
             cmp x6, #1
             beq WriteEnd
-            st1 {v10.4s}, [x11]
-            add x11, x11, #24
-            st1 {v11.2s}, [x19]
-            add x19, x19, #24
+            st1 {v10.4s}, [x11], #16
+            st1 {v11.2s}, [x19], #8
             cmp x6, #2
             beq WriteEnd
-            st1 {v12.4s}, [x11]
-            add x11, x11, #24
-            st1 {v13.2s}, [x19]
-            add x19, x19, #24
+            st1 {v12.4s}, [x11], #16
+            st1 {v13.2s}, [x19], #8
             cmp x6, #3
             beq WriteEnd
-            st1 {v14.4s}, [x11]
-            add x11, x11, #24
-            st1 {v15.2s}, [x19]
-            add x19, x19, #24
+            st1 {v14.4s}, [x11], #16
+            st1 {v15.2s}, [x19], #8
             cmp x6, #4
             beq WriteEnd
-            st1 {v16.4s}, [x11]
-            add x11, x11, #24
-            st1 {v17.2s}, [x19]
-            add x19, x19, #24
+            st1 {v16.4s}, [x11], #16
+            st1 {v17.2s}, [x19], #8
             cmp x6, #5
             beq WriteEnd
-            st1 {v18.4s}, [x11]
-            add x11, x11, #24
-            st1 {v19.2s}, [x19]
-            add x19, x19, #24
+            st1 {v18.4s}, [x11], #16
+            st1 {v19.2s}, [x19], #8
             cmp x6, #6
             beq WriteEnd
-            st1 {v20.4s}, [x11]
-            add x11, x11, #24
-            st1 {v21.2s}, [x19]
-            add x19, x19, #24
+            st1 {v20.4s}, [x11], #16
+            st1 {v21.2s}, [x19], #8
             cmp x6, #7
             beq WriteEnd
-            st1 {v22.4s}, [x11]
-            st1 {v23.2s}, [x19]
+            st1 {v22.4s}, [x11], #16
+            st1 {v23.2s}, [x19], #8
             b WriteEnd
         C4Write7:
-            add x19, x11, #16
-            add x16, x11, #24
-            mov x10, #28
-            st1 {v8.4s}, [x11], x10
-            st1 {v9.2s}, [x19], x10
-            st1 {v9.s}[2], [x16], x10
+            add x19, x11, x8
+            add x16, x19, #8
+            mov x15, #12
+            st1 {v8.4s}, [x11], #16
+            st1 {v9.2s}, [x19], x15
+            st1 {v9.s}[2], [x16], x15
             cmp x6, #1
             beq WriteEnd
-            st1 {v10.4s}, [x11], x10
-            st1 {v11.2s}, [x19], x10
-            st1 {v11.s}[2], [x16], x10
+            st1 {v10.4s}, [x11], #16
+            st1 {v11.2s}, [x19], x15
+            st1 {v11.s}[2], [x16], x15
             cmp x6, #2
             beq WriteEnd
-            st1 {v12.4s}, [x11], x10
-            st1 {v13.2s}, [x19], x10
-            st1 {v13.s}[2], [x16], x10
+            st1 {v12.4s}, [x11], #16
+            st1 {v13.2s}, [x19], x15
+            st1 {v13.s}[2], [x16], x15
             cmp x6, #3
             beq WriteEnd
-            st1 {v14.4s}, [x11], x10
-            st1 {v15.2s}, [x19], x10
-            st1 {v15.s}[2], [x16], x10
+            st1 {v14.4s}, [x11], #16
+            st1 {v15.2s}, [x19], x15
+            st1 {v15.s}[2], [x16], x15
             cmp x6, #4
             beq WriteEnd
-            st1 {v16.4s}, [x11], x10
-            st1 {v17.2s}, [x19], x10
-            st1 {v17.s}[2], [x16], x10
+            st1 {v16.4s}, [x11], #16
+            st1 {v17.2s}, [x19], x15
+            st1 {v17.s}[2], [x16], x15
             cmp x6, #5
             beq WriteEnd
-            st1 {v18.4s}, [x11], x10
-            st1 {v19.2s}, [x19], x10
-            st1 {v19.s}[2], [x16], x10
+            st1 {v18.4s}, [x11], #16
+            st1 {v19.2s}, [x19], x15
+            st1 {v19.s}[2], [x16], x15
             cmp x6, #6
             beq WriteEnd
-            st1 {v20.4s}, [x11], x10
-            st1 {v21.2s}, [x19], x10
-            st1 {v21.s}[2], [x16], x10
+            st1 {v20.4s}, [x11], #16
+            st1 {v21.2s}, [x19], x15
+            st1 {v21.s}[2], [x16], x15
             cmp x6, #7
             beq WriteEnd
-            st1 {v22.4s}, [x11], x10
-            st1 {v23.2s}, [x19], x10
-            st1 {v23.s}[2], [x16], x10
+            st1 {v22.4s}, [x11], #16
+            st1 {v23.2s}, [x19], x15
+            st1 {v23.s}[2], [x16], x15
             b WriteEnd
         C4Write8:
             add x19, x11, x8
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp32/convolution_1x1_fp32.cc b/mindspore/lite/src/runtime/kernel/arm/fp32/convolution_1x1_fp32.cc
index 7fff594c1ab..bef73c917d6 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp32/convolution_1x1_fp32.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/fp32/convolution_1x1_fp32.cc
@@ -195,13 +195,14 @@ int Convolution1x1CPUKernel::DoConv1x1Hw(int task_id) {
       MatMulOpt(thread_pack_input, reinterpret_cast<float *>(packed_weight_), cur_output,
                 reinterpret_cast<float *>(bias_data_), matmul_param_->act_type_, matmul_param_->deep_, cur_rows,
                 matmul_param_->col_, matmul_param_->col_, OutType_Nhwc);
+      cur_output += row_tile_ * matmul_param_->col_;
     } else {
       MatMulOpt(thread_pack_input, reinterpret_cast<float *>(packed_weight_), cur_output,
                 reinterpret_cast<float *>(bias_data_), matmul_param_->act_type_, matmul_param_->deep_, cur_rows,
                 matmul_param_->col_, matmul_param_->row_, OutType_NC4HW4);
+      cur_output += row_tile_ * MSMIN(matmul_param_->col_, C4NUM);
     }
     cur_intput += row_tile_ * matmul_param_->deep_;
-    cur_output += row_tile_ * matmul_param_->col_;
   }
 
   return RET_OK;