Update CPU kernels to remove x19 and w19
Resolves: COMPMID-5805
Change-Id: Idf720bbb136474810086f5089c5ed23b3f79835a
Signed-off-by: Michael Tyler <michael.tyler@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9081
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
diff --git a/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_16.hpp b/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_16.hpp
index dd1bd50..ca4dc20 100644
--- a/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_16.hpp
+++ b/src/core/NEON/kernels/arm_gemm/transforms/a64_transpose_interleave_16.hpp
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2021 Arm Limited.
+ * Copyright (c) 2021, 2023 Arm Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -10,16 +10,16 @@
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
*/
#pragma once
@@ -36,86 +36,86 @@
"cmp %x[height], #0x4\n"
"blt 6f\n"
"1:" // Main row loop: Head
- "mov x24, %x[in]\n"
- "mov x23, %x[out]\n"
- "add x22, x24, %x[in_stride]\n"
- "add x21, x22, %x[in_stride]\n"
- "add x20, x21, %x[in_stride]\n"
+ "mov x25, %x[in]\n"
+ "mov x24, %x[width]\n"
+ "add x23, x25, %x[in_stride]\n"
+ "add x22, x23, %x[in_stride]\n"
+ "add x20, x22, %x[in_stride]\n"
+ "cmp x24, #0x4\n"
"add %x[in], x20, %x[in_stride]\n"
+ "mov x21, %x[out]\n"
"sub %x[height], %x[height], #0x4\n"
- "mov x19, %x[width]\n"
- "cmp x19, #0x4\n"
"blt 3f\n"
"2:" // Main row loop: Column loop
- "ldr q19, [x24], #0x10\n"
- "sub x19, x19, #0x4\n"
- "ldr q18, [x22], #0x10\n"
- "cmp x19, #0x4\n"
- "ldr q17, [x21], #0x10\n"
+ "ldr q19, [x25], #0x10\n"
+ "ldr q18, [x23], #0x10\n"
+ "sub x24, x24, #0x4\n"
+ "cmp x24, #0x4\n"
+ "ldr q17, [x22], #0x10\n"
"ldr q16, [x20], #0x10\n"
- "str q19, [x23, #0x0]\n"
- "str q18, [x23, #0x10]\n"
- "str q17, [x23, #0x20]\n"
- "str q16, [x23, #0x30]\n"
- "add x23, x23, %x[out_stride]\n"
+ "str q19, [x21, #0x0]\n"
+ "str q18, [x21, #0x10]\n"
+ "str q17, [x21, #0x20]\n"
+ "str q16, [x21, #0x30]\n"
+ "add x21, x21, %x[out_stride]\n"
"bge 2b\n"
"3:" // Main row loop: Column loop skip
- "cmp x19, #0x1\n"
+ "cmp x24, #0x1\n"
"blt 5f\n"
"4:" // Main row loop: width 1 loop: loop
- "ldr s19, [x24], #0x4\n"
- "sub x19, x19, #0x1\n"
- "ldr s18, [x22], #0x4\n"
- "cmp x19, #0x1\n"
- "ldr s17, [x21], #0x4\n"
+ "ldr s19, [x25], #0x4\n"
+ "ldr s18, [x23], #0x4\n"
+ "sub x24, x24, #0x1\n"
+ "cmp x24, #0x1\n"
+ "ldr s17, [x22], #0x4\n"
"ldr s16, [x20], #0x4\n"
- "str s19, [x23, #0x0]\n"
- "str s18, [x23, #0x10]\n"
- "str s17, [x23, #0x20]\n"
- "str s16, [x23, #0x30]\n"
- "add x23, x23, #0x4\n"
+ "str s19, [x21, #0x0]\n"
+ "str s18, [x21, #0x10]\n"
+ "str s17, [x21, #0x20]\n"
+ "str s16, [x21, #0x30]\n"
+ "add x21, x21, #0x4\n"
"bge 4b\n"
"5:" // Main row loop: width 1 loop: skip
- "add %x[out], %x[out], #0x40\n"
"cmp %x[height], #0x4\n"
+ "add %x[out], %x[out], #0x40\n"
"bge 1b\n"
"cbz %x[height], 12f\n"
"6:" // Main loop skip
"7:" // Tail row loop: Head
- "mov x24, %x[in]\n"
- "mov x23, %x[out]\n"
- "add %x[in], x24, %x[in_stride]\n"
+ "mov x20, %x[width]\n"
+ "mov x25, %x[in]\n"
+ "cmp x20, #0x4\n"
+ "add %x[in], x25, %x[in_stride]\n"
+ "mov x21, %x[out]\n"
"sub %x[height], %x[height], #0x1\n"
- "mov x19, %x[width]\n"
- "cmp x19, #0x4\n"
"blt 9f\n"
"8:" // Tail row loop: Column loop
- "ldr q16, [x24], #0x10\n"
- "sub x19, x19, #0x4\n"
- "cmp x19, #0x4\n"
- "str q16, [x23, #0x0]\n"
- "add x23, x23, %x[out_stride]\n"
+ "ldr q16, [x25], #0x10\n"
+ "sub x20, x20, #0x4\n"
+ "cmp x20, #0x4\n"
+ "str q16, [x21, #0x0]\n"
+ "add x21, x21, %x[out_stride]\n"
"bge 8b\n"
"9:" // Tail row loop: Column loop skip
- "cmp x19, #0x1\n"
+ "cmp x20, #0x1\n"
"blt 11f\n"
"10:" // Tail row loop: width 1 loop: loop
- "ldr s16, [x24], #0x4\n"
- "sub x19, x19, #0x1\n"
- "cmp x19, #0x1\n"
- "str s16, [x23, #0x0]\n"
- "add x23, x23, #0x4\n"
+ "ldr s16, [x25], #0x4\n"
+ "sub x20, x20, #0x1\n"
+ "cmp x20, #0x1\n"
+ "str s16, [x21, #0x0]\n"
+ "add x21, x21, #0x4\n"
"bge 10b\n"
"11:" // Tail row loop: width 1 loop: skip
- "add %x[out], %x[out], #0x10\n"
"cmp %x[height], #0x1\n"
+ "add %x[out], %x[out], #0x10\n"
"bge 7b\n"
"12:" // Done
: [height] "+&r" (height), [in] "+&r" (in), [out] "+&r" (out)
: [in_stride] "r" (in_stride), [out_stride] "r" (out_stride), [width] "r" (width)
- : "cc", "memory", "v16", "v17", "v18", "v19", "x19", "x20", "x21", "x22", "x23", "x24"
+ : "cc", "memory", "v16", "v17", "v18", "v19", "x20", "x21", "x22", "x23", "x24", "x25"
);
}