diff options
author | Michael Tyler <michael.tyler@arm.com> | 2023-04-12 17:43:17 +0100 |
---|---|---|
committer | michael.tyler <michael.tyler@arm.com> | 2023-06-05 15:57:58 +0000 |
commit | 74921eee924625426429044decefe3673561b174 (patch) | |
tree | 654da1a95e3d42d6af8ad1ff27bb40d77b1fd8c5 /src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_block4_u8_u8.hpp | |
parent | df5d9878008be9b60586df97ebfff197abb5195e (diff) | |
download | ComputeLibrary-74921eee924625426429044decefe3673561b174.tar.gz |
Update CPU kernel implementations and guard directives
Resolves COMPMID-6023
Change-Id: I868975d14c4f98af6716726feda22405a6a4c891
Signed-off-by: Michael Tyler <michael.tyler@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9686
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_block4_u8_u8.hpp')
-rw-r--r-- | src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_block4_u8_u8.hpp | 18 |
1 files changed, 8 insertions, 10 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_block4_u8_u8.hpp b/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_block4_u8_u8.hpp index 683a315a96..ce9a0065c7 100644 --- a/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_block4_u8_u8.hpp +++ b/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_block4_u8_u8.hpp @@ -22,16 +22,14 @@ * SOFTWARE. */ -#if defined(__ARM_FEATURE_SVE) +#if defined(ARM_COMPUTE_ENABLE_SME) template <> void interleave_block<2, 4, VLType::SME, false>( uint8_t * &out, const uint8_t * const *in, - size_t width, size_t height, size_t row_offset, bool first + size_t width, size_t height, size_t row_offset, bool ) { - ARM_COMPUTE_UNUSED(first); - __asm__ __volatile__( ".inst 0xd503477f // SMSTART ZA\n" "cntb x21\n" @@ -248,13 +246,13 @@ void interleave_block<2, 4, VLType::SME, false>( ".inst 0xe0bf82a0 // st1w { za0v.s[x12] }, p0/Z, [x21, XZR, LSL #2]\n" ".inst 0x25306d20 // psel p0.s, p11.s/Z, p9.s[w12]\n" ".inst 0xe0b082a4 // st1w { za1v.s[x12] }, p0/Z, [x21, x16, LSL #2]\n" - "ldr x9, [x11, #0x0]\n" + "ldr x20, [x11, #0x0]\n" ".inst 0x25356140 // psel p0.b, p8.b/Z, p10.b[w13, #2]\n" - ".inst 0xe0162122 // ld1b { za0h.b[x13, #2] }, p0/Z, [x9, x22]\n" - "ldr x26, [x11, x16, LSL #0x3]\n" + ".inst 0xe0162282 // ld1b { za0h.b[x13, #2] }, p0/Z, [x20, x22]\n" + "ldr x20, [x11, x16, LSL #0x3]\n" "add x12, x12, #0x1\n" ".inst 0x253d6140 // psel p0.b, p8.b/Z, p10.b[w13, #3]\n" - ".inst 0xe0162343 // ld1b { za0h.b[x13, #3] }, p0/Z, [x26, x22]\n" + ".inst 0xe0162283 // ld1b { za0h.b[x13, #3] }, p0/Z, [x20, x22]\n" "cmp x12, x16\n" "add x11, x11, #0x8\n" "addvl x21, x21, #2\n" @@ -274,7 +272,7 @@ void interleave_block<2, 4, VLType::SME, false>( "addvl x21, x21, #2\n" "add x20, x20, #0x4\n" "blt 10b\n" - "whilelt p9.b, x14, %x[width]\n" + "whilelt p8.b, x14, %x[width]\n" "b 13f\n" "11:" // K loop: Tails: Odd "mov x12, #0x0\n" @@ -296,4 +294,4 @@ void interleave_block<2, 4, VLType::SME, false>( ); } -#endif // defined(__ARM_FEATURE_SVE) +#endif // defined(ARM_COMPUTE_ENABLE_SME) |