aboutsummaryrefslogtreecommitdiff
path: root/src/core/NEON/kernels/arm_gemm/interleave_indirect.cpp
diff options
context:
space:
mode:
authorMichalis Spyrou <michalis.spyrou@arm.com>2021-06-07 14:23:57 +0100
committerGeorgios Pinitas <georgios.pinitas@arm.com>2021-06-23 12:25:50 +0000
commit20fca524baf99402f742ce38c538f2fd07d5fff9 (patch)
treeb63d98383d1ba22bb3ca59d393e4ab9d47a9c762 /src/core/NEON/kernels/arm_gemm/interleave_indirect.cpp
parent1d359279e22874121def2ce4bfdb633d94ea5ade (diff)
downloadComputeLibrary-20fca524baf99402f742ce38c538f2fd07d5fff9.tar.gz
Create core library using high priority operators
A smaller core library is created using a subset of the operators. Changed the structure of filelist.json in order to include more information about the kernels and make the selection easier. Resolves: COMPMID-4514 Change-Id: I079ca7d8e64346174eebdd13b834e1dd4dc36ca2 Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5786 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/interleave_indirect.cpp')
-rw-r--r--src/core/NEON/kernels/arm_gemm/interleave_indirect.cpp16
1 files changed, 8 insertions, 8 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/interleave_indirect.cpp b/src/core/NEON/kernels/arm_gemm/interleave_indirect.cpp
index 0d56b46e19..a6b1269927 100644
--- a/src/core/NEON/kernels/arm_gemm/interleave_indirect.cpp
+++ b/src/core/NEON/kernels/arm_gemm/interleave_indirect.cpp
@@ -320,12 +320,12 @@ template void IndirectInterleave<8, 1, VLType::None>(float *, const float * cons
template void ConvolutionInterleave<8, 1, VLType::None>(float *, const float *, size_t, const convolver<float> &, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
template void Interleave<8, 1, VLType::None>(float *, const float *, size_t, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
-#if defined(__ARM_FEATURE_SVE) && defined(MMLA_FP32)
+#if defined(ARM_COMPUTE_ENABLE_SVE) && defined(ARM_COMPUTE_ENABLE_SVEF32MM)
/* FMMLA */
template void IndirectInterleave<8, 2, VLType::None>(float *, const float * const * const *, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
template void ConvolutionInterleave<8, 2, VLType::None>(float *, const float *, size_t, const convolver<float> &, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
template void Interleave<8, 2, VLType::None>(float *, const float *, size_t, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
-#endif // SVE && MMLA_FP32
+#endif // ARM_COMPUTE_ENABLE_SVE && ARM_COMPUTE_ENABLE_SVEF32MM
/* FP16 */
#if defined(FP16_KERNELS) || defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC)
@@ -340,7 +340,7 @@ template void Interleave<8, 1, VLType::None>(float *, const __fp16 *, size_t, un
/* BF16 */
/* Arm® Neon™/SVE BFDOT */
-#ifdef V8P6_BF
+#ifdef ARM_COMPUTE_ENABLE_BF16
template void IndirectInterleave<8, 2, VLType::None>(bfloat16 *, const bfloat16 * const * const *, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
template void ConvolutionInterleave<8, 2, VLType::None>(bfloat16 *, const bfloat16 *, size_t, const convolver<bfloat16> &, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
template void Interleave<8, 2, VLType::None>(bfloat16 *, const bfloat16 *, size_t, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
@@ -348,7 +348,7 @@ template void Interleave<8, 2, VLType::None>(bfloat16 *, const bfloat16 *, size_
template void IndirectInterleave<8, 4, VLType::None>(bfloat16 *, const bfloat16 * const * const *, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
template void ConvolutionInterleave<8, 4, VLType::None>(bfloat16 *, const bfloat16 *, size_t, const convolver<bfloat16> &, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
template void Interleave<8, 4, VLType::None>(bfloat16 *, const bfloat16 *, size_t, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
-#endif // V8P6_BF
+#endif // ARM_COMPUTE_ENABLE_BF16
/* Arm® Neon™/SVE using FP32 kernel */
template void IndirectInterleave<8, 1, VLType::None>(float *, const bfloat16 * const * const *, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
@@ -375,12 +375,12 @@ template void IndirectInterleave<8, 4, VLType::None>(int8_t *, const int8_t * co
template void ConvolutionInterleave<8, 4, VLType::None>(int8_t *, const int8_t *, size_t, const convolver<int8_t> &, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
template void Interleave<8, 4, VLType::None>(int8_t *, const int8_t *, size_t, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
-#ifdef MMLA_INT8
+#ifdef ARM_COMPUTE_ENABLE_I8MM
/* MMLA SMMLA (height 8, block 8) */
template void IndirectInterleave<8, 8, VLType::None>(int8_t *, const int8_t * const * const *, unsigned int, unsigned int, unsigned int y0, unsigned int ymax, unsigned int k0, unsigned int kmax, bool, int32_t);
template void ConvolutionInterleave<8, 8, VLType::None>(int8_t *, const int8_t *, size_t, const convolver<int8_t> &, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
template void Interleave<8, 8, VLType::None>(int8_t *, const int8_t *, size_t, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
-#endif // MMLA_INT8
+#endif // ARM_COMPUTE_ENABLE_I8MM
/* Arm® Neon™ SDOT (height 8, block 1) */
template void IndirectInterleave<8, 1, VLType::None>(int16_t *, const int8_t * const * const *, unsigned int, unsigned int, unsigned int y0, unsigned int ymax, unsigned int k0, unsigned int kmax, bool, int32_t);
@@ -397,12 +397,12 @@ template void IndirectInterleave<8, 4, VLType::None>(uint8_t *, const uint8_t *
template void ConvolutionInterleave<8, 4, VLType::None>(uint8_t *, const uint8_t *, size_t, const convolver<uint8_t> &, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
template void Interleave<8, 4, VLType::None>(uint8_t *, const uint8_t *, size_t, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
-#ifdef MMLA_INT8
+#ifdef ARM_COMPUTE_ENABLE_I8MM
/* MMLA SMMLA (height 8, block 8) */
template void IndirectInterleave<8, 8, VLType::None>(uint8_t *, const uint8_t * const * const *, unsigned int, unsigned int, unsigned int y0, unsigned int ymax, unsigned int k0, unsigned int kmax, bool, int32_t);
template void ConvolutionInterleave<8, 8, VLType::None>(uint8_t *, const uint8_t *, size_t, const convolver<uint8_t> &, unsigned int, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
template void Interleave<8, 8, VLType::None>(uint8_t *, const uint8_t *, size_t, unsigned int, unsigned int, unsigned int, unsigned int, bool, int32_t);
-#endif // MMLA_INT8
+#endif // ARM_COMPUTE_ENABLE_I8MM
/* Arm® Neon™ 16-bit (height 8, block 1) */
template void IndirectInterleave<8, 1, VLType::None>(uint16_t *, const uint8_t * const * const *, unsigned int, unsigned int, unsigned int y0, unsigned int ymax, unsigned int k0, unsigned int kmax, bool, int32_t);