diff options
| author | Michalis Spyrou <michalis.spyrou@arm.com> | 2021-06-07 14:23:57 +0100 |
|---|---|---|
| committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2021-06-23 12:25:50 +0000 |
| commit | 20fca524baf99402f742ce38c538f2fd07d5fff9 (patch) | |
| tree | b63d98383d1ba22bb3ca59d393e4ab9d47a9c762 /src/core/NEON/kernels/arm_gemm/gemm_fp32.cpp | |
| parent | 1d359279e22874121def2ce4bfdb633d94ea5ade (diff) | |
| download | ComputeLibrary-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/gemm_fp32.cpp')
| -rw-r--r-- | src/core/NEON/kernels/arm_gemm/gemm_fp32.cpp | 16 |
1 files changed, 8 insertions, 8 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/gemm_fp32.cpp b/src/core/NEON/kernels/arm_gemm/gemm_fp32.cpp index d94814fb4c..1632e301ac 100644 --- a/src/core/NEON/kernels/arm_gemm/gemm_fp32.cpp +++ b/src/core/NEON/kernels/arm_gemm/gemm_fp32.cpp @@ -59,7 +59,7 @@ static const GemmImplementation<float, float> gemm_fp32_methods[] = [](const GemmArgs &args) { return new GemvBatched<float, float>(args); } }, #ifdef __aarch64__ -#ifdef __ARM_FEATURE_SVE +#ifdef ARM_COMPUTE_ENABLE_SVE { GemmMethod::GEMM_HYBRID, "sve_gemv_fp32_mla_8VL", @@ -77,17 +77,17 @@ static const GemmImplementation<float, float> gemm_fp32_methods[] = }, // MMLA next due to higher throughput (SVE only) -#if defined(__ARM_FEATURE_SVE) && defined(MMLA_FP32) +#if defined(ARM_COMPUTE_ENABLE_SVE) && defined(ARM_COMPUTE_ENABLE_SVEF32MM) { GemmMethod::GEMM_INTERLEAVED, "sve_interleaved_fp32_mmla_8x3VL", - [](const GemmArgs &args) { return args._ci->has_sve() && (args._Ksize>4); }, + [](const GemmArgs &args) { return args._ci->has_svef32mm() && (args._Ksize>4); }, [](const GemmArgs &args) { return args._ci->get_cpu_model() != CPUModel::KLEIN; }, [](const GemmArgs &args) { return new GemmInterleaved<cls_sve_interleaved_fp32_mmla_8x3VL, float, float>(args); } }, -#endif // __ARM_FEATURE_SVE && MMLA_FP32 +#endif // ARM_COMPUTE_ENABLE_SVE && ARM_COMPUTE_ENABLE_SVEF32MM -#ifdef __ARM_FEATURE_SVE +#ifdef ARM_COMPUTE_ENABLE_SVE // SVE smallk / hybrid methods { GemmMethod::GEMM_HYBRID, @@ -110,7 +110,7 @@ static const GemmImplementation<float, float> gemm_fp32_methods[] = [](const GemmArgs &args) { return args._ci->get_cpu_model() != CPUModel::KLEIN && (((args._Ksize <= 256) && (args._Nsize <= 256)) || ((args._nmulti > 1) && ((args._Msize / args._maxthreads) < 8))); }, [](const GemmArgs &args) { return new GemmHybridIndirect<cls_sve_hybrid_fp32_mla_6x4VL, float, float>(args); } }, -#endif // __ARM_FEATURE_SVE +#endif // ARM_COMPUTE_ENABLE_SVE // Cortex-A35 specific kernel - use for any problem on A35, and never in any other cases. { GemmMethod::GEMM_INTERLEAVED, @@ -148,7 +148,7 @@ GemmImplementation<float, float>::with_estimate( [](const GemmArgs &args) { return GemmHybridIndirect<cls_a64_hybrid_fp32_mla_6x16, float, float>::estimate_cycles(args, cls_a64_hybrid_fp32_mla_6x16::get_performance_parameters(args._ci)); }, [](const GemmArgs &args) { return new GemmHybridIndirect<cls_a64_hybrid_fp32_mla_6x16, float, float>(args); } ), -#ifdef __ARM_FEATURE_SVE +#ifdef ARM_COMPUTE_ENABLE_SVE { GemmMethod::GEMM_INTERLEAVED, "sve_interleaved_fp32_mla_8x3VL", @@ -156,7 +156,7 @@ GemmImplementation<float, float>::with_estimate( [](const GemmArgs &args) { return args._ci->get_cpu_model() != CPUModel::KLEIN; }, [](const GemmArgs &args) { return new GemmInterleaved<cls_sve_interleaved_fp32_mla_8x3VL, float, float>(args); } }, -#endif // __ARM_FEATURE_SVE +#endif // ARM_COMPUTE_ENABLE_SVE GemmImplementation<float, float>::with_estimate( GemmMethod::GEMM_INTERLEAVED, "a64_sgemm_8x12", |