From f2cdce30ca782cac7caebc43c7e67caf677b7358 Mon Sep 17 00:00:00 2001 From: Georgios Pinitas Date: Mon, 9 Dec 2019 18:35:57 +0000 Subject: COMPMID-2841: Enable aarch32 builds Decouples the execution state from the architecture. Now architectures can be set as (armv7a, armv8a, etc) and execution state using the `estate` flag with the following options (auto, 32, 64). Change-Id: Ie7f757b3565495a39c7e20fb350a72fd9c5a2a4f Signed-off-by: Georgios Pinitas Reviewed-on: https://review.mlplatform.org/c/2438 Reviewed-by: Pablo Marquez Comments-Addressed: Arm Jenkins Tested-by: Arm Jenkins --- src/core/NEON/kernels/NESelectKernel.cpp | 4 ++-- src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp | 4 ++++ 2 files changed, 6 insertions(+), 2 deletions(-) (limited to 'src/core') diff --git a/src/core/NEON/kernels/NESelectKernel.cpp b/src/core/NEON/kernels/NESelectKernel.cpp index 72afe4f054..191d182002 100644 --- a/src/core/NEON/kernels/NESelectKernel.cpp +++ b/src/core/NEON/kernels/NESelectKernel.cpp @@ -197,7 +197,7 @@ void NESelectKernel::configure(const ITensor *c, const ITensor *x, const ITensor }; #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC map_function["op_F16"] = &select_op_16; -#endif /* ARM_COMPUTE_AARCH64_V8_2 */ +#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ } else { @@ -213,7 +213,7 @@ void NESelectKernel::configure(const ITensor *c, const ITensor *x, const ITensor }; #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC map_function["op_F16"] = &select_op_not_same_rank; -#endif /* ARM_COMPUTE_AARCH64_V8_2 */ +#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ } auto it = map_function.find(function_to_call); diff --git a/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp b/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp index 9409646818..16bdbb5986 100644 --- a/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp +++ b/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp @@ -83,12 +83,16 @@ void MergeResults<8, 6, false>(float *out, const float *in, const int ldout, con switch ((y + 5) - ymax) { case 4: outptr1 = dummyres; + /* fall through */ case 3: outptr2 = dummyres; + /* fall through */ case 2: outptr3 = dummyres; + /* fall through */ case 1: outptr4 = dummyres; + /* fall through */ case 0: outptr5 = dummyres; break; -- cgit v1.2.1