[ONCPUML-951] Variable weight support for Convolution.

API changes for NEGEMMConvolutionLayer and CpuGemmConv2d

Built with:

    scons neon=1 opencl=0 os=linux arch=armv8.2-a multi_isa=1 \
        build=native -j32 Werror=false validation_tests=1 build_dir=opt \
        standalone=1 asserts=1 experimental_fixed_format_kernels=1 .

Tested with:

    ./build/opt/tests/arm_compute_validation

Hardware where the test executable was run:

Neoverse N1

Test coverage:

* NEGEMMConvolutionLayer, CpuGemmConv2d
* NHWC (the only one supported by the fixed-format kernels)
* F16, F32
* Shapes: RunSmall

Change-Id: I4fd3e495a7cbf61210ea02d37440ba9652934e99
Signed-off-by: Francesco Petrogalli <francesco.petrogalli@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/7632
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 9 insertions, 9 deletions

diff --git a/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp b/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp
index 2796b0d204..d749dce98d 100644
--- a/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp
+++ b/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp
@@ -34,17 +34,17 @@
 #include "gemm_interleaved.hpp"
 
 #include "kernels/a32_sgemm_8x6.hpp"
-#ifdef ENABLE_FIXED_FORMAT_KERNELS
+#ifdef ARM_COMPUTE_ENABLE_FIXED_FORMAT_KERNELS
 #include "kernels/a64_ffhybrid_fp16_mla_6x32.hpp"
 #include "kernels/a64_ffinterleaved_fp16_mla_8x24.hpp"
-#endif // ENABLE_FIXED_FORMAT_KERNELS
+#endif // ARM_COMPUTE_ENABLE_FIXED_FORMAT_KERNELS
 #include "kernels/a64_hgemm_8x24.hpp"
 #include "kernels/a64_hybrid_fp16_mla_6x32.hpp"
 #include "kernels/a64_sgemm_8x12.hpp"
-#ifdef ENABLE_FIXED_FORMAT_KERNELS
+#ifdef ARM_COMPUTE_ENABLE_FIXED_FORMAT_KERNELS
 #include "kernels/sve_ffhybrid_fp16_mla_6x4VL.hpp"
 #include "kernels/sve_ffinterleaved_fp16_mla_8x3VL.hpp"
-#endif // ENABLE_FIXED_FORMAT_KERNELS
+#endif // ARM_COMPUTE_ENABLE_FIXED_FORMAT_KERNELS
 #include "kernels/sve_hybrid_fp16_mla_6x4VL.hpp"
 #include "kernels/sve_interleaved_fp16_mla_8x3VL.hpp"
 
@@ -66,7 +66,7 @@ GemmImplementation<__fp16, __fp16>::with_estimate(
     [](const GemmArgs &args) { return GemmInterleaved<cls_sve_interleaved_fp16_mla_8x3VL, __fp16, __fp16>::estimate_cycles<__fp16>(args); },
     [](const GemmArgs &args) { return new GemmInterleaved<cls_sve_interleaved_fp16_mla_8x3VL, __fp16, __fp16>(args); }
 ),
-#ifdef ENABLE_FIXED_FORMAT_KERNELS
+#ifdef ARM_COMPUTE_ENABLE_FIXED_FORMAT_KERNELS
 GemmImplementation<__fp16, __fp16>::with_estimate(
     GemmMethod::GEMM_INTERLEAVED,
     "sve_ffinterleaved_fp16_mla_8x3VL",
@@ -83,7 +83,7 @@ GemmImplementation<__fp16, __fp16>::with_estimate(
     [](const GemmArgs &args) { return GemmHybridIndirectFixedFormat<cls_sve_ffhybrid_fp16_mla_6x4VL, __fp16, __fp16>::estimate_cycles<__fp16>(args); },
     [](const GemmArgs &args) { return new GemmHybridIndirectFixedFormat<cls_sve_ffhybrid_fp16_mla_6x4VL, __fp16, __fp16>(args); }
 ),
-#endif // ENABLE_FIXED_FORMAT_KERNELS
+#endif // ARM_COMPUTE_ENABLE_FIXED_FORMAT_KERNELS
 #endif // ARM_COMPUTE_ENABLE_SVE
 #if defined(__aarch64__)
 GemmImplementation<__fp16, __fp16>::with_estimate(
@@ -100,7 +100,7 @@ GemmImplementation<__fp16, __fp16>::with_estimate(
     [](const GemmArgs &args) { return GemmInterleaved<cls_a64_hgemm_8x24, __fp16, __fp16>::estimate_cycles<__fp16>(args); },
     [](const GemmArgs &args) { return new GemmInterleaved<cls_a64_hgemm_8x24, __fp16, __fp16>(args); }
 ),
-#ifdef ENABLE_FIXED_FORMAT_KERNELS
+#ifdef ARM_COMPUTE_ENABLE_FIXED_FORMAT_KERNELS
 GemmImplementation<__fp16, __fp16>::with_estimate(
     GemmMethod::GEMM_INTERLEAVED,
     "a64_ffinterleaved_fp16_mla_8x24",
@@ -117,7 +117,7 @@ GemmImplementation<__fp16, __fp16>::with_estimate(
     [](const GemmArgs &args) { return GemmHybridIndirectFixedFormat<cls_a64_ffhybrid_fp16_mla_6x32, __fp16, __fp16>::estimate_cycles<__fp16>(args); },
     [](const GemmArgs &args) { return new GemmHybridIndirectFixedFormat<cls_a64_ffhybrid_fp16_mla_6x32, __fp16, __fp16>(args); }
 ),
-#endif // ENABLE_FIXED_FORMAT_KERNELS
+#endif // ARM_COMPUTE_ENABLE_FIXED_FORMAT_KERNELS
 {
     GemmMethod::GEMM_INTERLEAVED,
     "a64_sgemm_8x12",
@@ -152,7 +152,7 @@ const GemmImplementation<__fp16, __fp16> *gemm_implementation_list<__fp16, __fp1
 
 /* Explicitly instantiate the external functions for these types. */
 template UniqueGemmCommon<__fp16, __fp16> gemm<__fp16, __fp16, Nothing>(const GemmArgs &args, const Nothing &);
-template bool has_opt_gemm<__fp16, __fp16, Nothing>(const GemmArgs &args, const Nothing &);
+template bool has_opt_gemm<__fp16, __fp16, Nothing>(WeightFormat &weight_format, const GemmArgs &args, const Nothing &);
 template KernelDescription get_gemm_method<__fp16, __fp16, Nothing>(const GemmArgs &args, const Nothing &);
 template std::vector<KernelDescription> get_compatible_kernels<__fp16, __fp16, Nothing>(const GemmArgs &args, const Nothing &);