diff options
Diffstat (limited to 'src/dynamic_fusion/sketch/gpu/operators/GpuPool2d.cpp')
| -rw-r--r-- | src/dynamic_fusion/sketch/gpu/operators/GpuPool2d.cpp | 43 |
1 files changed, 25 insertions, 18 deletions
diff --git a/src/dynamic_fusion/sketch/gpu/operators/GpuPool2d.cpp b/src/dynamic_fusion/sketch/gpu/operators/GpuPool2d.cpp index 7ecfa0158b..55c604aacc 100644 --- a/src/dynamic_fusion/sketch/gpu/operators/GpuPool2d.cpp +++ b/src/dynamic_fusion/sketch/gpu/operators/GpuPool2d.cpp @@ -22,20 +22,21 @@ * SOFTWARE. */ +#include "arm_compute/dynamic_fusion/sketch/gpu/operators/GpuPool2d.h" + #include "arm_compute/core/CL/CLCompileContext.h" -#include "arm_compute/core/Validate.h" #include "arm_compute/core/experimental/Types.h" #include "arm_compute/core/utils/misc/ShapeCalculator.h" +#include "arm_compute/core/Validate.h" #include "arm_compute/dynamic_fusion/sketch/gpu/GpuWorkloadContext.h" - #include "arm_compute/dynamic_fusion/sketch/gpu/GpuWorkloadSketch.h" -#include "arm_compute/dynamic_fusion/sketch/gpu/operators/GpuPool2d.h" + #include "src/common/utils/Log.h" #include "src/core/helpers/AutoConfiguration.h" #include "src/dynamic_fusion/sketch/ArgumentPack.h" +#include "src/dynamic_fusion/sketch/gpu/components/cl/ClComponentPool2d.h" #include "src/dynamic_fusion/sketch/gpu/GpuWorkloadSketchImpl.h" #include "src/dynamic_fusion/sketch/gpu/GpuWorkloadSourceCode.h" -#include "src/dynamic_fusion/sketch/gpu/components/cl/ClComponentPool2d.h" #include "src/dynamic_fusion/utils/Utils.h" namespace arm_compute @@ -46,11 +47,15 @@ namespace dynamic_fusion { namespace { -void calculate_and_init_dst_if_empty(ITensorInfo *dst, const ITensorInfo *src, const Pool2dAttributes &attributes, const GpuPool2dSettings &settings) +void calculate_and_init_dst_if_empty(ITensorInfo *dst, + const ITensorInfo *src, + const Pool2dAttributes &attributes, + const GpuPool2dSettings &settings) { - if(dst->total_size() == 0U) + if (dst->total_size() == 0U) { - auto shape = misc::shape_calculator::compute_pool_shape(*src, convert_pool_attr_to_pool_info(attributes, settings.mixed_precision())); + auto shape = misc::shape_calculator::compute_pool_shape( + *src, convert_pool_attr_to_pool_info(attributes, settings.mixed_precision())); auto_init_if_empty(*dst, src->clone()->set_tensor_shape(shape)); } } @@ -82,7 +87,7 @@ bool GpuPool2dSettings::use_inf_as_limit() const Status GpuPool2d::validate_op(const GpuWorkloadSketch &sketch, const ITensorInfo *src, - const Pool2dAttributes &attributes, + const Pool2dAttributes &attributes, const GpuPool2dSettings &settings) { ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src); @@ -110,7 +115,7 @@ Status GpuPool2d::validate_op(const GpuWorkloadSketch &sketch, Status GpuPool2d::is_supported_op(const GpuWorkloadContext &context, const ITensorInfo *src, const Pool2dAttributes &attributes, - const GpuPool2dSettings &settings) + const GpuPool2dSettings &settings) { ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(src); // Data type @@ -118,7 +123,8 @@ Status GpuPool2d::is_supported_op(const GpuWorkloadContext &context, // Data layout ARM_COMPUTE_RETURN_ERROR_ON_DATA_LAYOUT_NOT_IN(src, DataLayout::NHWC); // Check exclude padding is not false - ARM_COMPUTE_RETURN_ERROR_ON_MSG(!attributes.exclude_padding(), "Exclude padding must be set to true in Attributes!"); + ARM_COMPUTE_RETURN_ERROR_ON_MSG(!attributes.exclude_padding(), + "Exclude padding must be set to true in Attributes!"); // Auto initialize dst tensor info TensorInfo dst_info_to_validate; @@ -126,14 +132,15 @@ Status GpuPool2d::is_supported_op(const GpuWorkloadContext &context, calculate_and_init_dst_if_empty(&dst_info_to_validate, src, attributes, settings); // Check components - if(context.gpu_language() == GpuLanguage::OpenCL) + if (context.gpu_language() == GpuLanguage::OpenCL) { const auto cl_compile_ctx = context.cl_compile_context(); ARM_COMPUTE_RETURN_ERROR_ON(cl_compile_ctx == nullptr); // Validate Component { - const KernelProperties properties = IGpuKernelComponent::Properties().stage(UnitWorkloadStage{ UnitWorkloadStage::Stage::Run }); + const KernelProperties properties = + IGpuKernelComponent::Properties().stage(UnitWorkloadStage{UnitWorkloadStage::Stage::Run}); ArgumentPack<ITensorInfo> arguments; arguments.add_const_tensor(ACL_SRC_0, src); @@ -148,10 +155,10 @@ Status GpuPool2d::is_supported_op(const GpuWorkloadContext &context, return Status{}; } -ITensorInfo *GpuPool2d::create_op(GpuWorkloadSketch &sketch, - ITensorInfo *src, - const Pool2dAttributes &attributes, - const GpuPool2dSettings &settings) +ITensorInfo *GpuPool2d::create_op(GpuWorkloadSketch &sketch, + ITensorInfo *src, + const Pool2dAttributes &attributes, + const GpuPool2dSettings &settings) { // Assert validation ARM_COMPUTE_ERROR_THROW_ON(GpuPool2d::validate_op(sketch, src, attributes, settings)); @@ -168,7 +175,7 @@ ITensorInfo *GpuPool2d::create_op(GpuWorkloadSketch &sketch, const auto sketch_ctx = sketch.implementation().context(); - if(sketch_ctx->gpu_language() == GpuLanguage::OpenCL) + if (sketch_ctx->gpu_language() == GpuLanguage::OpenCL) { const auto cl_compile_ctx = sketch_ctx->cl_compile_context(); ARM_COMPUTE_UNUSED(cl_compile_ctx); @@ -177,7 +184,7 @@ ITensorInfo *GpuPool2d::create_op(GpuWorkloadSketch &sketch, // Add Component { auto properties = IGpuKernelComponent::Properties(); - properties.stage(UnitWorkloadStage{ UnitWorkloadStage::Stage::Run }); + properties.stage(UnitWorkloadStage{UnitWorkloadStage::Stage::Run}); ArgumentPack<ITensorInfo> arguments; arguments.add_const_tensor(ACL_SRC_0, src); |