From 6b9f388f719dc9ff1181c9a43a41140f19e15ec8 Mon Sep 17 00:00:00 2001 From: Michele Di Giorgio Date: Mon, 1 Jul 2019 16:37:04 +0100 Subject: COMPMID-2336: Fix enable tuning of kernels targeted in COMPUTE-10611 Commit fffbdbcc52 added config_id to some CL kernels but did not provide lws_hint() to the enqueue method. Change-Id: I27ba5f39e76e22441c5a3deb3e80d4756189b109 Signed-off-by: Michele Di Giorgio Reviewed-on: https://review.mlplatform.org/c/1457 Tested-by: Arm Jenkins Reviewed-by: Pablo Marquez Comments-Addressed: Arm Jenkins --- src/core/CL/kernels/CLCannyEdgeKernel.cpp | 6 +++--- src/core/CL/kernels/CLColorConvertKernel.cpp | 8 ++++---- src/core/CL/kernels/CLConvolutionKernel.cpp | 2 +- src/core/CL/kernels/CLDerivativeKernel.cpp | 2 +- src/core/CL/kernels/CLFastCornersKernel.cpp | 4 ++-- src/core/CL/kernels/CLGaussianPyramidKernel.cpp | 4 ++-- src/core/CL/kernels/CLHOGDescriptorKernel.cpp | 4 ++-- src/core/CL/kernels/CLHOGDetectorKernel.cpp | 2 +- src/core/CL/kernels/CLHarrisCornersKernel.cpp | 2 +- src/core/CL/kernels/CLIntegralImageKernel.cpp | 2 +- src/core/CL/kernels/CLMagnitudePhaseKernel.cpp | 2 +- src/core/CL/kernels/CLSobel3x3Kernel.cpp | 2 +- src/core/CL/kernels/CLSobel5x5Kernel.cpp | 4 ++-- src/core/CL/kernels/CLSobel7x7Kernel.cpp | 4 ++-- 14 files changed, 24 insertions(+), 24 deletions(-) diff --git a/src/core/CL/kernels/CLCannyEdgeKernel.cpp b/src/core/CL/kernels/CLCannyEdgeKernel.cpp index 6e290ac148..b7b1dfcef6 100644 --- a/src/core/CL/kernels/CLCannyEdgeKernel.cpp +++ b/src/core/CL/kernels/CLCannyEdgeKernel.cpp @@ -102,7 +102,7 @@ void CLGradientKernel::run(const Window &window, cl::CommandQueue &queue) add_2D_tensor_argument(idx, _gy, slice); add_2D_tensor_argument(idx, _magnitude, slice); add_2D_tensor_argument(idx, _phase, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } @@ -181,7 +181,7 @@ void CLEdgeNonMaxSuppressionKernel::run(const Window &window, cl::CommandQueue & add_2D_tensor_argument(idx, _magnitude, slice); add_2D_tensor_argument(idx, _phase, slice); add_2D_tensor_argument(idx, _output, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } @@ -286,7 +286,7 @@ void CLEdgeTraceKernel::run(const Window &window, cl::CommandQueue &queue) add_2D_tensor_argument(idx, _l1_stack, slice); add_2D_tensor_argument(idx, _l1_stack_counter, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLColorConvertKernel.cpp b/src/core/CL/kernels/CLColorConvertKernel.cpp index c32c08f339..f3b93282e8 100644 --- a/src/core/CL/kernels/CLColorConvertKernel.cpp +++ b/src/core/CL/kernels/CLColorConvertKernel.cpp @@ -443,7 +443,7 @@ void CLColorConvertKernel::run(const Window &window, cl::CommandQueue &queue) unsigned int idx = 0; add_2D_tensor_argument(idx, _input, slice); add_2D_tensor_argument(idx, _output, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } @@ -466,7 +466,7 @@ void CLColorConvertKernel::run(const Window &window, cl::CommandQueue &queue) { add_2D_tensor_argument(idx, _multi_output->cl_plane(i), win_uv); } - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } @@ -491,7 +491,7 @@ void CLColorConvertKernel::run(const Window &window, cl::CommandQueue &queue) add_2D_tensor_argument(idx, _multi_input->cl_plane(i), win_uv); } add_2D_tensor_argument(idx, _output, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } @@ -528,7 +528,7 @@ void CLColorConvertKernel::run(const Window &window, cl::CommandQueue &queue) { add_2D_tensor_argument(idx, _multi_output->cl_plane(i), win_out_uv); } - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLConvolutionKernel.cpp b/src/core/CL/kernels/CLConvolutionKernel.cpp index ef2629d208..98c34970ec 100644 --- a/src/core/CL/kernels/CLConvolutionKernel.cpp +++ b/src/core/CL/kernels/CLConvolutionKernel.cpp @@ -348,7 +348,7 @@ void CLConvolutionRectangleKernel::run(const Window &window, cl::CommandQueue &q unsigned int idx = 0; add_2D_tensor_argument(idx, _input, slice); add_2D_tensor_argument(idx, _output, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLDerivativeKernel.cpp b/src/core/CL/kernels/CLDerivativeKernel.cpp index f55a4cc464..0a247f9071 100644 --- a/src/core/CL/kernels/CLDerivativeKernel.cpp +++ b/src/core/CL/kernels/CLDerivativeKernel.cpp @@ -153,7 +153,7 @@ void CLDerivativeKernel::run(const Window &window, cl::CommandQueue &queue) add_2D_tensor_argument(idx, _output_y, slice); } - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLFastCornersKernel.cpp b/src/core/CL/kernels/CLFastCornersKernel.cpp index ca3d528142..e3b5b817ca 100644 --- a/src/core/CL/kernels/CLFastCornersKernel.cpp +++ b/src/core/CL/kernels/CLFastCornersKernel.cpp @@ -120,7 +120,7 @@ void CLFastCornersKernel::run(const Window &window, cl::CommandQueue &queue) unsigned int idx = 0; add_2D_tensor_argument(idx, _input, slice); add_2D_tensor_argument(idx, _output, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } @@ -194,7 +194,7 @@ void CLCopyToArrayKernel::run(const Window &window, cl::CommandQueue &queue) { unsigned int idx = 0; add_2D_tensor_argument(idx, _input, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLGaussianPyramidKernel.cpp b/src/core/CL/kernels/CLGaussianPyramidKernel.cpp index 27b73b6d32..5153396880 100644 --- a/src/core/CL/kernels/CLGaussianPyramidKernel.cpp +++ b/src/core/CL/kernels/CLGaussianPyramidKernel.cpp @@ -132,7 +132,7 @@ void CLGaussianPyramidHorKernel::run(const Window &window, cl::CommandQueue &que unsigned int idx = 0; add_2D_tensor_argument(idx, _input, slice_in); add_2D_tensor_argument(idx, _output, slice_out); - enqueue(queue, *this, slice_out); + enqueue(queue, *this, slice_out, lws_hint()); } while(win_in.slide_window_slice_2D(slice_in) && win_out.slide_window_slice_2D(slice_out)); } @@ -229,7 +229,7 @@ void CLGaussianPyramidVertKernel::run(const Window &window, cl::CommandQueue &qu unsigned int idx = 0; add_2D_tensor_argument(idx, _input, slice_in); add_2D_tensor_argument(idx, _output, slice_out); - enqueue(queue, *this, slice_out); + enqueue(queue, *this, slice_out, lws_hint()); } while(win_in.slide_window_slice_2D(slice_in) && win_out.slide_window_slice_2D(slice_out)); } diff --git a/src/core/CL/kernels/CLHOGDescriptorKernel.cpp b/src/core/CL/kernels/CLHOGDescriptorKernel.cpp index abdac504f5..a531cf70d8 100644 --- a/src/core/CL/kernels/CLHOGDescriptorKernel.cpp +++ b/src/core/CL/kernels/CLHOGDescriptorKernel.cpp @@ -126,7 +126,7 @@ void CLHOGOrientationBinningKernel::run(const Window &window, cl::CommandQueue & add_2D_tensor_argument(idx, _input_phase, slice_mag_phase); add_2D_tensor_argument(idx, _output, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } @@ -222,7 +222,7 @@ void CLHOGBlockNormalizationKernel::run(const Window &window, cl::CommandQueue & add_2D_tensor_argument(idx, _input, slice_in); add_2D_tensor_argument(idx, _output, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLHOGDetectorKernel.cpp b/src/core/CL/kernels/CLHOGDetectorKernel.cpp index 160be988f9..51883426d7 100644 --- a/src/core/CL/kernels/CLHOGDetectorKernel.cpp +++ b/src/core/CL/kernels/CLHOGDetectorKernel.cpp @@ -134,7 +134,7 @@ void CLHOGDetectorKernel::run(const Window &window, cl::CommandQueue &queue) unsigned int idx = 0; add_2D_tensor_argument(idx, _input, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLHarrisCornersKernel.cpp b/src/core/CL/kernels/CLHarrisCornersKernel.cpp index 695e15d4b1..ab40265ed5 100644 --- a/src/core/CL/kernels/CLHarrisCornersKernel.cpp +++ b/src/core/CL/kernels/CLHarrisCornersKernel.cpp @@ -136,7 +136,7 @@ void CLHarrisScoreKernel::run(const Window &window, cl::CommandQueue &queue) add_2D_tensor_argument(idx, _input1, slice); add_2D_tensor_argument(idx, _input2, slice); add_2D_tensor_argument(idx, _output, slice); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLIntegralImageKernel.cpp b/src/core/CL/kernels/CLIntegralImageKernel.cpp index a1548a1223..507d7ef22d 100644 --- a/src/core/CL/kernels/CLIntegralImageKernel.cpp +++ b/src/core/CL/kernels/CLIntegralImageKernel.cpp @@ -130,7 +130,7 @@ void CLIntegralImageVertKernel::run(const Window &window, cl::CommandQueue &queu unsigned int idx = 0; add_2D_tensor_argument(idx, _in_out, slice); _kernel.setArg(idx++, height); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLMagnitudePhaseKernel.cpp b/src/core/CL/kernels/CLMagnitudePhaseKernel.cpp index 934e2d64a3..354113bba5 100644 --- a/src/core/CL/kernels/CLMagnitudePhaseKernel.cpp +++ b/src/core/CL/kernels/CLMagnitudePhaseKernel.cpp @@ -172,7 +172,7 @@ void CLMagnitudePhaseKernel::run(const Window &window, cl::CommandQueue &queue) add_2D_tensor_argument(idx, _phase, slice); } - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLSobel3x3Kernel.cpp b/src/core/CL/kernels/CLSobel3x3Kernel.cpp index 4439b3073c..1c2fbb0176 100644 --- a/src/core/CL/kernels/CLSobel3x3Kernel.cpp +++ b/src/core/CL/kernels/CLSobel3x3Kernel.cpp @@ -139,7 +139,7 @@ void CLSobel3x3Kernel::run(const Window &window, cl::CommandQueue &queue) add_2D_tensor_argument(idx, _output_y, slice); } - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLSobel5x5Kernel.cpp b/src/core/CL/kernels/CLSobel5x5Kernel.cpp index 1f4b496c75..f07c1c9098 100644 --- a/src/core/CL/kernels/CLSobel5x5Kernel.cpp +++ b/src/core/CL/kernels/CLSobel5x5Kernel.cpp @@ -138,7 +138,7 @@ void CLSobel5x5HorKernel::run(const Window &window, cl::CommandQueue &queue) add_2D_tensor_argument(idx, _output_y, slice); } - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } @@ -252,7 +252,7 @@ void CLSobel5x5VertKernel::run(const Window &window, cl::CommandQueue &queue) _kernel.setArg(idx++, 0 /*dummy*/); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } diff --git a/src/core/CL/kernels/CLSobel7x7Kernel.cpp b/src/core/CL/kernels/CLSobel7x7Kernel.cpp index cd221fc4aa..8c0b6a239f 100644 --- a/src/core/CL/kernels/CLSobel7x7Kernel.cpp +++ b/src/core/CL/kernels/CLSobel7x7Kernel.cpp @@ -140,7 +140,7 @@ void CLSobel7x7HorKernel::run(const Window &window, cl::CommandQueue &queue) add_2D_tensor_argument(idx, _output_y, slice); } - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } @@ -255,7 +255,7 @@ void CLSobel7x7VertKernel::run(const Window &window, cl::CommandQueue &queue) _kernel.setArg(idx++, 0 /*dummy*/); - enqueue(queue, *this, slice); + enqueue(queue, *this, slice, lws_hint()); } while(window.slide_window_slice_2D(slice)); } -- cgit v1.2.1