aboutsummaryrefslogtreecommitdiff
path: root/src/runtime/CL/functions
AgeCommit message (Collapse)Author
2021-07-27Fix memory lifetime issueGeorgios Pinitas
B matrix was getting released while was still needed for a Native GEMM execution Resolves: COMPMID-4705 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I1e490abbe1feabd451d10fa785de21e725f1d6e0 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6001 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-07-27Port CLGEMMConvolutionLayerManuel Bottini
Details: port CLWeightsReshapeKernel to ClWeightsReshapeKernel port CLGEMMConvolutionLayer to ClGemmConvolution Resolves: COMPMID-4515 Change-Id: I7d5b4ec72db2742f6eb9f3ffc88f717c35b4f2a3 Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5983 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-07-27Dispatch Conv2d using the Direct method when necessaryGeorgios Pinitas
Direct Convolution will be preferred for large kernel sizes (>=5) and when input feature maps are greater or equal to output feature maps. Resolves: COMPMID-4708 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Ic9fe8b178a58bd95e7c47597221e47d344f96c79 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5996 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-07-22Port ClGemmLowp to new APIGeorgios Pinitas
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Icef9ca564e61a00a3f4fd4ae7f465a711ff8c51d Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5939 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-07-22Update GEMM assembly kernelsGeorgios Pinitas
- Introduce Fp32 kernels with internal calculations in Bfloat16 when fast_mode is enabled - Improve kernel selection heuristics Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I68a9e7e862b6fd2721b46e0d7cc791091c4ab279 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5965 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-07-16Avoid multiple Rhs matrix transformation on ClGemmGeorgios Pinitas
ClWinogradConv2d was performing Rhs transformation on every step impacting the performance. Adds scope logging support through ARM_COMPUTE_LOG_MSG_WITH_FUNCNAME Resolves: COMPMID-4596 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Ib329d3bc8d8aa21abae9fabfe61de35cc84d4819 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5925 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-07-16Port CLIm2ColKernel to ClIm2ColKernelManuel Bottini
Resolves: COMPMID-4516 Change-Id: I6a6db66797fa801dfe1238fceca413277241d2ec Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5946 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-07-15Port CLCol2ImKernel to ClCol2ImKernelManuel Bottini
Resolves: COMPMID-4517 Change-Id: I50cb02116a1ab86fc29200371944c4774e830746 Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5949 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-07-08Change CLConvolution selection to run Direct approach on large kernelsGeorgios Pinitas
Consider selection of the Direct Conv2d approach on GPU for floating point inputs when kernel size is greater or equal to 7x7. Resolves: COMPMID-4595 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: If851b2e5241dab51127bf9051aa79d210d3f4421 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5894 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-07-02Rework OpenCL Depthwise ConvolutionGian Marco Iodice
- Remove dedicated kernels for NCHW. Now we only use NHWC with permute - Remove specialized kernels for 3x3 NHWC - Simplify CLDepthwiseConvolutionLayer.cpp to call just the native implementation for both floating-point and quantized data types - Develop two parametric opencl kernels for depthwise convolution layer NHWC (floating-point and quantized) - Add support to export the weights to cl_image - Extend test for depthwise convolution on opencl Resolves COMPMID-4417 Change-Id: Ibe533f79c2860f9cac8e921895d5a8f947753a5c Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5893 Reviewed-by: Giorgio Arena <giorgio.arena@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-07-02Implement FP GPU depthwise convolution 1x1 kernel for in-place computationSiCongLi
* Implement in-place graph node mutator for 1x1 depthwise convolution * Add in-place to validation fixture except for DepthwiseConvolutionLayerNativeValidationFixture as it would be a duplicate test otherwise (DepthwiseConvolutionLayerNative test tests the underlying kernel) Resolves: COMPMID-4432 Change-Id: Id7f10f5ebdce7d49f550c0b62dbaaab7f5b59d29 Signed-off-by: SiCongLi <sicong.li@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5874 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2021-06-30Port ClGemmLowpOutputStage operator to new interfaceGeorgios Pinitas
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I96630b45887eaba16ef358b95f3d9ac0b9045157 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5882 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-06-30Revert "Rework OpenCL Depthwise Convolution"Gian Marco Iodice
This reverts commit 561c176598cd14245e2e7918fdf136d1c888d1da. Reason for revert: <validation> Change-Id: I6f2d61c27520439bb538e9265736532104b24cf8 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5127 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-06-29Port the ClGemmLowp kernels to the new APIGeorgios Pinitas
Ported kernels: - CLGEMMLowpMatrixMultiplyNativeKernel - CLGEMMLowpMatrixMultiplyReshapedKernel - CLGEMMLowpMatrixMultiplyReshapedOnlyRHSKernel - CLGEMMLowpOffsetContributionKernel - CLGEMMLowpOffsetContributionOutputStageKernel - CLGEMMLowpQuantizeDownInt32ScaleByFixedPointKernel - CLGEMMLowpQuantizeDownInt32ScaleByFloatKernel - CLGEMMLowpQuantizeDownInt32ScaleKernel Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I9d5a744d6a2dd2f2726fdfb291bad000b6970de2 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5870 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-06-25Port NEGEMMConv2d to memory injecting interfaceMichele Di Giorgio
Resolves: COMPMID-4506, COMPMID-4570 Change-Id: I6d37a06da141f1fcfcaa8525322a319cb0234791 Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5824 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-06-24Rework OpenCL Depthwise ConvolutionGian Marco Iodice
- Remove dedicated kernels for NCHW. Now we only use NHWC with permute - Remove specialized kernels for 3x3 NHWC - Simplify CLDepthwiseConvolutionLayer.cpp to call just the native implementation for both floating-point and quantized data types - Develop two parametric opencl kernels for depthwise convolution layer NHWC (floating-point and quantized) - Add support to export the weights to cl_image - Extend test for depthwise convolution on opencl Resolves COMPMID-4417 Change-Id: I253dd5d959a70783c82e62b1771a5e9f91621cb0 Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5806 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>
2021-06-22Fix Winograd heuristic in CLConvolutionLayerGian Marco Iodice
- Dispath CLWinogradConvolutionLayer only for IFM >= 16 Resolves COMPMID-4591 Change-Id: Id314cd0003a0d938c48b49a1936ff7c6b7f8391f Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5841 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Georgios Pinitas <georgios.pinitas@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-06-22Add FP16 support to CLRemapFreddie Liardet
Add FP16 support to CLRemap when data layout is NHWC. Add relevant tests for FP16 and validation. Update NERemap function level to be consistent with CLRemap. Add depreciation notice for uint_8 only function level methods. Resolves: COMPMID-4335 Signed-off-by: Freddie Liardet <frederick.liardet@arm.com> Change-Id: If05f06801aef7a169b73ff1ebe760a42f11ca05c Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5816 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-06-21Rework the CLConvolutionLayer heuristicGian Marco Iodice
- Call winograd for large kernel size (> 5x5) Resolves COMPMID-4582 Change-Id: I78a187d91bc8099bec7b5c6fa941223cfc1423ba Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5814 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2021-06-18Integrate improved CPU depthwise convolution kernelsMichele Di Giorgio
* Replace assembly kernels for depthwise convolution with more optimized ones. * Add int8 assembly kernels. * Fix implicit padding on optimized kernels Resolves: COMPMID-3867, COMPMID-4361 Change-Id: I0b0867e05f61be4f368f62190d55e14d0ab3ebf2 Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5622 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2021-06-15Port CLWinogradConvolutionLayer with ClWinogradConv2dManuel Bottini
Port CLWinogradInputTransformKernel Port CLWinogradFilterTransformKernel Port CLWinogradOutputTransformKernel Resolves: COMPMID-4504 Change-Id: I3177dda0b9c2f56b36cb317027e94abe8d47229e Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5680 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-06-15Fix incorrect memory handling in ported functionsManuel Bottini
Details of the functions: - ClSoftmax - CpuSoftmax - CpuPool2d Change-Id: Icd2c14d5df010c3b2301e2693ce6f414d7c61916 Resolves: COMPMID-4404 Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5797 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-06-15Add NHWC support to CLRemapFrederick Liardet
Add NHWC support to CLRemap, also add relevant tests. Partially resolves COMPMID-4335. Change-Id: I119bea99be497fb85d5cd83a10f8d4e8e1f97f17 Signed-off-by: Freddie Liardet <frederick.liardet@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5773 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-06-11Fix errata in documentationJakub Sujak
This patch addresses the following errata found in the project documentation: * Common typos. * Missing use of trademarks. * Incomplete operator descriptions. * Examples of code that have since been removed from the library. * Plus clarification over the usage of `All` category for data types and layouts. In addition, the Operator list was not generated properly due to: * Non-matching cases in the filenames (i.e. `Elementwise` and `ElementWise`). For consistency, all usages of the latter have been renamed to the former. * Extra data layout tables in the headers for the `NESlice` and `NEStridedSlice` functions (note: not present in CL counterpart) meant documentation for those functions was generated twice. Resolves: COMPMID-4561, COMPMID-4562, COMPMID-4563 Change-Id: I1eb24559545397749e636ffbf927727fb1bc6201 Signed-off-by: Jakub Sujak <jakub.sujak@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5769 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Sheri Zhang <sheri.zhang@arm.com> Reviewed-by: SiCong Li <sicong.li@arm.com>
2021-06-07Fix WeightRetention tests on multiple calls of FullyConnectedGeorgios Pinitas
Alters CLGEMM to update only the mutable elements of the execution pack when reconfiguring the operator with weight retention. Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I993e3a486e1b025bf6172a7935a10b3a86db106c Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5771 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-06-01Rename ported functionsManuel Bottini
Rename CpuPooling to CpuPool2d Rename CpuPoolingKernel to CpuPool2dKernel Rename CpuPoolingAssemblyWrapperKernel to CpuPool2dAssemblyWrapperKernel Move CpuPool2dAssemblyWrapperKernel in internal subfolder Rename CpuDepthwiseConvolutionNativeKernel to CpuDepthwiseConv2dNativeKernel Rename CpuDepthwiseConvolutionAssemblyDispatch to CpuDepthwiseConv2dAssemblyDispatch Rename CpuDepthwiseConvolution to CpuDepthwiseConv2d Rename CpuDirectConvolutionKernel to CpuDirectConv2dKernel Rename CpuDirectConvolutionOutputStageKernel to CpuDirectConv2dOutputStageKernel Rename CpuDirectConvolution to CpuDirectConv2d Rename ClPoolingKernel to ClPool2dKernel Rename ClPooling to ClPool2d Rename ClDirectConvolutionKernel to ClDirectConv2dKernel Resolves: COMPMID-4405 Change-Id: I8e48f015e4e492a76a7512f5679cb3eb0cd028f6 Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5708 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-05-20Add support for dynamic weights in CL FullyConnected layerGiorgio Arena
Make GEMM use its native version if weights are dynamic. This ensures no reshape gets performed on the weights tensor Enable dynamic weights tests for the OpenCL backend Resolve COMPMID-4223 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Change-Id: Iccc4806701772cede23e24df09c786914d00034c Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5652 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
2021-05-19Port DepthConvert to new ApiGeorgios Pinitas
- Renames DepthConvert to Cast - Ports both NEDepthConverLayer and CLDepthConvert variants - Removes legacy shift capability from DepthConvert, allowing only shifts of 0 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I806a0f8eb23d23502b632c529fda7edde19c8176 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5565 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-05-18Port CLGEMM to memory injecting interfaceGeorgios Pinitas
Moves the following kernels: - CLGEMMMatrixMultiplyKernel - CLGEMMMatrixMultiplyNativeKernel - CLGEMMMatrixMultipluReshapedKernel - CLGEMMMatrixMultiplyReshapedOnlyRHSKernel Moves the following functions - CLGEMM Introduces facilities to easy handling of auxiliary temporary buffers under then new run interface. Such are: - CLAuxTensorHandler: That allows wrapping of workspace buffers memory to CLBuffer objects - Ability to inject TensorInfo to allocator without transferring ownership. This reduce the copy overhead if needed. Resolves: COMPMID-4188 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I7055435d831b05b749b26302082e4ac45f26dfb0 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5498 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-05-18Port CLFlattenLayer to a memory injecting interfaceGeorgios Pinitas
Resolves: COMPMID-4502 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I00621a772135f65a92aa70e6a19dd5f743915378 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5649 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-05-17Fix oclgrind error on CLGEMMLowp reshaped only RHS quantized per channelGiorgio Arena
Fix corner case in which the quantization is per channel and OFM == 1 The function can safely set the per_channel quantization flag to false since there is only one output channel This way, the kernel will avoid adding useless padding to the output multipliers and shifts Resolve COMPMID-4384 Change-Id: Ic03452bfaf52d1be536cd371721adedd2e580a08 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5648 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2021-05-14Fix the OpenCL convolution layer heuristic for int8Gian Marco Iodice
- Call GEMM-based convolution rather than Direct convolution for quantized data types Resolves COMPMID-4497 Change-Id: Idde5808b8a9de69c5196c56538520c0e3d52ba76 Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5618 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2021-05-07Update heuristic for CLConvolutionLayerGian Marco Iodice
- Call direct convolution when filter size height is greater than or equal to 5 Resolves COMPMID-4439 Change-Id: Ie8ccccc0629eb4c74bd62c4bb4ced47f6898a945 Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5589 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-05-05Rename Quantization/Dequantization kernels/operators to imperative moodGeorgios Pinitas
Renames the following kernels/functions - [Cl|Cpu]DequantizationKernel -> [Cl|Cpu]DequantizeKernel - [Cl|Cpu]Dequantization -> [Cl|Cpu]CpuDequantize - [Cl|Cpu]QuantizationKernel -> [Cl|Cpu]QuantizeKernel - [Cl|Cpu]Quantization -> [Cl|Cpu]Quantize Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Ic3c5eb3b7fe28f807294d159830eef99c2dd6219 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5566 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-05-04Rename PixelwiseMultiplications to Mul for simplicityGeorgios Pinitas
Changes the names of the following: - PixelWiseMultiplicationKernel to MulKernel for all backends - PixelWiseMultiplication to Mul for all backends Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I88108c2d22c888fce37ea1028863026160b9da97 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5534 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-04-29Remove OpenCL padding: CLReductionOperationKernelGiorgio Arena
Change the parallel implementation across the X, now every thread computes one row Add missing test for MEAN_SUM Make reduction on any axis != 0 work with num_channels > 1 Resolve COMPMID-3917 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Change-Id: Ib0f99540104e3c253bcd1ea637833db533f5e76e Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5522 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-04-28Fix for performance regression on G71 (mate9) for direct convolutionAleksandr Nikolaev
This variant with GEMM solves regression problem only partially, but guarantees precision. Resolves: COMPMID-4387 Signed-off-by: Aleksandr Nikolaev <aleksandr.nikolaev@arm.com> Change-Id: I7fdb7d6ecf9d67926d994a9f666a80aeac3147a5 Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/317792 Tested-by: bsgcomp <bsgcomp@arm.com> Reviewed-by: Michele DiGiorgio <michele.digiorgio@arm.com> Comments-Addressed: bsgcomp <bsgcomp@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5470 Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-04-28Add per-channel quantization support for CLDeconvolutionLayerFreddie Liardet
Add QSYMM8_PER_CHANNEL support on weight input for CLDeconvolutionLayer. When weights are per-channel quantized type "Direct" method is always used. Also reduce number of QSYMM8_PER_CHANNEL tests for NEDeconvolutionLayer. Resolves: COMPMID-3438 Signed-off-by: Freddie Liardet <frederick.liardet@arm.com> Change-Id: I1330cac5142e19d21e322574fb8d912558745b02 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5484 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-by: Giorgio Arena <giorgio.arena@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-04-27Fixed CTS failures CLInstanceNormPablo Tello
* Resolves COMPMID-4400 Change-Id: I54c33a017c735194fbf4437d1c7df465208bc0ca Signed-off-by: Pablo Tello <pablo.tello@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5505 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Sheri Zhang <sheri.zhang@arm.com>
2021-04-20CLDepthwiseConvolutionLayer rework - Part 1Giorgio Arena
Remove the reshaped variant for CLDepthwiseConvolutionLayer 3x3 NHWC Quantized - Remove kernel selection by GPUTarget - Remove unused quantized support from the NHWC kernel - Remove CLDepthwiseConvolutionLayerReshapeWeightsKernel - Remove OpenCL kernels for reshaped dwc 3x3 quantized and weights reshape - Remove the "_bifrost" suffix in common OpenCL kernel - Remove the ICLDepthwiseConvolutionLayer3x3Kernel common interface Resolve COMPMID-3864, COMPMID-3907 Change-Id: Icfac0fb6c00e214985beb05dad7c0cdbbee7d830 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5447 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-04-20Remove OpenCL padding: CLPixelWiseMultiplicationKernelGiorgio Arena
- Change kernel's vec_size to 16 / sizeof(output) - Change ICLKernel.cpp to handle broadcast without padding Resolve COMPMID-3913 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Change-Id: I03e884b250ef5784dc109bff8cf2c96b345d119f Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5450 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
2021-04-19CLInstanceNormalizationLayer NHWC optimisationPablo Marquez Tello
* Make changes to split the workload into two kernels. One kernel precomputes mean and variance and the second kernel just loads these precomputed values. * The new approach runs %30 faster than the original code for NHWC workloads like 32x192x256. * Resolves MLCE-337 Change-Id: I8356fcefa2d131ab4dcb32268ce7142421d073e4 Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5355 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
2021-04-19Port DepthwiseConvolution to new APIMichalis Spyrou
Resolves: COMPMID-4185 Change-Id: Ib5f22356356a022d567bb18d44ea272b62d10ebf Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5424 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-04-14Remove OpenCL padding: CLNormalizationLayerKernelManuel Bottini
Only for NHWC data layout Resolves: COMPMID-3910 Change-Id: Ie2d71482b3e3b55ac155e9af152032a5de8bbd50 Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5388 Reviewed-by: Giorgio Arena <giorgio.arena@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-04-13Port CLConvertFullyConnectedWeights to new APITeresa Charlin
* Replace ICLKernel by IClKernel in other unrelated kernels Resolves partially: COMPMID-4187 Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com> Change-Id: I173b8f2ac645dbfd7d412f4b058c5c9655c229ee Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5402 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-04-12Fix CLDepthwiseConvolutionLayer QSYMM8_PER_CHANNEL mismatchesGiorgio Arena
Resolve COMPMID-4367 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Change-Id: I5e65b62c2ca52cf65950c9c343864ef55b7122c3 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5407 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-04-12Add support for cl_image in CLDirectConvolutionLayerGian Marco Iodice
- The cl_image object can be used for the weights - cl_image can only work for f32/f16 - Fix the implicit padding on the first dimension X Resolves COMPMID-4341 Change-Id: I04e0901c69e7765c42afceca38c4a840645b9123 Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5393 Reviewed-by: Giorgio Arena <giorgio.arena@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-04-08Substitute CLFullyConnectedLayerReshapeWeights by CLTransposeTeresa Charlin
Resolves partially: COMPMID-4359 (1/2) Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com> Change-Id: Id1859f3cd530eb05f027226e2004cf518778147e Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5377 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-04-06Mixed data-layout testing on high priority operatorsManuel Bottini
Change data layouts after the configure in validation tests for: - Scale - Pooling - FullyConnected - DepthwiseConvolution - DirectConvolution - FFTConvolution - WinogradConvolution - GEMMConvolution (Indirect GEMM included) Extending fixtures Fixes for new mixed data layout tests Resolves: COMPMID-4162 Change-Id: I2f2eb2075f7e24ab3872249d88cadb57b82c5dde Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5326 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2021-04-01Added Qasymm8 datatype support to CLROIPoolingLayer with TestsSuhail Munshi
Also fixes RoiPoolingLayer not matching reference with Float32 datatype Issue Tests added to check ROIPooling Layer against reference with both Float32 and Qasymm8 input. Resolves : COMPMID-2320 Change-Id: Ib86d2e6b3803e74f922a545ea573da02c28e54cc Signed-off-by: Suhail Munshi <MohammedSuhail.Munshi@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5332 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
generated by cgit v1.2.1 (git 2.23.0) at 2024-05-23 01:44:59 +0000