aboutsummaryrefslogtreecommitdiff
path: root/src/runtime
AgeCommit message (Collapse)Author
2021-11-12Fix PostOp dependencySiCongLi
In general src headers should not be included in any public header of other modules. Since there are modules (graph, tests) that rely on specific PostOp definitions in the previous src/core/experimental/PostOp.h, export it to the public arm_compute header Resolves COMPMID-4974 Signed-off-by: SiCongLi <sicong.li@arm.com> Change-Id: I0fa4da5108a34fe6bfff1e9d57839da4e51dc314 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6673 Reviewed-by: Gunes Bayir <gunes.bayir@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-11-01Add PostOp support to GEMM and CLGEMM operators and functions Part 2SiCongLi
* Implement PostOp interface changes * Remove spaces around "=" in TypePrinter Partially resolves COMPMID-4435 Signed-off-by: SiCongLi <sicong.li@arm.com> Change-Id: If1e2280554030a0f635e73339a2e86987f6dc41b Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6484 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Sheri Zhang <sheri.zhang@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-10-28Add experimental PostOp interface to ClGemmMatrixMultiplyReshapedKernel Part 1SiCongLi
This interface supports the fusion of multiple elementwise operations Partially resolves: COMPMID-4435 Change-Id: If68dd7dd98dcf239fde7cb1f0a4a6d4d1e899a6f Signed-off-by: SiCongLi <sicong.li@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6483 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-10-18DirectConv3d support refineSheri Zhang
- Decouple data support of CpuDirectConv3dKernel - Update documentation for Conv3d Signed-off-by: Sheri Zhang <sheri.zhang@arm.com> Change-Id: I1d94aa28f821f45a1a3d39cc3335c8faeee89f0d Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6453 Reviewed-by: Giorgio Arena <giorgio.arena@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-10-18Remove legacy GeMM kernels on OpenCLGian Marco Iodice
Resolves COMPMID-4446 Change-Id: I1d3c2391b67681f4d3af440826aa95b47a1288a6 Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6444 Reviewed-by: Giorgio Arena <giorgio.arena@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-10-15Conv3d supportSheri Zhang
* Add CpuDirectConv3d support for fp32 and fp16 * Dilation is not supported * Need decouple Partially resolve: COMPMID-4661 Signed-off-by: Sheri Zhang <sheri.zhang@arm.com> Change-Id: Ib1865b9ff328b684d131512b1baf77bc2f10318f Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6430 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>
2021-10-14Implement CLDirectConv3D f32/f16Giorgio Arena
Resolve COMPMID-4660 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Change-Id: Ibd66ec1eb6faa60086981b1e3a9c12561df3445f Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6420 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
2021-10-07Per-operator build dependenciesGeorgios Pinitas
Creates a list of operators their respective dependencies. Alters the build system to walk-through them resolve the dependencies and build Compute Library. Removes the following unused kernels/functions: -[NE|CL]MinMaxLayerKernel -CLFillBorder Resolves: COMPMID-4695,COMPMID-4696 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I35ebeef38dac25ec5459cfe9c5f7c9a708621124 Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/357914 Tested-by: bsgcomp <bsgcomp@arm.com> Reviewed-by: Michele DiGiorgio <michele.digiorgio@arm.com> Comments-Addressed: bsgcomp <bsgcomp@arm.com> Signed-off-by: Freddie Liardet <frederick.liardet@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6295 Reviewed-by: Gunes Bayir <gunes.bayir@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-10-05Fix of linking errors to find CPPSplit symbols when compiling android nn driverramelg01
Resolves: COMPMID-4835 Signed-off-by: Ramy Elgammal <ramy.elgammal@arm.com> Change-Id: I1e768b3f3130316e71b994a9b3fbc81d518d5361 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6368 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Sheri Zhang <sheri.zhang@arm.com>
2021-10-04Provide logging for configure functions in all CPP functionsramelg01
- Moving impl of CPPSplit template to src/runtime/CPP to allow including of Log.h from src/common. - Fix logging of vector<ITensor*> to print contained tensor's info not their ptrs. Partially-Resovles: COMPMID-4718 Signed-off-by: Ramy Elgammal <ramy.elgammal@arm.com> Change-Id: Idec81665b2a7c0cfae5248803109c6e2edc520a1 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6362 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-10-01Provide logging for configure functions in all CL functionsramelg01
Partially Resolves: COMPMID-4718 Signed-off-by: Ramy Elgammal <ramy.elgammal@arm.com> Change-Id: I59943081de0f3105bdbfc78af53971a0de8cd53e Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6332 Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-09-22Provide logging for configure functions in all NEON functionsramelg01
Partially Resolves: COMPMID-4718 Signed-off-by: Ramy Elgammal <ramy.elgammal@arm.com> Change-Id: I655268c57fa126d9c99981c49d345a3aac75646e Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6286 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com> Reviewed-by: SiCong Li <sicong.li@arm.com>
2021-09-06OpenCL job chaining supportGiorgio Arena
Resolve COMPMID-4714 Change-Id: I53b74956da10a8d7ffa2c5681f3b7f74acaa3201 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6215 Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-08-25Move CPU/GPU files from Core/Runtime to the respective backend foldersGeorgios Pinitas
Legacy structure contained two libraries core/runtime with two backends in each. We reduce the core/runtime libraries to a single library thus merging the backend files Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I69545765fe7a730368105cdbd067d3135ec7a174 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6155 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-08-24Re-use auxiliary memory withing CpuWinogradConv2d operatorsGeorgios Pinitas
Input/Output transformation operations are independent and done in different time-steps of the algorithm, this memory can be re-used between this transformation stages. Moreover, reduce the allocation when extracting workspace sizes for Winograd trasformations. There is a mix return of sizes in bytes and elements, thus ensure the correct is in place. storage_size() member functions return elements while working_space() function bytes. Resolves: COMPMID-4781 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I705445ba7ca818cead48369db3cacd49684c7192 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6145 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
2021-08-23Remove padding from ClScaleKernelGiorgio Arena
- Merge quantized kernels with fp for bilinear interpolation (both NCHW and NHWC) - Pass dimensions at compile time rather than at run time - Use tile-based approach to rework the NCHW kernels - Remove unused functions/files Resolve COMPMID-4723 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Change-Id: Ifcdf02beb9daa9f318395751b3c85eb2fe874082 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6138 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
2021-08-20Rename [Cl|Cpu]GemmConvolution to [Cl|Gpu]GemmConv2dGeorgios Pinitas
Renaming the gemm-based convolution operators to accomodate for new operators with higher convolution dimensonality Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Id2f2cf11404221f0e87baa0e5d08ad5d63eaf78e Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6113 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-08-19Address comments on avoiding releasing weights if used by multiple functionsGiorgio Arena
Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Change-Id: I0b59c5326f5fcbc322fbeb864197ea999de6bd56 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6112 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-by: SiCong Li <sicong.li@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-08-18Enable fast_math on CpuGemmConvolutionGeorgios Pinitas
Fast-math mode enables faster GEMM implementations to be used assuming an accuracy impact is acceptable. Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I0d0f23a2843426cc37a31f63022c790249b71c69 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6122 Reviewed-by: SiCong Li <sicong.li@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-08-18Update the heuristic to call direct convolution in clConv2dGian Marco Iodice
Resolves COMPMID-4773 Change-Id: Id0e4126bd28025d13011a14cba99022a1c88322a Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6123 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
2021-08-18Retain weights in ClGemm when reconfiguring the operator with retentionGeorgios Pinitas
- Reduces the size of the WeightRetentions tests - Ensure the weights are retained the only the Src,Dst are updated Resolves: COMPMID-4775 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: If12daa4bd4bf89ec28faa743fb7291895cbe7b7c Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6121 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-08-13Avoid releasing weights if they are used by multiple functionsGeorgios Pinitas
Resolves: COMPMID-4769 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Iccadcbd68b0fd84ed3bf212e358a4ea944084a40 Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/349845 Tested-by: bsgcomp <bsgcomp@arm.com> Reviewed-by: Giorgio Arena <giorgio.arena@arm.com> Comments-Addressed: bsgcomp <bsgcomp@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6107 Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-by: SiCong Li <sicong.li@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-08-13Ensure correct transformed matrices are used in CpuGemmConvolutionGeorgios Pinitas
Resolves: COMPMID-4763 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Iae2e093cfb7d2c7172603897afe1c6a2e5d1caa3 Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/349725 Tested-by: bsgcomp <bsgcomp@arm.com> Reviewed-by: Pablo Tello <pablo.tello@arm.com> Comments-Addressed: bsgcomp <bsgcomp@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6101 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
2021-08-12Ensure that correct transformed matrices are used in CpuFullyConnectedGeorgios Pinitas
Execution pack of CpuFullyConnected was altered explicitly with local objects that were getting out of scope. Leading to incorrect results or memory related issues. Track transformed weights and register the weights matrix explicitly during execution honoring the object lifetime scope. Resolves: COMPMID-4762, COMPMID-4764 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I53449c377fb1cfccdf5e6f9505d963518748c318 Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/349345 Tested-by: bsgcomp <bsgcomp@arm.com> Reviewed-by: Pablo Tello <pablo.tello@arm.com> Comments-Addressed: bsgcomp <bsgcomp@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6092 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-by: SiCong Li <sicong.li@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-08-11Fix performance regression due to clFinish()Gian Marco Iodice
- In ClGemmLowpMatrixMultiplyCore::prepare we always called clFinish() also when the workload was already prepared Resolves COMPMID-4707 Change-Id: Icdcee528590e2c5efb75325a80c2a45ec84993d1 Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6082 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>
2021-08-10Fix compiler error in CLActivationLayerPablo Marquez Tello
* GCC 7.4 + Ubuntu 16 * Resolves COMPMID-4761. Change-Id: Id259f00c860b23c4a20881dd0c6ab6468cf17385 Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6070 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Nikhil Raj Arm <nikhil.raj@arm.com>
2021-08-06Fix compiler error in GCC 7.4 + Ubuntu 16Pablo Marquez Tello
* Resolves COMPMID-4761. Change-Id: I849836fa507e8fa2b01737ace16e0cb30a9eb42c Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6061 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Nikhil Raj Arm <nikhil.raj@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-08-04Remove 21.08 deprecated codeFreddie Liardet
Also correct some deprecation macros Signed-off-by: Freddie Liardet <frederick.liardet@arm.com> Change-Id: Ib28d42571f55d7b01563a0877b92b8f3128831ba Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6036 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-08-04Report error for unsupported non-constant weights in CpuFullyConnectedMichele Di Giorgio
Change-Id: I434844b9194997028e1666f4bce58e69839b886d Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6038 Reviewed-by: Sheri Zhang <sheri.zhang@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-08-04Avoid over-allocation of temporary buffers within CpuWinogradConv2dGeorgios Pinitas
Resolves: COMPMID-4716 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Ie036d2bb7a243301a62f089b3920ebee0f409190 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6028 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
2021-08-04Implement Operator APIGeorgios Pinitas
Resolves: COMPMID-4512 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Id12130365fa3fe2261160931dcc7affb6b467186 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6031 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-08-02Port CLConvolutionLayerSheri Zhang
Resolves: COMPMID-4508 Signed-off-by: Sheri Zhang <sheri.zhang@arm.com> Change-Id: Ifcb6d8ab290655dc8f47a067f23324cf227736f1 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6015 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-07-30Port ClFullyConnected to new APIGeorgios Pinitas
Resolves: COMPMID-4391 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Idcd5e22ed6e901c7f4c7530e5547ea6a7814ae59 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6025 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-07-30Reintroduce implementation of NEConvolutionLayer::get_convolution_methodMichele Di Giorgio
Issue caused by the porting of the function to the new API. The method will call down to the new CpuConv2d implementation. Change-Id: I650ad1f17c8b89a637b589e452ca785b5d14e975 Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6027 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Sheri Zhang <sheri.zhang@arm.com>
2021-07-29Port NEConvolutionLayerMichalis Spyrou
Resolves: COMPMID-4507 Change-Id: I9557026ec0052b5585994f7a1300a14565c976d0 Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5964 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2021-07-28Reduce binary footprint of CpuConvertFullyConnectedWeightsKernelMichele Di Giorgio
Binary size reduction for this kernel is almost 50%. Also remove unused NEConvertFullyConnectedWeightsManaged. Change-Id: Ia46a1342a0737397b4aac2578d963c2ebb7446e3 Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6011 Reviewed-by: Giorgio Arena <giorgio.arena@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
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>
2021-07-26Fix allocation of prepare tensor on ClWinogradConv2dGeorgios Pinitas
Preparation tensor was allocated during execution. Avoid if GEMM used by the Winograd function is reshaped. Resolves: COMPMID-4640 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Id5b7e6bbfe9f69661baa9d42698cdc2cc1b422b2 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5991 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-07-23Avoid allocation of auxiliary memory in CpuGemmConvolutionGeorgios Pinitas
Resolves: COMPMID-4690 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I2d44fd59fc66e2d3e80acffd1a130f6d3fab5c57 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5990 Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-07-23Port NEFullyConnectedLayer to memory injecting interfaceMichele Di Giorgio
Resolves: COMPMID-4501 Change-Id: Ib61b3d06974009e501b3fb86467735427e13a94a Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5931 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2021-07-23Pass fast math flag for correct GEMM3D validation supportGeorgios Pinitas
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Ic96bac64680e052e39c21d2e0eb9e618bcf259b4 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5989 Tested-by: Arm Jenkins <bsgcomp@arm.com>
2021-07-22Expose fast_math mode for GEMM through BFloat16Georgios Pinitas
Resolves: COMPMID-4641 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I7ccc663b2692d40c370794caa906b5be8fd25a32 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5977 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2021-07-22Inject temporary tensors to pack in they don't exist in CpuSoftmaxGeorgios Pinitas
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Ief95ad8c97271f9042534e149fbccc9199318c2a Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5975 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>
2021-07-15Disabled DirectConv method for NCHW and kernel > 15Pablo Marquez Tello
* We prefer GEMM method for kernel size > 15 when data layout is NCHW because direct convolution does not support this. * Resolves COMPMID-4581 Change-Id: Ie18cc96bc6b446fd59e8c8ebb10c3af5ca02c3bb Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5935 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
generated by cgit v1.2.1 (git 2.23.0) at 2024-06-28 06:48:01 +0000