Age | Commit message (Collapse) | Author |
|
Partially Resolves: MLCE-1255
Change-Id: Ibadcfedd43530232c65f05e571bc8b4568a63e67
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11499
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
|
|
Partially Resolves: ONCPUML-1444, MLINFSW-439
Change-Id: Ic7498d6944df2848f3e82eaf4e11cc5cb6ef5754
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11424
Reviewed-by: Anitha Raj <Anitha.Raj@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
- Add support for QASYMM_SIGNED*QASYMM8_SIGNED->F32 in
CpuGemmLowpMatrixMultiplyCore
- Add s8f32 kernel using existing s8->s32 kernels with a new
DequantizeFloat OutputStage, the structure is similar to Requantize32
but the opposite way around.
- Add SME s8f32 kernels with integrated support for DequantizeFloat.
- Add scale to CpuGemmLowpOffsetContributionKernel.
- Add virtual dequantize scale to gemm_common, only implemented for
gemm_interleaved.
- Update year to 2024 in generate_build_files.
- Add dynamic flag to QuantizationInfo which signals to operators that
it can change after configuration
- Add support for dynamic quantization in NEGEMMLowpMatrixMultiplyCore
- Add dynamic quantization fixture by extending
GEMMLowpGenericMatrixMultiplyCoreValidationFixture
- Add GEMMLowpDequantizedMatrixMultiplyValidationFixture
- Store k (number of cols of A) rather than k_offset in the offset
contribution kernels so that we can recompute it when the other
offsets change
relates to: ONCPUML-1444 MLINFSW-439
Co-authored-by: Milos Puzovic <Milos.Puzovic@arm.com>
Co-authored-by: David Mansell <David.Mansell@arm.com>
Change-Id: I58a3acf2c09289a303e52eea6b336a696a5bc8da
Signed-off-by: Jonathan Deakin <jonathan.deakin@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11022
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Partially Resolves: ONCPUML-1442
Signed-off-by: Radu Salavat <radu.salavat@arm.com>
Change-Id: I681df5e9c399996fbc7dc362b906af151588ca44
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11416
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
|
|
Add checks for bf16 support for bf16 fixed format tests.
This ensures tests pass in multi_isa setting where library was compiled
with bf16 support, even on systems that do not support bf16.
Also adds runtime check to GEMMConvolutionLayer/Float/BFLOAT16/RunSmall.
Resolves: COMPMID-6922
Signed-off-by: David Svantesson-Yeung <david.svantesson-yeung@arm.com>
Change-Id: Ic0f09ba34b5a2c64be8bfc848a4457a6b1c4d1c3
Signed-off-by: David Svantesson-Yeung <david.svantesson-yeung@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11408
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
aarch32. This patch guards the relevant tests.
Partially Resolves: ONCPUML-1442
Signed-off-by: Radu Salavat <radu.salavat@arm.com>
Change-Id: I8eed80db4b522185c3c50c13f0f701aa48961057
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11410
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
|
|
In addition to the softmax kernel, this patch fixes minor issues in the fp32 implementation.
Resolves: COMPMID-6920
Change-Id: Ibbd9f0af5f2a93fba0e92d72ba437279c34149d3
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11402
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
|
|
Instead of dispatching the sum postop for GEMM kernels to a
separate kernel + add, that requires an extra destination sized
allocation, plus 3 extra load/stores per element,
just do it in the GEMM kernel.
Resolves: ONCPUML-1442
Signed-off-by: Radu Salavat <radu.salavat@arm.com>
Co-authored-by: Milos Puzovic <milos.puzovic@arm.com>
Change-Id: I7a1f2da3300875fa1ac88b705a34390969518077
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11298
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
ARM_COMPUTE_ENABLE_FIXED_FORMAT_KERNELS
Change-Id: I6a01fe1e19a9d3e38908309d766fe7fc43775490
Signed-off-by: Renato Arantes <renato.arantes@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11338
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
|
|
PyTorch® autocast() function
The full range of tests must be added with [MLINFSW-482] epic due to the lack of reordering kernels implemented in Acl.
Co-Authored-By: David Mansell <David.Mansell@arm.com>
Change-Id: I820d316295a1ec94fdc89c37e4144a268f914c36
Signed-off-by: Renato Arantes <renato.arantes@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11169
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>
|
|
- Neon(TM) implementation converts integers to float and performs the division because there is no vector integer division instructions. However, leftover loop still uses integer division, which makes results inconsistent depending on where we are in the tensor.
- SVE path does it in integer domain.
- OpenCL(TM) does it similar to Neon(TM) vector path.
- Reference implementation does it in integer domain.
These differences cause intermittent mismatches. This patch ensures all follow the same logic.
On the other hand, the provided Neon(TM) implementation is faster than the Fp32 converted version.
Resolves: COMPMID-6925
Change-Id: Ia12606d57f40a7d331b9b698f87fd4321496b275
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11316
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Instead of increasing the tolerance amount, we increase the number of elements we tolerate to 0.02 % of the whole tensor. This ensures we do not affect the tolerance for smaller tests.
This amount is set according to the number of elements above the threshold. it was 1 over 512 elements, 1/512 ~ 0.02 %.
Resolves: COMPMID-6932
Change-Id: I9d3ce29a3972aa8b9daea5288005a0a41a266328
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11321
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
|
|
Tolerance for quantized tests is better to be 1 due to possible rounding differences between the Acl and reference implementations.
Resolves: COMPMID-6929
Change-Id: I6f317631322b702e6a9579593befff65bbf46151
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11319
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
* Validate output shape in CpuPool2dAssemblyWrapperKernel
* Resolves ARMCL-625
Change-Id: I4fd91c1b15ecb17efc39fd3e82a92210e4f182b2
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11290
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Indirect GEMM uses optimized assembly path while Direct Conv uses the fallback Acl kernel for convolution.
In certain cases, where the input tensor is large and filter size is greater than 7 (e.g. 9x9 filters), heuristics fall back to Direct Conv algorithm where it could still prefer the assembly path if the data layout is NHWC. This is more important when SME2 kernels are present.
Resolves: COMPMID-6900
Change-Id: Ia611c975eee0423615113fcaeaa8f9eef0421456
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11254
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Anitha Raj <Anitha.Raj@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
|
|
Int8 data types should be compared with Int8 tolerances. Same holds for UInt8 as well. O/w the differences between reference and target can be larger than it is because of casting.
Resolves: COMPMID-6930
Change-Id: I4940d821b7fecc21cf6b167e161dffceb764b909
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11269
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: <felixjohnny.thomasmathibalan@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves: [COMPMID-6681]
Signed-off-by: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>
Change-Id: I325b9d478dd1d04a45533bb7708cf76e98ee0cee
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11058
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves: COMPMID-6483
Signed-off-by: Omar Al Khatib <omar.alkhatib@arm.com>
Change-Id: I512102f5e27743098168101b5e02382f4ad4a22a
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11068
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves: [COMPMID-6791]
Signed-off-by: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>
Change-Id: Idae6ddb0e9655ec096f25917f0a44eb57aaef908
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11076
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
|
|
BF16 kernels are not expected to have the same tolerance/accuracy standards as full float kernels. The reference implementation is a standard floating point implementation, thus resulting in small mismatches.
We increase the tolerance of the MatMul BF16 tests, and add more tests to cover more shapes. Previously, the only tested bf16 kernel was a64_hybrid_fp32bf16fp32_mmla_4x24. With the inclusion of new shapes, heuristics also choose a64_hybrid_fp32bf16fp32_mmla_6x16 and stress this kernel as well, covering every implementation.
Resolves: COMPMID-6654
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Change-Id: I15342606912013c123b94c7e0ea2e6bbb25680d7
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11014
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
This patch calculates the output quantization info based on the inputs'
quantization information.
The previous approach was using the same quantization information for
input, weights and output.
Remove QSYMM8_PER_CHANNEL path from the fixture as there are no
related tests
Remove repeated shapes from the dataset now that we get rid of the
quantization info from the dataset.
Combine signed and unsigned SmallGEMMLowpFusedBatchedMatMulDataset
into one as they become identical
Resolves COMPMID-6481, COMPMID-6634
Change-Id: I9f5a20f4bb45c3e5adab388564135ae8a5c0a9ea
Signed-off-by: SiCong Li <sicong.li@arm.com>
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10680
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves COMPMID-6728
Signed-off-by: SiCong Li <sicong.li@arm.com>
Change-Id: Ic0682550a09db9aa420057a90ee65386e16e6034
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10853
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Implement a single kernel instead of having two consecutive ones. In the previous setup, one kernel was calculating the maximum value in the axis, and this maximum was being subtracted from each data while calculating the softmax, i.e.
softmax(x_i) = exp(x_i - max) / sum_i( exp(x_i - max) )
This patch integrates these two stages into a single kernel for Neon™ for all data types. This will save some memory because we don't need to hold the max values in a separate auxiliary tensor.
It also introduces some other optimizations that will ease memory pressure when the data type is float/half, by using the dst tensor as temporary storage for already exponentiated inputs.
It removes the references to SVE and SVE2 implementations, and most of the associated files; but, it leaves the implementations as these may be used in the future.
Resolves: COMPMID-6500
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Change-Id: Icff9976d1214c4c6cbe15a62ca60b8a77d3784cc
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10688
Reviewed-by: SiCong Li <sicong.li@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
When weight has no holes, we can replace CpuWeightsReshapeKernel with:
- Collapse by reinterpreting weight's 3 spatial dimensions
- Perform CpuTranspose
For more details see the documentation in
src/cpu/operators/CpuGemmConv2d.cpp
This is one optimization since the CpuTranspose is better performing
than CpuWeightsReshapeKernel
A second optimization is to fuse this transpose with other weight
transformations (e.g. pretranspose_B_array in CpuGemmAssemblyDispatch)
However this second optimization depends on how the underlying gemm
methods (the fall back path: CpuGemmMatrixMultiplyKernel or the assembly
path: CpuGemmAssemblyDispatch) chooses to fuse the transpose.
Therefore, this patch moves the transpose down from CpuGemmConv2d, to
the individual gemm operators where the fusion decision needs to be
made, by passing an extra "transpose_b" flag to CpuGemm
New transpose_b flag in different scopes (they are all the same, but
with different names because pretranspose_b has a different meaning in
GemmAssemblyDispatch):
GEMMInfo::pretranspose_B -> AsmGemmInfo::transpose_b
New auxilliary tensors holding the transposed b result:
- CpuGemm optimized path: CpuGemmAssemblyDispatch::PrePretransposedB
- CpuGemm fallback path: CpuGemm::PreTransposedRHS
Note that this patch does not yet have the second optimization
(COMPMID-6595), but it prepares for it.
Relates to COMPMID-6595
Resolves COMPMID-6499
Change-Id: I999a2da9da4b2b15369a3cc06d7872c86e0190ea
Signed-off-by: SiCong Li <sicong.li@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10526
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Anitha Raj <Anitha.Raj@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
This patch calculates the output quantization info based on the inputs'
quantization information.
The previous approach was using the same quantization information for
input, weights and output.
This implementation does not cover the cases where we have fused
activation function.
Resolves: [COMPMID-6484]
Signed-off-by: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>
Change-Id: Ib58143165191e82ae8547e661ac7c8d077bda200
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10539
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
This patch calculates the output quantization info based on the inputs'
quantization information.
The previous approach was using the same quantization information for
input, weights and output.
This implementation does not cover the cases where we have fused
activation function.
Resolves: COMPMID-6482
Change-Id: I4a9d87cfef8ad18ef241d457d23f44c8519a1389
Signed-off-by: SiCong Li <sicong.li@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10541
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves: COMPMID-6581
Signed-off-by: Adnan AlSinan <adnan.alsinan@arm.com>
Change-Id: I0a634e064377e54b9190241c01fc75c212522ba7
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10481
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Optimize the stack operation in Cpu by leveraging block memcpy.
Resolves: COMPMID-6498
Change-Id: I49d79d179f0375a73d654edd59fb33072112569b
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10451
Reviewed-by: SiCong Li <sicong.li@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves: COMPMID-6465
Change-Id: I5bbf4596dd5e34e806dc51de9be14df9b6fa320a
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10452
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
|
|
* Added a new test to make sure we support the following configuration:
NCHW
InputInfo=Shape=2,2
WeightsInfo=Shape=3,3
OutputInfo=Shape=4,4,
PadStrideInfo=1,1;0,0,0,0'
* Fixed the validate() method to allow this configuration
* Resolves MLCE-1120
Change-Id: I6874ad57bb81384185984741b983bf5e19ba150c
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10417
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Several test optimizations have been introduced into Winograd tests for Gpu and Cpu backends. The testing strategy has been detailed as a comment header in the test design files.
In summary
- Very large shapes in the nightly are made smaller
- If the underlying kernel is the same for different data types, we only need to stress some key aspects of the kernels (e.g. read/write lengths in case of fp32/fp16).
- In case the underlying kernel is the same (OpenCL), Fp16 is tested on a subset of the shapes
- In Cpu, there is no need to test every combination for both NCHW and NHWC as we just permute the inputs and use NHWC kernels anyways
- All activations does not need to be tested for each and every shape
Resolves: COMPMID-6464
Change-Id: Ie25fded85c65b9c7386dc21b23f9b695b1e77b07
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10393
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
- Add support for negative axis values.
- Add option to use opposite ACL convention for dimension addressing.
- Add validation tests for the mentioned additions.
Resolves COMPMID-6497
Change-Id: I9174b201c3adc070766cc6cffcbe4ec1fe5ec1c3
Signed-off-by: Adnan AlSinan <adnan.alsinan@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10335
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves : [COMPMID-6212]
Signed-off-by: Omar Al Khatib <omar.alkhatib@arm.com>
Signed-off-by: Adnan AlSinan <adnan.alsinan@arm.com>
Change-Id: I29bbd9a3d96af462faf7f0ee13b9849f75e05356
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10319
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
|
|
* Illustrate the benefit by writing CPU MatMul test dataset
in a more readable way.
Part of: COMPMID-6353
Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Change-Id: Id5dbc13a051709237bbcc4dd88716d0b24ecfd5d
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10227
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
* When the output is S64 the function NEArgMinMaxLayer uses CpuCast to
convert the output to S64 and this is only supported on aarch64.
* Disable this test case for non aarch64 builds
* Resolves COMPMID-6536
Change-Id: I554c21ce9a029af086e9137b5369b7951b779997
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10212
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves COMPMID-5279
Change-Id: Id9b007eed62c200702bbfcc83b94dab7b5de1714
Signed-off-by: Anitha Raj <anitha.raj@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9962
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
* NEArgMinMaxLayer uses NEReductionOperation to compute its result in S32
* We need to call NECast to convert from S32 to S64
* Resolves MLCE-1089
Change-Id: I6fded869b6076d7af1b9b3e70eb384f4ee82fd8a
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10054
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves: COMPMID-6423
Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Change-Id: I9cec051a7d1a2956218f8a6d8263bd5424f6d389
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10072
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
* ArgMinMax output was fixed to S32, this patch makes the changes required
to allow other output types like U64/S64
* Made changes to the ArgMinMax fixture and tests to allow specifying output data type.
* Made changes to the reference reduction_operation to allow specifying the output type
* Added tests case to output S64 for the CL backend.
* Added missing test cases in the neon backend.
* Partially resolves MLCE-1089
Change-Id: I6f1cbc7093669d12c2a3aff6974cf19d83b2ecda
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10003
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
* The kernel now supports the following conversions:
S64 -> F32
U64 -> F32
* Resolves MLCE-1089
Change-Id: I277cf58b78d919fde25947520d2056e1412c7f82
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9935
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Makes a small difference to compile times and opens up other opportunities
to simplify code.
Change-Id: I232876910bbe4fa9719f4a0ce4a54c090faeb5ef
Signed-off-by: Matthew Bentham <Matthew.Bentham@arm.com>
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/532429
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9856
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves: COMPMID-6319
Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Change-Id: I49a17ff973efc88b7ce0334c47ecf076c03f4cc3
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9829
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
* Moving the code out of Types.h will help
with the compilation time.
* Added LUT support for all other activation functions.
* Resolves COMPMID-6292
Change-Id: I1b5f0b21f03237447163276b8796b2aeb3fdd45c
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9749
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Tolerance number was too strict for fp16
Resolves COMPMID-6254
Signed-off-by: SiCong Li <sicong.li@arm.com>
Change-Id: I42a5df21c2545c38ea7234497effd232b43aabf8
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9635
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Reviewed-by: Omar Al Khatib <omar.alkhatib@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves: COMPMID-6235
Change-Id: I7a094a23244286090415ee2788632cfa7bd6c037
Signed-off-by: Ramy Elgammal <ramy.elgammal@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9608
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
|
|
* In multi-isa build with v8a base arch, FP16 is only available
in SVE kernels, hence the support for FP16 depends on runtime
environment.
* To simplify the test, the data type test is disable on FP16
if the base arch doesn't have FP16 support.
Resolves: COMPMID-6232
Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Change-Id: Iee6179e1b87d3e7f4349e228479ff170285e4841
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9599
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
|
|
Resolves COMPMID-6151
Signed-off-by: David Svantesson <david.svantesson@arm.com>
Change-Id: I0e8c957f3460633c32ef57be0cdc44a53b8c3e88
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9553
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
Adds Reorder kernel exposing blocking reorders from arm_gemm
Resolves ONCPUML-1232
Change-Id: I42bf4166311fe1771565134d3ed7039fc8e30230
Signed-off-by: David Svantesson <david.svantesson@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9500
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
|
|
for Neon™ Matmul
Partially Resovles: COMPMID-6026
Signed-off-by: Ramy Elgammal <ramy.elgammal@arm.com>
Change-Id: I273b213abba275f1609eae33058e3acbee2a7146
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9489
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
|
|
- All the GeMM CPU assembly kernels for integer datatypes require aarch64
Resolves: COMPMID-6026
Signed-off-by: Ramy Elgammal <ramy.elgammal@arm.com>
Change-Id: I34bb0d5ca5cc3684b996df851227fcd0ad452586
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9481
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Mohmun02 <MohammedSuhail.Munshi@arm.com>
|