aboutsummaryrefslogtreecommitdiff
path: root/src
AgeCommit message (Collapse)Author
2020-01-15COMPMID-2819: Retain layout in NEIm2ColKernelGeorgios Pinitas
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I5af7f9f87cedcfc8a4d59e0376113d2f285754fd Reviewed-on: https://review.mlplatform.org/c/2592 Reviewed-by: Pablo Marquez <pablo.tello@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2020-01-13COMPMID-2819: Fix offset in NEReductionOperationKernel.Pablo Tello
Change-Id: Iacb57077fae9f70b24ecd1cdb4c8d017edd6f237 Signed-off-by: Pablo Tello <pablo.tello@arm.com> Reviewed-on: https://review.mlplatform.org/c/2562 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2020-01-10COMPMID-2819: Retain layout during configuration for multiple functions.Georgios Pinitas
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Ia528762dc5a93bebfd8fd037bf1f4e75d0b8a6de Reviewed-on: https://review.mlplatform.org/c/2566 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
2020-01-09COMPMID-2819: Fix data layout in NEScaleKernel for nhwc.Georgios Pinitas
Instead of retrieving the data layout from the tensors, fix it to NHWC as the selected function is already specialized for NWHC. Change-Id: I67d8ca8b911adb8575e4c4a6232731b423a3604b Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-on: https://review.mlplatform.org/c/2545 Reviewed-by: Pablo Marquez <pablo.tello@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-12-20COMPMID-2819: Retain data layout in NESpaceToDepth and NEDepthToSpaceGeorgios Pinitas
Retain layout from configuration to execution as following configuration might alter tensor meta-data in client code. Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: Ifdbb285809956b966ea534ab3ab231911a307cc0 Reviewed-on: https://review.mlplatform.org/c/2512 Reviewed-by: Pablo Marquez <pablo.tello@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-12-19COMPMID-2819: Retain configuration step data layout to avoid side-effects.Pablo Tello
Configuring functions serially can lead to side-effects in tensor attributes. One of them is the data layout changing in case functions share same IO tensors. Retain DataLayout used during configuration. Change-Id: Ic8594300dc428282f4f9b9196f0d64842b6c1868 Signed-off-by: Pablo Tello <pablo.tello@arm.com> Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/217277 Tested-by: bsgcomp <bsgcomp@arm.com> Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Comments-Addressed: bsgcomp <bsgcomp@arm.com> Reviewed-on: https://review.mlplatform.org/c/2505 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-12-19COMPMID-2819: Retain configuration step data layout to avoid side-effects.Georgios Pinitas
Configuring functions serially can lead to side-effects in tensor attributes. One of them is the data layout changing in case functions share same IO tensors. Retain DataLayout used during configuration. Change-Id: I17538ce08b86df6986f0fcf21fa6544fbd5bd74b Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-on: https://review.mlplatform.org/c/2470 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>
2019-12-17COMPMID-2819: Perform sqrt in double domain for L2 pooling.Georgios Pinitas
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Change-Id: I51d50ceda903c4322b659cd872d041d7db73c335 Reviewed-on: https://review.mlplatform.org/c/2481
2019-12-11COMPMID-2855: CLReduceMean throws error for invalid configsPablo Tello
Signed-off-by: Pablo Tello <pablo.tello@arm.com> Reviewed-on: https://review.mlplatform.org/c/2452 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Change-Id: I7cda1b67aa6c3541fdd7781be12288c8fc36ffeb
2019-12-11COMPMID-2855: NEReduceMean throws error for invalid configsPablo Tello
Change-Id: I600507d0de19d7da6c1a13edcfff0a11ea6b5264 Signed-off-by: Pablo Tello <pablo.tello@arm.com> Reviewed-on: https://review.mlplatform.org/c/2254 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: Giorgio Arena <giorgio.arena@arm.com> Reviewed-by: Giuseppe Rossini <giuseppe.rossini@arm.com> Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com> Reviewed-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-by: SiCong Li <sicong.li@arm.com>
2019-12-11COMPMID-2637 [CL] fix broadcast pixel-wise multiplication with 5D tensorsSang-Hoon Park
Broadcast pixel-wise multiplication with 5D tensors is fixed by adding information whether a dimension has been broadcasted to compute correct start offset when adding 3D tensor argument. The testcase that failed is added to the validation test suite. Change-Id: I320876f507012c27b39daae1316f9b69138ed204 Signed-off-by: Sang-Hoon Park <sang-hoon.park@arm.com> Reviewed-on: https://review.mlplatform.org/c/1994 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2019-08-31COMPMID-2640: Fix performance regression for Resnet101 Int8 on NEONGian Marco Iodice
Change-Id: I32c8b67c5ce0918cc5603807bad80952ea2fd097 Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/1848 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2019-08-26COMPMID-2336: Validate multiplier and offset calculation in LSTMQuantized ↵Michele Di Giorgio
and DepthwiseConvolution functions This patch also adds validation of internal functions in LSTMQuantizedLayer. Change-Id: Id8dbbfbb421f7d053410476b4bb4ef7d85e5f41e Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Signed-off-by: giuros01 <giuseppe.rossini@arm.com> Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/1794 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2019-08-23COMPMID-2336: Update GEMM codeMichalis Spyrou
Fix incorrect stride Change-Id: I5bb9660bd73c40d587fa869d2488b7ab5f5b1ea7 Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com> Reviewed-on: https://review.mlplatform.org/c/1807 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
2019-08-22COMPMID-2602: FullyConnected VTS failure on CL and Neon TENSOR_QUANT8_ASYMMGian Marco Iodice
Change-Id: I918d2b9e6d1297cf78c7e567b2a95406875f82ad Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/1782 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-21COMPMID-2596 MobilenetSSD produce wrong resultsMichalis Spyrou
Update GEMM assembly code. Change-Id: Id315c51a11aa89915727c4d388e9335982216a2d Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com> Reviewed-on: https://review.mlplatform.org/c/1774 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-20COMPMID-2492 Fix Segfault in NEDeconvolutionLayer for NHWCSiCong Li
Manually manage the tensor _permuted_weights instead of letting memory group manage it. This is because the operation _permute_weights occurs inside the prepare stage. Signed-off-by: SiCong Li <sicong.li@arm.com> Change-Id: Idb639dffcc9942b4acf1bb8dab05f516bdec60d1 Reviewed-on: https://review.mlplatform.org/c/1769 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
2019-08-20COMPMID-2604: (Nightly) Failing 'GC/ConvolutionLayer/ConfigurationGian Marco Iodice
Change-Id: I336d736f27c7113d09b6b94a92d929613a6b06e4 Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/1767 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-20COMPMID-2597: Checking bias type in NEDeconvolutionLayer::validate even when ↵Manuel Bottini
bias == nullptr Change-Id: If7fb2e6d11e4653208b58ebdbfcc284d6c5a2e9a Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/1761 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-20COMPMID-2590: Avg_Pooling2d Fails on CL NHWC FP16Michele Di Giorgio
For large tensors with large pooling sizes, accumulation on FP16 gets saturated. This patch provides accumulation on F32 to overcome this issue. Change-Id: I5b2985c801970c8f2cb0053442296cc69f854186 Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/1762 Reviewed-by: Giorgio Arena <giorgio.arena@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>
2019-08-19COMPMID-2568: NEON Convolution layer failuregiuros01
Fixing the output tile size used when one of height or width is less than 4. Also added a test case that stresses this out. Change-Id: I99bb689f26aef713f8206c7d702f9fcf1017af58 Signed-off-by: giuros01 <giuseppe.rossini@arm.com> Reviewed-on: https://review.mlplatform.org/c/1744 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-19COMPMID-2336: Fix call to validate in NEDeconvolutionLayerMichele Di Giorgio
Change-Id: I15c16490f77b4346d3559f8edd0ca7bfd82d1961 Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/1749 Reviewed-by: Pablo Marquez <pablo.tello@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-19COMPMID-2595: (Nightly) Mismatches in NEArgMinMax for FP16Michele Di Giorgio
Results for FP16 were correctly computed but not stored to the output. Change-Id: I8a85e325ee42600e0450307562c344b48eb1c679 Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/1757 Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-14COMPMID-2582: Disable fuse batch normalization on OpenCL for FP16Gian Marco Iodice
Change-Id: I53bf35ff9f6d8c7e3664983ebcc94552a4ebf90d Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/1732 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Pablo Marquez <pablo.tello@arm.com>
2019-08-13COMPMID-2569: Fix access window in NEGEMMTranspose1xWKernel and ↵Michele Di Giorgio
NEGEMMLowpMatrixMultiplyKernel A lot of padding was unnecessarily added by NEGEMMLowpMatrixMultiplyKernel because the wrong number of elements accessed in the X direction was set to the access window of the second input. Change-Id: I1dac0a55dbf2183540d4fed31f097ef6d6243a5b Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/1728 Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-13COMPMID-2580: Fix out of bound read in Depthwise Convolution layer (OpenCL)Gian Marco Iodice
Change-Id: I00e39ed21cc30034aa10ac58b64d533e833eafc8 Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/1727 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-09COMPMID-2547: CLSpaceToBatchLayer causes NN Test Failures on QUANT8_ASYMM ↵Michele Di Giorgio
Data Type Change-Id: I47c9d057e50fa624f9b9e3fd79724e4fa7d0fd82 Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-on: https://review.mlplatform.org/c/1713 Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-08COMPMID-2336 Extend validation for depthwise native and fix same pad calculatorGiorgio Arena
Change-Id: I9f5cc95bc0cbd94869ac13064ffa0aa0f52a7a61 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Reviewed-on: https://review.mlplatform.org/c/1684 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
2019-08-06COMPMID-2526: Disabled Winograd when padding > half_kernel_sizePablo Tello
Change-Id: Id7b2d5dd4a0af2f914ee2bb35f27c07a43e2baca Signed-off-by: Pablo Tello <pablo.tello@arm.com> Reviewed-on: https://review.mlplatform.org/c/1688 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
2019-08-06COMPMID-1849: Implement CPPDetectionPostProcessLayerIsabella Gottardi
* Add DetectionPostProcessLayer * Add DetectionPostProcessLayer at the graph Change-Id: I7e56f6cffc26f112d26dfe74853085bb8ec7d849 Signed-off-by: Isabella Gottardi <isabella.gottardi@arm.com> Reviewed-on: https://review.mlplatform.org/c/1639 Reviewed-by: Giuseppe Rossini <giuseppe.rossini@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-06COMPMID-2336 Fix valgrind error for BatchNormalizationLayer on NEON with ↵Giorgio Arena
NHWC layout Change-Id: I9ed2d0647ae3c33bce6290acfdac356ffffcb709 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Reviewed-on: https://review.mlplatform.org/c/1697 Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-05COMPMID-2545: Reduce tests required by GEMM (OpenCL)Gian Marco Iodice
Removed FP16 tests from the new GEMM functions (GEMMNative, GEMMReshaped and GEMMReshapedOnlyRHS) since not called by CLGEMM Change-Id: Id52281fc9557d45e29db0a74964d4bdec55d8f46 Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/1695 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-02COMPMID-2539: Add bias addition check in CLGEMM validationGian Marco Iodice
Change-Id: Ib33574662d2b62ce80dd7f74a656199ed64225bc Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/1676 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-01COMPMID-2527: Remove activation function validation in ↵Gian Marco Iodice
CLWinogradConvolutionLayer Change-Id: I38c70b9e10e43d319a81ce9dbade771153e2446f Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/1672 Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-01COMPMID-2336: Fix build issues.Georgios Pinitas
Change-Id: I0932dc9ca4649f0825950ed9d6d249212bc6971e Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-on: https://review.mlplatform.org/c/1671 Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-08-01COMPMID-2055: Fusion of ConvolutionLayer with BatchNormalization at graph ↵Manuel Bottini
only for CL Change-Id: Ibdd0c56c1f59205dcaf15ca504b6e9c7fb30f34a Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/1618 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Comments-Addressed: Gian Marco Iodice <gianmarco.iodice@arm.com>
2019-08-01COMPMID-1979: Fuse Activation Function in CLGEMM - part 4Gian Marco Iodice
Fused activation function in CLGEMM Change-Id: I644fdf09349325c0b3a2cd5fef2a3ea2c974149d Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/1640 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-07-31COMPMID-2493: Update qs8 in Depthwise assemblyGeorgios Pinitas
Introduces minor optimisation for qasymm8 for depthwise convolution. Change-Id: I1b88b1475f8f1ef34c3a7c5580cdeef8b032a100 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-on: https://review.mlplatform.org/c/1647 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2019-07-31COMPMID-2500: Report error in Int8 Conv2d if multiplier > 1.Georgios Pinitas
Change-Id: I7d0263eddfb4f9cf0145e94b35d5f9e18737cd2d Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-on: https://review.mlplatform.org/c/1653 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Pablo Marquez <pablo.tello@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2019-07-31COMPMID-2336: Fix GEMMLowp 32bit checkGeorgios Pinitas
Fix invalid check for performing fused output stage in the intrinsic fallback path of NEGEMMLowpCore Change-Id: I9fa5a2d32376500fcb3d74e31dc5753b677c826a Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-on: https://review.mlplatform.org/c/1652 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2019-07-30COMPMID-2467 Include IP information to the tuning configGiorgio Arena
Change-Id: I00e29f30eda0edc76ae393bef14f0092e57a5195 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Reviewed-on: https://review.mlplatform.org/c/1634 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-07-29COMPMID-2430: Adding FusedDepthwiseConvolutionBatchNormalizationLayer ↵Manuel Bottini
condition in NEON function factory Change-Id: Id7ce542374b1625edbf586ef9688c94ea42f1bb1 Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/1645 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-07-29COMPMID-2336: Guard quint8 gemm files for aarch64 only.Georgios Pinitas
gemm_quint8 is only supported for 64-bit thus guarding to avoid any build related issues. Change-Id: Id8784dbacc467780318bd340f895a5abbd383182 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-on: https://review.mlplatform.org/c/1638 Reviewed-by: Pablo Marquez <pablo.tello@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-07-29COMPMID-1849: Add DetectorPostProcess operatorIsabella Gottardi
Part1 - Rework of CPPNonMaximumSuppression Change-Id: I2b34fbd12188db49b0ac050a12312494eeefd819 Signed-off-by: Isabella Gottardi <isabella.gottardi@arm.com> Reviewed-on: https://review.mlplatform.org/c/1585 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
2019-07-29COMPMID-2336: Rename the new generic depthwise convolution on NEONGian Marco Iodice
Change-Id: I45cacf75b08bb9d867343037507e56f200ad6ac0 Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Reviewed-on: https://review.mlplatform.org/c/1637 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Giorgio Arena <giorgio.arena@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
2019-07-26COMPMID-2179 New generic depthwise convolution for NEON F32 NHWCGiorgio Arena
Change-Id: I2b883785c0500d4bdb6ee4700382ee058be2cd36 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Reviewed-on: https://review.mlplatform.org/c/1538 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
2019-07-26COMPMID-2490: (Nightly) Valgrind error in NEON ScaleMichalis Spyrou
Border was not added in the case of UNDEFINED border mode. This resulted in reading memory out of bounds. Change-Id: I0f2ceceb0c3269ff560185d192a79bfb422c64ff Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com> Reviewed-on: https://review.mlplatform.org/c/1626 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
2019-07-26COMPMID-2178: Update GEMM assembly code.Georgios Pinitas
Perform offset reduction and requantization within the assembly wrapper. Change-Id: I5d5b3e1f6f9ef4c71805362c57f88ff199c027a3 Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-on: https://review.mlplatform.org/c/1541 Comments-Addressed: Pablo Marquez <pablo.tello@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-07-26COMPMID-2336: Set the correct valid region to the concatenation layer outputIsabella Gottardi
Change-Id: I1ff84054758a9a189f33d3180b5d81b3615963c7 Signed-off-by: Isabella Gottardi <isabella.gottardi@arm.com> Reviewed-on: https://review.mlplatform.org/c/1620 Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
2019-07-23Use .inst directives instead of .int directives.Ramana Radhakrishnan
Has a couple of benefits, one is a disassembler that actually understands dot product will start showing the dot product instruction for what it is rather than just a random .word. For the interested parties in actually why compilers and toolchains manage to disassemble this , please go and look up mapping symbols from toolchains. Secondly .word is a data directive and if you ever have a customer run Arm compute library on big endian, on AArch64 this will not work. This is because data on big endian is well, big endian but the code section is not big endian but just little endian. Admittedly there will be many other things that need to be fixed for big endian to work reliably. Eyeballed satisfactorily with a simple case. If someone could run this through a test run with the CI that would be great. Thanks, Ramana Change-Id: I0b9573ecbed298afc967d675b0542a6fe72b4c52 Signed-off-by: Ramana Radhakrishnan <ramana.radhakrishnan@arm.com> Reviewed-on: https://review.mlplatform.org/c/1588 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
generated by cgit v1.2.1 (git 2.23.0) at 2024-05-03 01:46:07 +0000