- a7077e9 Updateable weights in depthwise convolution by Milos Puzovic · 1 year, 10 months ago
- 199982f Add threshold for floating-point SOFT_RELU activation by Milos Puzovic · 1 year, 10 months ago
- 4b5f6ef Add check for Batch Matmul in GemmAssemblyDispatch by Mohammed Suhail Munshi · 1 year, 10 months ago
- f44bbc5 Rewrite dynamic fusion by SiCong Li · 2 years ago
- 3394f3e Rework direct convolution heuristic on OpenCL by Gian Marco Iodice · 2 years ago
- 910e3f9 Fix fixed-point quantized addition by Viet-Hoa Do · 1 year, 11 months ago
- 2bc8cfe Add FP16 tanh based on rational approximation by Jonathan Deakin · 1 year, 11 months ago
- 9fc0b5c Update reinterpret tensor as 1D for CPU add by Viet-Hoa Do · 1 year, 10 months ago
- 6782452 Add test in GEMMLowp for batch matmul by Mohammed Suhail Munshi · 1 year, 11 months ago
- 0a36f58 Fix FFTConvolutionLayer test by Viet-Hoa Do · 1 year, 10 months ago
- fa79fda Optimize Neon™ Logistic Activation by Mohammed Suhail Munshi · 2 years ago
- c8cc024 Adding documentation section explaining how BF16 is used by Ramy Elgammal · 1 year, 11 months ago
- 6670413 Fix LUT-based activation layer by Viet-Hoa Do · 1 year, 11 months ago
- 2ffab6d Workaround CL compiler issue on FP16 by Viet-Hoa Do · 1 year, 11 months ago
- 842ad21 Optimize Neon™ SUB operator by squashing execution window by Jakub Sujak · 2 years ago
- ad9a7ed Rework DepthwiseConvolution heuristic on OpenCL by Gian Marco Iodice · 2 years ago
- 3bedd2f Improve start-up time in gemmlowp reshaped rhs only. by Adnan AlSinan · 1 year, 11 months ago
- db14af6 Update GEMM reshaped rhs only heuristic by Gian Marco Iodice · 1 year, 11 months ago
- b5368fb Force CL kernel compilation with 64 registers by Viet-Hoa Do · 2 years ago
- 304dfdb Fix Batch Matmul nightly failure by Adnan AlSinan · 1 year, 11 months ago
- 40b4419 Optimize CPU add layer on quantized data by Viet-Hoa Do · 1 year, 11 months ago
- ff81de5 Fix overflow in NEActivationLayer for FP16 type by Pablo Marquez Tello · 1 year, 11 months ago
- d6b8a71 Add FP32 Neon™ swish activation by Jonathan Deakin · 2 years ago
- ce79ac6 CPU GEMM: Fix overreads in SVE merges. by David Mansell · 1 year, 11 months ago
- ead4d11 Fix unresolved symbol for target armv7a + Android by Pablo Marquez Tello · 2 years ago
- 047e5d0 Fix validation in validate_image2d_support_on_rhs by Gian Marco Iodice · 2 years ago
- 622b8ad Fix bug in QASYMM8_SIGNED to F32 cast layer by Viet-Hoa Do · 2 years ago
- c4f2743 Optimize Quantized/Integer Bilinear Scale for Neon™ by Gunes Bayir · 2 years ago
- 0d05b66 Interpreting tensor as 1D for CPU multiplication by Viet-Hoa Do · 2 years ago
- 5687e55 Fix invalid memory access for dynamically fused Cl Elementwise kernels by SiCong Li · 2 years ago
- 926f502 Adding GELU activation by Murray Kornelsen · 2 years, 1 month ago
- 6e09e14 INT8 Quantized MeanStdDevNorm (LayerNorm) by Murray Kornelsen · 2 years, 1 month ago
- 26c9d1a Add test for NEGEMM to test a batched matrix multiplication with variable input tensors by Adnan AlSinan · 2 years ago
- 4478e1c Rework heuristic in ClConv2d by Gian Marco Iodice · 2 years ago
- 0eed305 Optimize FP32/16 Bilinear Scale Kernel for Neon™ by Gunes Bayir · 2 years ago
- d11de98 Add a macro guard in all OpenCL kernels in gemmlowp.cl by Gian Marco Iodice · 2 years ago
- e4e3b2e Disable Winograd on fp16 if fast-math = false by Ramy Elgammal · 2 years ago
- 211a55d Optimize depthwise convolution on OpenCL by Gian Marco Iodice · 2 years ago
- 552fe4c F16 Specialization for MeanStdDevNorm by Murray Kornelsen · 2 years, 1 month ago
- 1257131 Enable Winograd-based conv2d when IFM>=8 on Gpu by Gian Marco Iodice · 2 years ago
- cd083cf Use parent buffer in CLSubTensor. This avoids calling enqueueMapBuffer repeatedly when mapping multiple children of the same parent. by Murray Kornelsen · 2 years, 1 month ago
- a331e48 Fix add for tensors with non-matching strides by Jonathan Deakin · 2 years ago
- 8bab588 Fix validation problem in CLQLSTMLayer by Pablo Marquez Tello · 2 years ago
- a81fc84 Fix macos build errors by Pablo Marquez Tello · 2 years ago
- 53929b1 Use Neon™ kernels for FP Bilinear Resize for SVE by Gunes Bayir · 2 years, 1 month ago
- 9ee8a3e Revert "Fix performance regression in ClConv2D" by Ramy Elgammal · 2 years ago
- 29db3d2 Add LUT for quantized sigmoid function by Viet-Hoa Do · 2 years, 1 month ago
- e54d8c0 Fix performance regression in ClConv2D by Gian Marco Iodice · 2 years ago
- cec1af5 Fix performance regression in Conv2D on OpenCL by Adnan AlSinan · 2 years, 1 month ago
- cc03419 Disable unsafe FP optimizations in Winograd Output Transform by Gunes Bayir · 2 years, 1 month ago
- 52695f5 Fix CTS/SLTS failure related to Depthwise Convolution by Gunes Bayir · 2 years, 1 month ago
- 65c8db8 Fix for AI benchmark ResNet regression by Viet-Hoa Do · 2 years, 1 month ago
- 0105ce5 Fix LeNet-f16 convolution regression by Adnan AlSinan · 2 years, 1 month ago
- 93581a5 [ONCPUML-970] Fast math mode for fixed format kernels by Pablo Marquez Tello · 2 years, 1 month ago
- f67903b Add Dynamic Fusion Tests with BugFixes by Mohammed Suhail Munshi · 2 years, 2 months ago
- 13b623e [ONCPUML-968] Fixed format kernel support in additional APIs by Milos Puzovic · 2 years, 1 month ago
- 3c4d085 Update the GPUTarget list by Gian Marco Iodice · 2 years, 2 months ago
- 9b921be Optimize add layer by considering the input tensors as 1D array by Gunes Bayir · 2 years, 1 month ago
- 385dad2 Fix for OpenMP scheduler work breakdown by Milos Puzovic · 2 years, 1 month ago
- aa52b7d Fix compilation error rasied in Nightly_NEW by Ramy Elgammal · 2 years, 1 month ago
- 9178002 Fix for inclusion of "arm_gemm" from src into "Types.h" from core by Ramy Elgammal · 2 years, 1 month ago
- d208f4f Enable march=armv8.6-a in non multi-isa builds by Pablo Marquez Tello · 2 years, 1 month ago
- e572dff Add GemmLowp MMUL Reshaped Only Rhs Support for QASYMM8/QASYMM8_SIGNED by Freddie Liardet · 2 years, 3 months ago
- e871207 Update ClConv2D heuristic to use direct convolution by Adnan AlSinan · 2 years, 1 month ago
- c6b7913 Fix direct convolution cases that were failing on Odroid by Adnan AlSinan · 2 years, 1 month ago
- 553f695 [ONCPUML-951] Variable weight support for Convolution. by Francesco Petrogalli · 2 years, 2 months ago
- 1f86449 Fix Neoverse V1 heuristics for FP32 fast mode by ramelg01 · 2 years, 2 months ago
- a1f7851 Integrate new winograd APIs from MLTech by ramelg01 · 2 years, 2 months ago
- 4bfc70e Add Gemm MMUL Reshaped Only Rhs Support for FP32/FP16 by Gunes Bayir · 2 years, 9 months ago
- 29cab36 Fixed clang-cl errors on Windows native builds. by Pablo Tello · 2 years, 6 months ago
- 2cc50b3 Extended direct conv 2d interface for tuning the OpenCl kernel by Gian Marco Iodice · 2 years, 3 months ago
- 22dd8b9 Add missing flag when building cl graph examples and fix by Michalis Spyrou · 2 years, 2 months ago
- a3cf241 Add G57 to GPUTarget by SiCong Li · 2 years, 2 months ago
- e417ff1 Fix build errors on armv8.6 SVE2 with NDK 23 and 24 by Michalis Spyrou · 2 years, 2 months ago
- 84a0941 Fix OpenBSD build errors by Pablo Marquez Tello · 2 years, 2 months ago
- 16aa474 Wrong arguments for running activation function in CpuGemmDirectConv2d by Michalis Spyrou · 2 years, 2 months ago
- b042e39 Add LUT-based leaky relu for QASYMM8 on CPU by Viet-Hoa Do · 2 years, 2 months ago
- d8804fd Fix OpenCL Winograd output transform by Gian Marco Iodice · 2 years, 2 months ago
- b1fcefd Implement new Elementwise Dynamic Fusion Operators: Div, Floor by Michalis Spyrou · 2 years, 2 months ago
- 41eb2d9 Improve LUT Neon Hard-Swish by Pablo Marquez Tello · 2 years, 2 months ago
- 700b913 Select neon LUT Hard-Swish kernel on all devices by Pablo Marquez Tello · 2 years, 2 months ago
- 3efdfb3 Enable loading OpenCL symbols in Android App native code, NDK environment. by ohadagoogle · 2 years, 2 months ago
- f1f7779 Fix SVE2 implementation of quantized SoftMax 1D by Viet-Hoa Do · 2 years, 2 months ago
- 586fff9 Fix build error v8.2-a-sve by Pablo Marquez Tello · 2 years, 3 months ago
- f8bb092 Add support OpenCL 3.0 non-uniform workgroup by Viet-Hoa Do · 2 years, 3 months ago
- 0c68704 Fix performance regression in Winograd Output Transform (OpenCL) by Gian Marco Iodice · 2 years, 2 months ago
- 894659a Add support for 2d and 3d indices for axis 1 by Pablo Marquez Tello · 2 years, 4 months ago
- c3bc093 Fix crash in CpuActivationKernel by Pablo Marquez Tello · 2 years, 3 months ago
- d75cd8a Compute Hard-Swish with a Lookup table for qasymm8. by Pablo Marquez Tello · 2 years, 3 months ago
- 82169b3 Add cl_khr_integer_dot_product extension support by Viet-Hoa Do · 2 years, 3 months ago
- f28abef Fix compilation error in 32-bit with logging build by Viet-Hoa Do · 2 years, 3 months ago
- fdb5342 Disable unsafe FP optimizations causing accuracy issues by Gunes Bayir · 2 years, 3 months ago
- 5fcf22d [arm_gemm] Import fixed-format kernels from gemm_linux. by Francesco.Petrogalli@arm.com · 2 years, 5 months ago
- 6fd75b9 DepthwiseConv reports full assembly kernel name by Pablo Marquez Tello · 2 years, 9 months ago
- dc4f276 Revert "Add support for 2d and 3d indices for axis 0" by Mohammed Suhail Munshi · 2 years, 4 months ago
- 4e9f568 Fix inclusion guard for dynamic fusion module by SiCong Li · 2 years, 4 months ago
- 920f2b6 Add support for 2d and 3d indices for axis 0 by Pablo Marquez Tello · 2 years, 4 months ago
- 06adbc5 Mismatches in dynamically fused direct conv2d + add kernel by Michalis Spyrou · 2 years, 4 months ago
- b63b119 Integrate Dynamic Fusion patches by SiCong Li · 2 years, 7 months ago
- 9d18738 Updating a64_gemm_u8 a64_gemm_s8 kernels headers by ramelg01 · 2 years, 4 months ago