Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / google/XNNPACK issues and pull requests
#5344 - GEMM testers allow random to also test when first and last element match
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5343 - QD8-F32-QC8W tester type fixes
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5342 - I8MM QS8 & QS8-QC8W IGEMM microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5339 - Eliminate redundant pthreadpool_get_threads_count calls
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5335 - QD8_F32_QC4W WASM scalar GEMM microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5289 - QS8 I8MM microkernel use ld2r to load initial bias
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5286 - F32-GEMM avx broadcast use python to remove + 0 from offsets
Pull Request -
State: open - Opened by copybara-service[bot] over 1 year ago
#5285 - Avoid using implicitly specified higher-level ISA in CMake builds
Pull Request -
State: open - Opened by copybara-service[bot] over 1 year ago
#5284 - Run generate-enum script to regenerate src/enums/operator-type.c
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5283 - Avoid using implicitly specified higher-level ISA in Bazel builds
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5282 - F32-QC8W WASMSIMD use wasm_i16x8_load8x8
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5281 - Remove F32-QC4W and F32-QC8W VEX128 GEMM Microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5280 - F32-QC4W SSE DUP microkernels use float magic conversion
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5279 - Create `reshape_dynamic_igemm` to run igemm without persistent indirection buffer
Pull Request -
State: open - Opened by copybara-service[bot] over 1 year ago
#5278 - Refactor `xnn_indirection_init_conv2d` to take lower-level arguments instead of xnn_operator
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5277 - Define TRANSIENT_INDIRECTION_BUFFER flag
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5276 - fix: error message typo
Pull Request -
State: closed - Opened by 0o001 over 1 year ago
#5275 - Bump PThreadPool version
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5274 - Scaled Dot Product Attention takes thread index as argument
Pull Request -
State: open - Opened by copybara-service[bot] over 1 year ago
#5273 - Subgraphs support for FP16 Scaled Dot Product Attention
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5272 - Rename to Scaled Dot Product Attention operator
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5271 - Consolidate constants for specifying max inputs/outputs in subgraph
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5270 - F32-QC4W params add magic for scalar and SSE4 and neon for QD8
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5269 - Enable 4x16 QD8-F32-QC8W GEMM for I8MM
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5268 - Add IGEMM splat WASM JIT kernel
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5267 - Fix NR=2 config check in F32 NHWC Convolution
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5266 - F32-QC4W GEMM Neon and Scalar use 32 bit zero point
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5265 - Add GEMM splat WASM JIT kernel
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5264 - Support different number of channels for values
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5263 - Fix typo in convolution-nhwc.c
Pull Request -
State: closed - Opened by bhbruce over 1 year ago
- 2 comments
#5262 - Scaled Dot Product Attention subgraph support
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5261 - Restrict cpuinfo dependency in config and util targets
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5260 - Restrict cpuinfo dependency in eval targets
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5259 - Refactor enabling/disabling build options in Bazel
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5258 - Remove xnn_enable_q[s/u]8_explicit_[true/false] Bazel options
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5257 - Remove deprecated inference flags
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5256 - Minor refactoring in F32 VLRELU and F32 VSQRT benchmarks
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5255 - Benchmarks for F32 VRND microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5254 - Add a unit-test for igemm loadsplat WASM JIT x8 kernel
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5253 - Add the missing ARM JIT kernel to the `generate-f32-gemm.sh` bash script
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5252 - Bump PThreadPool version
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5251 - Limit cpuinfo dependency in Bazel to AArch32, AArch64, and x86
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5250 - Refactor rounding benchmark
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5249 - Move calculation of batch size in binary ops from create to reshape
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5248 - Fix typo I8MM is C8 not C4 in generator
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5247 - Enable 8x16C8 I8MM GEMM microkernel
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5246 - I8MM GEMM replace zip with LD2 lane to transpose
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5245 - Enable X8-PACKW batch size of 2
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5244 - X8-PACKW batch size of 2
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5243 - Reformat Wasm assembler to 120 line length
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5242 - Add missing JIT microkernel generator to script
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5241 - Bump PThreadPool version
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5240 - Move calculation of batch size from create to reshape
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5239 - Add Multi Query Attention support
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5238 - Split out number of heads from batch_size in Scaled Dot Attention
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5237 - Enable QD8-F32-QC8W NEONI8MM GEMM microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5236 - Polyfill performance.now in Wasm builds
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5235 - NEONDOT C8 QC8/QD8/QS8 GEMM microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5234 - Change workspace to be zero-allocated
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5233 - FP16 Scaled Dot Product Attention operator
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5232 - Enable F32-QC8W AVX and FMA3 GEMM microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5231 - N-remainder specialization for S4 GEMM WASM JIT kernel
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5230 - Refactor scaled dot attention to be type agnostic
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5229 - Change rmax config ukernel to be type agnostic
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5228 - Use packing function specified in GEMM config instead of hard coding it
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5227 - Fix bad naming of fields and incorrect copying of params in Dot Attention
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5226 - Add scale argument to packing functions which support quantized datatypes
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5225 - Add benchmarks to compare scaled dot attention operator and batch matrix multiply
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5224 - F32-QC8W FMA3 and AVX broadcast GEMM microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5223 - Enable F32-QC4W AVX and FMA3 GEMM microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5222 - F32-QC4W AVX FMA3 microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5221 - Enable QC4W AVX512SKX 7x16 microkernel
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5220 - E2E benchmark for S4 WASM JIT kernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5219 - Set sr=4 for the WASM JIT S4 gemm/igem benchmarks
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5218 - Add missing #includes for <iomanip> and <ios>
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5217 - Fix function name shadowing typename
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5216 - Remove std::is_integral_v as it was added in c++17
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5215 - Temporarily disable browser test for f32_igemm_jit_test
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5214 - Fix open source build
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5213 - Fix incorrect version of pthreadpool in WORKSPACE
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5212 - Fix weird formatting
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5211 - Update pthreadpool version
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5210 - Add missing headers to compute.h
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5209 - Enable QC4W AVX2 3x16 microkernel with magic float conversion
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5208 - F32-QC4W change AVX params to full vectors
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5207 - AVX microkernels cast pointers to m128i instead of void
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5206 - F32-QC4W AVX GEMM microkernels using float magic conversion
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5205 - GEMM/IGEMM S4 WASM JIT kernel loop unroll
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5204 - Move post-operation to be part of JIT, it is only used by JIT library
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5203 - F32-QC4W AVX512 GEMM microkernels using _mm256_cvtepi8_epi32
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5202 - Use _mm256_cvtepu8_epi32 in F32-QC4W AVX2 GEMM
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5201 - xnn_init_f32_qc4w_minmax_avx_params use scalar values
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5200 - F32 GEMM AVX512 with NR=16 avoid masking NC for final remainder
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5199 - Reorder kernel_zero_point argument in xnn_create_fully_connected_nc_f32_qc4w
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5198 - QD8/QS8 1x16C8 and 1x8C8 NEONI8MM GEMM microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5197 - F32-GEMM avx512 use mask with nc & 15 to allow larger NC
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5196 - Generate xnn_node_type enum
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5195 - Fix I8MM compilation options
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5194 - Enable AVX2 QD8-F32-QC8W GEMM microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago
#5193 - Add xnn_f32_rminmax_ukernel__wasm_x4_acc4 to prod microkernels
Pull Request -
State: closed - Opened by copybara-service[bot] over 1 year ago