google/XNNPACK

View on GitHub

High-efficiency floating-point neural network inference operators for mobile, server, and Web

Stars: 2.5k
Forks: 560
Open beginner issues: 0
Indexed issues: 73
Avg merge: 1d 6h
Merged PRs (30d): 163

Dominant language: C
License: No license data
Last GitHub push: Sep 15, 2026
Latest indexed: Sep 16, 2026
Contributing guide: Contributing guide
Code of conduct: Code of conduct
Beginner labels: No beginner labels indexed

Open All

73 open issues indexed

2D resize nearest neighbor interpolation instead of bi-linear interpolation? Open

google/XNNPACK#1032 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Build crashes on ios_armv7 with bazel Open

google/XNNPACK#1124 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
CMake build error when using XNNPACK_USE_SYSTEM_LIBS=ON Open

google/XNNPACK#1543 · 8 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Openvino Benchmarks Open

google/XNNPACK#1803 · 0 comments · 1 reaction · 0 assignees · Indexed Sep 16, 2026
Performance drops on deeper convolution layers? (Pixel 4) Open

google/XNNPACK#1905 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Support int8 transposed convolutions with per-channel weight quantization Open

google/XNNPACK#2762 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Support int8 / qs8 `AVERAGE_POOL_2D` Open

google/XNNPACK#3040 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Add CMake support for Emscripten builds Open

google/XNNPACK#3455 · 10 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
armeabi-v7a assembler error Open

google/XNNPACK#4348 · 11 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
error: array type has incomplete element type xnn_timestamp (struct timespec branch) Open

google/XNNPACK#4581 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
CMAKE Parameters for armv6+vfp build Open

google/XNNPACK#4636 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Doesn't it support windows x86 (32 bit)? Open

google/XNNPACK#4637 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
RISC-V cpuinfo build error Open

google/XNNPACK#4650 · 7 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Unable to replicate RPi0 tests Open

google/XNNPACK#4701 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
clang14.0.7 from Android NDK crashes when compiling bf16-gemm-1x4c8-minmax-neonbf16-bfdot.c Open

google/XNNPACK#4775 · 14 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
pthreadpool cfi-icall check failure for XNNPACK function pointers casting Open

google/XNNPACK#4971 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
will you Plan to support int8 perchannel quantize for linear op? Open

google/XNNPACK#5312 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Build error when including XNNPACK using FetchContent Open

google/XNNPACK#5669 · 4 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
experiments-config.h is hiding the xnnpack.h header Open

google/XNNPACK#5713 · 1 comment · 1 reaction · 0 assignees · Indexed Sep 16, 2026
When running `build -c opt --config android_arm64 :end2end_bench` under bench, `sys/system_properties.h` is missing Open

google/XNNPACK#6146 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Xnnpack still builds with `+dotprod` and `+fp16` with `-DXNNPACK_ENABLE_ARM_DOTPROD=OFF -DXNNPACK_ENABLE_ARM_FP16_SCALAR=OFF -DXNNPACK_ENABLE_ARM_FP16_VECTOR=OFF` Open

google/XNNPACK#6165 · 10 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
ARMv7 (with NEON) can not support on Linux but only support ARMv7 (with NEON) on Android Open

google/XNNPACK#6210 · 5 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
`xnn_weights_cache_provider` look_up doesn't work? Open

google/XNNPACK#6257 · 4 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
How can I parallelize the execution of this benchmark? (https://github.com/google/XNNPACK/blob/master/bench/spmm-benchmark.h) Open

google/XNNPACK#6268 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
cmake build failure with XNNPACK_BUILD_TESTS=ON and XNNPACK_LIBRARY_TYPE=shared Open

google/XNNPACK#6285 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
test/sigmoid_nc_test fails on Hexagon simulator Open

google/XNNPACK#6305 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Load-from-misaligned-address failures on Hexagon simulator Open

google/XNNPACK#6306 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
XNNPACK tests that use mmap() fail on Hexagon devices Open

google/XNNPACK#6343 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
unsupported instruction `vpdpbusd' Open

google/XNNPACK#6389 · 5 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Enable QC8/QS8 GEMM/IGEMM for Wasm relaxed integer dot product instruction on x64 Open

google/XNNPACK#6454 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
4x16s4 fp32-gemm kernel have better performance than default(5x16) kernel for meteor lake Open

google/XNNPACK#6480 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Failed to compile XNNPACK on WoA(Windows on ARM) device. Open

google/XNNPACK#6558 · 6 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
f32-spmm missing tests Open

google/XNNPACK#7062 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
how to build with enabled sme arm kleidai Open

google/XNNPACK#7324 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Why does the benchmark program run with only one thread? Open

google/XNNPACK#7554 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
The purpose of different implementations of qs8/qu8-f32-vcvt kernels. Open

google/XNNPACK#7644 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Possible heap corruption in qs8-dwconv-bench with primary_tile=25 Open

google/XNNPACK#7657 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
The top level CMakeLists.txt should call enable_language(ASM_MARMASM) on Windows Open

google/XNNPACK#7704 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
How to build XNNPACK by Bazel 6.5 Open

google/XNNPACK#7720 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
In function ‘xnn_qs8_qc8w_gemm_minmax_fp32_ukernel_1x2c4__armsimd32’: error: unknown type name ‘int8x4_t’ Open

google/XNNPACK#7729 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
In function ‘xnn_u8_vclamp_ukernel__neon_u64’: passing argument 1 of ‘vld1q_dup_u8’ from incompatible pointer type Open

google/XNNPACK#7733 · 4 comments · 3 reactions · 0 assignees · Indexed Sep 16, 2026
How to build XNNPACK by Bazel 6.5 in windows, but without avx Open

google/XNNPACK#7781 · 5 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
When linking my own project to XNNPACK's library, I get an error: undefined reference to 'xnn_f32_vadd_ukernel__neon_u8' Open

google/XNNPACK#7827 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
tags please Open

google/XNNPACK#7930 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Running XNNPACK on constrained environment Open

google/XNNPACK#7966 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
xN-transpose-test fails when cross-compiling XNNPACK for riscv64 on x86 platform Open

google/XNNPACK#8052 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Bazel can't cross-compile the XNNPACK tests Open

google/XNNPACK#8211 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
[Request] Can't add new parameters to the kernel generated by 6x16-aarch64-neonfp16arith-cortex-a75.S.in Open

google/XNNPACK#8489 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Github runners are available for Windows and Linux on ARM64 Open

google/XNNPACK#8537 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Failed to compile XNNPACK with Emscripten Open

google/XNNPACK#8586 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026

google/XNNPACK

Get new issues in your inbox