-
Notifications
You must be signed in to change notification settings - Fork 526
Pull requests: google/XNNPACK
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Add xnn_pack_qs8_to_qu8_qc4uw_gemm_gio_w
#10760
opened Jul 16, 2026 by
copybara-service
Bot
Loading…
Fix LUT fusion treating an already-fused LUT node as a plain unary op
#10756
opened Jul 15, 2026 by
sachinhambar
Loading…
Fixed static tensor behavior with ynnpack to match XNNPACK. Old behavior didn't allow for external tensors with static weights.
#10755
opened Jul 15, 2026 by
copybara-service
Bot
Loading…
validate runtime input channels in convolution reshape paths
#10748
opened Jul 14, 2026 by
aizu-m
Contributor
Loading…
Add the
golden_parity global configuration for architecture-neutral golden tests.
#10747
opened Jul 14, 2026 by
copybara-service
Bot
Loading…
Inject explicit architecture-neutral arithmetic flags and golden parity defines directly into the ARCore unit test BuildableUnit.
#10745
opened Jul 14, 2026 by
copybara-service
Bot
Loading…
Support enforcing architecture-neutral baseline data cache sizes for Eigen GEMM operations.
#10744
opened Jul 14, 2026 by
copybara-service
Bot
Loading…
Re-enable packw optimized kernels and refine ISA guards in GEMM configs.
#10743
opened Jul 14, 2026 by
copybara-service
Bot
Loading…
Fix support for explicitly disabling x86 SIMD and matrix multiplication architecture extensions.
#10742
opened Jul 14, 2026 by
copybara-service
Bot
Loading…
Implement a transformer layer for Gemma 4.
#10741
opened Jul 13, 2026 by
copybara-service
Bot
Loading…
Implement an XNNPack runner for Gemma 4.
#10738
opened Jul 13, 2026 by
copybara-service
Bot
Loading…
Implement Rotary Position Embedding (RoPE) operation for Gemma 4.
#10734
opened Jul 13, 2026 by
copybara-service
Bot
Loading…
Add AVX512 FP32 dot kernel specialized for k=16
#10713
opened Jul 12, 2026 by
copybara-service
Bot
Loading…
Fix integer overflow in NHWC convolution kernel size calculation
#10710
opened Jul 11, 2026 by
YellowJune
Loading…
litert/tensor/arithmetic.h: bounds-check axis and index values before vector subscript
#10702
opened Jul 10, 2026 by
destro4evr-rgb
Contributor
Loading…
Refactor to avoid using
xnn_define_static_broadcast
#10688
opened Jul 8, 2026 by
copybara-service
Bot
Loading…
KFI-543 Integrate SME2 kernels matmul_clamp_f16_qai8dxp_qsi8cxp
#10686
opened Jul 8, 2026 by
Colm-in-Arm
Contributor
Loading…
Skip splitting dimensions with extent 1 in runtime.cc.
#10676
opened Jul 7, 2026 by
copybara-service
Bot
Loading…
Previous Next
ProTip!
Follow long discussions with comments:>50.