fbarchard

The problem was older iOS that supported 32 bit didnt have dot product. Newer iOS dropped 32 bit but added dot product for 64 bit. But the older tool chain...

The issue was reported as XNNPack not building with the -Wimplicit-fallthrough flag https://github.com/google/XNNPACK/issues/6047 But when I apply the flag in XNNPack build the pytorch/cpuinfo has 2 build warnings. The first...

libyuv issue #970 is resolved? If you're hitting a new bug in libyuv, please report with more details

As there are no code reviews / bugs for libyuv, I think you're asking me to review your fuzz code? I'm not really qualified to review it, but it seems...

For the x8-packw that calls x32-packw I made a hack PR https://github.com/google/XNNPACK/pull/6356 where you can see the idea. But I think instead of calling a common function, it will need...

Thanks for catching the alignment issue ``` Was vst1_lane_u32((uint32_t*) or, vreinterpret_u32_f16(vaccr_lo), 0); or += 2; 9d740: 04 00 10 e3 tst r0, #4 9d744: 03 00 00 0a beq 0x9d758...

end2end builds okay for me, but I'm on an old pytorch/cpuinfo Perhaps test if pytorch/cpuinfo is simply broken on android_arm64?

I can confirm the issue for both Qualcomm and Exynos S22 Affects both AArch32 and AArch64 builds. On Samsung S20 Exynos it reports correctly SoC name: Samsung Exynos 9830 Microarchitectures:...

For me the Samsung S22 detects Cortex X2 in AArch64 mode but not AArch32. Can you confirm you did AArch64 builds? And in both the SoC is unknown. Many new...

The review is approved but requires someone with write access to merge/land the PR