You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@serge-sans-paille I'm afraid this issue is not proved obsolete by the new CI job. It's about building with the MSVC compiler targeting arm64, not about building with gcc/mingw. Could you please reopen the issue?
Some feedback: I cana ctually reproduce your issue in #1250, which is somehow good. I had great hope that /Zc:arm64-aliased-neon-types- would help, but that doesn't seem to be enough, unfortunately.
This issue actually got worse with #1254. Before, including "xsimd" was basicly a no-op. I was even able to use the emulated SIMD xsimd::emulated as fallback. Now it detects neon (which is correct), but so this error here breaks building the emulated fallback.
The project now builds successfully with MSVC on Windows on ARM, with all tests passing. Additionally, a CI script has been added to support builds using LLVM Clang and MSYS2 CLANGARM64. Refer below PR #1285
Activity
niyas-sait commentedon Oct 19, 2021
The main issue seems to be due to MSVC arm/neon intrinsic vector type using the same underlying type for the neon vectors.
e.g: int8x16_t and uint8x16_t maps to __n128 type
This causes a couple of issues
And other issues are
niyas-sait commentedon Oct 19, 2021
Few approaches I can think of
niyas-sait commentedon Oct 19, 2021
Pull Request for approach (A)
stwiegand commentedon Sep 25, 2024
Any news on this issue, since nowadays ARM64 on Windows is a thing?
JohanMabille commentedon Oct 13, 2025
We should probably revive #612 , but I cannot give a deadline for now. Any help on that would be appreciated.
serge-sans-paille commentedon Jan 24, 2026
proved obsolete by #1248.
jgiannuzzi commentedon Jan 25, 2026
@serge-sans-paille I'm afraid this issue is not proved obsolete by the new CI job. It's about building with the MSVC compiler targeting arm64, not about building with gcc/mingw. Could you please reopen the issue?
serge-sans-paille commentedon Jan 25, 2026
As noted in #1248 (comment), I mistyped the label, but it's indeed msvc being run.
And this build: https://github.com/xtensor-stack/xsimd/actions/runs/21322768338/job/61375021654 makes me think everything is good?
...
or so I thought, but no, It's targeting amd64, my bad. Reopening
serge-sans-paille commentedon Jan 25, 2026
@jgiannuzzi https://github.com/xtensor-stack/xsimd/actions/runs/21334459960/job/61404179357?pr=1250 from #1250 looks good, right?
jgiannuzzi commentedon Jan 25, 2026
@serge-sans-paille yes that looks good to me! Thank you!
serge-sans-paille commentedon Feb 4, 2026
Some feedback: I cana ctually reproduce your issue in #1250, which is somehow good. I had great hope that
/Zc:arm64-aliased-neon-types-would help, but that doesn't seem to be enough, unfortunately.degasus commentedon Mar 30, 2026
This issue actually got worse with #1254. Before, including "xsimd" was basicly a no-op. I was even able to use the emulated SIMD
xsimd::emulatedas fallback. Now it detects neon (which is correct), but so this error here breaks building the emulated fallback.DiamonDinoia commentedon Mar 30, 2026
can you try something like this in the meantime? https://github.com/flatironinstitute/finufft/blob/e137ec92f1cc1e91191f50072460a80307195304/include/finufft/simd.hpp#L8
then you can check if NEON is detected and force emulated.
pdeep854 commentedon Apr 1, 2026
The project now builds successfully with MSVC on Windows on ARM, with all tests passing. Additionally, a CI script has been added to support builds using LLVM Clang and MSYS2 CLANGARM64. Refer below PR
#1285