Summary: | media-libs/embree - unmask CPU_FLAGS_ARM="neon" for arm64 | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Leonid Kopylov <leonchik1976> |
Component: | Current packages | Assignee: | Sebastian Parborg <darkdefende> |
Status: | UNCONFIRMED --- | ||
Severity: | normal | CC: | arm64, galiven, proxy-maint |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | ARM64 | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- |
Description
Leonid Kopylov
2023-05-10 06:18:13 UTC
i have to add: media-libs/embree -cpu_flags_arm_neon to /etc/portage/profile/package.use.mask for package to build To the arm64 maintainers: Is there any specific reason why the neon flag is masked in the arm64 use.mask? The file only has this comment which is three years old at this point: # Aaron Bauman <bman@gentoo.org> (2019-12-27) # Mask cpu_flags_arm_neon for all of arm64 # A neon64 USE is being discussed cpu_flags_arm_neon I don't have an ARM CPU, so I can't really test any of this. But to me I don't see any reason for this besides maybe that support for neon was flaky for the arm64 at some point (or perhaps still is). But as I didn't find any information about this, I just have to assume that it might be time to remove the flag mask instead of having a white list of packages. (In reply to Sebastian Parborg from comment #2) > To the arm64 maintainers: > > Is there any specific reason why the neon flag is masked in the arm64 > use.mask? > > The file only has this comment which is three years old at this point: > # Aaron Bauman <bman@gentoo.org> (2019-12-27) > # Mask cpu_flags_arm_neon for all of arm64 > # A neon64 USE is being discussed > cpu_flags_arm_neon > > I don't have an ARM CPU, so I can't really test any of this. But to me I > don't see any reason for this besides maybe that support for neon was flaky > for the arm64 at some point (or perhaps still is). But as I didn't find any > information about this, I just have to assume that it might be time to > remove the flag mask instead of having a white list of packages. i have gentoo on Apple M1 Max, VM on Parallels Desktop, so i can test anything for you if required Hmm, I would have liked to get some feedback from the arm64 maintainers first. However I guess you could try to compile every package with the `neon` arm cpu flag and see if it works for you. If it does, I don't see any reason to keep the flag masked. Or if only a few fails, we can do a blacklist for the flag instead of a white list. same for media-libs/openpgl Just noticed the global arm64 neon mask on a different package after checking that the new version of cpuid2cpuflags still happily reports that neon should be in CPU_FLAGS_ARM. We can test things on armv8 cortex-a72.cortex-a53 big.LITTLE RK3399 based RockPro64. |