[AArch64][GISel] Handle bitreverse intrinsic for some vector types

[madhur13490]

Parent issue: #115133
Godbolt link: https://godbolt.org/z/nP8jssKh8

This intrinsic is missing support for v2i64, v4i16 etc. as shown by the above godbolt link.

[madhur13490]

@aemerson Are you planning to handle this as you seem to have the recent patch around this? 49ecd66
If not, I will take a look.

[llvmbot]

@llvm/issue-subscribers-backend-aarch64

Author: Madhur Amilkanthwar (madhur13490)

Parent issue: https://github.com//issues/115133 Godbolt link: https://godbolt.org/z/nP8jssKh8

This intrinsic is missing support for v2i64, v4i16 etc. as shown by the above godbolt link.

[aemerson]

@aemerson Are you planning to handle this as you seem to have the recent patch around this? 49ecd66 If not, I will take a look.

I wasn't looking at this issue in particular, it was one of the more common fallbacks I saw when building CTMark with -O3.

I added some more tests to show the other types are now working, although there's still some room for improvement on some: 3bb0c73

Will close this issue as it seems to be done from a functionality standpoint.

[madhur13490]

2 x s32, 2 x i64, 4 x s16, 4 x s32, 8 x s16 are still falling back as I see on here

[aemerson]

2 x s32, 2 x i64, 4 x s16, 4 x s32, 8 x s16 are still falling back as I see on here

I think it takes some time (I don't know how much exactly) for godbolt to pick up the latest build of LLVM.

[madhur13490]

Yes, confirmed locally. Filed #120878 for optimization.
Thanks!