add vcvt, vcvta, vcvtn, vcvtm, vcvtp neon instructions #1084
Conversation
|
r? @Amanieu (rust-highfive has picked a reviewer for you, use r? to override) |
crates/stdarch-gen/neon.spec
Outdated
| name = vcvt | ||
| double-suffixes | ||
| fn = simd_cast | ||
| a = -1.0, 2.0, -3.0, 4.0 |
There was a problem hiding this comment.
Wouldn't it make more sense to use .5 and co. to actually test the rounding (for all of these instructions)?
There was a problem hiding this comment.
You are right, but I don’t have a good way to test the results. How can we simulate rounding towards to even?
There was a problem hiding this comment.
e.g.:
a = -1.5, 2.5, -3.5, 4.5
validate -2, 2, -4, 4
I'm confused what the problem is?
There was a problem hiding this comment.
Sorry I forgot that this is rounding to an integer. I will correct accordingly. Thanks!
| #[inline] | ||
| #[target_feature(enable = "neon")] | ||
| #[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))] | ||
| #[cfg_attr(all(test, target_arch = "arm"), assert_instr(vcvt))] | ||
| pub unsafe fn vcvt_s32_f32(a: float32x2_t) -> int32x2_t { |
There was a problem hiding this comment.
It doesn't seem like there's any assert_instr for aarch64 for these?!
There was a problem hiding this comment.
There are different implementations for arm and aarch64. You can use https://godbolt.org/ to verify
There was a problem hiding this comment.
Well it's fcvtzs and co., so idk why those aren't being asserted for here. https://rust.godbolt.org/z/Evce1T
There was a problem hiding this comment.
I think you are right, thanks a lot
Add vcvt, vcvta, vcvtn, vcvtm, vcvtp neon instructions(Floating-point convert to fixed-point with different rounding methods).