Restructure x86 signed pack instructions #1443
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
bjorn3
reviewed
Jan 2, 2024
This reduces the amount of duplicated code and the chance for bugs.
I validated the new code for correctness against LLVM using the
following script. It found many bugs in the implementation until I was
finally able to get it correct and passing.
```rust
//! Test for x86 pack instructions. Prints deterministic results, use it to compare backends.
use std::arch::x86_64::{self, __m128i, __m256i};
use rand::{rngs::SmallRng, Rng, SeedableRng};
fn main() {
let rng = &mut SmallRng::seed_from_u64(123);
for _ in 0..100_000 {
unsafe {
sse_test(rng);
avx_test(rng);
}
}
}
unsafe fn sse_test(rng: &mut SmallRng) {
print_sse_8(x86_64::_mm_packus_epi16(sse16(rng), sse16(rng)));
print_sse_8(x86_64::_mm_packs_epi16(sse16(rng), sse16(rng)));
print_sse_16(x86_64::_mm_packus_epi32(sse32(rng), sse32(rng)));
print_sse_16(x86_64::_mm_packs_epi32(sse32(rng), sse32(rng)));
}
unsafe fn avx_test(rng: &mut SmallRng) {
print_avx_8(x86_64::_mm256_packs_epi16(avx16(rng), avx16(rng)));
print_avx_8(x86_64::_mm256_packs_epi16(avx16(rng), avx16(rng)));
print_avx_16(x86_64::_mm256_packus_epi32(avx32(rng), avx32(rng)));
print_avx_16(x86_64::_mm256_packs_epi32(avx32(rng), avx32(rng)));
}
fn print_sse_8(t: __m128i) {
let ints = unsafe { std::mem::transmute::<_, [i8; 16]>(t) };
println!("{ints:?}");
}
fn print_sse_16(t: __m128i) {
let ints = unsafe { std::mem::transmute::<_, [i16; 8]>(t) };
println!("{ints:?}");
}
fn print_avx_8(t: __m256i) {
let ints = unsafe { std::mem::transmute::<_, [i8; 32]>(t) };
println!("{ints:?}");
}
fn print_avx_16(t: __m256i) {
let ints = unsafe { std::mem::transmute::<_, [i16; 16]>(t) };
println!("{ints:?}");
}
fn sse16(rand: &mut SmallRng) -> __m128i {
unsafe { std::mem::transmute([(); 8].map(|()| i16(rand))) }
}
fn sse32(rand: &mut SmallRng) -> __m128i {
unsafe { std::mem::transmute([(); 4].map(|()| i32(rand))) }
}
fn avx16(rand: &mut SmallRng) -> __m256i {
unsafe { std::mem::transmute([(); 16].map(|()| i16(rand))) }
}
fn avx32(rand: &mut SmallRng) -> __m256i {
unsafe { std::mem::transmute([(); 8].map(|()| i32(rand))) }
}
fn i16(rand: &mut SmallRng) -> i16 {
if rand.gen() {
rand.gen::<i16>()
} else {
rand.gen::<i8>() as i16
}
}
fn i32(rand: &mut SmallRng) -> i32 {
if rand.gen() {
rand.gen::<i32>()
} else {
rand.gen::<i16>() as i32
}
}
```
Noratrieb
force-pushed
the
x86-signed-pack
branch
from
3378308
to
22019db
Compare
|
Thanks! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This reduces the amount of duplicated code and the chance for bugs.
Sorta replaces #1442. I'm not very happy with how it turned out, but it looks mostly decent I think.
Also adds new support for
_mm_packs_epi16,_mm256_packs_epi16and_mm256_packus_epi32.For reference, here's a table of all the instructions: