user0/rust
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Add commented implementation of _mm_cvtepi8_epi64

Browse source
This commit is contained in:
André Oliveira 2017-11-03 14:45:50 +00:00 • committed by Andrew Gallant
parent 37396f3471
commit fa240f2477
1 changed files with 27 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

28
library/stdarch/src/x86/sse41.rs
View file

@ -4,7 +4,7 @@ use std::mem;
#[cfg(test)]
use stdsimd_test::assert_instr;
use simd_llvm::{simd_cast, simd_shuffle4, simd_shuffle8};
use simd_llvm::{simd_cast, simd_shuffle2, simd_shuffle4, simd_shuffle8};
use v128::*;
@ -261,6 +261,7 @@ pub unsafe fn _mm_cmpeq_epi64(a: i64x2, b: i64x2) -> i64x2 {
}
/// Sign extend packed 8-bit integers in `a` to packed 16-bit integers
#[inline(always)]
#[target_feature = "+sse4.1"]
#[cfg_attr(test, assert_instr(pmovsxbw))]
pub unsafe fn _mm_cvtepi8_epi16(a: i8x16) -> i16x8 {
@ -268,6 +269,7 @@ pub unsafe fn _mm_cvtepi8_epi16(a: i8x16) -> i16x8 {
}
/// Sign extend packed 8-bit integers in `a` to packed 32-bit integers
#[inline(always)]
#[target_feature = "+sse4.1"]
#[cfg_attr(test, assert_instr(pmovsxbd))]
pub unsafe fn _mm_cvtepi8_epi32(a: i8x16) -> i32x4 {
@ -275,6 +277,16 @@ pub unsafe fn _mm_cvtepi8_epi32(a: i8x16) -> i32x4 {
simd_shuffle4(cast, cast, [0, 1, 2, 3])
}
/// Sign extend packed 8-bit integers in the low 8 bytes of `a` to packed 64-bit integers
/*
#[inline(always)]
#[target_feature = "+sse4.1"]
#[cfg_attr(test, assert_instr(pmovsxbq))]
pub unsafe fn _mm_cvtepi8_epi64(a: i8x16) -> i64x2 {
simd_cast::<::v16::i8x2, _>(simd_shuffle2(a, a, [0, 1]))
}
*/
/// Returns the dot product of two f64x2 vectors.
///
/// `imm8[1:0]` is the broadcast mask, and `imm8[5:4]` is the condition mask.
@ -799,6 +811,20 @@ mod tests {
assert_eq!(r, e);
}
/*
#[simd_test = "sse4.1"]
unsafe fn _mm_cvtepi8_epi64() {
let a = i8x16::splat(10);
let r = sse41::_mm_cvtepi8_epi64(a);
let e = i64x2::splat(10);
assert_eq!(r, e);
let a = i8x16::splat(-10);
let r = sse41::_mm_cvtepi8_epi64(a);
let e = i64x2::splat(-10);
assert_eq!(r, e);
}
*/
#[simd_test = "sse4.1"]
unsafe fn _mm_dp_pd() {
let a = f64x2::new(2.0, 3.0);