cessen/psychopath
1
0
Fork 0
Code Issues 1 Pull Requests Projects Releases Wiki Activity

Make Sobol SIMD code work on all x86-64 platforms.

Browse Source 
Before this it needed SSE 4.1, which is not strictly present on
all x86-64 platforms.  This will still compile the faster path if
SSE 4.1 is available, but has an alternate path as well for all
x86-64 platforms.
This commit is contained in:
Nathan Vegdahl 2020-05-01 15:32:18 +09:00
parent fd75a72655
commit 78acaa7b63
1 changed files with 32 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

60
sub_crates/sobol/src/wide.rs
View File

@ -4,30 +4,9 @@
// #[cfg(all(target_arch = "x86_64", target_feature = "sse4.1"))]
pub(crate) mod sse {
use core::arch::x86_64::{
__m128i,
// SSE2 or less
_mm_add_epi32,
_mm_and_si128,
_mm_cvtepi32_ps,
_mm_mul_ps,
_mm_or_si128,
_mm_set1_epi32,
_mm_set1_ps,
_mm_setzero_si128,
_mm_slli_epi32,
_mm_srli_epi32,
_mm_xor_si128,
};
use core::arch::x86_64::{
// SSE3 / SSE4.1
// Note: these aren't necessarily actually available on all
// x86_64 platforms, so their use here isn't quite correct
// with the platform guard above.
// TODO: fix this at some point.
_mm_loadu_si128,
_mm_mullo_epi32,
__m128i, _mm_add_epi32, _mm_and_si128, _mm_cvtepi32_ps, _mm_mul_ps, _mm_or_si128,
_mm_set1_epi32, _mm_set1_ps, _mm_set_epi32, _mm_setzero_si128, _mm_slli_epi32,
_mm_srli_epi32, _mm_xor_si128,
};
#[derive(Debug, Copy, Clone)]
@ -124,9 +103,34 @@ pub(crate) mod sse {
impl std::ops::MulAssign for Int4 {
#[inline(always)]
fn mul_assign(&mut self, other: Self) {
*self = Int4 {
v: unsafe { _mm_mullo_epi32(self.v, other.v) },
};
// This only works with SSE 4.1 support.
#[cfg(target_feature = "sse4.1")]
unsafe {
use core::arch::x86_64::_mm_mullo_epi32;
*self = Int4 {
v: _mm_mullo_epi32(self.v, other.v),
};
}
// This works on all x86-64 chips.
#[cfg(not(target_feature = "sse4.1"))]
unsafe {
use core::arch::x86_64::{_mm_mul_epu32, _mm_shuffle_epi32};
let a = _mm_and_si128(
_mm_mul_epu32(self.v, other.v),
_mm_set_epi32(0, 0xffffffffu32 as i32, 0, 0xffffffffu32 as i32),
);
let b = _mm_and_si128(
_mm_mul_epu32(
_mm_shuffle_epi32(self.v, 0b11_11_01_01),
_mm_shuffle_epi32(other.v, 0b11_11_01_01),
),
_mm_set_epi32(0, 0xffffffffu32 as i32, 0, 0xffffffffu32 as i32),
);
*self = Int4 {
v: _mm_or_si128(a, _mm_shuffle_epi32(b, 0b10_11_00_01)),
};
}
}
}
@ -152,7 +156,7 @@ pub(crate) mod sse {
#[inline(always)]
fn from(v: [u32; 4]) -> Self {
Int4 {
v: unsafe { _mm_loadu_si128(std::mem::transmute(&v as *const u32)) },
v: unsafe { std::mem::transmute(v) },
}
}
}