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

Add popcount Arm intrinsics. (#935)

Browse source
This commit is contained in:
Adam Hillier 2020-10-26 18:21:24 +00:00 • committed by GitHub
parent d39189a4c9
commit 8f8df056ca
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 125 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

125
library/stdarch/crates/core_arch/src/arm/neon/mod.rs
View file

@ -195,6 +195,13 @@ extern "C" {
#[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vmins.v4f32")]
#[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.fmin.v4f32")]
fn vminq_f32_(a: float32x4_t, b: float32x4_t) -> float32x4_t;
#[cfg_attr(target_arch = "arm", link_name = "llvm.ctpop.v8i8")]
#[cfg_attr(target_arch = "aarch64", link_name = "llvm.ctpop.v8i8")]
fn vcnt_s8_(a: int8x8_t) -> int8x8_t;
#[cfg_attr(target_arch = "arm", link_name = "llvm.ctpop.v16i8")]
#[cfg_attr(target_arch = "aarch64", link_name = "llvm.ctpop.v16i8")]
fn vcntq_s8_(a: int8x16_t) -> int8x16_t;
}
#[cfg(target_arch = "arm")]
@ -1863,6 +1870,61 @@ pub unsafe fn vmaxq_f32(a: float32x4_t, b: float32x4_t) -> float32x4_t {
vmaxq_f32_(a, b)
}
/// Population count per byte.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vcnt))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(cnt))]
pub unsafe fn vcnt_s8(a: int8x8_t) -> int8x8_t {
vcnt_s8_(a)
}
/// Population count per byte.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vcnt))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(cnt))]
pub unsafe fn vcntq_s8(a: int8x16_t) -> int8x16_t {
vcntq_s8_(a)
}
/// Population count per byte.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vcnt))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(cnt))]
pub unsafe fn vcnt_u8(a: uint8x8_t) -> uint8x8_t {
transmute(vcnt_s8_(transmute(a)))
}
/// Population count per byte.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vcnt))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(cnt))]
pub unsafe fn vcntq_u8(a: uint8x16_t) -> uint8x16_t {
transmute(vcntq_s8_(transmute(a)))
}
/// Population count per byte.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vcnt))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(cnt))]
pub unsafe fn vcnt_p8(a: poly8x8_t) -> poly8x8_t {
transmute(vcnt_s8_(transmute(a)))
}
/// Population count per byte.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vcnt))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(cnt))]
pub unsafe fn vcntq_p8(a: poly8x16_t) -> poly8x16_t {
transmute(vcntq_s8_(transmute(a)))
}
#[cfg(test)]
mod tests {
use super::*;
@ -4357,6 +4419,69 @@ mod tests {
let r: f32x4 = transmute(vmaxq_f32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vcnt_s8() {
let a: i8x8 = transmute(u8x8::new(
0b11001000, 0b11111111, 0b00000000, 0b11011111, 0b10000001, 0b10101001, 0b00001000,
0b00111111,
));
let e = i8x8::new(3, 8, 0, 7, 2, 4, 1, 6);
let r: i8x8 = transmute(vcnt_s8(transmute(a)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vcntq_s8() {
let a: i8x16 = transmute(u8x16::new(
0b11001000, 0b11111111, 0b00000000, 0b11011111, 0b10000001, 0b10101001, 0b00001000,
0b00111111, 0b11101110, 0b00000000, 0b11111111, 0b00100001, 0b11111111, 0b10010111,
0b11100000, 0b00010000,
));
let e = i8x16::new(3, 8, 0, 7, 2, 4, 1, 6, 6, 0, 8, 2, 8, 5, 3, 1);
let r: i8x16 = transmute(vcntq_s8(transmute(a)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vcnt_u8() {
let a = u8x8::new(
0b11001000, 0b11111111, 0b00000000, 0b11011111, 0b10000001, 0b10101001, 0b00001000,
0b00111111,
);
let e = u8x8::new(3, 8, 0, 7, 2, 4, 1, 6);
let r: u8x8 = transmute(vcnt_u8(transmute(a)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vcntq_u8() {
let a = u8x16::new(
0b11001000, 0b11111111, 0b00000000, 0b11011111, 0b10000001, 0b10101001, 0b00001000,
0b00111111, 0b11101110, 0b00000000, 0b11111111, 0b00100001, 0b11111111, 0b10010111,
0b11100000, 0b00010000,
);
let e = u8x16::new(3, 8, 0, 7, 2, 4, 1, 6, 6, 0, 8, 2, 8, 5, 3, 1);
let r: u8x16 = transmute(vcntq_u8(transmute(a)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vcnt_p8() {
let a = u8x8::new(
0b11001000, 0b11111111, 0b00000000, 0b11011111, 0b10000001, 0b10101001, 0b00001000,
0b00111111,
);
let e = u8x8::new(3, 8, 0, 7, 2, 4, 1, 6);
let r: u8x8 = transmute(vcnt_p8(transmute(a)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vcntq_p8() {
let a = u8x16::new(
0b11001000, 0b11111111, 0b00000000, 0b11011111, 0b10000001, 0b10101001, 0b00001000,
0b00111111, 0b11101110, 0b00000000, 0b11111111, 0b00100001, 0b11111111, 0b10010111,
0b11100000, 0b00010000,
);
let e = u8x16::new(3, 8, 0, 7, 2, 4, 1, 6, 6, 0, 8, 2, 8, 5, 3, 1);
let r: u8x16 = transmute(vcntq_p8(transmute(a)));
assert_eq!(r, e);
}
}
#[cfg(test)]