add vtrn1 and vtrn2 neon instructions

This commit is contained in:
SparrowLii
2021-03-22 19:18:52 +08:00
committed by Amanieu d'Antras
parent 7599fd5310
commit 69bb74bca6
3 changed files with 819 additions and 2 deletions

View File

@@ -2023,6 +2023,358 @@ pub unsafe fn vrecpeq_f64(a: float64x2_t) -> float64x2_t {
vrecpeq_f64_(a) vrecpeq_f64_(a)
} }
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1_s8(a: int8x8_t, b: int8x8_t) -> int8x8_t {
simd_shuffle8(a, b, [0, 8, 2, 10, 4, 12, 6, 14])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_s8(a: int8x16_t, b: int8x16_t) -> int8x16_t {
simd_shuffle16(a, b, [0, 16, 2, 18, 4, 20, 6, 22, 8, 24, 10, 26, 12, 28, 14, 30])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1_s16(a: int16x4_t, b: int16x4_t) -> int16x4_t {
simd_shuffle4(a, b, [0, 4, 2, 6])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_s16(a: int16x8_t, b: int16x8_t) -> int16x8_t {
simd_shuffle8(a, b, [0, 8, 2, 10, 4, 12, 6, 14])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1_s32(a: int32x2_t, b: int32x2_t) -> int32x2_t {
simd_shuffle2(a, b, [0, 2])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_s32(a: int32x4_t, b: int32x4_t) -> int32x4_t {
simd_shuffle4(a, b, [0, 4, 2, 6])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_s64(a: int64x2_t, b: int64x2_t) -> int64x2_t {
simd_shuffle2(a, b, [0, 2])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1_u8(a: uint8x8_t, b: uint8x8_t) -> uint8x8_t {
simd_shuffle8(a, b, [0, 8, 2, 10, 4, 12, 6, 14])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_u8(a: uint8x16_t, b: uint8x16_t) -> uint8x16_t {
simd_shuffle16(a, b, [0, 16, 2, 18, 4, 20, 6, 22, 8, 24, 10, 26, 12, 28, 14, 30])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1_u16(a: uint16x4_t, b: uint16x4_t) -> uint16x4_t {
simd_shuffle4(a, b, [0, 4, 2, 6])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_u16(a: uint16x8_t, b: uint16x8_t) -> uint16x8_t {
simd_shuffle8(a, b, [0, 8, 2, 10, 4, 12, 6, 14])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1_u32(a: uint32x2_t, b: uint32x2_t) -> uint32x2_t {
simd_shuffle2(a, b, [0, 2])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_u32(a: uint32x4_t, b: uint32x4_t) -> uint32x4_t {
simd_shuffle4(a, b, [0, 4, 2, 6])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_u64(a: uint64x2_t, b: uint64x2_t) -> uint64x2_t {
simd_shuffle2(a, b, [0, 2])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1_p8(a: poly8x8_t, b: poly8x8_t) -> poly8x8_t {
simd_shuffle8(a, b, [0, 8, 2, 10, 4, 12, 6, 14])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_p8(a: poly8x16_t, b: poly8x16_t) -> poly8x16_t {
simd_shuffle16(a, b, [0, 16, 2, 18, 4, 20, 6, 22, 8, 24, 10, 26, 12, 28, 14, 30])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1_p16(a: poly16x4_t, b: poly16x4_t) -> poly16x4_t {
simd_shuffle4(a, b, [0, 4, 2, 6])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_p16(a: poly16x8_t, b: poly16x8_t) -> poly16x8_t {
simd_shuffle8(a, b, [0, 8, 2, 10, 4, 12, 6, 14])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_p64(a: poly64x2_t, b: poly64x2_t) -> poly64x2_t {
simd_shuffle2(a, b, [0, 2])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1_f32(a: float32x2_t, b: float32x2_t) -> float32x2_t {
simd_shuffle2(a, b, [0, 2])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_f32(a: float32x4_t, b: float32x4_t) -> float32x4_t {
simd_shuffle4(a, b, [0, 4, 2, 6])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn1))]
pub unsafe fn vtrn1q_f64(a: float64x2_t, b: float64x2_t) -> float64x2_t {
simd_shuffle2(a, b, [0, 2])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2_s8(a: int8x8_t, b: int8x8_t) -> int8x8_t {
simd_shuffle8(a, b, [1, 9, 3, 11, 5, 13, 7, 15])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_s8(a: int8x16_t, b: int8x16_t) -> int8x16_t {
simd_shuffle16(a, b, [1, 17, 3, 19, 5, 21, 7, 23, 9, 25, 11, 27, 13, 29, 15, 31])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2_s16(a: int16x4_t, b: int16x4_t) -> int16x4_t {
simd_shuffle4(a, b, [1, 5, 3, 7])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_s16(a: int16x8_t, b: int16x8_t) -> int16x8_t {
simd_shuffle8(a, b, [1, 9, 3, 11, 5, 13, 7, 15])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2_s32(a: int32x2_t, b: int32x2_t) -> int32x2_t {
simd_shuffle2(a, b, [1, 3])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_s32(a: int32x4_t, b: int32x4_t) -> int32x4_t {
simd_shuffle4(a, b, [1, 5, 3, 7])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_s64(a: int64x2_t, b: int64x2_t) -> int64x2_t {
simd_shuffle2(a, b, [1, 3])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2_u8(a: uint8x8_t, b: uint8x8_t) -> uint8x8_t {
simd_shuffle8(a, b, [1, 9, 3, 11, 5, 13, 7, 15])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_u8(a: uint8x16_t, b: uint8x16_t) -> uint8x16_t {
simd_shuffle16(a, b, [1, 17, 3, 19, 5, 21, 7, 23, 9, 25, 11, 27, 13, 29, 15, 31])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2_u16(a: uint16x4_t, b: uint16x4_t) -> uint16x4_t {
simd_shuffle4(a, b, [1, 5, 3, 7])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_u16(a: uint16x8_t, b: uint16x8_t) -> uint16x8_t {
simd_shuffle8(a, b, [1, 9, 3, 11, 5, 13, 7, 15])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2_u32(a: uint32x2_t, b: uint32x2_t) -> uint32x2_t {
simd_shuffle2(a, b, [1, 3])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_u32(a: uint32x4_t, b: uint32x4_t) -> uint32x4_t {
simd_shuffle4(a, b, [1, 5, 3, 7])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_u64(a: uint64x2_t, b: uint64x2_t) -> uint64x2_t {
simd_shuffle2(a, b, [1, 3])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2_p8(a: poly8x8_t, b: poly8x8_t) -> poly8x8_t {
simd_shuffle8(a, b, [1, 9, 3, 11, 5, 13, 7, 15])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_p8(a: poly8x16_t, b: poly8x16_t) -> poly8x16_t {
simd_shuffle16(a, b, [1, 17, 3, 19, 5, 21, 7, 23, 9, 25, 11, 27, 13, 29, 15, 31])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2_p16(a: poly16x4_t, b: poly16x4_t) -> poly16x4_t {
simd_shuffle4(a, b, [1, 5, 3, 7])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_p16(a: poly16x8_t, b: poly16x8_t) -> poly16x8_t {
simd_shuffle8(a, b, [1, 9, 3, 11, 5, 13, 7, 15])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_p64(a: poly64x2_t, b: poly64x2_t) -> poly64x2_t {
simd_shuffle2(a, b, [1, 3])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2_f32(a: float32x2_t, b: float32x2_t) -> float32x2_t {
simd_shuffle2(a, b, [1, 3])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_f32(a: float32x4_t, b: float32x4_t) -> float32x4_t {
simd_shuffle4(a, b, [1, 5, 3, 7])
}
/// Transpose vectors
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(test, assert_instr(trn2))]
pub unsafe fn vtrn2q_f64(a: float64x2_t, b: float64x2_t) -> float64x2_t {
simd_shuffle2(a, b, [1, 3])
}
#[cfg(test)] #[cfg(test)]
mod test { mod test {
use super::*; use super::*;
@@ -3822,4 +4174,400 @@ mod test {
let r: f64x2 = transmute(vrecpeq_f64(transmute(a))); let r: f64x2 = transmute(vrecpeq_f64(transmute(a)));
assert_eq!(r, e); assert_eq!(r, e);
} }
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1_s8() {
let a: i8x8 = i8x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: i8x8 = i8x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: i8x8 = i8x8::new(0, 1, 4, 5, 8, 9, 12, 13);
let r: i8x8 = transmute(vtrn1_s8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_s8() {
let a: i8x16 = i8x16::new(0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30);
let b: i8x16 = i8x16::new(1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31);
let e: i8x16 = i8x16::new(0, 1, 4, 5, 8, 9, 12, 13, 16, 17, 20, 21, 24, 25, 28, 29);
let r: i8x16 = transmute(vtrn1q_s8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1_s16() {
let a: i16x4 = i16x4::new(0, 2, 4, 6);
let b: i16x4 = i16x4::new(1, 3, 5, 7);
let e: i16x4 = i16x4::new(0, 1, 4, 5);
let r: i16x4 = transmute(vtrn1_s16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_s16() {
let a: i16x8 = i16x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: i16x8 = i16x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: i16x8 = i16x8::new(0, 1, 4, 5, 8, 9, 12, 13);
let r: i16x8 = transmute(vtrn1q_s16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1_s32() {
let a: i32x2 = i32x2::new(0, 2);
let b: i32x2 = i32x2::new(1, 3);
let e: i32x2 = i32x2::new(0, 1);
let r: i32x2 = transmute(vtrn1_s32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_s32() {
let a: i32x4 = i32x4::new(0, 2, 4, 6);
let b: i32x4 = i32x4::new(1, 3, 5, 7);
let e: i32x4 = i32x4::new(0, 1, 4, 5);
let r: i32x4 = transmute(vtrn1q_s32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_s64() {
let a: i64x2 = i64x2::new(0, 2);
let b: i64x2 = i64x2::new(1, 3);
let e: i64x2 = i64x2::new(0, 1);
let r: i64x2 = transmute(vtrn1q_s64(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1_u8() {
let a: u8x8 = u8x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: u8x8 = u8x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: u8x8 = u8x8::new(0, 1, 4, 5, 8, 9, 12, 13);
let r: u8x8 = transmute(vtrn1_u8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_u8() {
let a: u8x16 = u8x16::new(0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30);
let b: u8x16 = u8x16::new(1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31);
let e: u8x16 = u8x16::new(0, 1, 4, 5, 8, 9, 12, 13, 16, 17, 20, 21, 24, 25, 28, 29);
let r: u8x16 = transmute(vtrn1q_u8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1_u16() {
let a: u16x4 = u16x4::new(0, 2, 4, 6);
let b: u16x4 = u16x4::new(1, 3, 5, 7);
let e: u16x4 = u16x4::new(0, 1, 4, 5);
let r: u16x4 = transmute(vtrn1_u16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_u16() {
let a: u16x8 = u16x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: u16x8 = u16x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: u16x8 = u16x8::new(0, 1, 4, 5, 8, 9, 12, 13);
let r: u16x8 = transmute(vtrn1q_u16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1_u32() {
let a: u32x2 = u32x2::new(0, 2);
let b: u32x2 = u32x2::new(1, 3);
let e: u32x2 = u32x2::new(0, 1);
let r: u32x2 = transmute(vtrn1_u32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_u32() {
let a: u32x4 = u32x4::new(0, 2, 4, 6);
let b: u32x4 = u32x4::new(1, 3, 5, 7);
let e: u32x4 = u32x4::new(0, 1, 4, 5);
let r: u32x4 = transmute(vtrn1q_u32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_u64() {
let a: u64x2 = u64x2::new(0, 2);
let b: u64x2 = u64x2::new(1, 3);
let e: u64x2 = u64x2::new(0, 1);
let r: u64x2 = transmute(vtrn1q_u64(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1_p8() {
let a: i8x8 = i8x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: i8x8 = i8x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: i8x8 = i8x8::new(0, 1, 4, 5, 8, 9, 12, 13);
let r: i8x8 = transmute(vtrn1_p8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_p8() {
let a: i8x16 = i8x16::new(0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30);
let b: i8x16 = i8x16::new(1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31);
let e: i8x16 = i8x16::new(0, 1, 4, 5, 8, 9, 12, 13, 16, 17, 20, 21, 24, 25, 28, 29);
let r: i8x16 = transmute(vtrn1q_p8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1_p16() {
let a: i16x4 = i16x4::new(0, 2, 4, 6);
let b: i16x4 = i16x4::new(1, 3, 5, 7);
let e: i16x4 = i16x4::new(0, 1, 4, 5);
let r: i16x4 = transmute(vtrn1_p16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_p16() {
let a: i16x8 = i16x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: i16x8 = i16x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: i16x8 = i16x8::new(0, 1, 4, 5, 8, 9, 12, 13);
let r: i16x8 = transmute(vtrn1q_p16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_p64() {
let a: i64x2 = i64x2::new(0, 2);
let b: i64x2 = i64x2::new(1, 3);
let e: i64x2 = i64x2::new(0, 1);
let r: i64x2 = transmute(vtrn1q_p64(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1_f32() {
let a: f32x2 = f32x2::new(0., 2.);
let b: f32x2 = f32x2::new(1., 3.);
let e: f32x2 = f32x2::new(0., 1.);
let r: f32x2 = transmute(vtrn1_f32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_f32() {
let a: f32x4 = f32x4::new(0., 2., 4., 6.);
let b: f32x4 = f32x4::new(1., 3., 5., 7.);
let e: f32x4 = f32x4::new(0., 1., 4., 5.);
let r: f32x4 = transmute(vtrn1q_f32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn1q_f64() {
let a: f64x2 = f64x2::new(0., 2.);
let b: f64x2 = f64x2::new(1., 3.);
let e: f64x2 = f64x2::new(0., 1.);
let r: f64x2 = transmute(vtrn1q_f64(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2_s8() {
let a: i8x8 = i8x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: i8x8 = i8x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: i8x8 = i8x8::new(2, 3, 6, 7, 10, 11, 14, 15);
let r: i8x8 = transmute(vtrn2_s8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_s8() {
let a: i8x16 = i8x16::new(0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30);
let b: i8x16 = i8x16::new(1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31);
let e: i8x16 = i8x16::new(2, 3, 6, 7, 10, 11, 14, 15, 18, 19, 22, 23, 26, 27, 30, 31);
let r: i8x16 = transmute(vtrn2q_s8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2_s16() {
let a: i16x4 = i16x4::new(0, 2, 4, 6);
let b: i16x4 = i16x4::new(1, 3, 5, 7);
let e: i16x4 = i16x4::new(2, 3, 6, 7);
let r: i16x4 = transmute(vtrn2_s16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_s16() {
let a: i16x8 = i16x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: i16x8 = i16x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: i16x8 = i16x8::new(2, 3, 6, 7, 10, 11, 14, 15);
let r: i16x8 = transmute(vtrn2q_s16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2_s32() {
let a: i32x2 = i32x2::new(0, 2);
let b: i32x2 = i32x2::new(1, 3);
let e: i32x2 = i32x2::new(2, 3);
let r: i32x2 = transmute(vtrn2_s32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_s32() {
let a: i32x4 = i32x4::new(0, 2, 4, 6);
let b: i32x4 = i32x4::new(1, 3, 5, 7);
let e: i32x4 = i32x4::new(2, 3, 6, 7);
let r: i32x4 = transmute(vtrn2q_s32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_s64() {
let a: i64x2 = i64x2::new(0, 2);
let b: i64x2 = i64x2::new(1, 3);
let e: i64x2 = i64x2::new(2, 3);
let r: i64x2 = transmute(vtrn2q_s64(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2_u8() {
let a: u8x8 = u8x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: u8x8 = u8x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: u8x8 = u8x8::new(2, 3, 6, 7, 10, 11, 14, 15);
let r: u8x8 = transmute(vtrn2_u8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_u8() {
let a: u8x16 = u8x16::new(0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30);
let b: u8x16 = u8x16::new(1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31);
let e: u8x16 = u8x16::new(2, 3, 6, 7, 10, 11, 14, 15, 18, 19, 22, 23, 26, 27, 30, 31);
let r: u8x16 = transmute(vtrn2q_u8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2_u16() {
let a: u16x4 = u16x4::new(0, 2, 4, 6);
let b: u16x4 = u16x4::new(1, 3, 5, 7);
let e: u16x4 = u16x4::new(2, 3, 6, 7);
let r: u16x4 = transmute(vtrn2_u16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_u16() {
let a: u16x8 = u16x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: u16x8 = u16x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: u16x8 = u16x8::new(2, 3, 6, 7, 10, 11, 14, 15);
let r: u16x8 = transmute(vtrn2q_u16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2_u32() {
let a: u32x2 = u32x2::new(0, 2);
let b: u32x2 = u32x2::new(1, 3);
let e: u32x2 = u32x2::new(2, 3);
let r: u32x2 = transmute(vtrn2_u32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_u32() {
let a: u32x4 = u32x4::new(0, 2, 4, 6);
let b: u32x4 = u32x4::new(1, 3, 5, 7);
let e: u32x4 = u32x4::new(2, 3, 6, 7);
let r: u32x4 = transmute(vtrn2q_u32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_u64() {
let a: u64x2 = u64x2::new(0, 2);
let b: u64x2 = u64x2::new(1, 3);
let e: u64x2 = u64x2::new(2, 3);
let r: u64x2 = transmute(vtrn2q_u64(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2_p8() {
let a: i8x8 = i8x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: i8x8 = i8x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: i8x8 = i8x8::new(2, 3, 6, 7, 10, 11, 14, 15);
let r: i8x8 = transmute(vtrn2_p8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_p8() {
let a: i8x16 = i8x16::new(0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30);
let b: i8x16 = i8x16::new(1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31);
let e: i8x16 = i8x16::new(2, 3, 6, 7, 10, 11, 14, 15, 18, 19, 22, 23, 26, 27, 30, 31);
let r: i8x16 = transmute(vtrn2q_p8(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2_p16() {
let a: i16x4 = i16x4::new(0, 2, 4, 6);
let b: i16x4 = i16x4::new(1, 3, 5, 7);
let e: i16x4 = i16x4::new(2, 3, 6, 7);
let r: i16x4 = transmute(vtrn2_p16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_p16() {
let a: i16x8 = i16x8::new(0, 2, 4, 6, 8, 10, 12, 14);
let b: i16x8 = i16x8::new(1, 3, 5, 7, 9, 11, 13, 15);
let e: i16x8 = i16x8::new(2, 3, 6, 7, 10, 11, 14, 15);
let r: i16x8 = transmute(vtrn2q_p16(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_p64() {
let a: i64x2 = i64x2::new(0, 2);
let b: i64x2 = i64x2::new(1, 3);
let e: i64x2 = i64x2::new(2, 3);
let r: i64x2 = transmute(vtrn2q_p64(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2_f32() {
let a: f32x2 = f32x2::new(0., 2.);
let b: f32x2 = f32x2::new(1., 3.);
let e: f32x2 = f32x2::new(2., 3.);
let r: f32x2 = transmute(vtrn2_f32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_f32() {
let a: f32x4 = f32x4::new(0., 2., 4., 6.);
let b: f32x4 = f32x4::new(1., 3., 5., 7.);
let e: f32x4 = f32x4::new(2., 3., 6., 7.);
let r: f32x4 = transmute(vtrn2q_f32(transmute(a), transmute(b)));
assert_eq!(r, e);
}
#[simd_test(enable = "neon")]
unsafe fn test_vtrn2q_f64() {
let a: f64x2 = f64x2::new(0., 2.);
let b: f64x2 = f64x2::new(1., 3.);
let e: f64x2 = f64x2::new(2., 3.);
let r: f64x2 = transmute(vtrn2q_f64(transmute(a), transmute(b)));
assert_eq!(r, e);
}
} }

View File

@@ -1247,3 +1247,43 @@ generate float64x*_t
arm = vrecpe arm = vrecpe
link-arm = vrecpe._EXT_ link-arm = vrecpe._EXT_
generate float*_t generate float*_t
/// Transpose vectors
name = vtrn1
multi_fn = simd_shuffle-in_len-noext, a, b, {transpose-1-in_len}
a = 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30
b = 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31
validate 0, 1, 4, 5, 8, 9, 12, 13, 16, 17, 20, 21, 24, 25, 28, 29
aarch64 = trn1
generate int*_t, int64x2_t, uint*_t, uint64x2_t, poly8x8_t:poly8x8_t, poly8x16_t:poly8x16_t, poly16x4_t:poly16x4_t, poly16x8_t:poly16x8_t, poly64x2_t:poly64x2_t
/// Transpose vectors
name = vtrn1
multi_fn = simd_shuffle-in_len-noext, a, b, {transpose-1-in_len}
a = 0., 2., 4., 6., 8., 10., 12., 14.
b = 1., 3., 5., 7., 9., 11., 13., 15.
validate 0., 1., 4., 5., 8., 9., 12., 13.
aarch64 = trn1
generate float32x2_t:float32x2_t, float32x4_t:float32x4_t, float64x2_t:float64x2_t
/// Transpose vectors
name = vtrn2
multi_fn = simd_shuffle-in_len-noext, a, b, {transpose-2-in_len}
a = 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30
b = 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31
validate 2, 3, 6, 7, 10, 11, 14, 15, 18, 19, 22, 23, 26, 27, 30, 31
aarch64 = trn2
generate int*_t, int64x2_t, uint*_t, uint64x2_t, poly8x8_t:poly8x8_t, poly8x16_t:poly8x16_t, poly16x4_t:poly16x4_t, poly16x8_t:poly16x8_t, poly64x2_t:poly64x2_t
/// Transpose vectors
name = vtrn2
multi_fn = simd_shuffle-in_len-noext, a, b, {transpose-2-in_len}
a = 0., 2., 4., 6., 8., 10., 12., 14.
b = 1., 3., 5., 7., 9., 11., 13., 15.
validate 2., 3., 6., 7., 10., 11., 14., 15.
aarch64 = trn2
generate float32x2_t:float32x2_t, float32x4_t:float32x4_t, float64x2_t:float64x2_t

View File

@@ -109,6 +109,8 @@ fn type_to_suffix(t: &str) -> &str {
"float64x2_t" => "q_f64", "float64x2_t" => "q_f64",
"poly8x8_t" => "_p8", "poly8x8_t" => "_p8",
"poly8x16_t" => "q_p8", "poly8x16_t" => "q_p8",
"poly16x4_t" => "_p16",
"poly16x8_t" => "q_p16",
"poly64x1_t" => "_p64", "poly64x1_t" => "_p64",
"poly64x2_t" => "q_p64", "poly64x2_t" => "q_p64",
_ => panic!("unknown type: {}", t), _ => panic!("unknown type: {}", t),
@@ -325,6 +327,26 @@ fn type_to_half(t: &str) -> &str {
} }
} }
fn transpose1(x: usize) -> &'static str {
match x {
2 => "[0, 2]",
4 => "[0, 4, 2, 6]",
8 => "[0, 8, 2, 10, 4, 12, 6, 14]",
16 => "[0, 16, 2, 18, 4, 20, 6, 22, 8, 24, 10, 26, 12, 28, 14, 30]",
_ => panic!("unknown transpose order of len {}", x),
}
}
fn transpose2(x: usize) -> &'static str {
match x {
2 => "[1, 3]",
4 => "[1, 5, 3, 7]",
8 => "[1, 9, 3, 11, 5, 13, 7, 15]",
16 => "[1, 17, 3, 19, 5, 21, 7, 23, 9, 25, 11, 27, 13, 29, 15, 31]",
_ => panic!("unknown transpose order of len {}", x),
}
}
fn values(t: &str, vs: &[String]) -> String { fn values(t: &str, vs: &[String]) -> String {
if vs.len() == 1 && !t.contains('x') { if vs.len() == 1 && !t.contains('x') {
format!(": {} = {}", t, vs[0]) format!(": {} = {}", t, vs[0])
@@ -1044,6 +1066,12 @@ fn get_call(
}); });
return format!(r#"[{}]"#, &half[..half.len() - 2]); return format!(r#"[{}]"#, &half[..half.len() - 2]);
} }
if fn_name == "transpose-1-in_len" {
return transpose1(type_len(in_t[1])).to_string();
}
if fn_name == "transpose-2-in_len" {
return transpose2(type_len(in_t[1])).to_string();
}
if fn_name.contains('-') { if fn_name.contains('-') {
let fn_format: Vec<_> = fn_name.split('-').map(|v| v.to_string()).collect(); let fn_format: Vec<_> = fn_name.split('-').map(|v| v.to_string()).collect();
assert_eq!(fn_format.len(), 3); assert_eq!(fn_format.len(), 3);
@@ -1302,7 +1330,7 @@ mod test {
tests_aarch64.push('}'); tests_aarch64.push('}');
tests_aarch64.push('\n'); tests_aarch64.push('\n');
let arm_out_path: PathBuf = PathBuf::from(env::var("OUT_DIR").unwrap()) let arm_out_path: PathBuf = PathBuf::from("./crates/core_arch")
.join("src") .join("src")
.join("arm") .join("arm")
.join("neon"); .join("neon");
@@ -1312,7 +1340,8 @@ mod test {
file_arm.write_all(out_arm.as_bytes())?; file_arm.write_all(out_arm.as_bytes())?;
file_arm.write_all(tests_arm.as_bytes())?; file_arm.write_all(tests_arm.as_bytes())?;
let aarch64_out_path: PathBuf = PathBuf::from(env::var("OUT_DIR").unwrap()) //let aarch64_out_path: PathBuf = PathBuf::from(env::var("OUT_DIR").unwrap())
let aarch64_out_path: PathBuf = PathBuf::from("./crates/core_arch")
.join("src") .join("src")
.join("aarch64") .join("aarch64")
.join("neon"); .join("neon");