Skip to content

sse 4.1 #183

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
gnzlbg merged 13 commits into from
Nov 17, 2017
Merged

sse 4.1 #183

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
13 commits
Select commit Hold shift + click to select a range
73b21f5
Add _mm_cvtepi32_epi64 and fix typo
p32blo Nov 6, 2017
f8cc9f6
Add _mm_cvtepu8_epi{16, 32, 64}
p32blo Nov 6, 2017
33a3819
Add remaining _mm_cvtep* intrinsics
p32blo Nov 6, 2017
f630b1f
Add _mm_mul_epi32 and _mm_mullo_epi32
p32blo Nov 6, 2017
cdf0c24
Add _mm_testz_si128, _mm_testc_si128 and _mm_testnzc_si128
p32blo Nov 6, 2017
c8a5739
Add documentation for testz, testc and testnzc
p32blo Nov 7, 2017
19b2aed
Add _mm_test_all_zeros, _mm_test_all_ones and _mm_test_mix_ones_zeros
p32blo Nov 7, 2017
f7f2a23
Add _mm_minpos_epu16
p32blo Nov 7, 2017
6b718d9
Add _mm_mpsadbw_epu8
p32blo Nov 7, 2017
0d406d0
Change _mm_mpsadbw_epu8 to work with unsigned integers
p32blo Nov 7, 2017
e6b9e9d
Make test intrinsics use __m128i
p32blo Nov 7, 2017
e6cb465
Formatting
p32blo Nov 7, 2017
35f71a8
Merge pull request #1 from gnzlbg/p32blo_sse41
p32blo Nov 17, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 11 additions & 3 deletions src/lib.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -169,14 +169,20 @@ mod v32 {

define_ty! { i16x2, i16, i16 }
define_impl! { i16x2, i16, 2, i16x2, x0, x1 }
define_ty! { u16x2, u16, u16 }
define_impl! { u16x2, u16, 2, i16x2, x0, x1 }

define_ty! { i8x4, i8, i8, i8, i8 }
define_impl! { i8x4, i8, 4, i8x4, x0, x1, x2, x3 }

define_ty! { u8x4, u8, u8, u8, u8 }
define_impl! { u8x4, u8, 4, i8x4, x0, x1, x2, x3 }

define_casts!((i8x4, i32x4, as_i32x4), (i16x2, i64x2, as_i64x2));
define_casts!(
(i16x2, i64x2, as_i64x2),
(u16x2, i64x2, as_i64x2),
(i8x4, i32x4, as_i32x4),
(u8x4, i32x4, as_i32x4)
);
}

/// 16-bit wide vector tpyes
Expand All @@ -185,8 +191,10 @@ mod v16 {

define_ty! { i8x2, i8, i8 }
define_impl! { i8x2, i8, 2, i8x2, x0, x1 }
define_ty! { u8x2, u8, u8 }
define_impl! { u8x2, u8, 2, i8x2, x0, x1 }

define_casts!((i8x2, i64x2, as_i64x2));
define_casts!((i8x2, i64x2, as_i64x2), (u8x2, i64x2, as_i64x2));
}

#[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
Expand Down
8 changes: 4 additions & 4 deletions src/macros.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -485,7 +485,7 @@ macro_rules! test_arithmetic_ {

#[cfg(test)]
#[macro_export]
macro_rules! test_neg_ {
macro_rules! test_neg_ {
($tn:ident, $zero:expr, $one:expr, $two:expr, $four:expr) => {
{
let z = $tn::splat($zero);
Expand Down Expand Up @@ -573,7 +573,7 @@ macro_rules! test_bit_arithmetic_ {

#[cfg(test)]
#[macro_export]
macro_rules! test_ops_si {
macro_rules! test_ops_si {
($($tn:ident),+) => {
$(
test_arithmetic_!($tn, 0, 1, 2, 4);
Expand All @@ -585,7 +585,7 @@ macro_rules! test_bit_arithmetic_ {

#[cfg(test)]
#[macro_export]
macro_rules! test_ops_ui {
macro_rules! test_ops_ui {
($($tn:ident),+) => {
$(
test_arithmetic_!($tn, 0, 1, 2, 4);
Expand All @@ -596,7 +596,7 @@ macro_rules! test_bit_arithmetic_ {

#[cfg(test)]
#[macro_export]
macro_rules! test_ops_f {
macro_rules! test_ops_f {
($($tn:ident),+) => {
$(
test_arithmetic_!($tn, 0., 1., 2., 4.);
Expand Down
5 changes: 4 additions & 1 deletion src/v64.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,11 +60,14 @@ define_casts!(
(u8x8, i8x8, as_i8x8),
(i8x8, u8x8, as_u8x8),
(i8x8, i16x8, as_i16x8),
(u8x8, i16x8, as_i16x8),
(i16x4, i32x4, as_i32x4),
(i32x2, i64x2, as_i64x2),
(u8x8, u16x8, as_u16x8),
(u16x4, u32x4, as_u32x4),
(u32x2, u64x2, as_u64x2)
(u16x4, i32x4, as_i32x4),
(u32x2, u64x2, as_u64x2),
(u32x2, i64x2, as_i64x2)
);

#[cfg(test)]
Expand Down
16 changes: 16 additions & 0 deletions src/x86/macros.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -328,6 +328,22 @@ macro_rules! constify_imm4 {
}
}

macro_rules! constify_imm3 {
($imm8:expr, $expand:ident) => {
#[allow(overflowing_literals)]
match $imm8 & 0b111 {
0 => $expand!(0),
1 => $expand!(1),
2 => $expand!(2),
3 => $expand!(3),
4 => $expand!(4),
5 => $expand!(5),
6 => $expand!(6),
_ => $expand!(7),
}
}
}

macro_rules! constify_imm2 {
($imm8:expr, $expand:ident) => {
#[allow(overflowing_literals)]
Expand Down
Loading