diff options
Diffstat (limited to 'tests/codegen/simd-intrinsic')
22 files changed, 2272 insertions, 0 deletions
diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-abs.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-abs.rs new file mode 100644 index 000000000..e7bb2327a --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-abs.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_fabs<T>(x: T) -> T; +} + +// CHECK-LABEL: @fabs_32x2 +#[no_mangle] +pub unsafe fn fabs_32x2(a: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.fabs.v2f32 + simd_fabs(a) +} + +// CHECK-LABEL: @fabs_32x4 +#[no_mangle] +pub unsafe fn fabs_32x4(a: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.fabs.v4f32 + simd_fabs(a) +} + +// CHECK-LABEL: @fabs_32x8 +#[no_mangle] +pub unsafe fn fabs_32x8(a: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.fabs.v8f32 + simd_fabs(a) +} + +// CHECK-LABEL: @fabs_32x16 +#[no_mangle] +pub unsafe fn fabs_32x16(a: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.fabs.v16f32 + simd_fabs(a) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @fabs_64x4 +#[no_mangle] +pub unsafe fn fabs_64x4(a: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.fabs.v4f64 + simd_fabs(a) +} + +// CHECK-LABEL: @fabs_64x2 +#[no_mangle] +pub unsafe fn fabs_64x2(a: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.fabs.v2f64 + simd_fabs(a) +} + +// CHECK-LABEL: @fabs_64x8 +#[no_mangle] +pub unsafe fn fabs_64x8(a: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.fabs.v8f64 + simd_fabs(a) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-ceil.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-ceil.rs new file mode 100644 index 000000000..e33482d75 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-ceil.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_ceil<T>(x: T) -> T; +} + +// CHECK-LABEL: @ceil_32x2 +#[no_mangle] +pub unsafe fn ceil_32x2(a: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.ceil.v2f32 + simd_ceil(a) +} + +// CHECK-LABEL: @ceil_32x4 +#[no_mangle] +pub unsafe fn ceil_32x4(a: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.ceil.v4f32 + simd_ceil(a) +} + +// CHECK-LABEL: @ceil_32x8 +#[no_mangle] +pub unsafe fn ceil_32x8(a: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.ceil.v8f32 + simd_ceil(a) +} + +// CHECK-LABEL: @ceil_32x16 +#[no_mangle] +pub unsafe fn ceil_32x16(a: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.ceil.v16f32 + simd_ceil(a) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @ceil_64x4 +#[no_mangle] +pub unsafe fn ceil_64x4(a: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.ceil.v4f64 + simd_ceil(a) +} + +// CHECK-LABEL: @ceil_64x2 +#[no_mangle] +pub unsafe fn ceil_64x2(a: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.ceil.v2f64 + simd_ceil(a) +} + +// CHECK-LABEL: @ceil_64x8 +#[no_mangle] +pub unsafe fn ceil_64x8(a: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.ceil.v8f64 + simd_ceil(a) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-cos.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-cos.rs new file mode 100644 index 000000000..0f52952bc --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-cos.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_fcos<T>(x: T) -> T; +} + +// CHECK-LABEL: @fcos_32x2 +#[no_mangle] +pub unsafe fn fcos_32x2(a: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.cos.v2f32 + simd_fcos(a) +} + +// CHECK-LABEL: @fcos_32x4 +#[no_mangle] +pub unsafe fn fcos_32x4(a: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.cos.v4f32 + simd_fcos(a) +} + +// CHECK-LABEL: @fcos_32x8 +#[no_mangle] +pub unsafe fn fcos_32x8(a: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.cos.v8f32 + simd_fcos(a) +} + +// CHECK-LABEL: @fcos_32x16 +#[no_mangle] +pub unsafe fn fcos_32x16(a: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.cos.v16f32 + simd_fcos(a) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @fcos_64x4 +#[no_mangle] +pub unsafe fn fcos_64x4(a: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.cos.v4f64 + simd_fcos(a) +} + +// CHECK-LABEL: @fcos_64x2 +#[no_mangle] +pub unsafe fn fcos_64x2(a: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.cos.v2f64 + simd_fcos(a) +} + +// CHECK-LABEL: @fcos_64x8 +#[no_mangle] +pub unsafe fn fcos_64x8(a: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.cos.v8f64 + simd_fcos(a) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-exp.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-exp.rs new file mode 100644 index 000000000..1154acf69 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-exp.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_fexp<T>(x: T) -> T; +} + +// CHECK-LABEL: @exp_32x2 +#[no_mangle] +pub unsafe fn exp_32x2(a: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.exp.v2f32 + simd_fexp(a) +} + +// CHECK-LABEL: @exp_32x4 +#[no_mangle] +pub unsafe fn exp_32x4(a: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.exp.v4f32 + simd_fexp(a) +} + +// CHECK-LABEL: @exp_32x8 +#[no_mangle] +pub unsafe fn exp_32x8(a: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.exp.v8f32 + simd_fexp(a) +} + +// CHECK-LABEL: @exp_32x16 +#[no_mangle] +pub unsafe fn exp_32x16(a: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.exp.v16f32 + simd_fexp(a) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @exp_64x4 +#[no_mangle] +pub unsafe fn exp_64x4(a: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.exp.v4f64 + simd_fexp(a) +} + +// CHECK-LABEL: @exp_64x2 +#[no_mangle] +pub unsafe fn exp_64x2(a: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.exp.v2f64 + simd_fexp(a) +} + +// CHECK-LABEL: @exp_64x8 +#[no_mangle] +pub unsafe fn exp_64x8(a: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.exp.v8f64 + simd_fexp(a) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-exp2.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-exp2.rs new file mode 100644 index 000000000..929dc9ac8 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-exp2.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_fexp2<T>(x: T) -> T; +} + +// CHECK-LABEL: @exp2_32x2 +#[no_mangle] +pub unsafe fn exp2_32x2(a: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.exp2.v2f32 + simd_fexp2(a) +} + +// CHECK-LABEL: @exp2_32x4 +#[no_mangle] +pub unsafe fn exp2_32x4(a: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.exp2.v4f32 + simd_fexp2(a) +} + +// CHECK-LABEL: @exp2_32x8 +#[no_mangle] +pub unsafe fn exp2_32x8(a: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.exp2.v8f32 + simd_fexp2(a) +} + +// CHECK-LABEL: @exp2_32x16 +#[no_mangle] +pub unsafe fn exp2_32x16(a: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.exp2.v16f32 + simd_fexp2(a) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @exp2_64x4 +#[no_mangle] +pub unsafe fn exp2_64x4(a: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.exp2.v4f64 + simd_fexp2(a) +} + +// CHECK-LABEL: @exp2_64x2 +#[no_mangle] +pub unsafe fn exp2_64x2(a: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.exp2.v2f64 + simd_fexp2(a) +} + +// CHECK-LABEL: @exp2_64x8 +#[no_mangle] +pub unsafe fn exp2_64x8(a: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.exp2.v8f64 + simd_fexp2(a) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-floor.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-floor.rs new file mode 100644 index 000000000..56ca644f6 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-floor.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_floor<T>(x: T) -> T; +} + +// CHECK-LABEL: @floor_32x2 +#[no_mangle] +pub unsafe fn floor_32x2(a: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.floor.v2f32 + simd_floor(a) +} + +// CHECK-LABEL: @floor_32x4 +#[no_mangle] +pub unsafe fn floor_32x4(a: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.floor.v4f32 + simd_floor(a) +} + +// CHECK-LABEL: @floor_32x8 +#[no_mangle] +pub unsafe fn floor_32x8(a: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.floor.v8f32 + simd_floor(a) +} + +// CHECK-LABEL: @floor_32x16 +#[no_mangle] +pub unsafe fn floor_32x16(a: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.floor.v16f32 + simd_floor(a) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @floor_64x4 +#[no_mangle] +pub unsafe fn floor_64x4(a: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.floor.v4f64 + simd_floor(a) +} + +// CHECK-LABEL: @floor_64x2 +#[no_mangle] +pub unsafe fn floor_64x2(a: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.floor.v2f64 + simd_floor(a) +} + +// CHECK-LABEL: @floor_64x8 +#[no_mangle] +pub unsafe fn floor_64x8(a: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.floor.v8f64 + simd_floor(a) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-fma.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-fma.rs new file mode 100644 index 000000000..fd65cb72b --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-fma.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_fma<T>(x: T, b: T, c: T) -> T; +} + +// CHECK-LABEL: @fma_32x2 +#[no_mangle] +pub unsafe fn fma_32x2(a: f32x2, b: f32x2, c: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.fma.v2f32 + simd_fma(a, b, c) +} + +// CHECK-LABEL: @fma_32x4 +#[no_mangle] +pub unsafe fn fma_32x4(a: f32x4, b: f32x4, c: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.fma.v4f32 + simd_fma(a, b, c) +} + +// CHECK-LABEL: @fma_32x8 +#[no_mangle] +pub unsafe fn fma_32x8(a: f32x8, b: f32x8, c: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.fma.v8f32 + simd_fma(a, b, c) +} + +// CHECK-LABEL: @fma_32x16 +#[no_mangle] +pub unsafe fn fma_32x16(a: f32x16, b: f32x16, c: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.fma.v16f32 + simd_fma(a, b, c) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @fma_64x4 +#[no_mangle] +pub unsafe fn fma_64x4(a: f64x4, b: f64x4, c: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.fma.v4f64 + simd_fma(a, b, c) +} + +// CHECK-LABEL: @fma_64x2 +#[no_mangle] +pub unsafe fn fma_64x2(a: f64x2, b: f64x2, c: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.fma.v2f64 + simd_fma(a, b, c) +} + +// CHECK-LABEL: @fma_64x8 +#[no_mangle] +pub unsafe fn fma_64x8(a: f64x8, b: f64x8, c: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.fma.v8f64 + simd_fma(a, b, c) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-fsqrt.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-fsqrt.rs new file mode 100644 index 000000000..adc191925 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-fsqrt.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_fsqrt<T>(x: T) -> T; +} + +// CHECK-LABEL: @fsqrt_32x2 +#[no_mangle] +pub unsafe fn fsqrt_32x2(a: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.sqrt.v2f32 + simd_fsqrt(a) +} + +// CHECK-LABEL: @fsqrt_32x4 +#[no_mangle] +pub unsafe fn fsqrt_32x4(a: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.sqrt.v4f32 + simd_fsqrt(a) +} + +// CHECK-LABEL: @fsqrt_32x8 +#[no_mangle] +pub unsafe fn fsqrt_32x8(a: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.sqrt.v8f32 + simd_fsqrt(a) +} + +// CHECK-LABEL: @fsqrt_32x16 +#[no_mangle] +pub unsafe fn fsqrt_32x16(a: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.sqrt.v16f32 + simd_fsqrt(a) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @fsqrt_64x4 +#[no_mangle] +pub unsafe fn fsqrt_64x4(a: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.sqrt.v4f64 + simd_fsqrt(a) +} + +// CHECK-LABEL: @fsqrt_64x2 +#[no_mangle] +pub unsafe fn fsqrt_64x2(a: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.sqrt.v2f64 + simd_fsqrt(a) +} + +// CHECK-LABEL: @fsqrt_64x8 +#[no_mangle] +pub unsafe fn fsqrt_64x8(a: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.sqrt.v8f64 + simd_fsqrt(a) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-log.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-log.rs new file mode 100644 index 000000000..c072519c0 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-log.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_flog<T>(x: T) -> T; +} + +// CHECK-LABEL: @log_32x2 +#[no_mangle] +pub unsafe fn log_32x2(a: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.log.v2f32 + simd_flog(a) +} + +// CHECK-LABEL: @log_32x4 +#[no_mangle] +pub unsafe fn log_32x4(a: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.log.v4f32 + simd_flog(a) +} + +// CHECK-LABEL: @log_32x8 +#[no_mangle] +pub unsafe fn log_32x8(a: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.log.v8f32 + simd_flog(a) +} + +// CHECK-LABEL: @log_32x16 +#[no_mangle] +pub unsafe fn log_32x16(a: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.log.v16f32 + simd_flog(a) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @log_64x4 +#[no_mangle] +pub unsafe fn log_64x4(a: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.log.v4f64 + simd_flog(a) +} + +// CHECK-LABEL: @log_64x2 +#[no_mangle] +pub unsafe fn log_64x2(a: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.log.v2f64 + simd_flog(a) +} + +// CHECK-LABEL: @log_64x8 +#[no_mangle] +pub unsafe fn log_64x8(a: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.log.v8f64 + simd_flog(a) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-log10.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-log10.rs new file mode 100644 index 000000000..5fd648995 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-log10.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_flog10<T>(x: T) -> T; +} + +// CHECK-LABEL: @log10_32x2 +#[no_mangle] +pub unsafe fn log10_32x2(a: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.log10.v2f32 + simd_flog10(a) +} + +// CHECK-LABEL: @log10_32x4 +#[no_mangle] +pub unsafe fn log10_32x4(a: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.log10.v4f32 + simd_flog10(a) +} + +// CHECK-LABEL: @log10_32x8 +#[no_mangle] +pub unsafe fn log10_32x8(a: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.log10.v8f32 + simd_flog10(a) +} + +// CHECK-LABEL: @log10_32x16 +#[no_mangle] +pub unsafe fn log10_32x16(a: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.log10.v16f32 + simd_flog10(a) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @log10_64x4 +#[no_mangle] +pub unsafe fn log10_64x4(a: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.log10.v4f64 + simd_flog10(a) +} + +// CHECK-LABEL: @log10_64x2 +#[no_mangle] +pub unsafe fn log10_64x2(a: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.log10.v2f64 + simd_flog10(a) +} + +// CHECK-LABEL: @log10_64x8 +#[no_mangle] +pub unsafe fn log10_64x8(a: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.log10.v8f64 + simd_flog10(a) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-log2.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-log2.rs new file mode 100644 index 000000000..35175f0ca --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-log2.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_flog2<T>(x: T) -> T; +} + +// CHECK-LABEL: @log2_32x2 +#[no_mangle] +pub unsafe fn log2_32x2(a: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.log2.v2f32 + simd_flog2(a) +} + +// CHECK-LABEL: @log2_32x4 +#[no_mangle] +pub unsafe fn log2_32x4(a: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.log2.v4f32 + simd_flog2(a) +} + +// CHECK-LABEL: @log2_32x8 +#[no_mangle] +pub unsafe fn log2_32x8(a: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.log2.v8f32 + simd_flog2(a) +} + +// CHECK-LABEL: @log2_32x16 +#[no_mangle] +pub unsafe fn log2_32x16(a: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.log2.v16f32 + simd_flog2(a) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @log2_64x4 +#[no_mangle] +pub unsafe fn log2_64x4(a: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.log2.v4f64 + simd_flog2(a) +} + +// CHECK-LABEL: @log2_64x2 +#[no_mangle] +pub unsafe fn log2_64x2(a: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.log2.v2f64 + simd_flog2(a) +} + +// CHECK-LABEL: @log2_64x8 +#[no_mangle] +pub unsafe fn log2_64x8(a: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.log2.v8f64 + simd_flog2(a) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-minmax.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-minmax.rs new file mode 100644 index 000000000..4e0abed78 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-minmax.rs @@ -0,0 +1,29 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_fmin<T>(x: T, y: T) -> T; + fn simd_fmax<T>(x: T, y: T) -> T; +} + +// CHECK-LABEL: @fmin +#[no_mangle] +pub unsafe fn fmin(a: f32x4, b: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.minnum.v4f32 + simd_fmin(a, b) +} + +// CHECK-LABEL: @fmax +#[no_mangle] +pub unsafe fn fmax(a: f32x4, b: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.maxnum.v4f32 + simd_fmax(a, b) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-pow.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-pow.rs new file mode 100644 index 000000000..3b8d611ab --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-pow.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_fpow<T>(x: T, b: T) -> T; +} + +// CHECK-LABEL: @fpow_32x2 +#[no_mangle] +pub unsafe fn fpow_32x2(a: f32x2, b: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.pow.v2f32 + simd_fpow(a, b) +} + +// CHECK-LABEL: @fpow_32x4 +#[no_mangle] +pub unsafe fn fpow_32x4(a: f32x4, b: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.pow.v4f32 + simd_fpow(a, b) +} + +// CHECK-LABEL: @fpow_32x8 +#[no_mangle] +pub unsafe fn fpow_32x8(a: f32x8, b: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.pow.v8f32 + simd_fpow(a, b) +} + +// CHECK-LABEL: @fpow_32x16 +#[no_mangle] +pub unsafe fn fpow_32x16(a: f32x16, b: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.pow.v16f32 + simd_fpow(a, b) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @fpow_64x4 +#[no_mangle] +pub unsafe fn fpow_64x4(a: f64x4, b: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.pow.v4f64 + simd_fpow(a, b) +} + +// CHECK-LABEL: @fpow_64x2 +#[no_mangle] +pub unsafe fn fpow_64x2(a: f64x2, b: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.pow.v2f64 + simd_fpow(a, b) +} + +// CHECK-LABEL: @fpow_64x8 +#[no_mangle] +pub unsafe fn fpow_64x8(a: f64x8, b: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.pow.v8f64 + simd_fpow(a, b) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-powi.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-powi.rs new file mode 100644 index 000000000..e80c50c10 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-powi.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_fpowi<T>(x: T, b: i32) -> T; +} + +// CHECK-LABEL: @fpowi_32x2 +#[no_mangle] +pub unsafe fn fpowi_32x2(a: f32x2, b: i32) -> f32x2 { + // CHECK: call <2 x float> @llvm.powi.v2f32 + simd_fpowi(a, b) +} + +// CHECK-LABEL: @fpowi_32x4 +#[no_mangle] +pub unsafe fn fpowi_32x4(a: f32x4, b: i32) -> f32x4 { + // CHECK: call <4 x float> @llvm.powi.v4f32 + simd_fpowi(a, b) +} + +// CHECK-LABEL: @fpowi_32x8 +#[no_mangle] +pub unsafe fn fpowi_32x8(a: f32x8, b: i32) -> f32x8 { + // CHECK: call <8 x float> @llvm.powi.v8f32 + simd_fpowi(a, b) +} + +// CHECK-LABEL: @fpowi_32x16 +#[no_mangle] +pub unsafe fn fpowi_32x16(a: f32x16, b: i32) -> f32x16 { + // CHECK: call <16 x float> @llvm.powi.v16f32 + simd_fpowi(a, b) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @fpowi_64x4 +#[no_mangle] +pub unsafe fn fpowi_64x4(a: f64x4, b: i32) -> f64x4 { + // CHECK: call <4 x double> @llvm.powi.v4f64 + simd_fpowi(a, b) +} + +// CHECK-LABEL: @fpowi_64x2 +#[no_mangle] +pub unsafe fn fpowi_64x2(a: f64x2, b: i32) -> f64x2 { + // CHECK: call <2 x double> @llvm.powi.v2f64 + simd_fpowi(a, b) +} + +// CHECK-LABEL: @fpowi_64x8 +#[no_mangle] +pub unsafe fn fpowi_64x8(a: f64x8, b: i32) -> f64x8 { + // CHECK: call <8 x double> @llvm.powi.v8f64 + simd_fpowi(a, b) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-float-sin.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-float-sin.rs new file mode 100644 index 000000000..9e3fab49a --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-float-sin.rs @@ -0,0 +1,92 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x2(pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x16(pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32, + pub f32, pub f32, pub f32, pub f32); + +extern "platform-intrinsic" { + fn simd_fsin<T>(x: T) -> T; +} + +// CHECK-LABEL: @fsin_32x2 +#[no_mangle] +pub unsafe fn fsin_32x2(a: f32x2) -> f32x2 { + // CHECK: call <2 x float> @llvm.sin.v2f32 + simd_fsin(a) +} + +// CHECK-LABEL: @fsin_32x4 +#[no_mangle] +pub unsafe fn fsin_32x4(a: f32x4) -> f32x4 { + // CHECK: call <4 x float> @llvm.sin.v4f32 + simd_fsin(a) +} + +// CHECK-LABEL: @fsin_32x8 +#[no_mangle] +pub unsafe fn fsin_32x8(a: f32x8) -> f32x8 { + // CHECK: call <8 x float> @llvm.sin.v8f32 + simd_fsin(a) +} + +// CHECK-LABEL: @fsin_32x16 +#[no_mangle] +pub unsafe fn fsin_32x16(a: f32x16) -> f32x16 { + // CHECK: call <16 x float> @llvm.sin.v16f32 + simd_fsin(a) +} + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x2(pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x4(pub f64, pub f64, pub f64, pub f64); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f64x8(pub f64, pub f64, pub f64, pub f64, + pub f64, pub f64, pub f64, pub f64); + +// CHECK-LABEL: @fsin_64x4 +#[no_mangle] +pub unsafe fn fsin_64x4(a: f64x4) -> f64x4 { + // CHECK: call <4 x double> @llvm.sin.v4f64 + simd_fsin(a) +} + +// CHECK-LABEL: @fsin_64x2 +#[no_mangle] +pub unsafe fn fsin_64x2(a: f64x2) -> f64x2 { + // CHECK: call <2 x double> @llvm.sin.v2f64 + simd_fsin(a) +} + +// CHECK-LABEL: @fsin_64x8 +#[no_mangle] +pub unsafe fn fsin_64x8(a: f64x8) -> f64x8 { + // CHECK: call <8 x double> @llvm.sin.v8f64 + simd_fsin(a) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-generic-arithmetic-saturating.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-arithmetic-saturating.rs new file mode 100644 index 000000000..6fb0ceb40 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-arithmetic-saturating.rs @@ -0,0 +1,692 @@ +// compile-flags: -C no-prepopulate-passes +// + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] +#![deny(unused)] + +// signed integer types + +#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x2(i8, i8); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x4(i8, i8, i8, i8); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x8( + i8, i8, i8, i8, i8, i8, i8, i8, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x16( + i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x32( + i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, + i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x64( + i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, + i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, + i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, + i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, +); + +#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x2(i16, i16); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x4(i16, i16, i16, i16); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x8( + i16, i16, i16, i16, i16, i16, i16, i16, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x16( + i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x32( + i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, + i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, +); + +#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x2(i32, i32); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x4(i32, i32, i32, i32); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x8( + i32, i32, i32, i32, i32, i32, i32, i32, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x16( + i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, +); + +#[repr(simd)] #[derive(Copy, Clone)] pub struct i64x2(i64, i64); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i64x4(i64, i64, i64, i64); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i64x8( + i64, i64, i64, i64, i64, i64, i64, i64, +); + +#[repr(simd)] #[derive(Copy, Clone)] pub struct i128x2(i128, i128); +#[repr(simd)] #[derive(Copy, Clone)] pub struct i128x4(i128, i128, i128, i128); + +// unsigned integer types + +#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x2(u8, u8); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x4(u8, u8, u8, u8); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x8( + u8, u8, u8, u8, u8, u8, u8, u8, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x16( + u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x32( + u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, + u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x64( + u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, + u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, + u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, + u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, +); + +#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x2(u16, u16); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x4(u16, u16, u16, u16); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x8( + u16, u16, u16, u16, u16, u16, u16, u16, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x16( + u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x32( + u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, + u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, +); + +#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x2(u32, u32); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x4(u32, u32, u32, u32); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x8( + u32, u32, u32, u32, u32, u32, u32, u32, +); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x16( + u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, +); + +#[repr(simd)] #[derive(Copy, Clone)] pub struct u64x2(u64, u64); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u64x4(u64, u64, u64, u64); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u64x8( + u64, u64, u64, u64, u64, u64, u64, u64, +); + +#[repr(simd)] #[derive(Copy, Clone)] pub struct u128x2(u128, u128); +#[repr(simd)] #[derive(Copy, Clone)] pub struct u128x4(u128, u128, u128, u128); + +extern "platform-intrinsic" { + fn simd_saturating_add<T>(x: T, y: T) -> T; + fn simd_saturating_sub<T>(x: T, y: T) -> T; +} + +// NOTE(eddyb) `%{{x|_3}}` is used because on some targets (e.g. WASM) +// SIMD vectors are passed directly, resulting in `%x` being a vector, +// while on others they're passed indirectly, resulting in `%x` being +// a pointer to a vector, and `%_3` a vector loaded from that pointer. +// This is controlled by the target spec option `simd_types_indirect`. +// The same applies to `%{{y|_4}}` as well. + +// CHECK-LABEL: @sadd_i8x2 +#[no_mangle] +pub unsafe fn sadd_i8x2(x: i8x2, y: i8x2) -> i8x2 { + // CHECK: %{{[0-9]+}} = call <2 x i8> @llvm.sadd.sat.v2i8(<2 x i8> %{{x|_3}}, <2 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i8x4 +#[no_mangle] +pub unsafe fn sadd_i8x4(x: i8x4, y: i8x4) -> i8x4 { + // CHECK: %{{[0-9]+}} = call <4 x i8> @llvm.sadd.sat.v4i8(<4 x i8> %{{x|_3}}, <4 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i8x8 +#[no_mangle] +pub unsafe fn sadd_i8x8(x: i8x8, y: i8x8) -> i8x8 { + // CHECK: %{{[0-9]+}} = call <8 x i8> @llvm.sadd.sat.v8i8(<8 x i8> %{{x|_3}}, <8 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i8x16 +#[no_mangle] +pub unsafe fn sadd_i8x16(x: i8x16, y: i8x16) -> i8x16 { + // CHECK: %{{[0-9]+}} = call <16 x i8> @llvm.sadd.sat.v16i8(<16 x i8> %{{x|_3}}, <16 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i8x32 +#[no_mangle] +pub unsafe fn sadd_i8x32(x: i8x32, y: i8x32) -> i8x32 { + // CHECK: %{{[0-9]+}} = call <32 x i8> @llvm.sadd.sat.v32i8(<32 x i8> %{{x|_3}}, <32 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i8x64 +#[no_mangle] +pub unsafe fn sadd_i8x64(x: i8x64, y: i8x64) -> i8x64 { + // CHECK: %{{[0-9]+}} = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %{{x|_3}}, <64 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i16x2 +#[no_mangle] +pub unsafe fn sadd_i16x2(x: i16x2, y: i16x2) -> i16x2 { + // CHECK: %{{[0-9]+}} = call <2 x i16> @llvm.sadd.sat.v2i16(<2 x i16> %{{x|_3}}, <2 x i16> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i16x4 +#[no_mangle] +pub unsafe fn sadd_i16x4(x: i16x4, y: i16x4) -> i16x4 { + // CHECK: %{{[0-9]+}} = call <4 x i16> @llvm.sadd.sat.v4i16(<4 x i16> %{{x|_3}}, <4 x i16> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i16x8 +#[no_mangle] +pub unsafe fn sadd_i16x8(x: i16x8, y: i16x8) -> i16x8 { + // CHECK: %{{[0-9]+}} = call <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16> %{{x|_3}}, <8 x i16> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i16x16 +#[no_mangle] +pub unsafe fn sadd_i16x16(x: i16x16, y: i16x16) -> i16x16 { + // CHECK: %{{[0-9]+}} = call <16 x i16> @llvm.sadd.sat.v16i16(<16 x i16> %{{x|_3}}, <16 x i16> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i16x32 +#[no_mangle] +pub unsafe fn sadd_i16x32(x: i16x32, y: i16x32) -> i16x32 { + // CHECK: %{{[0-9]+}} = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %{{x|_3}}, <32 x i16> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i32x2 +#[no_mangle] +pub unsafe fn sadd_i32x2(x: i32x2, y: i32x2) -> i32x2 { + // CHECK: %{{[0-9]+}} = call <2 x i32> @llvm.sadd.sat.v2i32(<2 x i32> %{{x|_3}}, <2 x i32> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i32x4 +#[no_mangle] +pub unsafe fn sadd_i32x4(x: i32x4, y: i32x4) -> i32x4 { + // CHECK: %{{[0-9]+}} = call <4 x i32> @llvm.sadd.sat.v4i32(<4 x i32> %{{x|_3}}, <4 x i32> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i32x8 +#[no_mangle] +pub unsafe fn sadd_i32x8(x: i32x8, y: i32x8) -> i32x8 { + // CHECK: %{{[0-9]+}} = call <8 x i32> @llvm.sadd.sat.v8i32(<8 x i32> %{{x|_3}}, <8 x i32> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i32x16 +#[no_mangle] +pub unsafe fn sadd_i32x16(x: i32x16, y: i32x16) -> i32x16 { + // CHECK: %{{[0-9]+}} = call <16 x i32> @llvm.sadd.sat.v16i32(<16 x i32> %{{x|_3}}, <16 x i32> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i64x2 +#[no_mangle] +pub unsafe fn sadd_i64x2(x: i64x2, y: i64x2) -> i64x2 { + // CHECK: %{{[0-9]+}} = call <2 x i64> @llvm.sadd.sat.v2i64(<2 x i64> %{{x|_3}}, <2 x i64> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i64x4 +#[no_mangle] +pub unsafe fn sadd_i64x4(x: i64x4, y: i64x4) -> i64x4 { + // CHECK: %{{[0-9]+}} = call <4 x i64> @llvm.sadd.sat.v4i64(<4 x i64> %{{x|_3}}, <4 x i64> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i64x8 +#[no_mangle] +pub unsafe fn sadd_i64x8(x: i64x8, y: i64x8) -> i64x8 { + // CHECK: %{{[0-9]+}} = call <8 x i64> @llvm.sadd.sat.v8i64(<8 x i64> %{{x|_3}}, <8 x i64> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i128x2 +#[no_mangle] +pub unsafe fn sadd_i128x2(x: i128x2, y: i128x2) -> i128x2 { + // CHECK: %{{[0-9]+}} = call <2 x i128> @llvm.sadd.sat.v2i128(<2 x i128> %{{x|_3}}, <2 x i128> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @sadd_i128x4 +#[no_mangle] +pub unsafe fn sadd_i128x4(x: i128x4, y: i128x4) -> i128x4 { + // CHECK: %{{[0-9]+}} = call <4 x i128> @llvm.sadd.sat.v4i128(<4 x i128> %{{x|_3}}, <4 x i128> %{{y|_4}}) + simd_saturating_add(x, y) +} + + + +// CHECK-LABEL: @uadd_u8x2 +#[no_mangle] +pub unsafe fn uadd_u8x2(x: u8x2, y: u8x2) -> u8x2 { + // CHECK: %{{[0-9]+}} = call <2 x i8> @llvm.uadd.sat.v2i8(<2 x i8> %{{x|_3}}, <2 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u8x4 +#[no_mangle] +pub unsafe fn uadd_u8x4(x: u8x4, y: u8x4) -> u8x4 { + // CHECK: %{{[0-9]+}} = call <4 x i8> @llvm.uadd.sat.v4i8(<4 x i8> %{{x|_3}}, <4 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u8x8 +#[no_mangle] +pub unsafe fn uadd_u8x8(x: u8x8, y: u8x8) -> u8x8 { + // CHECK: %{{[0-9]+}} = call <8 x i8> @llvm.uadd.sat.v8i8(<8 x i8> %{{x|_3}}, <8 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u8x16 +#[no_mangle] +pub unsafe fn uadd_u8x16(x: u8x16, y: u8x16) -> u8x16 { + // CHECK: %{{[0-9]+}} = call <16 x i8> @llvm.uadd.sat.v16i8(<16 x i8> %{{x|_3}}, <16 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u8x32 +#[no_mangle] +pub unsafe fn uadd_u8x32(x: u8x32, y: u8x32) -> u8x32 { + // CHECK: %{{[0-9]+}} = call <32 x i8> @llvm.uadd.sat.v32i8(<32 x i8> %{{x|_3}}, <32 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u8x64 +#[no_mangle] +pub unsafe fn uadd_u8x64(x: u8x64, y: u8x64) -> u8x64 { + // CHECK: %{{[0-9]+}} = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %{{x|_3}}, <64 x i8> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u16x2 +#[no_mangle] +pub unsafe fn uadd_u16x2(x: u16x2, y: u16x2) -> u16x2 { + // CHECK: %{{[0-9]+}} = call <2 x i16> @llvm.uadd.sat.v2i16(<2 x i16> %{{x|_3}}, <2 x i16> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u16x4 +#[no_mangle] +pub unsafe fn uadd_u16x4(x: u16x4, y: u16x4) -> u16x4 { + // CHECK: %{{[0-9]+}} = call <4 x i16> @llvm.uadd.sat.v4i16(<4 x i16> %{{x|_3}}, <4 x i16> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u16x8 +#[no_mangle] +pub unsafe fn uadd_u16x8(x: u16x8, y: u16x8) -> u16x8 { + // CHECK: %{{[0-9]+}} = call <8 x i16> @llvm.uadd.sat.v8i16(<8 x i16> %{{x|_3}}, <8 x i16> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u16x16 +#[no_mangle] +pub unsafe fn uadd_u16x16(x: u16x16, y: u16x16) -> u16x16 { + // CHECK: %{{[0-9]+}} = call <16 x i16> @llvm.uadd.sat.v16i16(<16 x i16> %{{x|_3}}, <16 x i16> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u16x32 +#[no_mangle] +pub unsafe fn uadd_u16x32(x: u16x32, y: u16x32) -> u16x32 { + // CHECK: %{{[0-9]+}} = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %{{x|_3}}, <32 x i16> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u32x2 +#[no_mangle] +pub unsafe fn uadd_u32x2(x: u32x2, y: u32x2) -> u32x2 { + // CHECK: %{{[0-9]+}} = call <2 x i32> @llvm.uadd.sat.v2i32(<2 x i32> %{{x|_3}}, <2 x i32> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u32x4 +#[no_mangle] +pub unsafe fn uadd_u32x4(x: u32x4, y: u32x4) -> u32x4 { + // CHECK: %{{[0-9]+}} = call <4 x i32> @llvm.uadd.sat.v4i32(<4 x i32> %{{x|_3}}, <4 x i32> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u32x8 +#[no_mangle] +pub unsafe fn uadd_u32x8(x: u32x8, y: u32x8) -> u32x8 { + // CHECK: %{{[0-9]+}} = call <8 x i32> @llvm.uadd.sat.v8i32(<8 x i32> %{{x|_3}}, <8 x i32> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u32x16 +#[no_mangle] +pub unsafe fn uadd_u32x16(x: u32x16, y: u32x16) -> u32x16 { + // CHECK: %{{[0-9]+}} = call <16 x i32> @llvm.uadd.sat.v16i32(<16 x i32> %{{x|_3}}, <16 x i32> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u64x2 +#[no_mangle] +pub unsafe fn uadd_u64x2(x: u64x2, y: u64x2) -> u64x2 { + // CHECK: %{{[0-9]+}} = call <2 x i64> @llvm.uadd.sat.v2i64(<2 x i64> %{{x|_3}}, <2 x i64> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u64x4 +#[no_mangle] +pub unsafe fn uadd_u64x4(x: u64x4, y: u64x4) -> u64x4 { + // CHECK: %{{[0-9]+}} = call <4 x i64> @llvm.uadd.sat.v4i64(<4 x i64> %{{x|_3}}, <4 x i64> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u64x8 +#[no_mangle] +pub unsafe fn uadd_u64x8(x: u64x8, y: u64x8) -> u64x8 { + // CHECK: %{{[0-9]+}} = call <8 x i64> @llvm.uadd.sat.v8i64(<8 x i64> %{{x|_3}}, <8 x i64> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u128x2 +#[no_mangle] +pub unsafe fn uadd_u128x2(x: u128x2, y: u128x2) -> u128x2 { + // CHECK: %{{[0-9]+}} = call <2 x i128> @llvm.uadd.sat.v2i128(<2 x i128> %{{x|_3}}, <2 x i128> %{{y|_4}}) + simd_saturating_add(x, y) +} + +// CHECK-LABEL: @uadd_u128x4 +#[no_mangle] +pub unsafe fn uadd_u128x4(x: u128x4, y: u128x4) -> u128x4 { + // CHECK: %{{[0-9]+}} = call <4 x i128> @llvm.uadd.sat.v4i128(<4 x i128> %{{x|_3}}, <4 x i128> %{{y|_4}}) + simd_saturating_add(x, y) +} + + + + + +// CHECK-LABEL: @ssub_i8x2 +#[no_mangle] +pub unsafe fn ssub_i8x2(x: i8x2, y: i8x2) -> i8x2 { + // CHECK: %{{[0-9]+}} = call <2 x i8> @llvm.ssub.sat.v2i8(<2 x i8> %{{x|_3}}, <2 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i8x4 +#[no_mangle] +pub unsafe fn ssub_i8x4(x: i8x4, y: i8x4) -> i8x4 { + // CHECK: %{{[0-9]+}} = call <4 x i8> @llvm.ssub.sat.v4i8(<4 x i8> %{{x|_3}}, <4 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i8x8 +#[no_mangle] +pub unsafe fn ssub_i8x8(x: i8x8, y: i8x8) -> i8x8 { + // CHECK: %{{[0-9]+}} = call <8 x i8> @llvm.ssub.sat.v8i8(<8 x i8> %{{x|_3}}, <8 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i8x16 +#[no_mangle] +pub unsafe fn ssub_i8x16(x: i8x16, y: i8x16) -> i8x16 { + // CHECK: %{{[0-9]+}} = call <16 x i8> @llvm.ssub.sat.v16i8(<16 x i8> %{{x|_3}}, <16 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i8x32 +#[no_mangle] +pub unsafe fn ssub_i8x32(x: i8x32, y: i8x32) -> i8x32 { + // CHECK: %{{[0-9]+}} = call <32 x i8> @llvm.ssub.sat.v32i8(<32 x i8> %{{x|_3}}, <32 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i8x64 +#[no_mangle] +pub unsafe fn ssub_i8x64(x: i8x64, y: i8x64) -> i8x64 { + // CHECK: %{{[0-9]+}} = call <64 x i8> @llvm.ssub.sat.v64i8(<64 x i8> %{{x|_3}}, <64 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i16x2 +#[no_mangle] +pub unsafe fn ssub_i16x2(x: i16x2, y: i16x2) -> i16x2 { + // CHECK: %{{[0-9]+}} = call <2 x i16> @llvm.ssub.sat.v2i16(<2 x i16> %{{x|_3}}, <2 x i16> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i16x4 +#[no_mangle] +pub unsafe fn ssub_i16x4(x: i16x4, y: i16x4) -> i16x4 { + // CHECK: %{{[0-9]+}} = call <4 x i16> @llvm.ssub.sat.v4i16(<4 x i16> %{{x|_3}}, <4 x i16> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i16x8 +#[no_mangle] +pub unsafe fn ssub_i16x8(x: i16x8, y: i16x8) -> i16x8 { + // CHECK: %{{[0-9]+}} = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> %{{x|_3}}, <8 x i16> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i16x16 +#[no_mangle] +pub unsafe fn ssub_i16x16(x: i16x16, y: i16x16) -> i16x16 { + // CHECK: %{{[0-9]+}} = call <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16> %{{x|_3}}, <16 x i16> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i16x32 +#[no_mangle] +pub unsafe fn ssub_i16x32(x: i16x32, y: i16x32) -> i16x32 { + // CHECK: %{{[0-9]+}} = call <32 x i16> @llvm.ssub.sat.v32i16(<32 x i16> %{{x|_3}}, <32 x i16> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i32x2 +#[no_mangle] +pub unsafe fn ssub_i32x2(x: i32x2, y: i32x2) -> i32x2 { + // CHECK: %{{[0-9]+}} = call <2 x i32> @llvm.ssub.sat.v2i32(<2 x i32> %{{x|_3}}, <2 x i32> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i32x4 +#[no_mangle] +pub unsafe fn ssub_i32x4(x: i32x4, y: i32x4) -> i32x4 { + // CHECK: %{{[0-9]+}} = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> %{{x|_3}}, <4 x i32> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i32x8 +#[no_mangle] +pub unsafe fn ssub_i32x8(x: i32x8, y: i32x8) -> i32x8 { + // CHECK: %{{[0-9]+}} = call <8 x i32> @llvm.ssub.sat.v8i32(<8 x i32> %{{x|_3}}, <8 x i32> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i32x16 +#[no_mangle] +pub unsafe fn ssub_i32x16(x: i32x16, y: i32x16) -> i32x16 { + // CHECK: %{{[0-9]+}} = call <16 x i32> @llvm.ssub.sat.v16i32(<16 x i32> %{{x|_3}}, <16 x i32> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i64x2 +#[no_mangle] +pub unsafe fn ssub_i64x2(x: i64x2, y: i64x2) -> i64x2 { + // CHECK: %{{[0-9]+}} = call <2 x i64> @llvm.ssub.sat.v2i64(<2 x i64> %{{x|_3}}, <2 x i64> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i64x4 +#[no_mangle] +pub unsafe fn ssub_i64x4(x: i64x4, y: i64x4) -> i64x4 { + // CHECK: %{{[0-9]+}} = call <4 x i64> @llvm.ssub.sat.v4i64(<4 x i64> %{{x|_3}}, <4 x i64> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i64x8 +#[no_mangle] +pub unsafe fn ssub_i64x8(x: i64x8, y: i64x8) -> i64x8 { + // CHECK: %{{[0-9]+}} = call <8 x i64> @llvm.ssub.sat.v8i64(<8 x i64> %{{x|_3}}, <8 x i64> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i128x2 +#[no_mangle] +pub unsafe fn ssub_i128x2(x: i128x2, y: i128x2) -> i128x2 { + // CHECK: %{{[0-9]+}} = call <2 x i128> @llvm.ssub.sat.v2i128(<2 x i128> %{{x|_3}}, <2 x i128> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @ssub_i128x4 +#[no_mangle] +pub unsafe fn ssub_i128x4(x: i128x4, y: i128x4) -> i128x4 { + // CHECK: %{{[0-9]+}} = call <4 x i128> @llvm.ssub.sat.v4i128(<4 x i128> %{{x|_3}}, <4 x i128> %{{y|_4}}) + simd_saturating_sub(x, y) +} + + + +// CHECK-LABEL: @usub_u8x2 +#[no_mangle] +pub unsafe fn usub_u8x2(x: u8x2, y: u8x2) -> u8x2 { + // CHECK: %{{[0-9]+}} = call <2 x i8> @llvm.usub.sat.v2i8(<2 x i8> %{{x|_3}}, <2 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u8x4 +#[no_mangle] +pub unsafe fn usub_u8x4(x: u8x4, y: u8x4) -> u8x4 { + // CHECK: %{{[0-9]+}} = call <4 x i8> @llvm.usub.sat.v4i8(<4 x i8> %{{x|_3}}, <4 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u8x8 +#[no_mangle] +pub unsafe fn usub_u8x8(x: u8x8, y: u8x8) -> u8x8 { + // CHECK: %{{[0-9]+}} = call <8 x i8> @llvm.usub.sat.v8i8(<8 x i8> %{{x|_3}}, <8 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u8x16 +#[no_mangle] +pub unsafe fn usub_u8x16(x: u8x16, y: u8x16) -> u8x16 { + // CHECK: %{{[0-9]+}} = call <16 x i8> @llvm.usub.sat.v16i8(<16 x i8> %{{x|_3}}, <16 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u8x32 +#[no_mangle] +pub unsafe fn usub_u8x32(x: u8x32, y: u8x32) -> u8x32 { + // CHECK: %{{[0-9]+}} = call <32 x i8> @llvm.usub.sat.v32i8(<32 x i8> %{{x|_3}}, <32 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u8x64 +#[no_mangle] +pub unsafe fn usub_u8x64(x: u8x64, y: u8x64) -> u8x64 { + // CHECK: %{{[0-9]+}} = call <64 x i8> @llvm.usub.sat.v64i8(<64 x i8> %{{x|_3}}, <64 x i8> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u16x2 +#[no_mangle] +pub unsafe fn usub_u16x2(x: u16x2, y: u16x2) -> u16x2 { + // CHECK: %{{[0-9]+}} = call <2 x i16> @llvm.usub.sat.v2i16(<2 x i16> %{{x|_3}}, <2 x i16> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u16x4 +#[no_mangle] +pub unsafe fn usub_u16x4(x: u16x4, y: u16x4) -> u16x4 { + // CHECK: %{{[0-9]+}} = call <4 x i16> @llvm.usub.sat.v4i16(<4 x i16> %{{x|_3}}, <4 x i16> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u16x8 +#[no_mangle] +pub unsafe fn usub_u16x8(x: u16x8, y: u16x8) -> u16x8 { + // CHECK: %{{[0-9]+}} = call <8 x i16> @llvm.usub.sat.v8i16(<8 x i16> %{{x|_3}}, <8 x i16> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u16x16 +#[no_mangle] +pub unsafe fn usub_u16x16(x: u16x16, y: u16x16) -> u16x16 { + // CHECK: %{{[0-9]+}} = call <16 x i16> @llvm.usub.sat.v16i16(<16 x i16> %{{x|_3}}, <16 x i16> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u16x32 +#[no_mangle] +pub unsafe fn usub_u16x32(x: u16x32, y: u16x32) -> u16x32 { + // CHECK: %{{[0-9]+}} = call <32 x i16> @llvm.usub.sat.v32i16(<32 x i16> %{{x|_3}}, <32 x i16> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u32x2 +#[no_mangle] +pub unsafe fn usub_u32x2(x: u32x2, y: u32x2) -> u32x2 { + // CHECK: %{{[0-9]+}} = call <2 x i32> @llvm.usub.sat.v2i32(<2 x i32> %{{x|_3}}, <2 x i32> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u32x4 +#[no_mangle] +pub unsafe fn usub_u32x4(x: u32x4, y: u32x4) -> u32x4 { + // CHECK: %{{[0-9]+}} = call <4 x i32> @llvm.usub.sat.v4i32(<4 x i32> %{{x|_3}}, <4 x i32> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u32x8 +#[no_mangle] +pub unsafe fn usub_u32x8(x: u32x8, y: u32x8) -> u32x8 { + // CHECK: %{{[0-9]+}} = call <8 x i32> @llvm.usub.sat.v8i32(<8 x i32> %{{x|_3}}, <8 x i32> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u32x16 +#[no_mangle] +pub unsafe fn usub_u32x16(x: u32x16, y: u32x16) -> u32x16 { + // CHECK: %{{[0-9]+}} = call <16 x i32> @llvm.usub.sat.v16i32(<16 x i32> %{{x|_3}}, <16 x i32> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u64x2 +#[no_mangle] +pub unsafe fn usub_u64x2(x: u64x2, y: u64x2) -> u64x2 { + // CHECK: %{{[0-9]+}} = call <2 x i64> @llvm.usub.sat.v2i64(<2 x i64> %{{x|_3}}, <2 x i64> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u64x4 +#[no_mangle] +pub unsafe fn usub_u64x4(x: u64x4, y: u64x4) -> u64x4 { + // CHECK: %{{[0-9]+}} = call <4 x i64> @llvm.usub.sat.v4i64(<4 x i64> %{{x|_3}}, <4 x i64> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u64x8 +#[no_mangle] +pub unsafe fn usub_u64x8(x: u64x8, y: u64x8) -> u64x8 { + // CHECK: %{{[0-9]+}} = call <8 x i64> @llvm.usub.sat.v8i64(<8 x i64> %{{x|_3}}, <8 x i64> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u128x2 +#[no_mangle] +pub unsafe fn usub_u128x2(x: u128x2, y: u128x2) -> u128x2 { + // CHECK: %{{[0-9]+}} = call <2 x i128> @llvm.usub.sat.v2i128(<2 x i128> %{{x|_3}}, <2 x i128> %{{y|_4}}) + simd_saturating_sub(x, y) +} + +// CHECK-LABEL: @usub_u128x4 +#[no_mangle] +pub unsafe fn usub_u128x4(x: u128x4, y: u128x4) -> u128x4 { + // CHECK: %{{[0-9]+}} = call <4 x i128> @llvm.usub.sat.v4i128(<4 x i128> %{{x|_3}}, <4 x i128> %{{y|_4}}) + simd_saturating_sub(x, y) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-generic-bitmask.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-bitmask.rs new file mode 100644 index 000000000..4a98d797b --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-bitmask.rs @@ -0,0 +1,63 @@ +// compile-flags: -C no-prepopulate-passes +// + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone)] +pub struct u32x2(u32, u32); + +#[repr(simd)] +#[derive(Copy, Clone)] +pub struct i32x2(i32, i32); + +#[repr(simd)] +#[derive(Copy, Clone)] +pub struct i8x16( + i8, i8, i8, i8, i8, i8, i8, i8, + i8, i8, i8, i8, i8, i8, i8, i8, +); + + +extern "platform-intrinsic" { + fn simd_bitmask<T, U>(x: T) -> U; +} + +// NOTE(eddyb) `%{{x|_2}}` is used because on some targets (e.g. WASM) +// SIMD vectors are passed directly, resulting in `%x` being a vector, +// while on others they're passed indirectly, resulting in `%x` being +// a pointer to a vector, and `%_2` a vector loaded from that pointer. +// This is controlled by the target spec option `simd_types_indirect`. + +// CHECK-LABEL: @bitmask_int +#[no_mangle] +pub unsafe fn bitmask_int(x: i32x2) -> u8 { + // CHECK: [[A:%[0-9]+]] = lshr <2 x i32> %{{x|_2}}, <i32 31, i32 31> + // CHECK: [[B:%[0-9]+]] = trunc <2 x i32> [[A]] to <2 x i1> + // CHECK: [[C:%[0-9]+]] = bitcast <2 x i1> [[B]] to i2 + // CHECK: %{{[0-9]+}} = zext i2 [[C]] to i8 + simd_bitmask(x) +} + +// CHECK-LABEL: @bitmask_uint +#[no_mangle] +pub unsafe fn bitmask_uint(x: u32x2) -> u8 { + // CHECK: [[A:%[0-9]+]] = lshr <2 x i32> %{{x|_2}}, <i32 31, i32 31> + // CHECK: [[B:%[0-9]+]] = trunc <2 x i32> [[A]] to <2 x i1> + // CHECK: [[C:%[0-9]+]] = bitcast <2 x i1> [[B]] to i2 + // CHECK: %{{[0-9]+}} = zext i2 [[C]] to i8 + simd_bitmask(x) +} + +// CHECK-LABEL: @bitmask_int16 +#[no_mangle] +pub unsafe fn bitmask_int16(x: i8x16) -> u16 { + // CHECK: [[A:%[0-9]+]] = lshr <16 x i8> %{{x|_2}}, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7> + // CHECK: [[B:%[0-9]+]] = trunc <16 x i8> [[A]] to <16 x i1> + // CHECK: %{{[0-9]+}} = bitcast <16 x i1> [[B]] to i16 + // CHECK-NOT: zext + simd_bitmask(x) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-generic-extract-insert.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-extract-insert.rs new file mode 100644 index 000000000..b5b0b1330 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-extract-insert.rs @@ -0,0 +1,47 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics, min_const_generics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone)] +pub struct M(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone)] +pub struct S<const N: usize>([f32; N]); + +extern "platform-intrinsic" { + fn simd_extract<T, U>(x: T, idx: u32) -> U; + fn simd_insert<T, U>(x: T, idx: u32, b: U) -> T; +} + +// CHECK-LABEL: @extract_m +#[no_mangle] +pub unsafe fn extract_m(v: M, i: u32) -> f32 { + // CHECK: extractelement <4 x float> %{{v|_3}}, i32 %i + simd_extract(v, i) +} + +// CHECK-LABEL: @extract_s +#[no_mangle] +pub unsafe fn extract_s(v: S<4>, i: u32) -> f32 { + // CHECK: extractelement <4 x float> %{{v|_3}}, i32 %i + simd_extract(v, i) +} + +// CHECK-LABEL: @insert_m +#[no_mangle] +pub unsafe fn insert_m(v: M, i: u32, j: f32) -> M { + // CHECK: insertelement <4 x float> %{{v|_4}}, float %j, i32 %i + simd_insert(v, i, j) +} + +// CHECK-LABEL: @insert_s +#[no_mangle] +pub unsafe fn insert_s(v: S<4>, i: u32, j: f32) -> S<4> { + // CHECK: insertelement <4 x float> %{{v|_4}}, float %j, i32 %i + simd_insert(v, i, j) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-generic-gather.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-gather.rs new file mode 100644 index 000000000..cacc32f2f --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-gather.rs @@ -0,0 +1,36 @@ +// + +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct Vec2<T>(pub T, pub T); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct Vec4<T>(pub T, pub T, pub T, pub T); + +extern "platform-intrinsic" { + fn simd_gather<T, P, M>(value: T, pointers: P, mask: M) -> T; +} + +// CHECK-LABEL: @gather_f32x2 +#[no_mangle] +pub unsafe fn gather_f32x2(pointers: Vec2<*const f32>, mask: Vec2<i32>, + values: Vec2<f32>) -> Vec2<f32> { + // CHECK: call <2 x float> @llvm.masked.gather.v2f32.{{.+}}(<2 x {{float\*|ptr}}> {{.*}}, i32 {{.*}}, <2 x i1> {{.*}}, <2 x float> {{.*}}) + simd_gather(values, pointers, mask) +} + +// CHECK-LABEL: @gather_pf32x2 +#[no_mangle] +pub unsafe fn gather_pf32x2(pointers: Vec2<*const *const f32>, mask: Vec2<i32>, + values: Vec2<*const f32>) -> Vec2<*const f32> { + // CHECK: call <2 x {{float\*|ptr}}> @llvm.masked.gather.{{.+}}(<2 x {{float\*\*|ptr}}> {{.*}}, i32 {{.*}}, <2 x i1> {{.*}}, <2 x {{float\*|ptr}}> {{.*}}) + simd_gather(values, pointers, mask) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-generic-scatter.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-scatter.rs new file mode 100644 index 000000000..94ecaf609 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-scatter.rs @@ -0,0 +1,37 @@ +// + +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#![allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct Vec2<T>(pub T, pub T); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct Vec4<T>(pub T, pub T, pub T, pub T); + +extern "platform-intrinsic" { + fn simd_scatter<T, P, M>(value: T, pointers: P, mask: M); +} + +// CHECK-LABEL: @scatter_f32x2 +#[no_mangle] +pub unsafe fn scatter_f32x2(pointers: Vec2<*mut f32>, mask: Vec2<i32>, + values: Vec2<f32>) { + // CHECK: call void @llvm.masked.scatter.v2f32.v2p0{{.*}}(<2 x float> {{.*}}, <2 x {{float\*|ptr}}> {{.*}}, i32 {{.*}}, <2 x i1> {{.*}}) + simd_scatter(values, pointers, mask) +} + + +// CHECK-LABEL: @scatter_pf32x2 +#[no_mangle] +pub unsafe fn scatter_pf32x2(pointers: Vec2<*mut *const f32>, mask: Vec2<i32>, + values: Vec2<*const f32>) { + // CHECK: call void @llvm.masked.scatter.v2p0{{.*}}.v2p0{{.*}}(<2 x {{float\*|ptr}}> {{.*}}, <2 x {{float\*\*|ptr}}> {{.*}}, i32 {{.*}}, <2 x i1> {{.*}}) + simd_scatter(values, pointers, mask) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-generic-select.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-select.rs new file mode 100644 index 000000000..03bb22655 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-generic-select.rs @@ -0,0 +1,37 @@ +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![feature(repr_simd, platform_intrinsics)] +#[allow(non_camel_case_types)] + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x4(pub f32, pub f32, pub f32, pub f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct f32x8(f32, f32, f32, f32, f32, f32, f32, f32); + +#[repr(simd)] +#[derive(Copy, Clone, PartialEq, Debug)] +pub struct b8x4(pub i8, pub i8, pub i8, pub i8); + +extern "platform-intrinsic" { + fn simd_select<T, U>(x: T, a: U, b: U) -> U; + fn simd_select_bitmask<T, U>(x: T, a: U, b: U) -> U; +} + +// CHECK-LABEL: @select +#[no_mangle] +pub unsafe fn select(m: b8x4, a: f32x4, b: f32x4) -> f32x4 { + // CHECK: select <4 x i1> + simd_select(m, a, b) +} + +// CHECK-LABEL: @select_bitmask +#[no_mangle] +pub unsafe fn select_bitmask(m: i8, a: f32x8, b: f32x8) -> f32x8 { + // CHECK: select <8 x i1> + simd_select_bitmask(m, a, b) +} diff --git a/tests/codegen/simd-intrinsic/simd-intrinsic-transmute-array.rs b/tests/codegen/simd-intrinsic/simd-intrinsic-transmute-array.rs new file mode 100644 index 000000000..db5b60567 --- /dev/null +++ b/tests/codegen/simd-intrinsic/simd-intrinsic-transmute-array.rs @@ -0,0 +1,43 @@ +// +// compile-flags: -C no-prepopulate-passes + +#![crate_type = "lib"] + +#![allow(non_camel_case_types)] +#![feature(repr_simd, platform_intrinsics, min_const_generics)] + +#[repr(simd)] +#[derive(Copy, Clone)] +pub struct S<const N: usize>([f32; N]); + +#[repr(simd)] +#[derive(Copy, Clone)] +pub struct T([f32; 4]); + +#[repr(simd)] +#[derive(Copy, Clone)] +pub struct U(f32, f32, f32, f32); + +// CHECK-LABEL: @build_array_s +#[no_mangle] +pub fn build_array_s(x: [f32; 4]) -> S<4> { + // CHECK: call void @llvm.memcpy.{{.+}}({{.*}}, i{{[0-9]+}} 16, i1 false) + // CHECK: call void @llvm.memcpy.{{.+}}({{.*}}, i{{[0-9]+}} 16, i1 false) + S::<4>(x) +} + +// CHECK-LABEL: @build_array_t +#[no_mangle] +pub fn build_array_t(x: [f32; 4]) -> T { + // CHECK: call void @llvm.memcpy.{{.+}}({{.*}}, i{{[0-9]+}} 16, i1 false) + // CHECK: call void @llvm.memcpy.{{.+}}({{.*}}, i{{[0-9]+}} 16, i1 false) + T(x) +} + +// CHECK-LABEL: @build_array_u +#[no_mangle] +pub fn build_array_u(x: [f32; 4]) -> U { + // CHECK: call void @llvm.memcpy.{{.+}}({{.*}}, i{{[0-9]+}} 16, i1 false) + // CHECK: call void @llvm.memcpy.{{.+}}({{.*}}, i{{[0-9]+}} 16, i1 false) + unsafe { std::mem::transmute(x) } +} |