Merging upstream version 1.68.2+dfsg1.

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

36 files changed, 2525 insertions, 0 deletions

diff --git a/tests/ui/simd/intrinsic/float-math-pass.rs b/tests/ui/simd/intrinsic/float-math-pass.rs
new file mode 100644
index 000000000..7a4f74665
--- /dev/null
+++ b/tests/ui/simd/intrinsic/float-math-pass.rs
@@ -0,0 +1,114 @@
+// run-pass
+// ignore-emscripten
+// ignore-android
+
+// FIXME: this test fails on arm-android because the NDK version 14 is too old.
+// It needs at least version 18. We disable it on all android build bots because
+// there is no way in compile-test to disable it for an (arch,os) pair.
+
+// Test that the simd floating-point math intrinsics produce correct results.
+
+#![feature(repr_simd, platform_intrinsics)]
+#![allow(non_camel_case_types)]
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+struct f32x4(pub f32, pub f32, pub f32, pub f32);
+
+extern "platform-intrinsic" {
+    fn simd_fsqrt<T>(x: T) -> T;
+    fn simd_fabs<T>(x: T) -> T;
+    fn simd_fsin<T>(x: T) -> T;
+    fn simd_fcos<T>(x: T) -> T;
+    fn simd_fexp<T>(x: T) -> T;
+    fn simd_fexp2<T>(x: T) -> T;
+    fn simd_fma<T>(x: T, y: T, z: T) -> T;
+    fn simd_flog<T>(x: T) -> T;
+    fn simd_flog10<T>(x: T) -> T;
+    fn simd_flog2<T>(x: T) -> T;
+    fn simd_fpow<T>(x: T, y: T) -> T;
+    fn simd_fpowi<T>(x: T, y: i32) -> T;
+
+    // rounding functions
+    fn simd_ceil<T>(x: T) -> T;
+    fn simd_floor<T>(x: T) -> T;
+    fn simd_round<T>(x: T) -> T;
+    fn simd_trunc<T>(x: T) -> T;
+}
+
+macro_rules! assert_approx_eq_f32 {
+    ($a:expr, $b:expr) => ({
+        let (a, b) = (&$a, &$b);
+        assert!((*a - *b).abs() < 1.0e-6,
+                "{} is not approximately equal to {}", *a, *b);
+    })
+}
+macro_rules! assert_approx_eq {
+    ($a:expr, $b:expr) => ({
+        let a = $a;
+        let b = $b;
+        assert_approx_eq_f32!(a.0, b.0);
+        assert_approx_eq_f32!(a.1, b.1);
+        assert_approx_eq_f32!(a.2, b.2);
+        assert_approx_eq_f32!(a.3, b.3);
+    })
+}
+
+fn main() {
+    let x = f32x4(1.0, 1.0, 1.0, 1.0);
+    let y = f32x4(-1.0, -1.0, -1.0, -1.0);
+    let z = f32x4(0.0, 0.0, 0.0, 0.0);
+
+    let h = f32x4(0.5, 0.5, 0.5, 0.5);
+
+    unsafe {
+        let r = simd_fabs(y);
+        assert_approx_eq!(x, r);
+
+        let r = simd_fcos(z);
+        assert_approx_eq!(x, r);
+
+        let r = simd_fexp(z);
+        assert_approx_eq!(x, r);
+
+        let r = simd_fexp2(z);
+        assert_approx_eq!(x, r);
+
+        let r = simd_fma(x, h, h);
+        assert_approx_eq!(x, r);
+
+        let r = simd_fsqrt(x);
+        assert_approx_eq!(x, r);
+
+        let r = simd_flog(x);
+        assert_approx_eq!(z, r);
+
+        let r = simd_flog2(x);
+        assert_approx_eq!(z, r);
+
+        let r = simd_flog10(x);
+        assert_approx_eq!(z, r);
+
+        let r = simd_fpow(h, x);
+        assert_approx_eq!(h, r);
+
+        let r = simd_fpowi(h, 1);
+        assert_approx_eq!(h, r);
+
+        let r = simd_fsin(z);
+        assert_approx_eq!(z, r);
+
+        // rounding functions
+        let r = simd_floor(h);
+        assert_eq!(z, r);
+
+        let r = simd_ceil(h);
+        assert_eq!(x, r);
+
+        let r = simd_round(h);
+        assert_eq!(x, r);
+
+        let r = simd_trunc(h);
+        assert_eq!(z, r);
+    }
+}
diff --git a/tests/ui/simd/intrinsic/float-minmax-pass.rs b/tests/ui/simd/intrinsic/float-minmax-pass.rs
new file mode 100644
index 000000000..d79be61f9
--- /dev/null
+++ b/tests/ui/simd/intrinsic/float-minmax-pass.rs
@@ -0,0 +1,52 @@
+// run-pass
+// ignore-emscripten
+
+// Test that the simd_f{min,max} intrinsics produce the correct results.
+
+#![feature(repr_simd, platform_intrinsics)]
+#![allow(non_camel_case_types)]
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+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;
+}
+
+fn main() {
+    let x = f32x4(1.0, 2.0, 3.0, 4.0);
+    let y = f32x4(2.0, 1.0, 4.0, 3.0);
+
+    #[cfg(not(any(target_arch = "mips", target_arch = "mips64")))]
+    let nan = f32::NAN;
+    // MIPS hardware treats f32::NAN as SNAN. Clear the signaling bit.
+    // See https://github.com/rust-lang/rust/issues/52746.
+    #[cfg(any(target_arch = "mips", target_arch = "mips64"))]
+    let nan = f32::from_bits(f32::NAN.to_bits() - 1);
+
+    let n = f32x4(nan, nan, nan, nan);
+
+    unsafe {
+        let min0 = simd_fmin(x, y);
+        let min1 = simd_fmin(y, x);
+        assert_eq!(min0, min1);
+        let e = f32x4(1.0, 1.0, 3.0, 3.0);
+        assert_eq!(min0, e);
+        let minn = simd_fmin(x, n);
+        assert_eq!(minn, x);
+        let minn = simd_fmin(y, n);
+        assert_eq!(minn, y);
+
+        let max0 = simd_fmax(x, y);
+        let max1 = simd_fmax(y, x);
+        assert_eq!(max0, max1);
+        let e = f32x4(2.0, 2.0, 4.0, 4.0);
+        assert_eq!(max0, e);
+        let maxn = simd_fmax(x, n);
+        assert_eq!(maxn, x);
+        let maxn = simd_fmax(y, n);
+        assert_eq!(maxn, y);
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-arithmetic-2.rs b/tests/ui/simd/intrinsic/generic-arithmetic-2.rs
new file mode 100644
index 000000000..3576eed71
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-arithmetic-2.rs
@@ -0,0 +1,103 @@
+// build-fail
+
+#![feature(repr_simd, platform_intrinsics)]
+#![allow(non_camel_case_types)]
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct i32x4(pub i32, pub i32, pub i32, pub i32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct u32x4(pub u32, pub u32, pub u32, pub u32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
+
+extern "platform-intrinsic" {
+    fn simd_add<T>(x: T, y: T) -> T;
+    fn simd_sub<T>(x: T, y: T) -> T;
+    fn simd_mul<T>(x: T, y: T) -> T;
+    fn simd_div<T>(x: T, y: T) -> T;
+    fn simd_rem<T>(x: T, y: T) -> T;
+    fn simd_shl<T>(x: T, y: T) -> T;
+    fn simd_shr<T>(x: T, y: T) -> T;
+    fn simd_and<T>(x: T, y: T) -> T;
+    fn simd_or<T>(x: T, y: T) -> T;
+    fn simd_xor<T>(x: T, y: T) -> T;
+
+    fn simd_neg<T>(x: T) -> T;
+}
+
+fn main() {
+    let x = i32x4(0, 0, 0, 0);
+    let y = u32x4(0, 0, 0, 0);
+    let z = f32x4(0.0, 0.0, 0.0, 0.0);
+
+    unsafe {
+        simd_add(x, x);
+        simd_add(y, y);
+        simd_add(z, z);
+        simd_sub(x, x);
+        simd_sub(y, y);
+        simd_sub(z, z);
+        simd_mul(x, x);
+        simd_mul(y, y);
+        simd_mul(z, z);
+        simd_div(x, x);
+        simd_div(y, y);
+        simd_div(z, z);
+        simd_rem(x, x);
+        simd_rem(y, y);
+        simd_rem(z, z);
+
+        simd_shl(x, x);
+        simd_shl(y, y);
+        simd_shr(x, x);
+        simd_shr(y, y);
+        simd_and(x, x);
+        simd_and(y, y);
+        simd_or(x, x);
+        simd_or(y, y);
+        simd_xor(x, x);
+        simd_xor(y, y);
+
+        simd_neg(x);
+        simd_neg(z);
+
+
+        simd_add(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_sub(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_mul(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_div(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_shl(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_shr(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_and(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_or(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_xor(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+
+        simd_neg(0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+
+
+        simd_shl(z, z);
+//~^ ERROR unsupported operation on `f32x4` with element `f32`
+        simd_shr(z, z);
+//~^ ERROR unsupported operation on `f32x4` with element `f32`
+        simd_and(z, z);
+//~^ ERROR unsupported operation on `f32x4` with element `f32`
+        simd_or(z, z);
+//~^ ERROR unsupported operation on `f32x4` with element `f32`
+        simd_xor(z, z);
+//~^ ERROR unsupported operation on `f32x4` with element `f32`
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-arithmetic-2.stderr b/tests/ui/simd/intrinsic/generic-arithmetic-2.stderr
new file mode 100644
index 000000000..0f0a7ea66
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-arithmetic-2.stderr
@@ -0,0 +1,93 @@
+error[E0511]: invalid monomorphization of `simd_add` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-arithmetic-2.rs:69:9
+   |
+LL |         simd_add(0, 0);
+   |         ^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_sub` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-arithmetic-2.rs:71:9
+   |
+LL |         simd_sub(0, 0);
+   |         ^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_mul` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-arithmetic-2.rs:73:9
+   |
+LL |         simd_mul(0, 0);
+   |         ^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_div` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-arithmetic-2.rs:75:9
+   |
+LL |         simd_div(0, 0);
+   |         ^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shl` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-arithmetic-2.rs:77:9
+   |
+LL |         simd_shl(0, 0);
+   |         ^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shr` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-arithmetic-2.rs:79:9
+   |
+LL |         simd_shr(0, 0);
+   |         ^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_and` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-arithmetic-2.rs:81:9
+   |
+LL |         simd_and(0, 0);
+   |         ^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_or` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-arithmetic-2.rs:83:9
+   |
+LL |         simd_or(0, 0);
+   |         ^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_xor` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-arithmetic-2.rs:85:9
+   |
+LL |         simd_xor(0, 0);
+   |         ^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_neg` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-arithmetic-2.rs:88:9
+   |
+LL |         simd_neg(0);
+   |         ^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shl` intrinsic: unsupported operation on `f32x4` with element `f32`
+  --> $DIR/generic-arithmetic-2.rs:92:9
+   |
+LL |         simd_shl(z, z);
+   |         ^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shr` intrinsic: unsupported operation on `f32x4` with element `f32`
+  --> $DIR/generic-arithmetic-2.rs:94:9
+   |
+LL |         simd_shr(z, z);
+   |         ^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_and` intrinsic: unsupported operation on `f32x4` with element `f32`
+  --> $DIR/generic-arithmetic-2.rs:96:9
+   |
+LL |         simd_and(z, z);
+   |         ^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_or` intrinsic: unsupported operation on `f32x4` with element `f32`
+  --> $DIR/generic-arithmetic-2.rs:98:9
+   |
+LL |         simd_or(z, z);
+   |         ^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_xor` intrinsic: unsupported operation on `f32x4` with element `f32`
+  --> $DIR/generic-arithmetic-2.rs:100:9
+   |
+LL |         simd_xor(z, z);
+   |         ^^^^^^^^^^^^^^
+
+error: aborting due to 15 previous errors
+
+For more information about this error, try `rustc --explain E0511`.
diff --git a/tests/ui/simd/intrinsic/generic-arithmetic-pass.rs b/tests/ui/simd/intrinsic/generic-arithmetic-pass.rs
new file mode 100644
index 000000000..c507b8d31
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-arithmetic-pass.rs
@@ -0,0 +1,136 @@
+// run-pass
+#![allow(non_camel_case_types)]
+
+// ignore-emscripten FIXME(#45351) hits an LLVM assert
+
+#![feature(repr_simd, platform_intrinsics)]
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct i32x4(pub i32, pub i32, pub i32, pub i32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct U32<const N: usize>([u32; N]);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct f32x4(pub f32, pub f32, pub f32, pub f32);
+
+macro_rules! all_eq {
+    ($a: expr, $b: expr) => {{
+        let a = $a;
+        let b = $b;
+        assert!(a.0 == b.0 && a.1 == b.1 && a.2 == b.2 && a.3 == b.3);
+    }}
+}
+
+macro_rules! all_eq_ {
+    ($a: expr, $b: expr) => {{
+        let a = $a;
+        let b = $b;
+        assert!(a.0 == b.0);
+    }}
+}
+
+
+extern "platform-intrinsic" {
+    fn simd_add<T>(x: T, y: T) -> T;
+    fn simd_sub<T>(x: T, y: T) -> T;
+    fn simd_mul<T>(x: T, y: T) -> T;
+    fn simd_div<T>(x: T, y: T) -> T;
+    fn simd_rem<T>(x: T, y: T) -> T;
+    fn simd_shl<T>(x: T, y: T) -> T;
+    fn simd_shr<T>(x: T, y: T) -> T;
+    fn simd_and<T>(x: T, y: T) -> T;
+    fn simd_or<T>(x: T, y: T) -> T;
+    fn simd_xor<T>(x: T, y: T) -> T;
+
+    fn simd_neg<T>(x: T) -> T;
+}
+
+fn main() {
+    let x1 = i32x4(1, 2, 3, 4);
+    let y1 = U32::<4>([1, 2, 3, 4]);
+    let z1 = f32x4(1.0, 2.0, 3.0, 4.0);
+    let x2 = i32x4(2, 3, 4, 5);
+    let y2 = U32::<4>([2, 3, 4, 5]);
+    let z2 = f32x4(2.0, 3.0, 4.0, 5.0);
+
+    unsafe {
+        all_eq!(simd_add(x1, x2), i32x4(3, 5, 7, 9));
+        all_eq!(simd_add(x2, x1), i32x4(3, 5, 7, 9));
+        all_eq_!(simd_add(y1, y2), U32::<4>([3, 5, 7, 9]));
+        all_eq_!(simd_add(y2, y1), U32::<4>([3, 5, 7, 9]));
+        all_eq!(simd_add(z1, z2), f32x4(3.0, 5.0, 7.0, 9.0));
+        all_eq!(simd_add(z2, z1), f32x4(3.0, 5.0, 7.0, 9.0));
+
+        all_eq!(simd_mul(x1, x2), i32x4(2, 6, 12, 20));
+        all_eq!(simd_mul(x2, x1), i32x4(2, 6, 12, 20));
+        all_eq_!(simd_mul(y1, y2), U32::<4>([2, 6, 12, 20]));
+        all_eq_!(simd_mul(y2, y1), U32::<4>([2, 6, 12, 20]));
+        all_eq!(simd_mul(z1, z2), f32x4(2.0, 6.0, 12.0, 20.0));
+        all_eq!(simd_mul(z2, z1), f32x4(2.0, 6.0, 12.0, 20.0));
+
+        all_eq!(simd_sub(x2, x1), i32x4(1, 1, 1, 1));
+        all_eq!(simd_sub(x1, x2), i32x4(-1, -1, -1, -1));
+        all_eq_!(simd_sub(y2, y1), U32::<4>([1, 1, 1, 1]));
+        all_eq_!(simd_sub(y1, y2), U32::<4>([!0, !0, !0, !0]));
+        all_eq!(simd_sub(z2, z1), f32x4(1.0, 1.0, 1.0, 1.0));
+        all_eq!(simd_sub(z1, z2), f32x4(-1.0, -1.0, -1.0, -1.0));
+
+        all_eq!(simd_div(x1, x1), i32x4(1, 1, 1, 1));
+        all_eq!(simd_div(i32x4(2, 4, 6, 8), i32x4(2, 2, 2, 2)), x1);
+        all_eq_!(simd_div(y1, y1), U32::<4>([1, 1, 1, 1]));
+        all_eq_!(simd_div(U32::<4>([2, 4, 6, 8]), U32::<4>([2, 2, 2, 2])), y1);
+        all_eq!(simd_div(z1, z1), f32x4(1.0, 1.0, 1.0, 1.0));
+        all_eq!(simd_div(z1, z2), f32x4(1.0/2.0, 2.0/3.0, 3.0/4.0, 4.0/5.0));
+        all_eq!(simd_div(z2, z1), f32x4(2.0/1.0, 3.0/2.0, 4.0/3.0, 5.0/4.0));
+
+        all_eq!(simd_rem(x1, x1), i32x4(0, 0, 0, 0));
+        all_eq!(simd_rem(x2, x1), i32x4(0, 1, 1, 1));
+        all_eq_!(simd_rem(y1, y1), U32::<4>([0, 0, 0, 0]));
+        all_eq_!(simd_rem(y2, y1), U32::<4>([0, 1, 1, 1]));
+        all_eq!(simd_rem(z1, z1), f32x4(0.0, 0.0, 0.0, 0.0));
+        all_eq!(simd_rem(z1, z2), z1);
+        all_eq!(simd_rem(z2, z1), f32x4(0.0, 1.0, 1.0, 1.0));
+
+        all_eq!(simd_shl(x1, x2), i32x4(1 << 2, 2 << 3, 3 << 4, 4 << 5));
+        all_eq!(simd_shl(x2, x1), i32x4(2 << 1, 3 << 2, 4 << 3, 5 << 4));
+        all_eq_!(simd_shl(y1, y2), U32::<4>([1 << 2, 2 << 3, 3 << 4, 4 << 5]));
+        all_eq_!(simd_shl(y2, y1), U32::<4>([2 << 1, 3 << 2, 4 << 3, 5 << 4]));
+
+        // test right-shift by assuming left-shift is correct
+        all_eq!(simd_shr(simd_shl(x1, x2), x2), x1);
+        all_eq!(simd_shr(simd_shl(x2, x1), x1), x2);
+        all_eq_!(simd_shr(simd_shl(y1, y2), y2), y1);
+        all_eq_!(simd_shr(simd_shl(y2, y1), y1), y2);
+
+        // ensure we get logical vs. arithmetic shifts correct
+        let (a, b, c, d) = (-12, -123, -1234, -12345);
+        all_eq!(simd_shr(i32x4(a, b, c, d), x1), i32x4(a >> 1, b >> 2, c >> 3, d >> 4));
+        all_eq_!(simd_shr(U32::<4>([a as u32, b as u32, c as u32, d as u32]), y1),
+                U32::<4>([(a as u32) >> 1, (b as u32) >> 2, (c as u32) >> 3, (d as u32) >> 4]));
+
+        all_eq!(simd_and(x1, x2), i32x4(0, 2, 0, 4));
+        all_eq!(simd_and(x2, x1), i32x4(0, 2, 0, 4));
+        all_eq_!(simd_and(y1, y2), U32::<4>([0, 2, 0, 4]));
+        all_eq_!(simd_and(y2, y1), U32::<4>([0, 2, 0, 4]));
+
+        all_eq!(simd_or(x1, x2), i32x4(3, 3, 7, 5));
+        all_eq!(simd_or(x2, x1), i32x4(3, 3, 7, 5));
+        all_eq_!(simd_or(y1, y2), U32::<4>([3, 3, 7, 5]));
+        all_eq_!(simd_or(y2, y1), U32::<4>([3, 3, 7, 5]));
+
+        all_eq!(simd_xor(x1, x2), i32x4(3, 1, 7, 1));
+        all_eq!(simd_xor(x2, x1), i32x4(3, 1, 7, 1));
+        all_eq_!(simd_xor(y1, y2), U32::<4>([3, 1, 7, 1]));
+        all_eq_!(simd_xor(y2, y1), U32::<4>([3, 1, 7, 1]));
+
+        all_eq!(simd_neg(x1), i32x4(-1, -2, -3, -4));
+        all_eq!(simd_neg(x2), i32x4(-2, -3, -4, -5));
+        all_eq!(simd_neg(z1), f32x4(-1.0, -2.0, -3.0, -4.0));
+        all_eq!(simd_neg(z2), f32x4(-2.0, -3.0, -4.0, -5.0));
+
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-arithmetic-saturating-2.rs b/tests/ui/simd/intrinsic/generic-arithmetic-saturating-2.rs
new file mode 100644
index 000000000..9736d1b96
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-arithmetic-saturating-2.rs
@@ -0,0 +1,38 @@
+// build-fail
+// ignore-emscripten
+#![feature(repr_simd, platform_intrinsics)]
+#![allow(non_camel_case_types)]
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct i32x4(pub i32, pub i32, pub i32, pub i32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct x4<T>(pub T, pub T, pub T, pub T);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
+
+extern "platform-intrinsic" {
+    fn simd_saturating_add<T>(x: T, y: T) -> T;
+    fn simd_saturating_sub<T>(x: T, y: T) -> T;
+}
+
+fn main() {
+    let x = i32x4(0, 0, 0, 0);
+    let y = x4(0_usize, 0, 0, 0);
+    let z = f32x4(0.0, 0.0, 0.0, 0.0);
+
+    unsafe {
+        simd_saturating_add(x, x);
+        simd_saturating_add(y, y);
+        simd_saturating_sub(x, x);
+        simd_saturating_sub(y, y);
+
+        simd_saturating_add(z, z);
+        //~^ ERROR expected element type `f32` of vector type `f32x4` to be a signed or unsigned integer type
+        simd_saturating_sub(z, z);
+        //~^ ERROR expected element type `f32` of vector type `f32x4` to be a signed or unsigned integer type
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-arithmetic-saturating-2.stderr b/tests/ui/simd/intrinsic/generic-arithmetic-saturating-2.stderr
new file mode 100644
index 000000000..f349cb565
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-arithmetic-saturating-2.stderr
@@ -0,0 +1,15 @@
+error[E0511]: invalid monomorphization of `simd_saturating_add` intrinsic: expected element type `f32` of vector type `f32x4` to be a signed or unsigned integer type
+  --> $DIR/generic-arithmetic-saturating-2.rs:33:9
+   |
+LL |         simd_saturating_add(z, z);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_saturating_sub` intrinsic: expected element type `f32` of vector type `f32x4` to be a signed or unsigned integer type
+  --> $DIR/generic-arithmetic-saturating-2.rs:35:9
+   |
+LL |         simd_saturating_sub(z, z);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error: aborting due to 2 previous errors
+
+For more information about this error, try `rustc --explain E0511`.
diff --git a/tests/ui/simd/intrinsic/generic-arithmetic-saturating-pass.rs b/tests/ui/simd/intrinsic/generic-arithmetic-saturating-pass.rs
new file mode 100644
index 000000000..c11d14b99
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-arithmetic-saturating-pass.rs
@@ -0,0 +1,91 @@
+// run-pass
+// ignore-emscripten
+
+#![allow(non_camel_case_types)]
+#![feature(repr_simd, platform_intrinsics)]
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+struct u32x4(pub u32, pub u32, pub u32, pub u32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct I32<const N: usize>([i32; N]);
+
+extern "platform-intrinsic" {
+    fn simd_saturating_add<T>(x: T, y: T) -> T;
+    fn simd_saturating_sub<T>(x: T, y: T) -> T;
+}
+
+fn main() {
+    // unsigned
+    {
+        const M: u32 = u32::MAX;
+
+        let a = u32x4(1, 2, 3, 4);
+        let b = u32x4(2, 4, 6, 8);
+        let m = u32x4(M, M, M, M);
+        let m1 = u32x4(M - 1, M - 1, M - 1, M - 1);
+        let z = u32x4(0, 0, 0, 0);
+
+        unsafe {
+            assert_eq!(simd_saturating_add(z, z), z);
+            assert_eq!(simd_saturating_add(z, a), a);
+            assert_eq!(simd_saturating_add(b, z), b);
+            assert_eq!(simd_saturating_add(a, a), b);
+            assert_eq!(simd_saturating_add(a, m), m);
+            assert_eq!(simd_saturating_add(m, b), m);
+            assert_eq!(simd_saturating_add(m1, a), m);
+
+            assert_eq!(simd_saturating_sub(b, z), b);
+            assert_eq!(simd_saturating_sub(b, a), a);
+            assert_eq!(simd_saturating_sub(a, a), z);
+            assert_eq!(simd_saturating_sub(a, b), z);
+            assert_eq!(simd_saturating_sub(a, m1), z);
+            assert_eq!(simd_saturating_sub(b, m1), z);
+        }
+    }
+
+    // signed
+    {
+        const MIN: i32 = i32::MIN;
+        const MAX: i32 = i32::MAX;
+
+        let a = I32::<4>([1, 2, 3, 4]);
+        let b = I32::<4>([2, 4, 6, 8]);
+        let c = I32::<4>([-1, -2, -3, -4]);
+        let d = I32::<4>([-2, -4, -6, -8]);
+
+        let max = I32::<4>([MAX, MAX, MAX, MAX]);
+        let max1 = I32::<4>([MAX - 1, MAX - 1, MAX - 1, MAX - 1]);
+        let min = I32::<4>([MIN, MIN, MIN, MIN]);
+        let min1 = I32::<4>([MIN + 1, MIN + 1, MIN + 1, MIN + 1]);
+
+        let z = I32::<4>([0, 0, 0, 0]);
+
+        unsafe {
+            assert_eq!(simd_saturating_add(z, z).0, z.0);
+            assert_eq!(simd_saturating_add(z, a).0, a.0);
+            assert_eq!(simd_saturating_add(b, z).0, b.0);
+            assert_eq!(simd_saturating_add(a, a).0, b.0);
+            assert_eq!(simd_saturating_add(a, max).0, max.0);
+            assert_eq!(simd_saturating_add(max, b).0, max.0);
+            assert_eq!(simd_saturating_add(max1, a).0, max.0);
+            assert_eq!(simd_saturating_add(min1, z).0, min1.0);
+            assert_eq!(simd_saturating_add(min, z).0, min.0);
+            assert_eq!(simd_saturating_add(min1, c).0, min.0);
+            assert_eq!(simd_saturating_add(min, c).0, min.0);
+            assert_eq!(simd_saturating_add(min1, d).0, min.0);
+            assert_eq!(simd_saturating_add(min, d).0, min.0);
+
+            assert_eq!(simd_saturating_sub(b, z).0, b.0);
+            assert_eq!(simd_saturating_sub(b, a).0, a.0);
+            assert_eq!(simd_saturating_sub(a, a).0, z.0);
+            assert_eq!(simd_saturating_sub(a, b).0, c.0);
+            assert_eq!(simd_saturating_sub(z, max).0, min1.0);
+            assert_eq!(simd_saturating_sub(min1, z).0, min1.0);
+            assert_eq!(simd_saturating_sub(min1, a).0, min.0);
+            assert_eq!(simd_saturating_sub(min1, b).0, min.0);
+        }
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-as.rs b/tests/ui/simd/intrinsic/generic-as.rs
new file mode 100644
index 000000000..a975190a2
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-as.rs
@@ -0,0 +1,48 @@
+// run-pass
+
+#![feature(repr_simd, platform_intrinsics)]
+
+extern "platform-intrinsic" {
+    fn simd_as<T, U>(x: T) -> U;
+}
+
+#[derive(Copy, Clone)]
+#[repr(simd)]
+struct V<T>([T; 2]);
+
+fn main() {
+    unsafe {
+        let u = V::<u32>([u32::MIN, u32::MAX]);
+        let i: V<i16> = simd_as(u);
+        assert_eq!(i.0[0], u.0[0] as i16);
+        assert_eq!(i.0[1], u.0[1] as i16);
+    }
+
+    unsafe {
+        let f = V::<f32>([f32::MIN, f32::MAX]);
+        let i: V<i16> = simd_as(f);
+        assert_eq!(i.0[0], f.0[0] as i16);
+        assert_eq!(i.0[1], f.0[1] as i16);
+    }
+
+    unsafe {
+        let f = V::<f32>([f32::MIN, f32::MAX]);
+        let u: V<u8> = simd_as(f);
+        assert_eq!(u.0[0], f.0[0] as u8);
+        assert_eq!(u.0[1], f.0[1] as u8);
+    }
+
+    unsafe {
+        let f = V::<f64>([f64::MIN, f64::MAX]);
+        let i: V<isize> = simd_as(f);
+        assert_eq!(i.0[0], f.0[0] as isize);
+        assert_eq!(i.0[1], f.0[1] as isize);
+    }
+
+    unsafe {
+        let f = V::<f64>([f64::MIN, f64::MAX]);
+        let u: V<usize> = simd_as(f);
+        assert_eq!(u.0[0], f.0[0] as usize);
+        assert_eq!(u.0[1], f.0[1] as usize);
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-bitmask-pass.rs b/tests/ui/simd/intrinsic/generic-bitmask-pass.rs
new file mode 100644
index 000000000..8c436841b
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-bitmask-pass.rs
@@ -0,0 +1,62 @@
+// run-pass
+#![allow(non_camel_case_types)]
+
+// ignore-emscripten
+// ignore-endian-big behavior of simd_bitmask is endian-specific
+
+// Test that the simd_bitmask intrinsic produces correct results.
+
+#![feature(repr_simd, platform_intrinsics)]
+#[allow(non_camel_case_types)]
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+struct u32x4(pub u32, pub u32, pub u32, pub u32);
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+struct u8x4(pub u8, pub u8, pub u8, pub u8);
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+struct Tx4<T>(pub T, pub T, pub T, pub T);
+
+extern "platform-intrinsic" {
+    fn simd_bitmask<T, U>(x: T) -> U;
+}
+
+fn main() {
+    let z = u32x4(0, 0, 0, 0);
+    let ez = 0_u8;
+
+    let o = u32x4(!0, !0, !0, !0);
+    let eo = 0b_1111_u8;
+
+    let m0 = u32x4(!0, 0, !0, 0);
+    let e0 = 0b_0000_0101_u8;
+
+    // Check that the MSB is extracted:
+    let m = u8x4(0b_1000_0000, 0b_0100_0001, 0b_1100_0001, 0b_1111_1111);
+    let e = 0b_1101;
+
+    // Check usize / isize
+    let msize: Tx4<usize> = Tx4(usize::MAX, 0, usize::MAX, usize::MAX);
+
+    unsafe {
+        let r: u8 = simd_bitmask(z);
+        assert_eq!(r, ez);
+
+        let r: u8 = simd_bitmask(o);
+        assert_eq!(r, eo);
+
+        let r: u8 = simd_bitmask(m0);
+        assert_eq!(r, e0);
+
+        let r: u8 = simd_bitmask(m);
+        assert_eq!(r, e);
+
+        let r: u8 = simd_bitmask(msize);
+        assert_eq!(r, e);
+
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-bitmask.rs b/tests/ui/simd/intrinsic/generic-bitmask.rs
new file mode 100644
index 000000000..9a23dae77
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-bitmask.rs
@@ -0,0 +1,69 @@
+// build-fail
+
+// Test that the simd_bitmask intrinsic produces ok-ish error
+// messages when misused.
+
+#![feature(repr_simd, platform_intrinsics)]
+#![allow(non_camel_case_types)]
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct u32x2([u32; 2]);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct u32x4([u32; 4]);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct u8x8([u8; 8]);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct u8x16([u8; 16]);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct u8x32([u8; 32]);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct u8x64([u8; 64]);
+
+extern "platform-intrinsic" {
+    fn simd_bitmask<T, U>(x: T) -> U;
+}
+
+fn main() {
+    let m2 = u32x2([0; 2]);
+    let m4 = u32x4([0; 4]);
+    let m8 = u8x8([0; 8]);
+    let m16 = u8x16([0; 16]);
+    let m32 = u8x32([0; 32]);
+    let m64 = u8x64([0; 64]);
+
+    unsafe {
+        let _: u8 = simd_bitmask(m2);
+        let _: u8 = simd_bitmask(m4);
+        let _: u8 = simd_bitmask(m8);
+        let _: u16 = simd_bitmask(m16);
+        let _: u32 = simd_bitmask(m32);
+        let _: u64 = simd_bitmask(m64);
+
+        let _: u16 = simd_bitmask(m2);
+        //~^ ERROR invalid monomorphization of `simd_bitmask` intrinsic
+
+        let _: u16 = simd_bitmask(m8);
+        //~^ ERROR invalid monomorphization of `simd_bitmask` intrinsic
+
+        let _: u32 = simd_bitmask(m16);
+        //~^ ERROR invalid monomorphization of `simd_bitmask` intrinsic
+
+        let _: u64 = simd_bitmask(m32);
+        //~^ ERROR invalid monomorphization of `simd_bitmask` intrinsic
+
+        let _: u128 = simd_bitmask(m64);
+        //~^ ERROR invalid monomorphization of `simd_bitmask` intrinsic
+
+   }
+}
diff --git a/tests/ui/simd/intrinsic/generic-bitmask.stderr b/tests/ui/simd/intrinsic/generic-bitmask.stderr
new file mode 100644
index 000000000..0de3f8eea
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-bitmask.stderr
@@ -0,0 +1,33 @@
+error[E0511]: invalid monomorphization of `simd_bitmask` intrinsic: cannot return `u16`, expected `u8` or `[u8; 1]`
+  --> $DIR/generic-bitmask.rs:53:22
+   |
+LL |         let _: u16 = simd_bitmask(m2);
+   |                      ^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_bitmask` intrinsic: cannot return `u16`, expected `u8` or `[u8; 1]`
+  --> $DIR/generic-bitmask.rs:56:22
+   |
+LL |         let _: u16 = simd_bitmask(m8);
+   |                      ^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_bitmask` intrinsic: cannot return `u32`, expected `u16` or `[u8; 2]`
+  --> $DIR/generic-bitmask.rs:59:22
+   |
+LL |         let _: u32 = simd_bitmask(m16);
+   |                      ^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_bitmask` intrinsic: cannot return `u64`, expected `u32` or `[u8; 4]`
+  --> $DIR/generic-bitmask.rs:62:22
+   |
+LL |         let _: u64 = simd_bitmask(m32);
+   |                      ^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_bitmask` intrinsic: cannot return `u128`, expected `u64` or `[u8; 8]`
+  --> $DIR/generic-bitmask.rs:65:23
+   |
+LL |         let _: u128 = simd_bitmask(m64);
+   |                       ^^^^^^^^^^^^^^^^^
+
+error: aborting due to 5 previous errors
+
+For more information about this error, try `rustc --explain E0511`.
diff --git a/tests/ui/simd/intrinsic/generic-cast-pass.rs b/tests/ui/simd/intrinsic/generic-cast-pass.rs
new file mode 100644
index 000000000..15f232e2c
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-cast-pass.rs
@@ -0,0 +1,121 @@
+// run-pass
+#![allow(unused_must_use)]
+// ignore-emscripten FIXME(#45351) hits an LLVM assert
+
+#![feature(repr_simd, platform_intrinsics, concat_idents, test)]
+#![allow(non_camel_case_types)]
+
+extern crate test;
+
+#[repr(simd)]
+#[derive(PartialEq, Debug)]
+struct i32x4(i32, i32, i32, i32);
+#[repr(simd)]
+#[derive(PartialEq, Debug)]
+struct i8x4(i8, i8, i8, i8);
+
+#[repr(simd)]
+#[derive(PartialEq, Debug)]
+struct u32x4(u32, u32, u32, u32);
+#[repr(simd)]
+#[derive(PartialEq, Debug)]
+struct u8x4(u8, u8, u8, u8);
+
+#[repr(simd)]
+#[derive(PartialEq, Debug)]
+struct f32x4(f32, f32, f32, f32);
+
+#[repr(simd)]
+#[derive(PartialEq, Debug)]
+struct f64x4(f64, f64, f64, f64);
+
+
+extern "platform-intrinsic" {
+    fn simd_cast<T, U>(x: T) -> U;
+}
+
+const A: i32 = -1234567;
+const B: i32 = 12345678;
+const C: i32 = -123456789;
+const D: i32 = 1234567890;
+
+trait Foo {
+    fn is_float() -> bool { false }
+    fn in_range(x: i32) -> bool;
+}
+impl Foo for i32 {
+    fn in_range(_: i32) -> bool { true }
+}
+impl Foo for i8 {
+    fn in_range(x: i32) -> bool { -128 <= x && x < 128 }
+}
+impl Foo for u32 {
+    fn in_range(x: i32) -> bool { 0 <= x }
+}
+impl Foo for u8 {
+    fn in_range(x: i32) -> bool { 0 <= x && x < 128 }
+}
+impl Foo for f32 {
+    fn is_float() -> bool { true }
+    fn in_range(_: i32) -> bool { true }
+}
+impl Foo for f64 {
+    fn is_float() -> bool { true }
+    fn in_range(_: i32) -> bool { true }
+}
+
+fn main() {
+    macro_rules! test {
+        ($from: ident, $to: ident) => {{
+            // force the casts to actually happen, or else LLVM/rustc
+            // may fold them and get slightly different results.
+            let (a, b, c, d) = test::black_box((A as $from, B as $from, C as $from, D as $from));
+            // the SIMD vectors are all FOOx4, so we can concat_idents
+            // so we don't have to pass in the extra args to the macro
+            let mut from = simd_cast(concat_idents!($from, x4)(a, b, c, d));
+            let mut to = concat_idents!($to, x4)(a as $to,
+                                                 b as $to,
+                                                 c as $to,
+                                                 d as $to);
+            // assist type inference, it needs to know what `from` is
+            // for the `if` statements.
+            to == from;
+
+            // there are platform differences for some out of range
+            // casts, so we just normalize such things: it's OK for
+            // "invalid" calculations to result in nonsense answers.
+            // (e.g., negative float to unsigned integer goes through a
+            // library routine on the default i686 platforms, and the
+            // implementation of that routine differs on e.g., Linux
+            // vs. macOS, resulting in different answers.)
+            if $from::is_float() {
+                if !$to::in_range(A) { from.0 = 0 as $to; to.0 = 0 as $to; }
+                if !$to::in_range(B) { from.1 = 0 as $to; to.1 = 0 as $to; }
+                if !$to::in_range(C) { from.2 = 0 as $to; to.2 = 0 as $to; }
+                if !$to::in_range(D) { from.3 = 0 as $to; to.3 = 0 as $to; }
+            }
+
+            assert!(to == from,
+                    "{} -> {} ({:?} != {:?})", stringify!($from), stringify!($to),
+                    from, to);
+        }}
+    }
+    macro_rules! tests {
+        (: $($to: ident),*) => { () };
+        // repeating the list twice is easier than writing a cartesian
+        // product macro
+        ($from: ident $(, $from_: ident)*: $($to: ident),*) => {
+            fn $from() { unsafe { $( test!($from, $to); )* } }
+            tests!($($from_),*: $($to),*)
+        };
+        ($($types: ident),*) => {{
+            tests!($($types),* : $($types),*);
+            $($types();)*
+        }}
+    }
+
+    // test various combinations, including truncation,
+    // signed/unsigned extension, and floating point casts.
+    tests!(i32, i8, u32, u8, f32);
+    tests!(i32, u32, f32, f64)
+}
diff --git a/tests/ui/simd/intrinsic/generic-cast-pointer-width.rs b/tests/ui/simd/intrinsic/generic-cast-pointer-width.rs
new file mode 100644
index 000000000..b9382310d
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-cast-pointer-width.rs
@@ -0,0 +1,21 @@
+// run-pass
+#![feature(repr_simd, platform_intrinsics)]
+
+extern "platform-intrinsic" {
+    fn simd_cast<T, U>(x: T) -> U;
+}
+
+#[derive(Copy, Clone)]
+#[repr(simd)]
+struct V<T>([T; 4]);
+
+fn main() {
+    let u = V::<usize>([0, 1, 2, 3]);
+    let uu32: V<u32> = unsafe { simd_cast(u) };
+    let ui64: V<i64> = unsafe { simd_cast(u) };
+
+    for (u, (uu32, ui64)) in u.0.iter().zip(uu32.0.iter().zip(ui64.0.iter())) {
+        assert_eq!(*u as u32, *uu32);
+        assert_eq!(*u as i64, *ui64);
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-cast.rs b/tests/ui/simd/intrinsic/generic-cast.rs
new file mode 100644
index 000000000..4f4fa06b0
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-cast.rs
@@ -0,0 +1,43 @@
+// build-fail
+
+#![feature(repr_simd, platform_intrinsics)]
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct i32x4(i32, i32, i32, i32);
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct i32x8(i32, i32, i32, i32,
+             i32, i32, i32, i32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct f32x4(f32, f32, f32, f32);
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct f32x8(f32, f32, f32, f32,
+             f32, f32, f32, f32);
+
+
+extern "platform-intrinsic" {
+    fn simd_cast<T, U>(x: T) -> U;
+}
+
+fn main() {
+    let x = i32x4(0, 0, 0, 0);
+
+    unsafe {
+        simd_cast::<i32, i32>(0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_cast::<i32, i32x4>(0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_cast::<i32x4, i32>(x);
+        //~^ ERROR expected SIMD return type, found non-SIMD `i32`
+        simd_cast::<_, i32x8>(x);
+//~^ ERROR return type with length 4 (same as input type `i32x4`), found `i32x8` with length 8
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-cast.stderr b/tests/ui/simd/intrinsic/generic-cast.stderr
new file mode 100644
index 000000000..2226bbbe1
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-cast.stderr
@@ -0,0 +1,27 @@
+error[E0511]: invalid monomorphization of `simd_cast` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-cast.rs:34:9
+   |
+LL |         simd_cast::<i32, i32>(0);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_cast` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-cast.rs:36:9
+   |
+LL |         simd_cast::<i32, i32x4>(0);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_cast` intrinsic: expected SIMD return type, found non-SIMD `i32`
+  --> $DIR/generic-cast.rs:38:9
+   |
+LL |         simd_cast::<i32x4, i32>(x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_cast` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i32x8` with length 8
+  --> $DIR/generic-cast.rs:40:9
+   |
+LL |         simd_cast::<_, i32x8>(x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^
+
+error: aborting due to 4 previous errors
+
+For more information about this error, try `rustc --explain E0511`.
diff --git a/tests/ui/simd/intrinsic/generic-comparison-pass.rs b/tests/ui/simd/intrinsic/generic-comparison-pass.rs
new file mode 100644
index 000000000..da5c42a1a
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-comparison-pass.rs
@@ -0,0 +1,106 @@
+// run-pass
+// ignore-emscripten FIXME(#45351) hits an LLVM assert
+// revisions: mir thir
+// [thir]compile-flags: -Zthir-unsafeck
+
+#![feature(repr_simd, platform_intrinsics, concat_idents)]
+#![allow(non_camel_case_types)]
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct i32x4(i32, i32, i32, i32);
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct u32x4(pub u32, pub u32, pub u32, pub u32);
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct f32x4(pub f32, pub f32, pub f32, pub f32);
+
+extern "platform-intrinsic" {
+    fn simd_eq<T, U>(x: T, y: T) -> U;
+    fn simd_ne<T, U>(x: T, y: T) -> U;
+    fn simd_lt<T, U>(x: T, y: T) -> U;
+    fn simd_le<T, U>(x: T, y: T) -> U;
+    fn simd_gt<T, U>(x: T, y: T) -> U;
+    fn simd_ge<T, U>(x: T, y: T) -> U;
+}
+
+macro_rules! cmp {
+    ($method: ident($lhs: expr, $rhs: expr)) => {{
+        let lhs = $lhs;
+        let rhs = $rhs;
+        let e: u32x4 = concat_idents!(simd_, $method)($lhs, $rhs);
+        // assume the scalar version is correct/the behaviour we want.
+        assert!((e.0 != 0) == lhs.0 .$method(&rhs.0));
+        assert!((e.1 != 0) == lhs.1 .$method(&rhs.1));
+        assert!((e.2 != 0) == lhs.2 .$method(&rhs.2));
+        assert!((e.3 != 0) == lhs.3 .$method(&rhs.3));
+    }}
+}
+macro_rules! tests {
+    ($($lhs: ident, $rhs: ident;)*) => {{
+        $(
+            (|| {
+                cmp!(eq($lhs, $rhs));
+                cmp!(ne($lhs, $rhs));
+
+                // test both directions
+                cmp!(lt($lhs, $rhs));
+                cmp!(lt($rhs, $lhs));
+
+                cmp!(le($lhs, $rhs));
+                cmp!(le($rhs, $lhs));
+
+                cmp!(gt($lhs, $rhs));
+                cmp!(gt($rhs, $lhs));
+
+                cmp!(ge($lhs, $rhs));
+                cmp!(ge($rhs, $lhs));
+            })();
+            )*
+    }}
+}
+fn main() {
+    // 13 vs. -100 tests that we get signed vs. unsigned comparisons
+    // correct (i32: 13 > -100, u32: 13 < -100).    let i1 = i32x4(10, -11, 12, 13);
+    let i1 = i32x4(10, -11, 12, 13);
+    let i2 = i32x4(5, -5, 20, -100);
+    let i3 = i32x4(10, -11, 20, -100);
+
+    let u1 = u32x4(10, !11+1, 12, 13);
+    let u2 = u32x4(5, !5+1, 20, !100+1);
+    let u3 = u32x4(10, !11+1, 20, !100+1);
+
+    let f1 = f32x4(10.0, -11.0, 12.0, 13.0);
+    let f2 = f32x4(5.0, -5.0, 20.0, -100.0);
+    let f3 = f32x4(10.0, -11.0, 20.0, -100.0);
+
+    unsafe {
+        tests! {
+            i1, i1;
+            u1, u1;
+            f1, f1;
+
+            i1, i2;
+            u1, u2;
+            f1, f2;
+
+            i1, i3;
+            u1, u3;
+            f1, f3;
+        }
+    }
+
+    // NAN comparisons are special:
+    // -11 (*)    13
+    // -5        -100 (*)
+    let f4 = f32x4(f32::NAN, f1.1, f32::NAN, f2.3);
+
+    unsafe {
+        tests! {
+            f1, f4;
+            f2, f4;
+            f4, f4;
+        }
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-comparison.rs b/tests/ui/simd/intrinsic/generic-comparison.rs
new file mode 100644
index 000000000..3cd38042f
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-comparison.rs
@@ -0,0 +1,67 @@
+// build-fail
+
+#![feature(repr_simd, platform_intrinsics)]
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct i32x4(i32, i32, i32, i32);
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct i16x8(i16, i16, i16, i16,
+             i16, i16, i16, i16);
+
+extern "platform-intrinsic" {
+    fn simd_eq<T, U>(x: T, y: T) -> U;
+    fn simd_ne<T, U>(x: T, y: T) -> U;
+    fn simd_lt<T, U>(x: T, y: T) -> U;
+    fn simd_le<T, U>(x: T, y: T) -> U;
+    fn simd_gt<T, U>(x: T, y: T) -> U;
+    fn simd_ge<T, U>(x: T, y: T) -> U;
+}
+
+fn main() {
+    let x = i32x4(0, 0, 0, 0);
+
+    unsafe {
+        simd_eq::<i32, i32>(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_ne::<i32, i32>(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_lt::<i32, i32>(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_le::<i32, i32>(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_gt::<i32, i32>(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_ge::<i32, i32>(0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+
+        simd_eq::<_, i32>(x, x);
+        //~^ ERROR expected SIMD return type, found non-SIMD `i32`
+        simd_ne::<_, i32>(x, x);
+        //~^ ERROR expected SIMD return type, found non-SIMD `i32`
+        simd_lt::<_, i32>(x, x);
+        //~^ ERROR expected SIMD return type, found non-SIMD `i32`
+        simd_le::<_, i32>(x, x);
+        //~^ ERROR expected SIMD return type, found non-SIMD `i32`
+        simd_gt::<_, i32>(x, x);
+        //~^ ERROR expected SIMD return type, found non-SIMD `i32`
+        simd_ge::<_, i32>(x, x);
+        //~^ ERROR expected SIMD return type, found non-SIMD `i32`
+
+        simd_eq::<_, i16x8>(x, x);
+//~^ ERROR return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+        simd_ne::<_, i16x8>(x, x);
+//~^ ERROR return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+        simd_lt::<_, i16x8>(x, x);
+//~^ ERROR return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+        simd_le::<_, i16x8>(x, x);
+//~^ ERROR return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+        simd_gt::<_, i16x8>(x, x);
+//~^ ERROR return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+        simd_ge::<_, i16x8>(x, x);
+//~^ ERROR return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-comparison.stderr b/tests/ui/simd/intrinsic/generic-comparison.stderr
new file mode 100644
index 000000000..0eae2688b
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-comparison.stderr
@@ -0,0 +1,111 @@
+error[E0511]: invalid monomorphization of `simd_eq` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:28:9
+   |
+LL |         simd_eq::<i32, i32>(0, 0);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_ne` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:30:9
+   |
+LL |         simd_ne::<i32, i32>(0, 0);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_lt` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:32:9
+   |
+LL |         simd_lt::<i32, i32>(0, 0);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_le` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:34:9
+   |
+LL |         simd_le::<i32, i32>(0, 0);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_gt` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:36:9
+   |
+LL |         simd_gt::<i32, i32>(0, 0);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_ge` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:38:9
+   |
+LL |         simd_ge::<i32, i32>(0, 0);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_eq` intrinsic: expected SIMD return type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:41:9
+   |
+LL |         simd_eq::<_, i32>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_ne` intrinsic: expected SIMD return type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:43:9
+   |
+LL |         simd_ne::<_, i32>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_lt` intrinsic: expected SIMD return type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:45:9
+   |
+LL |         simd_lt::<_, i32>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_le` intrinsic: expected SIMD return type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:47:9
+   |
+LL |         simd_le::<_, i32>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_gt` intrinsic: expected SIMD return type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:49:9
+   |
+LL |         simd_gt::<_, i32>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_ge` intrinsic: expected SIMD return type, found non-SIMD `i32`
+  --> $DIR/generic-comparison.rs:51:9
+   |
+LL |         simd_ge::<_, i32>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_eq` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+  --> $DIR/generic-comparison.rs:54:9
+   |
+LL |         simd_eq::<_, i16x8>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_ne` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+  --> $DIR/generic-comparison.rs:56:9
+   |
+LL |         simd_ne::<_, i16x8>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_lt` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+  --> $DIR/generic-comparison.rs:58:9
+   |
+LL |         simd_lt::<_, i16x8>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_le` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+  --> $DIR/generic-comparison.rs:60:9
+   |
+LL |         simd_le::<_, i16x8>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_gt` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+  --> $DIR/generic-comparison.rs:62:9
+   |
+LL |         simd_gt::<_, i16x8>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_ge` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
+  --> $DIR/generic-comparison.rs:64:9
+   |
+LL |         simd_ge::<_, i16x8>(x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error: aborting due to 18 previous errors
+
+For more information about this error, try `rustc --explain E0511`.
diff --git a/tests/ui/simd/intrinsic/generic-elements-pass.rs b/tests/ui/simd/intrinsic/generic-elements-pass.rs
new file mode 100644
index 000000000..3c913c0ad
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-elements-pass.rs
@@ -0,0 +1,102 @@
+// run-pass
+// ignore-emscripten FIXME(#45351) hits an LLVM assert
+
+#![feature(repr_simd, platform_intrinsics)]
+#![feature(inline_const)]
+
+#[repr(simd)]
+#[derive(Copy, Clone, Debug, PartialEq)]
+#[allow(non_camel_case_types)]
+struct i32x2(i32, i32);
+#[repr(simd)]
+#[derive(Copy, Clone, Debug, PartialEq)]
+#[allow(non_camel_case_types)]
+struct i32x4(i32, i32, i32, i32);
+#[repr(simd)]
+#[derive(Copy, Clone, Debug, PartialEq)]
+#[allow(non_camel_case_types)]
+struct i32x8(i32, i32, i32, i32,
+             i32, i32, i32, i32);
+
+extern "platform-intrinsic" {
+    fn simd_insert<T, E>(x: T, idx: u32, y: E) -> T;
+    fn simd_extract<T, E>(x: T, idx: u32) -> E;
+
+    fn simd_shuffle2<T, U>(x: T, y: T, idx: [u32; 2]) -> U;
+    fn simd_shuffle4<T, U>(x: T, y: T, idx: [u32; 4]) -> U;
+    fn simd_shuffle8<T, U>(x: T, y: T, idx: [u32; 8]) -> U;
+}
+
+macro_rules! all_eq {
+    ($a: expr, $b: expr) => {{
+        let a = $a;
+        let b = $b;
+        // type inference works better with the concrete type on the
+        // left, but humans work better with the expected on the
+        // right.
+        assert!(b == a,
+                "{:?} != {:?}", a, b);
+    }}
+}
+
+fn main() {
+    let x2 = i32x2(20, 21);
+    let x4 = i32x4(40, 41, 42, 43);
+    let x8 = i32x8(80, 81, 82, 83, 84, 85, 86, 87);
+    unsafe {
+        all_eq!(simd_insert(x2, 0, 100), i32x2(100, 21));
+        all_eq!(simd_insert(x2, 1, 100), i32x2(20, 100));
+
+        all_eq!(simd_insert(x4, 0, 100), i32x4(100, 41, 42, 43));
+        all_eq!(simd_insert(x4, 1, 100), i32x4(40, 100, 42, 43));
+        all_eq!(simd_insert(x4, 2, 100), i32x4(40, 41, 100, 43));
+        all_eq!(simd_insert(x4, 3, 100), i32x4(40, 41, 42, 100));
+
+        all_eq!(simd_insert(x8, 0, 100), i32x8(100, 81, 82, 83, 84, 85, 86, 87));
+        all_eq!(simd_insert(x8, 1, 100), i32x8(80, 100, 82, 83, 84, 85, 86, 87));
+        all_eq!(simd_insert(x8, 2, 100), i32x8(80, 81, 100, 83, 84, 85, 86, 87));
+        all_eq!(simd_insert(x8, 3, 100), i32x8(80, 81, 82, 100, 84, 85, 86, 87));
+        all_eq!(simd_insert(x8, 4, 100), i32x8(80, 81, 82, 83, 100, 85, 86, 87));
+        all_eq!(simd_insert(x8, 5, 100), i32x8(80, 81, 82, 83, 84, 100, 86, 87));
+        all_eq!(simd_insert(x8, 6, 100), i32x8(80, 81, 82, 83, 84, 85, 100, 87));
+        all_eq!(simd_insert(x8, 7, 100), i32x8(80, 81, 82, 83, 84, 85, 86, 100));
+
+        all_eq!(simd_extract(x2, 0), 20);
+        all_eq!(simd_extract(x2, 1), 21);
+
+        all_eq!(simd_extract(x4, 0), 40);
+        all_eq!(simd_extract(x4, 1), 41);
+        all_eq!(simd_extract(x4, 2), 42);
+        all_eq!(simd_extract(x4, 3), 43);
+
+        all_eq!(simd_extract(x8, 0), 80);
+        all_eq!(simd_extract(x8, 1), 81);
+        all_eq!(simd_extract(x8, 2), 82);
+        all_eq!(simd_extract(x8, 3), 83);
+        all_eq!(simd_extract(x8, 4), 84);
+        all_eq!(simd_extract(x8, 5), 85);
+        all_eq!(simd_extract(x8, 6), 86);
+        all_eq!(simd_extract(x8, 7), 87);
+    }
+
+    let y2 = i32x2(120, 121);
+    let y4 = i32x4(140, 141, 142, 143);
+    let y8 = i32x8(180, 181, 182, 183, 184, 185, 186, 187);
+    unsafe {
+        all_eq!(simd_shuffle2(x2, y2, const { [3u32, 0] }), i32x2(121, 20));
+        all_eq!(simd_shuffle4(x2, y2, const { [3u32, 0, 1, 2] }), i32x4(121, 20, 21, 120));
+        all_eq!(simd_shuffle8(x2, y2, const { [3u32, 0, 1, 2, 1, 2, 3, 0] }),
+                i32x8(121, 20, 21, 120, 21, 120, 121, 20));
+
+        all_eq!(simd_shuffle2(x4, y4, const { [7u32, 2] }), i32x2(143, 42));
+        all_eq!(simd_shuffle4(x4, y4, const { [7u32, 2, 5, 0] }), i32x4(143, 42, 141, 40));
+        all_eq!(simd_shuffle8(x4, y4, const { [7u32, 2, 5, 0, 3, 6, 4, 1] }),
+                i32x8(143, 42, 141, 40, 43, 142, 140, 41));
+
+        all_eq!(simd_shuffle2(x8, y8, const { [11u32, 5] }), i32x2(183, 85));
+        all_eq!(simd_shuffle4(x8, y8, const { [11u32, 5, 15, 0] }), i32x4(183, 85, 187, 80));
+        all_eq!(simd_shuffle8(x8, y8, const { [11u32, 5, 15, 0, 3, 8, 12, 1] }),
+                i32x8(183, 85, 187, 80, 83, 180, 184, 81));
+    }
+
+}
diff --git a/tests/ui/simd/intrinsic/generic-elements.rs b/tests/ui/simd/intrinsic/generic-elements.rs
new file mode 100644
index 000000000..abde69163
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-elements.rs
@@ -0,0 +1,77 @@
+// build-fail
+
+#![feature(repr_simd, platform_intrinsics, rustc_attrs)]
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct i32x2(i32, i32);
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct i32x4(i32, i32, i32, i32);
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct i32x8(i32, i32, i32, i32,
+             i32, i32, i32, i32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct f32x2(f32, f32);
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct f32x4(f32, f32, f32, f32);
+#[repr(simd)]
+#[derive(Copy, Clone)]
+#[allow(non_camel_case_types)]
+struct f32x8(f32, f32, f32, f32,
+             f32, f32, f32, f32);
+
+extern "platform-intrinsic" {
+    fn simd_insert<T, E>(x: T, idx: u32, y: E) -> T;
+    fn simd_extract<T, E>(x: T, idx: u32) -> E;
+
+    fn simd_shuffle2<T, U>(x: T, y: T, idx: [u32; 2]) -> U;
+    fn simd_shuffle4<T, U>(x: T, y: T, idx: [u32; 4]) -> U;
+    fn simd_shuffle8<T, U>(x: T, y: T, idx: [u32; 8]) -> U;
+}
+
+fn main() {
+    let x = i32x4(0, 0, 0, 0);
+
+    unsafe {
+        simd_insert(0, 0, 0);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        simd_insert(x, 0, 1.0);
+        //~^ ERROR expected inserted type `i32` (element of input `i32x4`), found `f64`
+        simd_extract::<_, f32>(x, 0);
+        //~^ ERROR expected return type `i32` (element of input `i32x4`), found `f32`
+
+        const IDX2: [u32; 2] = [0; 2];
+        simd_shuffle2::<i32, i32>(0, 0, IDX2);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        const IDX4: [u32; 4] = [0; 4];
+        simd_shuffle4::<i32, i32>(0, 0, IDX4);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+        const IDX8: [u32; 8] = [0; 8];
+        simd_shuffle8::<i32, i32>(0, 0, IDX8);
+        //~^ ERROR expected SIMD input type, found non-SIMD `i32`
+
+        simd_shuffle2::<_, f32x2>(x, x, IDX2);
+//~^ ERROR element type `i32` (element of input `i32x4`), found `f32x2` with element type `f32`
+        simd_shuffle4::<_, f32x4>(x, x, IDX4);
+//~^ ERROR element type `i32` (element of input `i32x4`), found `f32x4` with element type `f32`
+        simd_shuffle8::<_, f32x8>(x, x, IDX8);
+//~^ ERROR element type `i32` (element of input `i32x4`), found `f32x8` with element type `f32`
+
+        simd_shuffle2::<_, i32x8>(x, x, IDX2);
+        //~^ ERROR expected return type of length 2, found `i32x8` with length 8
+        simd_shuffle4::<_, i32x8>(x, x, IDX4);
+        //~^ ERROR expected return type of length 4, found `i32x8` with length 8
+        simd_shuffle8::<_, i32x2>(x, x, IDX8);
+        //~^ ERROR expected return type of length 8, found `i32x2` with length 2
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-elements.stderr b/tests/ui/simd/intrinsic/generic-elements.stderr
new file mode 100644
index 000000000..5b423f704
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-elements.stderr
@@ -0,0 +1,75 @@
+error[E0511]: invalid monomorphization of `simd_insert` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-elements.rs:46:9
+   |
+LL |         simd_insert(0, 0, 0);
+   |         ^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_insert` intrinsic: expected inserted type `i32` (element of input `i32x4`), found `f64`
+  --> $DIR/generic-elements.rs:48:9
+   |
+LL |         simd_insert(x, 0, 1.0);
+   |         ^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_extract` intrinsic: expected return type `i32` (element of input `i32x4`), found `f32`
+  --> $DIR/generic-elements.rs:50:9
+   |
+LL |         simd_extract::<_, f32>(x, 0);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shuffle2` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-elements.rs:54:9
+   |
+LL |         simd_shuffle2::<i32, i32>(0, 0, IDX2);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shuffle4` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-elements.rs:57:9
+   |
+LL |         simd_shuffle4::<i32, i32>(0, 0, IDX4);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shuffle8` intrinsic: expected SIMD input type, found non-SIMD `i32`
+  --> $DIR/generic-elements.rs:60:9
+   |
+LL |         simd_shuffle8::<i32, i32>(0, 0, IDX8);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shuffle2` intrinsic: expected return element type `i32` (element of input `i32x4`), found `f32x2` with element type `f32`
+  --> $DIR/generic-elements.rs:63:9
+   |
+LL |         simd_shuffle2::<_, f32x2>(x, x, IDX2);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shuffle4` intrinsic: expected return element type `i32` (element of input `i32x4`), found `f32x4` with element type `f32`
+  --> $DIR/generic-elements.rs:65:9
+   |
+LL |         simd_shuffle4::<_, f32x4>(x, x, IDX4);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shuffle8` intrinsic: expected return element type `i32` (element of input `i32x4`), found `f32x8` with element type `f32`
+  --> $DIR/generic-elements.rs:67:9
+   |
+LL |         simd_shuffle8::<_, f32x8>(x, x, IDX8);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shuffle2` intrinsic: expected return type of length 2, found `i32x8` with length 8
+  --> $DIR/generic-elements.rs:70:9
+   |
+LL |         simd_shuffle2::<_, i32x8>(x, x, IDX2);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shuffle4` intrinsic: expected return type of length 4, found `i32x8` with length 8
+  --> $DIR/generic-elements.rs:72:9
+   |
+LL |         simd_shuffle4::<_, i32x8>(x, x, IDX4);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shuffle8` intrinsic: expected return type of length 8, found `i32x2` with length 2
+  --> $DIR/generic-elements.rs:74:9
+   |
+LL |         simd_shuffle8::<_, i32x2>(x, x, IDX8);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error: aborting due to 12 previous errors
+
+For more information about this error, try `rustc --explain E0511`.
diff --git a/tests/ui/simd/intrinsic/generic-gather-pass.rs b/tests/ui/simd/intrinsic/generic-gather-pass.rs
new file mode 100644
index 000000000..805caebe5
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-gather-pass.rs
@@ -0,0 +1,141 @@
+// run-pass
+// ignore-emscripten
+
+// Test that the simd_{gather,scatter} intrinsics produce the correct results.
+
+#![feature(repr_simd, platform_intrinsics)]
+#![allow(non_camel_case_types)]
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+struct x4<T>(pub T, pub T, pub T, pub T);
+
+extern "platform-intrinsic" {
+    fn simd_gather<T, U, V>(x: T, y: U, z: V) -> T;
+    fn simd_scatter<T, U, V>(x: T, y: U, z: V) -> ();
+}
+
+fn main() {
+    let mut x = [0_f32, 1., 2., 3., 4., 5., 6., 7.];
+
+    let default = x4(-3_f32, -3., -3., -3.);
+    let s_strided = x4(0_f32, 2., -3., 6.);
+    let mask = x4(-1_i32, -1, 0, -1);
+
+    // reading from *const
+    unsafe {
+        let pointer = &x[0] as *const f32;
+        let pointers =  x4(
+            pointer.offset(0) as *const f32,
+            pointer.offset(2),
+            pointer.offset(4),
+            pointer.offset(6)
+        );
+
+        let r_strided = simd_gather(default, pointers, mask);
+
+        assert_eq!(r_strided, s_strided);
+    }
+
+    // reading from *mut
+    unsafe {
+        let pointer = &mut x[0] as *mut f32;
+        let pointers = x4(
+            pointer.offset(0) as *mut f32,
+            pointer.offset(2),
+            pointer.offset(4),
+            pointer.offset(6)
+        );
+
+        let r_strided = simd_gather(default, pointers, mask);
+
+        assert_eq!(r_strided, s_strided);
+    }
+
+    // writing to *mut
+    unsafe {
+        let pointer = &mut x[0] as *mut f32;
+        let pointers = x4(
+            pointer.offset(0) as *mut f32,
+            pointer.offset(2),
+            pointer.offset(4),
+            pointer.offset(6)
+        );
+
+        let values = x4(42_f32, 43_f32, 44_f32, 45_f32);
+        simd_scatter(values, pointers, mask);
+
+        assert_eq!(x, [42., 1., 43., 3., 4., 5., 45., 7.]);
+    }
+
+    // test modifying array of *const f32
+    let mut y = [
+        &x[0] as *const f32,
+        &x[1] as *const f32,
+        &x[2] as *const f32,
+        &x[3] as *const f32,
+        &x[4] as *const f32,
+        &x[5] as *const f32,
+        &x[6] as *const f32,
+        &x[7] as *const f32
+    ];
+
+    let default = x4(y[0], y[0], y[0], y[0]);
+    let s_strided = x4(y[0], y[2], y[0], y[6]);
+
+    // reading from *const
+    unsafe {
+        let pointer = &y[0] as *const *const f32;
+        let pointers = x4(
+            pointer.offset(0) as *const *const f32,
+            pointer.offset(2),
+            pointer.offset(4),
+            pointer.offset(6)
+        );
+
+        let r_strided = simd_gather(default, pointers, mask);
+
+        assert_eq!(r_strided, s_strided);
+    }
+
+    // reading from *mut
+    unsafe {
+        let pointer = &mut y[0] as *mut *const f32;
+        let pointers = x4(
+            pointer.offset(0) as *mut *const f32,
+            pointer.offset(2),
+            pointer.offset(4),
+            pointer.offset(6)
+        );
+
+        let r_strided = simd_gather(default, pointers, mask);
+
+        assert_eq!(r_strided, s_strided);
+    }
+
+    // writing to *mut
+    unsafe {
+        let pointer = &mut y[0] as *mut *const f32;
+        let pointers = x4(
+            pointer.offset(0) as *mut *const f32,
+            pointer.offset(2),
+            pointer.offset(4),
+            pointer.offset(6)
+        );
+
+        let values = x4(y[7], y[6], y[5], y[1]);
+        simd_scatter(values, pointers, mask);
+
+        let s = [
+            &x[7] as *const f32,
+            &x[1] as *const f32,
+            &x[6] as *const f32,
+            &x[3] as *const f32,
+            &x[4] as *const f32,
+            &x[5] as *const f32,
+            &x[1] as *const f32,
+            &x[7] as *const f32
+        ];
+        assert_eq!(y, s);
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-reduction-pass.rs b/tests/ui/simd/intrinsic/generic-reduction-pass.rs
new file mode 100644
index 000000000..4a54afee8
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-reduction-pass.rs
@@ -0,0 +1,155 @@
+// run-pass
+#![allow(non_camel_case_types)]
+
+// ignore-emscripten
+
+// Test that the simd_reduce_{op} intrinsics produce the correct results.
+
+#![feature(repr_simd, platform_intrinsics)]
+#[allow(non_camel_case_types)]
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct i32x4(pub i32, pub i32, pub i32, pub i32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct u32x4(pub u32, pub u32, pub u32, pub u32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct f32x4(pub f32, pub f32, pub f32, pub f32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+struct b8x4(pub i8, pub i8, pub i8, pub i8);
+
+extern "platform-intrinsic" {
+    fn simd_reduce_add_unordered<T, U>(x: T) -> U;
+    fn simd_reduce_mul_unordered<T, U>(x: T) -> U;
+    fn simd_reduce_add_ordered<T, U>(x: T, acc: U) -> U;
+    fn simd_reduce_mul_ordered<T, U>(x: T, acc: U) -> U;
+    fn simd_reduce_min<T, U>(x: T) -> U;
+    fn simd_reduce_max<T, U>(x: T) -> U;
+    fn simd_reduce_min_nanless<T, U>(x: T) -> U;
+    fn simd_reduce_max_nanless<T, U>(x: T) -> U;
+    fn simd_reduce_and<T, U>(x: T) -> U;
+    fn simd_reduce_or<T, U>(x: T) -> U;
+    fn simd_reduce_xor<T, U>(x: T) -> U;
+    fn simd_reduce_all<T>(x: T) -> bool;
+    fn simd_reduce_any<T>(x: T) -> bool;
+}
+
+fn main() {
+    unsafe {
+        let x = i32x4(1, -2, 3, 4);
+        let r: i32 = simd_reduce_add_unordered(x);
+        assert_eq!(r, 6_i32);
+        let r: i32 = simd_reduce_mul_unordered(x);
+        assert_eq!(r, -24_i32);
+        let r: i32 = simd_reduce_add_ordered(x, -1);
+        assert_eq!(r, 5_i32);
+        let r: i32 = simd_reduce_mul_ordered(x, -1);
+        assert_eq!(r, 24_i32);
+
+        let r: i32 = simd_reduce_min(x);
+        assert_eq!(r, -2_i32);
+        let r: i32 = simd_reduce_max(x);
+        assert_eq!(r, 4_i32);
+
+        let x = i32x4(-1, -1, -1, -1);
+        let r: i32 = simd_reduce_and(x);
+        assert_eq!(r, -1_i32);
+        let r: i32 = simd_reduce_or(x);
+        assert_eq!(r, -1_i32);
+        let r: i32 = simd_reduce_xor(x);
+        assert_eq!(r, 0_i32);
+
+        let x = i32x4(-1, -1, 0, -1);
+        let r: i32 = simd_reduce_and(x);
+        assert_eq!(r, 0_i32);
+        let r: i32 = simd_reduce_or(x);
+        assert_eq!(r, -1_i32);
+        let r: i32 = simd_reduce_xor(x);
+        assert_eq!(r, -1_i32);
+    }
+
+    unsafe {
+        let x = u32x4(1, 2, 3, 4);
+        let r: u32 = simd_reduce_add_unordered(x);
+        assert_eq!(r, 10_u32);
+        let r: u32 = simd_reduce_mul_unordered(x);
+        assert_eq!(r, 24_u32);
+        let r: u32 = simd_reduce_add_ordered(x, 1);
+        assert_eq!(r, 11_u32);
+        let r: u32 = simd_reduce_mul_ordered(x, 2);
+        assert_eq!(r, 48_u32);
+
+        let r: u32 = simd_reduce_min(x);
+        assert_eq!(r, 1_u32);
+        let r: u32 = simd_reduce_max(x);
+        assert_eq!(r, 4_u32);
+
+        let t = u32::MAX;
+        let x = u32x4(t, t, t, t);
+        let r: u32 = simd_reduce_and(x);
+        assert_eq!(r, t);
+        let r: u32 = simd_reduce_or(x);
+        assert_eq!(r, t);
+        let r: u32 = simd_reduce_xor(x);
+        assert_eq!(r, 0_u32);
+
+        let x = u32x4(t, t, 0, t);
+        let r: u32 = simd_reduce_and(x);
+        assert_eq!(r, 0_u32);
+        let r: u32 = simd_reduce_or(x);
+        assert_eq!(r, t);
+        let r: u32 = simd_reduce_xor(x);
+        assert_eq!(r, t);
+    }
+
+    unsafe {
+        let x = f32x4(1., -2., 3., 4.);
+        let r: f32 = simd_reduce_add_unordered(x);
+        assert_eq!(r, 6_f32);
+        let r: f32 = simd_reduce_mul_unordered(x);
+        assert_eq!(r, -24_f32);
+        let r: f32 = simd_reduce_add_ordered(x, 0.);
+        assert_eq!(r, 6_f32);
+        let r: f32 = simd_reduce_mul_ordered(x, 1.);
+        assert_eq!(r, -24_f32);
+        let r: f32 = simd_reduce_add_ordered(x, 1.);
+        assert_eq!(r, 7_f32);
+        let r: f32 = simd_reduce_mul_ordered(x, 2.);
+        assert_eq!(r, -48_f32);
+
+        let r: f32 = simd_reduce_min(x);
+        assert_eq!(r, -2_f32);
+        let r: f32 = simd_reduce_max(x);
+        assert_eq!(r, 4_f32);
+        let r: f32 = simd_reduce_min_nanless(x);
+        assert_eq!(r, -2_f32);
+        let r: f32 = simd_reduce_max_nanless(x);
+        assert_eq!(r, 4_f32);
+    }
+
+    unsafe {
+        let x = b8x4(!0, !0, !0, !0);
+        let r: bool = simd_reduce_all(x);
+        assert_eq!(r, true);
+        let r: bool = simd_reduce_any(x);
+        assert_eq!(r, true);
+
+        let x = b8x4(!0, !0, 0, !0);
+        let r: bool = simd_reduce_all(x);
+        assert_eq!(r, false);
+        let r: bool = simd_reduce_any(x);
+        assert_eq!(r, true);
+
+        let x = b8x4(0, 0, 0, 0);
+        let r: bool = simd_reduce_all(x);
+        assert_eq!(r, false);
+        let r: bool = simd_reduce_any(x);
+        assert_eq!(r, false);
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-reduction.rs b/tests/ui/simd/intrinsic/generic-reduction.rs
new file mode 100644
index 000000000..ede4b26d1
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-reduction.rs
@@ -0,0 +1,58 @@
+// build-fail
+// ignore-emscripten
+
+// Test that the simd_reduce_{op} intrinsics produce ok-ish error
+// messages when misused.
+
+#![feature(repr_simd, platform_intrinsics)]
+#![allow(non_camel_case_types)]
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct u32x4(pub u32, pub u32, pub u32, pub u32);
+
+
+extern "platform-intrinsic" {
+    fn simd_reduce_add_ordered<T, U>(x: T, y: U) -> U;
+    fn simd_reduce_mul_ordered<T, U>(x: T, y: U) -> U;
+    fn simd_reduce_and<T, U>(x: T) -> U;
+    fn simd_reduce_or<T, U>(x: T) -> U;
+    fn simd_reduce_xor<T, U>(x: T) -> U;
+    fn simd_reduce_all<T>(x: T) -> bool;
+    fn simd_reduce_any<T>(x: T) -> bool;
+}
+
+fn main() {
+    let x = u32x4(0, 0, 0, 0);
+    let z = f32x4(0.0, 0.0, 0.0, 0.0);
+
+    unsafe {
+        simd_reduce_add_ordered(z, 0);
+        //~^ ERROR expected return type `f32` (element of input `f32x4`), found `i32`
+        simd_reduce_mul_ordered(z, 1);
+        //~^ ERROR expected return type `f32` (element of input `f32x4`), found `i32`
+
+        let _: f32 = simd_reduce_and(x);
+        //~^ ERROR expected return type `u32` (element of input `u32x4`), found `f32`
+        let _: f32 = simd_reduce_or(x);
+        //~^ ERROR expected return type `u32` (element of input `u32x4`), found `f32`
+        let _: f32 = simd_reduce_xor(x);
+        //~^ ERROR expected return type `u32` (element of input `u32x4`), found `f32`
+
+        let _: f32 = simd_reduce_and(z);
+        //~^ ERROR unsupported simd_reduce_and from `f32x4` with element `f32` to `f32`
+        let _: f32 = simd_reduce_or(z);
+        //~^ ERROR unsupported simd_reduce_or from `f32x4` with element `f32` to `f32`
+        let _: f32 = simd_reduce_xor(z);
+        //~^ ERROR unsupported simd_reduce_xor from `f32x4` with element `f32` to `f32`
+
+        let _: bool = simd_reduce_all(z);
+        //~^ ERROR unsupported simd_reduce_all from `f32x4` with element `f32` to `bool`
+        let _: bool = simd_reduce_any(z);
+        //~^ ERROR unsupported simd_reduce_any from `f32x4` with element `f32` to `bool`
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-reduction.stderr b/tests/ui/simd/intrinsic/generic-reduction.stderr
new file mode 100644
index 000000000..1028faf69
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-reduction.stderr
@@ -0,0 +1,63 @@
+error[E0511]: invalid monomorphization of `simd_reduce_add_ordered` intrinsic: expected return type `f32` (element of input `f32x4`), found `i32`
+  --> $DIR/generic-reduction.rs:34:9
+   |
+LL |         simd_reduce_add_ordered(z, 0);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_reduce_mul_ordered` intrinsic: expected return type `f32` (element of input `f32x4`), found `i32`
+  --> $DIR/generic-reduction.rs:36:9
+   |
+LL |         simd_reduce_mul_ordered(z, 1);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_reduce_and` intrinsic: expected return type `u32` (element of input `u32x4`), found `f32`
+  --> $DIR/generic-reduction.rs:39:22
+   |
+LL |         let _: f32 = simd_reduce_and(x);
+   |                      ^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_reduce_or` intrinsic: expected return type `u32` (element of input `u32x4`), found `f32`
+  --> $DIR/generic-reduction.rs:41:22
+   |
+LL |         let _: f32 = simd_reduce_or(x);
+   |                      ^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_reduce_xor` intrinsic: expected return type `u32` (element of input `u32x4`), found `f32`
+  --> $DIR/generic-reduction.rs:43:22
+   |
+LL |         let _: f32 = simd_reduce_xor(x);
+   |                      ^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_reduce_and` intrinsic: unsupported simd_reduce_and from `f32x4` with element `f32` to `f32`
+  --> $DIR/generic-reduction.rs:46:22
+   |
+LL |         let _: f32 = simd_reduce_and(z);
+   |                      ^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_reduce_or` intrinsic: unsupported simd_reduce_or from `f32x4` with element `f32` to `f32`
+  --> $DIR/generic-reduction.rs:48:22
+   |
+LL |         let _: f32 = simd_reduce_or(z);
+   |                      ^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_reduce_xor` intrinsic: unsupported simd_reduce_xor from `f32x4` with element `f32` to `f32`
+  --> $DIR/generic-reduction.rs:50:22
+   |
+LL |         let _: f32 = simd_reduce_xor(z);
+   |                      ^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_reduce_all` intrinsic: unsupported simd_reduce_all from `f32x4` with element `f32` to `bool`
+  --> $DIR/generic-reduction.rs:53:23
+   |
+LL |         let _: bool = simd_reduce_all(z);
+   |                       ^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_reduce_any` intrinsic: unsupported simd_reduce_any from `f32x4` with element `f32` to `bool`
+  --> $DIR/generic-reduction.rs:55:23
+   |
+LL |         let _: bool = simd_reduce_any(z);
+   |                       ^^^^^^^^^^^^^^^^^^
+
+error: aborting due to 10 previous errors
+
+For more information about this error, try `rustc --explain E0511`.
diff --git a/tests/ui/simd/intrinsic/generic-select-pass.rs b/tests/ui/simd/intrinsic/generic-select-pass.rs
new file mode 100644
index 000000000..b850cf975
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-select-pass.rs
@@ -0,0 +1,195 @@
+// run-pass
+#![allow(non_camel_case_types)]
+
+// ignore-emscripten
+// ignore-endian-big behavior of simd_select_bitmask is endian-specific
+
+// Test that the simd_select intrinsics produces correct results.
+
+#![feature(repr_simd, platform_intrinsics)]
+#[allow(non_camel_case_types)]
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+struct i32x4(pub i32, pub i32, pub i32, pub i32);
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+struct u32x4(pub u32, pub u32, pub u32, pub u32);
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+struct u32x8(u32, u32, u32, u32, u32, u32, u32, u32);
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+struct f32x4(pub f32, pub f32, pub f32, pub f32);
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq, Debug)]
+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;
+}
+
+fn main() {
+    let m0 = b8x4(!0, !0, !0, !0);
+    let m1 = b8x4(0, 0, 0, 0);
+    let m2 = b8x4(!0, !0, 0, 0);
+    let m3 = b8x4(0, 0, !0, !0);
+    let m4 = b8x4(!0, 0, !0, 0);
+
+    unsafe {
+        let a = i32x4(1, -2, 3, 4);
+        let b = i32x4(5, 6, -7, 8);
+
+        let r: i32x4 = simd_select(m0, a, b);
+        let e = a;
+        assert_eq!(r, e);
+
+        let r: i32x4 = simd_select(m1, a, b);
+        let e = b;
+        assert_eq!(r, e);
+
+        let r: i32x4 = simd_select(m2, a, b);
+        let e = i32x4(1, -2, -7, 8);
+        assert_eq!(r, e);
+
+        let r: i32x4 = simd_select(m3, a, b);
+        let e = i32x4(5, 6, 3, 4);
+        assert_eq!(r, e);
+
+        let r: i32x4 = simd_select(m4, a, b);
+        let e = i32x4(1, 6, 3, 8);
+        assert_eq!(r, e);
+    }
+
+    unsafe {
+        let a = u32x4(1, 2, 3, 4);
+        let b = u32x4(5, 6, 7, 8);
+
+        let r: u32x4 = simd_select(m0, a, b);
+        let e = a;
+        assert_eq!(r, e);
+
+        let r: u32x4 = simd_select(m1, a, b);
+        let e = b;
+        assert_eq!(r, e);
+
+        let r: u32x4 = simd_select(m2, a, b);
+        let e = u32x4(1, 2, 7, 8);
+        assert_eq!(r, e);
+
+        let r: u32x4 = simd_select(m3, a, b);
+        let e = u32x4(5, 6, 3, 4);
+        assert_eq!(r, e);
+
+        let r: u32x4 = simd_select(m4, a, b);
+        let e = u32x4(1, 6, 3, 8);
+        assert_eq!(r, e);
+    }
+
+    unsafe {
+        let a = f32x4(1., 2., 3., 4.);
+        let b = f32x4(5., 6., 7., 8.);
+
+        let r: f32x4 = simd_select(m0, a, b);
+        let e = a;
+        assert_eq!(r, e);
+
+        let r: f32x4 = simd_select(m1, a, b);
+        let e = b;
+        assert_eq!(r, e);
+
+        let r: f32x4 = simd_select(m2, a, b);
+        let e = f32x4(1., 2., 7., 8.);
+        assert_eq!(r, e);
+
+        let r: f32x4 = simd_select(m3, a, b);
+        let e = f32x4(5., 6., 3., 4.);
+        assert_eq!(r, e);
+
+        let r: f32x4 = simd_select(m4, a, b);
+        let e = f32x4(1., 6., 3., 8.);
+        assert_eq!(r, e);
+    }
+
+    unsafe {
+        let t = !0 as i8;
+        let f = 0 as i8;
+        let a = b8x4(t, f, t, f);
+        let b = b8x4(f, f, f, t);
+
+        let r: b8x4 = simd_select(m0, a, b);
+        let e = a;
+        assert_eq!(r, e);
+
+        let r: b8x4 = simd_select(m1, a, b);
+        let e = b;
+        assert_eq!(r, e);
+
+        let r: b8x4 = simd_select(m2, a, b);
+        let e = b8x4(t, f, f, t);
+        assert_eq!(r, e);
+
+        let r: b8x4 = simd_select(m3, a, b);
+        let e = b8x4(f, f, t, f);
+        assert_eq!(r, e);
+
+        let r: b8x4 = simd_select(m4, a, b);
+        let e = b8x4(t, f, t, t);
+        assert_eq!(r, e);
+    }
+
+    unsafe {
+        let a = u32x8(0, 1, 2, 3, 4, 5, 6, 7);
+        let b = u32x8(8, 9, 10, 11, 12, 13, 14, 15);
+
+        let r: u32x8 = simd_select_bitmask(0u8, a, b);
+        let e = b;
+        assert_eq!(r, e);
+
+        let r: u32x8 = simd_select_bitmask(0xffu8, a, b);
+        let e = a;
+        assert_eq!(r, e);
+
+        let r: u32x8 = simd_select_bitmask(0b01010101u8, a, b);
+        let e = u32x8(0, 9, 2, 11, 4, 13, 6, 15);
+        assert_eq!(r, e);
+
+        let r: u32x8 = simd_select_bitmask(0b10101010u8, a, b);
+        let e = u32x8(8, 1, 10, 3, 12, 5, 14, 7);
+        assert_eq!(r, e);
+
+        let r: u32x8 = simd_select_bitmask(0b11110000u8, a, b);
+        let e = u32x8(8, 9, 10, 11, 4, 5, 6, 7);
+        assert_eq!(r, e);
+    }
+
+    unsafe {
+        let a = u32x4(0, 1, 2, 3);
+        let b = u32x4(4, 5, 6, 7);
+
+        let r: u32x4 = simd_select_bitmask(0u8, a, b);
+        let e = b;
+        assert_eq!(r, e);
+
+        let r: u32x4 = simd_select_bitmask(0xfu8, a, b);
+        let e = a;
+        assert_eq!(r, e);
+
+        let r: u32x4 = simd_select_bitmask(0b0101u8, a, b);
+        let e = u32x4(0, 5, 2, 7);
+        assert_eq!(r, e);
+
+        let r: u32x4 = simd_select_bitmask(0b1010u8, a, b);
+        let e = u32x4(4, 1, 6, 3);
+        assert_eq!(r, e);
+
+        let r: u32x4 = simd_select_bitmask(0b1100u8, a, b);
+        let e = u32x4(4, 5, 2, 3);
+        assert_eq!(r, e);
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-select.rs b/tests/ui/simd/intrinsic/generic-select.rs
new file mode 100644
index 000000000..248e82ea2
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-select.rs
@@ -0,0 +1,63 @@
+// build-fail
+
+// Test that the simd_select intrinsic produces ok-ish error
+// messages when misused.
+
+#![feature(repr_simd, platform_intrinsics)]
+#![allow(non_camel_case_types)]
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct u32x4(pub u32, pub u32, pub u32, pub u32);
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq)]
+struct b8x4(pub i8, pub i8, pub i8, pub i8);
+
+#[repr(simd)]
+#[derive(Copy, Clone, PartialEq)]
+struct b8x8(pub i8, pub i8, pub i8, pub i8, 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;
+}
+
+fn main() {
+    let m4 = b8x4(0, 0, 0, 0);
+    let m8 = b8x8(0, 0, 0, 0, 0, 0, 0, 0);
+    let x = u32x4(0, 0, 0, 0);
+    let z = f32x4(0.0, 0.0, 0.0, 0.0);
+
+    unsafe {
+        simd_select(m4, x, x);
+
+        simd_select(m8, x, x);
+        //~^ ERROR mismatched lengths: mask length `8` != other vector length `4`
+
+        simd_select(x, x, x);
+        //~^ ERROR mask element type is `u32`, expected `i_`
+
+        simd_select(z, z, z);
+        //~^ ERROR mask element type is `f32`, expected `i_`
+
+        simd_select(m4, 0u32, 1u32);
+        //~^ ERROR found non-SIMD `u32`
+
+        simd_select_bitmask(0u16, x, x);
+        //~^ ERROR invalid bitmask `u16`, expected `u8` or `[u8; 1]`
+
+        simd_select_bitmask(0u8, 1u32, 2u32);
+        //~^ ERROR found non-SIMD `u32`
+
+        simd_select_bitmask(0.0f32, x, x);
+        //~^ ERROR invalid bitmask `f32`, expected `u8` or `[u8; 1]`
+
+        simd_select_bitmask("x", x, x);
+        //~^ ERROR invalid bitmask `&str`, expected `u8` or `[u8; 1]`
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-select.stderr b/tests/ui/simd/intrinsic/generic-select.stderr
new file mode 100644
index 000000000..d576f1bc7
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-select.stderr
@@ -0,0 +1,51 @@
+error[E0511]: invalid monomorphization of `simd_select` intrinsic: mismatched lengths: mask length `8` != other vector length `4`
+  --> $DIR/generic-select.rs:39:9
+   |
+LL |         simd_select(m8, x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_select` intrinsic: mask element type is `u32`, expected `i_`
+  --> $DIR/generic-select.rs:42:9
+   |
+LL |         simd_select(x, x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_select` intrinsic: mask element type is `f32`, expected `i_`
+  --> $DIR/generic-select.rs:45:9
+   |
+LL |         simd_select(z, z, z);
+   |         ^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_select` intrinsic: expected SIMD argument type, found non-SIMD `u32`
+  --> $DIR/generic-select.rs:48:9
+   |
+LL |         simd_select(m4, 0u32, 1u32);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_select_bitmask` intrinsic: invalid bitmask `u16`, expected `u8` or `[u8; 1]`
+  --> $DIR/generic-select.rs:51:9
+   |
+LL |         simd_select_bitmask(0u16, x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_select_bitmask` intrinsic: expected SIMD argument type, found non-SIMD `u32`
+  --> $DIR/generic-select.rs:54:9
+   |
+LL |         simd_select_bitmask(0u8, 1u32, 2u32);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_select_bitmask` intrinsic: invalid bitmask `f32`, expected `u8` or `[u8; 1]`
+  --> $DIR/generic-select.rs:57:9
+   |
+LL |         simd_select_bitmask(0.0f32, x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_select_bitmask` intrinsic: invalid bitmask `&str`, expected `u8` or `[u8; 1]`
+  --> $DIR/generic-select.rs:60:9
+   |
+LL |         simd_select_bitmask("x", x, x);
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+error: aborting due to 8 previous errors
+
+For more information about this error, try `rustc --explain E0511`.
diff --git a/tests/ui/simd/intrinsic/generic-shuffle.rs b/tests/ui/simd/intrinsic/generic-shuffle.rs
new file mode 100644
index 000000000..9611780ac
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-shuffle.rs
@@ -0,0 +1,33 @@
+// build-fail
+
+// Test that the simd_shuffle intrinsic produces ok-ish error
+// messages when misused.
+
+#![feature(repr_simd, platform_intrinsics)]
+
+#[repr(simd)]
+#[derive(Copy, Clone)]
+pub struct Simd<T, const N: usize>([T; N]);
+
+extern "platform-intrinsic" {
+    fn simd_shuffle<T, I, U>(a: T, b: T, i: I) -> U;
+}
+
+fn main() {
+    const I: [u32; 2] = [0; 2];
+    const I2: [f32; 2] = [0.; 2];
+    let v = Simd::<u32, 4>([0; 4]);
+
+    unsafe {
+        let _: Simd<u32, 2> = simd_shuffle(v, v, I);
+
+        let _: Simd<u32, 4> = simd_shuffle(v, v, I);
+        //~^ ERROR invalid monomorphization of `simd_shuffle` intrinsic
+
+        let _: Simd<f32, 2> = simd_shuffle(v, v, I);
+        //~^ ERROR invalid monomorphization of `simd_shuffle` intrinsic
+
+        let _: Simd<u32, 2> = simd_shuffle(v, v, I2);
+        //~^ ERROR invalid monomorphization of `simd_shuffle` intrinsic
+    }
+}
diff --git a/tests/ui/simd/intrinsic/generic-shuffle.stderr b/tests/ui/simd/intrinsic/generic-shuffle.stderr
new file mode 100644
index 000000000..81e641612
--- /dev/null
+++ b/tests/ui/simd/intrinsic/generic-shuffle.stderr
@@ -0,0 +1,21 @@
+error[E0511]: invalid monomorphization of `simd_shuffle` intrinsic: expected return type of length 2, found `Simd<u32, 4>` with length 4
+  --> $DIR/generic-shuffle.rs:24:31
+   |
+LL |         let _: Simd<u32, 4> = simd_shuffle(v, v, I);
+   |                               ^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shuffle` intrinsic: expected return element type `u32` (element of input `Simd<u32, 4>`), found `Simd<f32, 2>` with element type `f32`
+  --> $DIR/generic-shuffle.rs:27:31
+   |
+LL |         let _: Simd<f32, 2> = simd_shuffle(v, v, I);
+   |                               ^^^^^^^^^^^^^^^^^^^^^
+
+error[E0511]: invalid monomorphization of `simd_shuffle` intrinsic: simd_shuffle index must be an array of `u32`, got `[f32; 2]`
+  --> $DIR/generic-shuffle.rs:30:31
+   |
+LL |         let _: Simd<u32, 2> = simd_shuffle(v, v, I2);
+   |                               ^^^^^^^^^^^^^^^^^^^^^^
+
+error: aborting due to 3 previous errors
+
+For more information about this error, try `rustc --explain E0511`.
diff --git a/tests/ui/simd/intrinsic/inlining-issue67557-ice.rs b/tests/ui/simd/intrinsic/inlining-issue67557-ice.rs
new file mode 100644
index 000000000..7221b3ab7
--- /dev/null
+++ b/tests/ui/simd/intrinsic/inlining-issue67557-ice.rs
@@ -0,0 +1,26 @@
+// This used to cause an ICE for an internal index out of range due to simd_shuffle_indices being
+// passed the wrong Instance, causing issues with inlining. See #67557.
+//
+// run-pass
+// compile-flags: -Zmir-opt-level=4
+#![feature(platform_intrinsics, repr_simd)]
+
+extern "platform-intrinsic" {
+    fn simd_shuffle2<T, U>(x: T, y: T, idx: [u32; 2]) -> U;
+}
+
+#[repr(simd)]
+#[derive(Debug, PartialEq)]
+struct Simd2(u8, u8);
+
+fn main() {
+    unsafe {
+        let _: Simd2 = inline_me();
+    }
+}
+
+#[inline(always)]
+unsafe fn inline_me() -> Simd2 {
+    const IDX: [u32; 2] = [0, 3];
+    simd_shuffle2(Simd2(10, 11), Simd2(12, 13), IDX)
+}
diff --git a/tests/ui/simd/intrinsic/inlining-issue67557.rs b/tests/ui/simd/intrinsic/inlining-issue67557.rs
new file mode 100644
index 000000000..0d1542709
--- /dev/null
+++ b/tests/ui/simd/intrinsic/inlining-issue67557.rs
@@ -0,0 +1,42 @@
+// This used to cause assert_10_13 to unexpectingly fail, due to simd_shuffle_indices being passed
+// the wrong Instance, causing issues with inlining. See #67557.
+//
+// run-pass
+// compile-flags: -Zmir-opt-level=4
+#![feature(platform_intrinsics, repr_simd)]
+
+extern "platform-intrinsic" {
+    fn simd_shuffle2<T, U>(x: T, y: T, idx: [u32; 2]) -> U;
+}
+
+#[repr(simd)]
+#[derive(Debug, PartialEq)]
+struct Simd2(u8, u8);
+
+fn main() {
+    unsafe {
+        const IDX: [u32; 2] = [0, 1];
+        let p_res: Simd2 = simd_shuffle2(Simd2(10, 11), Simd2(12, 13), IDX);
+        let a_res: Simd2 = inline_me();
+
+        assert_10_11(p_res);
+        assert_10_13(a_res);
+    }
+}
+
+#[inline(never)]
+fn assert_10_11(x: Simd2) {
+    assert_eq!(x, Simd2(10, 11));
+}
+
+#[inline(never)]
+fn assert_10_13(x: Simd2) {
+    assert_eq!(x, Simd2(10, 13));
+}
+
+
+#[inline(always)]
+unsafe fn inline_me() -> Simd2 {
+    const IDX: [u32; 2] = [0, 3];
+    simd_shuffle2(Simd2(10, 11), Simd2(12, 13), IDX)
+}
diff --git a/tests/ui/simd/intrinsic/issue-85855.rs b/tests/ui/simd/intrinsic/issue-85855.rs
new file mode 100644
index 000000000..f276fbd66
--- /dev/null
+++ b/tests/ui/simd/intrinsic/issue-85855.rs
@@ -0,0 +1,19 @@
+// Check that appropriate errors are reported if an intrinsic is defined
+// with the wrong number of generic lifetime/type/const parameters, and
+// that no ICE occurs in these cases.
+
+#![feature(platform_intrinsics)]
+#![crate_type="lib"]
+
+extern "platform-intrinsic" {
+    fn simd_saturating_add<'a, T: 'a>(x: T, y: T);
+    //~^ ERROR: intrinsic has wrong number of lifetime parameters
+
+    fn simd_add<'a, T>(x: T, y: T) -> T;
+
+    fn simd_sub<T, U>(x: T, y: U);
+    //~^ ERROR: intrinsic has wrong number of type parameters
+
+    fn simd_mul<T, const N: usize>(x: T, y: T);
+    //~^ ERROR: intrinsic has wrong number of const parameters
+}
diff --git a/tests/ui/simd/intrinsic/issue-85855.stderr b/tests/ui/simd/intrinsic/issue-85855.stderr
new file mode 100644
index 000000000..fb2f1fbc5
--- /dev/null
+++ b/tests/ui/simd/intrinsic/issue-85855.stderr
@@ -0,0 +1,21 @@
+error[E0094]: intrinsic has wrong number of lifetime parameters: found 1, expected 0
+  --> $DIR/issue-85855.rs:9:27
+   |
+LL |     fn simd_saturating_add<'a, T: 'a>(x: T, y: T);
+   |                           ^^^^^^^^^^^ expected 0 lifetime parameters
+
+error[E0094]: intrinsic has wrong number of type parameters: found 2, expected 1
+  --> $DIR/issue-85855.rs:14:16
+   |
+LL |     fn simd_sub<T, U>(x: T, y: U);
+   |                ^^^^^^ expected 1 type parameter
+
+error[E0094]: intrinsic has wrong number of const parameters: found 1, expected 0
+  --> $DIR/issue-85855.rs:17:16
+   |
+LL |     fn simd_mul<T, const N: usize>(x: T, y: T);
+   |                ^^^^^^^^^^^^^^^^^^^ expected 0 const parameters
+
+error: aborting due to 3 previous errors
+
+For more information about this error, try `rustc --explain E0094`.
diff --git a/tests/ui/simd/intrinsic/ptr-cast.rs b/tests/ui/simd/intrinsic/ptr-cast.rs
new file mode 100644
index 000000000..1d13720bc
--- /dev/null
+++ b/tests/ui/simd/intrinsic/ptr-cast.rs
@@ -0,0 +1,33 @@
+// run-pass
+
+#![feature(repr_simd, platform_intrinsics)]
+
+extern "platform-intrinsic" {
+    fn simd_cast_ptr<T, U>(x: T) -> U;
+    fn simd_expose_addr<T, U>(x: T) -> U;
+    fn simd_from_exposed_addr<T, U>(x: T) -> U;
+}
+
+#[derive(Copy, Clone)]
+#[repr(simd)]
+struct V<T>([T; 2]);
+
+fn main() {
+    unsafe {
+        let mut foo = 4i8;
+        let ptr = &mut foo as *mut i8;
+
+        let ptrs = V::<*mut i8>([ptr, core::ptr::null_mut()]);
+
+        // change constness and type
+        let const_ptrs: V<*const u8> = simd_cast_ptr(ptrs);
+
+        let exposed_addr: V<usize> = simd_expose_addr(const_ptrs);
+
+        let from_exposed_addr: V<*mut i8> = simd_from_exposed_addr(exposed_addr);
+
+        assert!(const_ptrs.0 == [ptr as *const u8, core::ptr::null()]);
+        assert!(exposed_addr.0 == [ptr as usize, 0]);
+        assert!(from_exposed_addr.0 == ptrs.0);
+    }
+}