Auto merge of #52895 - draganmladjenovic:minmax_qnan, r=alexcrichton

run-pass/simd-intrinsic-float-minmax: Force use of qNaN on Mips

Workaround for #52746.
r? @gnzlbg

src/test/run-pass/simd-intrinsic-float-minmax.rs

@@ -29,7 +29,14 @@ extern "platform-intrinsic" {
 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 = ::std::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(::std::f32::NAN.to_bits() - 1);
+
     let n = f32x4(nan, nan, nan, nan);
 
     unsafe {