Rollup merge of rust-lang#128417 - tgross35:f16-f128-math, r=dtolnay

Add `f16` and `f128` math functions

This adds intrinsics and math functions for `f16` and `f128` floating point types. Support is quite limited and some things are broken so tests don't run on many platforms, but this provides a starting point.

compiler/rustc_codegen_llvm/src/context.rs

@@ -775,10 +775,10 @@ impl<'ll> CodegenCx<'ll, '_> {
         ifn!("llvm.debugtrap", fn() -> void);
         ifn!("llvm.frameaddress", fn(t_i32) -> ptr);
 
-        ifn!("llvm.powi.f16", fn(t_f16, t_i32) -> t_f16);
-        ifn!("llvm.powi.f32", fn(t_f32, t_i32) -> t_f32);
-        ifn!("llvm.powi.f64", fn(t_f64, t_i32) -> t_f64);
-        ifn!("llvm.powi.f128", fn(t_f128, t_i32) -> t_f128);
+        ifn!("llvm.powi.f16.i32", fn(t_f16, t_i32) -> t_f16);
+        ifn!("llvm.powi.f32.i32", fn(t_f32, t_i32) -> t_f32);
+        ifn!("llvm.powi.f64.i32", fn(t_f64, t_i32) -> t_f64);
+        ifn!("llvm.powi.f128.i32", fn(t_f128, t_i32) -> t_f128);
 
         ifn!("llvm.pow.f16", fn(t_f16, t_f16) -> t_f16);
         ifn!("llvm.pow.f32", fn(t_f32, t_f32) -> t_f32);

compiler/rustc_codegen_llvm/src/intrinsic.rs

@@ -35,10 +35,10 @@ fn get_simple_intrinsic<'ll>(
         sym::sqrtf64 => "llvm.sqrt.f64",
         sym::sqrtf128 => "llvm.sqrt.f128",
 
-        sym::powif16 => "llvm.powi.f16",
-        sym::powif32 => "llvm.powi.f32",
-        sym::powif64 => "llvm.powi.f64",
-        sym::powif128 => "llvm.powi.f128",
+        sym::powif16 => "llvm.powi.f16.i32",
+        sym::powif32 => "llvm.powi.f32.i32",
+        sym::powif64 => "llvm.powi.f64.i32",
+        sym::powif128 => "llvm.powi.f128.i32",
 
         sym::sinf16 => "llvm.sin.f16",
         sym::sinf32 => "llvm.sin.f32",