Add support for f128 integer exponentiation

Adds `__powitf2`.

README.md

@@ -274,7 +274,7 @@ of being added to Rust.
 - [ ] floatunsitf.c
 - [ ] floatuntitf.c
 - [x] multf3.c
-- [ ] powitf2.c
+- [x] powitf2.c
 - [x] subtf3.c
 - [x] truncdfhf2.c
 - [x] truncsfhf2.c

build.rs

@@ -527,7 +527,6 @@ mod c {
                 ("__floatunditf", "floatunditf.c"),
                 ("__floatunsitf", "floatunsitf.c"),
                 ("__divtf3", "divtf3.c"),
-                ("__powitf2", "powitf2.c"),
                 ("__fe_getround", "fp_mode.c"),
                 ("__fe_raise_inexact", "fp_mode.c"),
             ]);

src/float/pow.rs

@@ -35,4 +35,16 @@ intrinsics! {
     pub extern "C" fn __powidf2(a: f64, b: i32) -> f64 {
         pow(a, b)
     }
+
+    #[avr_skip]
+    #[cfg(not(any(feature = "no-f16-f128", target_arch = "powerpc", target_arch = "powerpc64")))]
+    pub extern "C" fn __powitf2(a: f128, b: i32) -> f128 {
+        pow(a, b)
+    }
+
+    #[avr_skip]
+    #[cfg(all(not(feature = "no-f16-f128"), any(target_arch = "powerpc", target_arch = "powerpc64")))]
+    pub extern "C" fn __powikf2(a: f128, b: i32) -> f128 {
+        pow(a, b)
+    }
 }

testcrate/tests/float_pow.rs

@@ -1,4 +1,5 @@
 #![allow(unused_macros)]
+#![cfg_attr(f128_enabled, feature(f128))]
 #![cfg(not(all(target_arch = "x86", not(target_feature = "sse"))))]
 
 use testcrate::*;
@@ -53,3 +54,17 @@ pow! {
     f32, 1e-4, __powisf2;
     f64, 1e-12, __powidf2;
 }
+
+#[cfg(f128_enabled)]
+// Windows can't link the required arithmetic functions. See
+// <https://github.com/rust-lang/compiler-builtins/pull/614#issuecomment-2118636613>
+#[cfg(not(any(target_arch = "powerpc", target_arch = "powerpc64")))]
+pow! {
+    f128, 1e-36, __powitf2;
+}
+
+#[cfg(f128_enabled)]
+#[cfg(any(target_arch = "powerpc", target_arch = "powerpc64"))]
+pow! {
+    f128, 1e-36, __powikf2;
+}