[clang] [CIR][CUDA] Implement NVVM math builtins (fabs, ex2_approx) (PR #195663)

[Andy Kaylor via cfe-commits]

================
@@ -0,0 +1,57 @@
+#include "Inputs/cuda.h"
+
+// RUN: %clang_cc1 -triple nvptx64-nvidia-cuda -target-cpu sm_80 -x cuda \
+// RUN:            -fcuda-is-device -fclangir -emit-cir %s -o %t.cir
+// RUN: FileCheck --check-prefix=CIR --input-file=%t.cir %s
+
+// RUN: %clang_cc1 -triple nvptx64-nvidia-cuda -target-cpu sm_80 -x cuda \
+// RUN:            -fcuda-is-device -fclangir -emit-llvm %s -o %t.ll
+// RUN: FileCheck --check-prefix=LLVM --input-file=%t.ll %s
----------------
andykaylor wrote:

Classic codegen was adding `contract`? That's OK. We don't have any support for fast-math flags in CIR yet, so this is an expected mismatch.

https://github.com/llvm/llvm-project/pull/195663