[llvm] [NVPTX] Improve support for rsqrt.approx (PR #89417)

[Artem Belevich via llvm-commits]

================
@@ -30,6 +30,10 @@ using namespace llvm;
 #define DEBUG_TYPE "nvptx-isel"
 #define PASS_NAME "NVPTX DAG->DAG Pattern Instruction Selection"
 
+static cl::opt<bool>
+    DisableRsqrtOpt("nvptx-disable-rsqrt-opt", cl::init(false), cl::Hidden,
----------------
Artem-B wrote:

We'll need someone familiar with FP nuances to take a look. 

Enabling sqrt->rsqrt.approx by default sounds risky to me.

How do other targets handle sqrt optimizations?


https://github.com/llvm/llvm-project/pull/89417