[libc-commits] [libc] [libc][math] Fix the FTZ/DAZ checks for log1p. (PR #174424)

[via libc-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-libc

Author: None (lntue)

<details>
<summary>Changes</summary>

The previous checks will fail with https://github.com/llvm/llvm-project/pull/174123.
See https://github.com/llvm/llvm-project/pull/174290#issuecomment-3708052716 for the discussion.

---
Full diff: https://github.com/llvm/llvm-project/pull/174424.diff


1 Files Affected:

- (modified) libc/src/math/generic/log1p.cpp (+2-2) 


``````````diff
diff --git a/libc/src/math/generic/log1p.cpp b/libc/src/math/generic/log1p.cpp
index 1595981ac309b..ddac3c2bacc05 100644
--- a/libc/src/math/generic/log1p.cpp
+++ b/libc/src/math/generic/log1p.cpp
@@ -935,7 +935,7 @@ LLVM_LIBC_FUNCTION(double, log1p, (double x)) {
       //   log(1 + x) = nextafter(x, -inf) for FE_DOWNWARD, or
       //                                       FE_TOWARDZERO and x > 0,
       //              = x                  otherwise.
-      if (x == 0.0)
+      if (x + x == 0.0)
         return x + x; // Handle FTZ/DAZ correctly.
 
       volatile float tp = 1.0f;
@@ -951,7 +951,7 @@ LLVM_LIBC_FUNCTION(double, log1p, (double x)) {
         return FPBits_t(x_u + 1).get_val();
       }
 
-      return (x + x == 0.0) ? x + x : x;
+      return x;
     }
     x_dd = fputil::exact_add(1.0, x);
   }

``````````

</details>


https://github.com/llvm/llvm-project/pull/174424