[libc][math][c23] fixing dfmal for rv32

[aaryanshukla]

No description provided.

[llvmbot]

@llvm/pr-subscribers-libc

Author: None (aaryanshukla)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/101763.diff

1 Files Affected:

• (modified) libc/test/src/math/smoke/FmaTest.h (+5-3)

diff --git a/libc/test/src/math/smoke/FmaTest.h b/libc/test/src/math/smoke/FmaTest.h
index f942de37654dd..bf6d06d698fde 100644
--- a/libc/test/src/math/smoke/FmaTest.h
+++ b/libc/test/src/math/smoke/FmaTest.h
@@ -67,12 +67,14 @@ class FmaTestTemplate : public LIBC_NAMESPACE::testing::FEnvSafeTest {
                          out.min_normal));
 
     if constexpr (sizeof(OutType) < sizeof(InType)) {
+      InFPBits tmp = InFPBits::one();
+      tmp.set_biased_exponent(InFPBits::EXP_BIAS - InFPBits::FRACTION_LEN - 1);
+      InType reciprocal_value = tmp.get_val();
+
       InType v = InFPBits(static_cast<InStorageType>(IN_MIN_NORMAL_U +
                                                      InStorageType(1)))
                      .get_val();
-      EXPECT_FP_EQ(
-          out.min_normal,
-          func(InType(1) / InType(IN_MIN_NORMAL_U << 1), v, out.min_normal));
+      EXPECT_FP_EQ(out.min_normal, func(reciprocal_value, v, out.min_normal));
     }
 
     // Test overflow.

[aaryanshukla]

@mikhailramalho let me know if this passes on your build

[mikhailramalho]

Thanks for the fix!

[mikhailramalho]

BTW, you can probably enable it in arm32 too