[libc] Remove specific nan payload in math functions

[gchatelet]

No description provided.

[llvmbot]

@llvm/pr-subscribers-libc

Author: Guillaume Chatelet (gchatelet)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/79165.diff

4 Files Affected:

• (modified) libc/src/math/generic/asinf.cpp (+1-1)
• (modified) libc/src/math/generic/sincosf.cpp (+1-3)
• (modified) libc/test/UnitTest/FPMatcher.h (+2-2)
• (modified) libc/test/src/__support/FPUtil/fpbits_test.cpp (+6-11)

diff --git a/libc/src/math/generic/asinf.cpp b/libc/src/math/generic/asinf.cpp
index ee8e853063644e..d673bdea6697ad 100644
--- a/libc/src/math/generic/asinf.cpp
+++ b/libc/src/math/generic/asinf.cpp
@@ -109,7 +109,7 @@ LLVM_LIBC_FUNCTION(float, asinf, (float x)) {
       fputil::set_errno_if_required(EDOM);
       fputil::raise_except_if_required(FE_INVALID);
     }
-    return x + FPBits::build_nan(Sign::POS, FPBits::FRACTION_MASK).get_val();
+    return x + FPBits::build_nan().get_val();
   }
 
   // Check for exceptional values
diff --git a/libc/src/math/generic/sincosf.cpp b/libc/src/math/generic/sincosf.cpp
index f12b93a0e69653..f1ad741f9be80e 100644
--- a/libc/src/math/generic/sincosf.cpp
+++ b/libc/src/math/generic/sincosf.cpp
@@ -148,9 +148,7 @@ LLVM_LIBC_FUNCTION(void, sincosf, (float x, float *sinp, float *cosp)) {
       fputil::set_errno_if_required(EDOM);
       fputil::raise_except_if_required(FE_INVALID);
     }
-    *sinp =
-        x +
-        FPBits::build_nan(fputil::Sign::POS, FPBits::FRACTION_MASK).get_val();
+    *sinp = x + FPBits::build_nan().get_val();
     *cosp = *sinp;
     return;
   }
diff --git a/libc/test/UnitTest/FPMatcher.h b/libc/test/UnitTest/FPMatcher.h
index 69da5387d382b5..7c81475ecec3ae 100644
--- a/libc/test/UnitTest/FPMatcher.h
+++ b/libc/test/UnitTest/FPMatcher.h
@@ -69,7 +69,7 @@ template <typename T> struct FPTest : public Test {
   static constexpr T zero = FPBits::zero(Sign::POS).get_val();
   static constexpr T neg_zero = FPBits::zero(Sign::NEG).get_val();
   static constexpr T aNaN = FPBits::build_quiet_nan().get_val();
-  static constexpr T sNaN = FPBits::build_nan(Sign::POS, 1).get_val();
+  static constexpr T sNaN = FPBits::build_nan().get_val();
   static constexpr T inf = FPBits::inf(Sign::POS).get_val();
   static constexpr T neg_inf = FPBits::inf(Sign::NEG).get_val();
   static constexpr T min_normal = FPBits::min_normal().get_val();
@@ -98,7 +98,7 @@ template <typename T> struct FPTest : public Test {
   const T zero = FPBits::zero(Sign::POS).get_val();                            \
   const T neg_zero = FPBits::zero(Sign::NEG).get_val();                        \
   const T aNaN = FPBits::build_quiet_nan().get_val();                          \
-  const T sNaN = FPBits::build_nan(Sign::POS, 1).get_val();                    \
+  const T sNaN = FPBits::build_nan().get_val();                                \
   const T inf = FPBits::inf(Sign::POS).get_val();                              \
   const T neg_inf = FPBits::inf(Sign::NEG).get_val();                          \
   const T min_normal = FPBits::min_normal().get_val();                         \
diff --git a/libc/test/src/__support/FPUtil/fpbits_test.cpp b/libc/test/src/__support/FPUtil/fpbits_test.cpp
index aadcd3fe92b9e5..a4324a071536f8 100644
--- a/libc/test/src/__support/FPUtil/fpbits_test.cpp
+++ b/libc/test/src/__support/FPUtil/fpbits_test.cpp
@@ -236,8 +236,7 @@ TEST(LlvmLibcFPBitsTest, FloatType) {
                "(+Infinity)");
   EXPECT_STREQ(LIBC_NAMESPACE::str(FloatBits::inf(Sign::NEG)).c_str(),
                "(-Infinity)");
-  EXPECT_STREQ(LIBC_NAMESPACE::str(FloatBits::build_nan(Sign::POS, 1)).c_str(),
-               "(NaN)");
+  EXPECT_STREQ(LIBC_NAMESPACE::str(FloatBits::build_nan()).c_str(), "(NaN)");
 
   FloatBits zero(0.0f);
   EXPECT_TRUE(zero.is_pos());
@@ -363,9 +362,8 @@ TEST(LlvmLibcFPBitsTest, X86LongDoubleType) {
                "(+Infinity)");
   EXPECT_STREQ(LIBC_NAMESPACE::str(LongDoubleBits::inf(Sign::NEG)).c_str(),
                "(-Infinity)");
-  EXPECT_STREQ(
-      LIBC_NAMESPACE::str(LongDoubleBits::build_nan(Sign::POS, 1)).c_str(),
-      "(NaN)");
+  EXPECT_STREQ(LIBC_NAMESPACE::str(LongDoubleBits::build_nan()).c_str(),
+               "(NaN)");
 
   LongDoubleBits zero(0.0l);
   EXPECT_TRUE(zero.is_pos());
@@ -444,9 +442,8 @@ TEST(LlvmLibcFPBitsTest, LongDoubleType) {
                "(+Infinity)");
   EXPECT_STREQ(LIBC_NAMESPACE::str(LongDoubleBits::inf(Sign::NEG)).c_str(),
                "(-Infinity)");
-  EXPECT_STREQ(
-      LIBC_NAMESPACE::str(LongDoubleBits::build_nan(Sign::POS, 1)).c_str(),
-      "(NaN)");
+  EXPECT_STREQ(LIBC_NAMESPACE::str(LongDoubleBits::build_nan()).c_str(),
+               "(NaN)");
 
   LongDoubleBits zero(0.0l);
   EXPECT_TRUE(zero.is_pos());
@@ -519,9 +516,7 @@ TEST(LlvmLibcFPBitsTest, Float128Type) {
                "(+Infinity)");
   EXPECT_STREQ(LIBC_NAMESPACE::str(Float128Bits::inf(Sign::NEG)).c_str(),
                "(-Infinity)");
-  EXPECT_STREQ(
-      LIBC_NAMESPACE::str(Float128Bits::build_nan(Sign::POS, 1)).c_str(),
-      "(NaN)");
+  EXPECT_STREQ(LIBC_NAMESPACE::str(Float128Bits::build_nan()).c_str(), "(NaN)");
 
   Float128Bits zero = Float128Bits::zero(Sign::POS);
   EXPECT_TRUE(zero.is_pos());

[gchatelet]

@lntue can you PTAL?
If this patch is correct this means we can completely remove FPBits::build_nan

edit: You can review 2fdf373 only, the other one you already reviewed but I had to push force (conflict -> rebase).

[lntue]

@lntue can you PTAL? If this patch is correct this means we can completely remove FPBits::build_nan

edit: You can review 2fdf373 only, the other one you already reviewed but I had to push force (conflict -> rebase).

Yes, most of the time we only need build_quiet_nan, but there might be some rare cases that we need signaling nan. I guess we can remove build_nan, and only use build_quiet_nan + build_signaling_nan to be explicit.