[C23] Add *_NORM_MAX macros to <float.h> #96643

AaronBallman · 2024-06-25T14:29:33Z

This adds the *_NORM_MAX macros to <float.h> for freestanding mode in Clang; the values were chosen based on the behavior seen coming from GCC and the values already produced for the *_MAX macros by Clang.

llvmbot · 2024-06-25T14:30:04Z

@llvm/pr-subscribers-clang

@llvm/pr-subscribers-backend-x86

Author: Aaron Ballman (AaronBallman)

Changes

This adds the *_NORM_MAX macros to <float.h> for freestanding mode in Clang; the values were chosen based on the behavior seen coming from GCC and the values already produced for the *_MAX macros by Clang.

Full diff: https://github.com/llvm/llvm-project/pull/96643.diff

4 Files Affected:

(modified) clang/docs/ReleaseNotes.rst (+3)
(modified) clang/lib/Frontend/InitPreprocessor.cpp (+6-1)
(modified) clang/lib/Headers/float.h (+13)
(modified) clang/test/Headers/float.c (+27)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index df579ae398c5e..8879a77ba4dd9 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -336,6 +336,9 @@ C23 Feature Support
 - Properly promote bit-fields of bit-precise integer types to the field's type
   rather than to ``int``. #GH87641
 
+- Added the ``FLT_NORM_MAX``, ``DBL_NORM_MAX``, and ``LDBL_NORM_MAX`` to the
+  freestanding implementation of ``<float.h>`` that ships with Clang.
+
 Non-comprehensive list of changes in this release
 -------------------------------------------------
 
diff --git a/clang/lib/Frontend/InitPreprocessor.cpp b/clang/lib/Frontend/InitPreprocessor.cpp
index 55ec460064830..afc4c16ab685b 100644
--- a/clang/lib/Frontend/InitPreprocessor.cpp
+++ b/clang/lib/Frontend/InitPreprocessor.cpp
@@ -113,7 +113,11 @@ static T PickFP(const llvm::fltSemantics *Sem, T IEEEHalfVal, T IEEESingleVal,
 
 static void DefineFloatMacros(MacroBuilder &Builder, StringRef Prefix,
                               const llvm::fltSemantics *Sem, StringRef Ext) {
-  const char *DenormMin, *Epsilon, *Max, *Min;
+  const char *DenormMin, *NormMax, *Epsilon, *Max, *Min;
+  NormMax = PickFP(Sem, "6.5504e+4", "3.40282347e+38",
+                   "1.7976931348623157e+308", "1.18973149535723176502e+4932",
+                   "1.79769313486231580793728971405301e+308",
+                   "1.18973149535723176508575932662800702e+4932");
   DenormMin = PickFP(Sem, "5.9604644775390625e-8", "1.40129846e-45",
                      "4.9406564584124654e-324", "3.64519953188247460253e-4951",
                      "4.94065645841246544176568792868221e-324",
@@ -144,6 +148,7 @@ static void DefineFloatMacros(MacroBuilder &Builder, StringRef Prefix,
   DefPrefix += "_";
 
   Builder.defineMacro(DefPrefix + "DENORM_MIN__", Twine(DenormMin)+Ext);
+  Builder.defineMacro(DefPrefix + "NORM_MAX__", Twine(NormMax)+Ext);
   Builder.defineMacro(DefPrefix + "HAS_DENORM__");
   Builder.defineMacro(DefPrefix + "DIG__", Twine(Digits));
   Builder.defineMacro(DefPrefix + "DECIMAL_DIG__", Twine(DecimalDigits));
diff --git a/clang/lib/Headers/float.h b/clang/lib/Headers/float.h
index 642c8f06cc938..61f65dad8debc 100644
--- a/clang/lib/Headers/float.h
+++ b/clang/lib/Headers/float.h
@@ -86,6 +86,12 @@
 #    undef DBL_HAS_SUBNORM
 #    undef LDBL_HAS_SUBNORM
 #  endif
+#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 202311L) ||              \
+    !defined(__STRICT_ANSI__)
+#    undef FLT_NORM_MAX
+#    undef DBL_NORM_MAX
+#    undef LDBL_NORM_MAX
+#endif
 #endif
 
 /* Characteristics of floating point types, C99 5.2.4.2.2 */
@@ -155,6 +161,13 @@
 #  define LDBL_HAS_SUBNORM __LDBL_HAS_DENORM__
 #endif
 
+#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 202311L) ||              \
+    !defined(__STRICT_ANSI__)
+#  define FLT_NORM_MAX __FLT_NORM_MAX__
+#  define DBL_NORM_MAX __DBL_NORM_MAX__
+#  define LDBL_NORM_MAX __LDBL_NORM_MAX__
+#endif
+
 #ifdef __STDC_WANT_IEC_60559_TYPES_EXT__
 #  define FLT16_MANT_DIG    __FLT16_MANT_DIG__
 #  define FLT16_DECIMAL_DIG __FLT16_DECIMAL_DIG__
diff --git a/clang/test/Headers/float.c b/clang/test/Headers/float.c
index 70c11b0537537..59cc0faa074db 100644
--- a/clang/test/Headers/float.c
+++ b/clang/test/Headers/float.c
@@ -1,17 +1,21 @@
 // RUN: %clang_cc1 -fsyntax-only -verify -std=c89 -ffreestanding %s
 // RUN: %clang_cc1 -fsyntax-only -verify -std=c99 -ffreestanding %s
 // RUN: %clang_cc1 -fsyntax-only -verify -std=c11 -ffreestanding %s
+// RUN: %clang_cc1 -fsyntax-only -verify -std=c23 -ffreestanding %s
 // RUN: %clang_cc1 -fsyntax-only -verify -xc++ -std=c++11 -ffreestanding %s
 // RUN: %clang_cc1 -fsyntax-only -verify -xc++ -std=c++14 -ffreestanding %s
 // RUN: %clang_cc1 -fsyntax-only -verify -xc++ -std=c++17 -ffreestanding %s
+// RUN: %clang_cc1 -fsyntax-only -verify -xc++ -std=c++23 -ffreestanding %s
 // expected-no-diagnostics
 
 /* Basic floating point conformance checks against:
+    - C23 Final Std.
     - N1570 draft of C11 Std.
     - N1256 draft of C99 Std.
     - http://port70.net/~nsz/c/c89/c89-draft.html draft of C89/C90 Std.
 */
 /*
+    C23,    5.2.5.3.3p21,   pp. 25
     C11,    5.2.4.2.2p11,   pp. 30
     C99,    5.2.4.2.2p9,    pp. 25
     C89,    2.2.4.2
@@ -207,6 +211,23 @@
     #error "Mandatory macros {FLT,DBL,LDBL}_MAX_10_EXP are invalid."
 #endif
 
+#if __STDC_VERSION__ >= 202311L || !defined(__STRICT_ANSI__)
+#ifndef FLT_NORM_MAX
+  #error "Mandatory macro FLT_NORM_MAX is missing."
+#else
+  _Static_assert(FLT_NORM_MAX >= 1.0E+37F, "Mandatory macro FLT_NORM_MAX is invalid.");
+#endif
+#ifndef DBL_NORM_MAX
+  #error "Mandatory macro DBL_NORM_MAX is missing."
+#else
+  _Static_assert(DBL_NORM_MAX >= 1.0E+37F, "Mandatory macro DBL_NORM_MAX is invalid.");
+#endif
+#ifndef LDBL_NORM_MAX
+  #error "Mandatory macro LDBL_NORM_MAX is missing."
+#else
+  _Static_assert(LDBL_NORM_MAX >= 1.0E+37F, "Mandatory macro LDBL_NORM_MAX is invalid.");
+#endif
+#endif
 
 /* Internal consistency checks */
 _Static_assert(FLT_RADIX == __FLT_RADIX__, "");
@@ -244,3 +265,9 @@ _Static_assert(LDBL_MAX_EXP == __LDBL_MAX_EXP__, "");
 _Static_assert(FLT_MAX_10_EXP == __FLT_MAX_10_EXP__, "");
 _Static_assert(DBL_MAX_10_EXP == __DBL_MAX_10_EXP__, "");
 _Static_assert(LDBL_MAX_10_EXP == __LDBL_MAX_10_EXP__, "");
+
+#if __STDC_VERSION__ >= 202311L || !defined(__STRICT_ANSI__)
+_Static_assert(FLT_NORM_MAX == __FLT_NORM_MAX__, "");
+_Static_assert(DBL_NORM_MAX == __DBL_NORM_MAX__, "");
+_Static_assert(LDBL_NORM_MAX == __LDBL_NORM_MAX__, "");
+#endif

github-actions · 2024-06-25T15:14:11Z

⚠️ C/C++ code formatter, clang-format found issues in your code. ⚠️

You can test this locally with the following command:

git-clang-format --diff 748a6ad68ba31d242e3f86eddd5b0fd7b2a02ff9 aee0da9de46bfadae00774ec9206f22bf0254d05 -- clang/lib/Frontend/InitPreprocessor.cpp clang/lib/Headers/float.h clang/test/Headers/float.c clang/test/Preprocessor/init-aarch64.c clang/test/Preprocessor/init.c

View the diff from clang-format here.

diff --git a/clang/lib/Frontend/InitPreprocessor.cpp b/clang/lib/Frontend/InitPreprocessor.cpp
index d40d78a385..4ab1e30e5e 100644
--- a/clang/lib/Frontend/InitPreprocessor.cpp
+++ b/clang/lib/Frontend/InitPreprocessor.cpp
@@ -148,7 +148,7 @@ static void DefineFloatMacros(MacroBuilder &Builder, StringRef Prefix,
   DefPrefix += "_";
 
   Builder.defineMacro(DefPrefix + "DENORM_MIN__", Twine(DenormMin)+Ext);
-  Builder.defineMacro(DefPrefix + "NORM_MAX__", Twine(NormMax)+Ext);
+  Builder.defineMacro(DefPrefix + "NORM_MAX__", Twine(NormMax) + Ext);
   Builder.defineMacro(DefPrefix + "HAS_DENORM__");
   Builder.defineMacro(DefPrefix + "DIG__", Twine(Digits));
   Builder.defineMacro(DefPrefix + "DECIMAL_DIG__", Twine(DecimalDigits));
diff --git a/clang/lib/Headers/float.h b/clang/lib/Headers/float.h
index e5c439a9d4..356c2cab78 100644
--- a/clang/lib/Headers/float.h
+++ b/clang/lib/Headers/float.h
@@ -88,9 +88,9 @@
 #  endif
 #if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 202311L) ||              \
     !defined(__STRICT_ANSI__)
-#    undef FLT_NORM_MAX
-#    undef DBL_NORM_MAX
-#    undef LDBL_NORM_MAX
+#undef FLT_NORM_MAX
+#undef DBL_NORM_MAX
+#undef LDBL_NORM_MAX
 #endif
 #endif
 
@@ -172,10 +172,10 @@
    /* C23 5.2.5.3.3p29-30 */
 #  define INFINITY (__builtin_inff())
 #  define NAN (__builtin_nanf(""))
-   /* C23 5.2.5.3.3p32 */
-#  define FLT_NORM_MAX __FLT_NORM_MAX__
-#  define DBL_NORM_MAX __DBL_NORM_MAX__
-#  define LDBL_NORM_MAX __LDBL_NORM_MAX__
+/* C23 5.2.5.3.3p32 */
+#define FLT_NORM_MAX __FLT_NORM_MAX__
+#define DBL_NORM_MAX __DBL_NORM_MAX__
+#define LDBL_NORM_MAX __LDBL_NORM_MAX__
 #endif
 
 #ifdef __STDC_WANT_IEC_60559_TYPES_EXT__

clang/test/Headers/float.c

jcranmer-intel · 2024-06-25T21:26:54Z

clang/lib/Frontend/InitPreprocessor.cpp

+  const char *DenormMin, *NormMax, *Epsilon, *Max, *Min;
+  NormMax = PickFP(Sem, "6.5504e+4", "3.40282347e+38",
+                   "1.7976931348623157e+308", "1.18973149535723176502e+4932",
+                   "1.79769313486231580793728971405301e+308",


I don't think this is the correct value for ppc double-double, since double-double is supposed to give a different result for NORM_MAX than MAX (indeed, that's explicitly the point of the new macros!)

Good catch, thank you! I see now that GCC uses the following values for long double:

#define __LDBL_MAX__ 1.79769313486231580793728971405301e+308L #define __LDBL_NORM_MAX__ 8.98846567431157953864652595394501e+307L

https://godbolt.org/z/M1zjxx7K5

so I should probably go with 8.98846567431157953864652595394501e+307 as the value? CC @hubert-reinterpretcast

Ping -- I think I've corrected the value (at least, I'm using the same value as GCC in this case).

I'm pretty sure this is correct, but I'd rather prefer someone who knows PPC to confirm this.

Ping @hubert-reinterpretcast @chenzheng1030 -- hoping to get this landed for Clang 19, just need verification that the value is correct here.

Yes, the value is correct for an interpretation where the LDBL_MANT_DIG is 106.

Thank you Hubert!

This value was gleaned from GCC's output

This adds the *_NORM_MAX macros to <float.h> for freestanding mode in Clang; the values were chosen based on the behavior seen coming from GCC and the values already produced for the *_MAX macros by Clang.

[C23] Add *_NORM_MAX macros to <float.h>

408c34b

This adds the *_NORM_MAX macros to <float.h> for freestanding mode in Clang; the values were chosen based on the behavior seen coming from GCC and the values already produced for the *_MAX macros by Clang.

AaronBallman added c23 clang:frontend Language frontend issues, e.g. anything involving "Sema" clang:headers Headers provided by Clang, e.g. for intrinsics floating-point Floating-point math labels Jun 25, 2024

AaronBallman assigned jrtc27, jyknight, andykaylor, jcranmer-intel and zahiraam Jun 25, 2024

llvmbot added clang Clang issues not falling into any other category backend:X86 labels Jun 25, 2024

Fix failing test cases

486d9fb

AaronBallman unassigned jrtc27, jyknight, andykaylor, jcranmer-intel and zahiraam Jun 25, 2024

AaronBallman requested review from jrtc27, jyknight, andykaylor, jcranmer-intel and zahiraam June 25, 2024 15:10

zahiraam reviewed Jun 25, 2024

View reviewed changes

clang/test/Headers/float.c Outdated Show resolved Hide resolved

jcranmer-intel reviewed Jun 25, 2024

View reviewed changes

AaronBallman added 3 commits June 26, 2024 08:19

Correct copy pasta in tests

cd2de3b

Correct value for PPC double double format

e8c7c83

This value was gleaned from GCC's output

Merge remote-tracking branch 'origin/main' into aballman/norm_max

aee0da9

AaronBallman merged commit 0e7590a into llvm:main Jul 10, 2024
5 of 7 checks passed

AaronBallman deleted the aballman/norm_max branch July 10, 2024 13:49

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[C23] Add *_NORM_MAX macros to <float.h> #96643

[C23] Add *_NORM_MAX macros to <float.h> #96643

Uh oh!

AaronBallman commented Jun 25, 2024

Uh oh!

llvmbot commented Jun 25, 2024 •

edited

Loading

Uh oh!

github-actions bot commented Jun 25, 2024 •

edited

Loading

Uh oh!

Uh oh!

jcranmer-intel Jun 25, 2024

Uh oh!

AaronBallman Jun 26, 2024 •

edited

Loading

Uh oh!

AaronBallman Jul 1, 2024

Uh oh!

jcranmer-intel Jul 3, 2024

Uh oh!

AaronBallman Jul 9, 2024

Uh oh!

hubert-reinterpretcast Jul 9, 2024

Uh oh!

AaronBallman Jul 10, 2024

Uh oh!

Uh oh!

Uh oh!

[C23] Add *_NORM_MAX macros to <float.h> #96643

[C23] Add *_NORM_MAX macros to <float.h> #96643

Uh oh!

Conversation

AaronBallman commented Jun 25, 2024

Uh oh!

llvmbot commented Jun 25, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Jun 25, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

jcranmer-intel Jun 25, 2024

Choose a reason for hiding this comment

Uh oh!

AaronBallman Jun 26, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

AaronBallman Jul 1, 2024

Choose a reason for hiding this comment

Uh oh!

jcranmer-intel Jul 3, 2024

Choose a reason for hiding this comment

Uh oh!

AaronBallman Jul 9, 2024

Choose a reason for hiding this comment

Uh oh!

hubert-reinterpretcast Jul 9, 2024

Choose a reason for hiding this comment

Uh oh!

AaronBallman Jul 10, 2024

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

llvmbot commented Jun 25, 2024 •

edited

Loading

github-actions bot commented Jun 25, 2024 •

edited

Loading

AaronBallman Jun 26, 2024 •

edited

Loading