![]() The following patch implements the compiler part of C++23 P1467R9 - Extended floating-point types and standard names compiler part by introducing _Float{16,32,64,128} as keywords and builtin types like they are implemented for C already since GCC 7, with DF{16,32,64,128}_ mangling. It also introduces _Float{32,64,128}x for C++ with the https://github.com/itanium-cxx-abi/cxx-abi/pull/147 proposed mangling of DF{32,64,128}x. The patch doesn't add anything for bfloat16_t support, as right now __bf16 type refuses all conversions and arithmetic operations. The patch wants to keep backwards compatibility with how __float128 has been handled in C++ before, both for mangling and behavior in binary operations, overload resolution etc. So, there are some backend changes where for C __float128 and _Float128 are the same type (float128_type_node and float128t_type_node are the same pointer), but for C++ they are distinct types which mangle differently and _Float128 is treated as extended floating-point type while __float128 is treated as non-standard floating point type. The various C++23 changes about how floating-point types are changed are actually implemented as written in the spec only if at least one of the types involved is _Float{16,32,64,128,32x,64x,128x} (_FloatNx are also treated as extended floating-point types) and kept previous behavior otherwise. For float/double/long double the rules are actually written that they behave the same as before. There is some backwards incompatibility at least on x86 regarding _Float16, because that type was already used by that name and with the DF16_ mangling (but only since GCC 12 and I think it isn't that widely used in the wild yet). E.g. config/i386/avx512fp16intrin.h shows the issues, where in C or in GCC 12 in C++ one could pass 0.0f to a builtin taking _Float16 argument, but with the changes that is not possible anymore, one needs to either use 0.0f16 or (_Float16) 0.0f. We have also a problem with glibc headers, where since glibc 2.27 math.h and complex.h aren't compilable with these changes. One gets errors like: In file included from /usr/include/math.h:43, from abc.c:1: /usr/include/bits/floatn.h:86:9: error: multiple types in one declaration 86 | typedef __float128 _Float128; | ^~~~~~~~~~ /usr/include/bits/floatn.h:86:20: error: declaration does not declare anything [-fpermissive] 86 | typedef __float128 _Float128; | ^~~~~~~~~ In file included from /usr/include/bits/floatn.h:119: /usr/include/bits/floatn-common.h:214:9: error: multiple types in one declaration 214 | typedef float _Float32; | ^~~~~ /usr/include/bits/floatn-common.h:214:15: error: declaration does not declare anything [-fpermissive] 214 | typedef float _Float32; | ^~~~~~~~ /usr/include/bits/floatn-common.h:251:9: error: multiple types in one declaration 251 | typedef double _Float64; | ^~~~~~ /usr/include/bits/floatn-common.h:251:16: error: declaration does not declare anything [-fpermissive] 251 | typedef double _Float64; | ^~~~~~~~ This is from snippets like: /* The remaining of this file provides support for older compilers. */ # if __HAVE_FLOAT128 /* The type _Float128 exists only since GCC 7.0. */ # if !__GNUC_PREREQ (7, 0) || defined __cplusplus typedef __float128 _Float128; # endif where it hardcodes that C++ doesn't have _Float{16,32,64,128,32x,64x,128x} support nor {f,F}{16,32,64,128}{,x} literal suffixes nor _Complex _Float{16,32,64,128,32x,64x,128x}. The patch fixincludes this for now and hopefully if this is committed, then glibc can change those. The patch changes those # if !__GNUC_PREREQ (7, 0) || defined __cplusplus conditions to # if !__GNUC_PREREQ (7, 0) || (defined __cplusplus && !__GNUC_PREREQ (13, 0)) Another thing is mangling, as said above, Itanium C++ ABI specifies DF <number> _ as _Float{16,32,64,128} mangling, but GCC was implementing a mangling incompatible with that starting with DF for fixed point types. Fixed point was never supported in C++ though, I believe the reason why the mangling has been added was that due to a bug it would leak into the C++ FE through decltype (0.0r) etc. But that has been shortly after the mangling was added fixed (I think in the same GCC release cycle), so we now reject 0.0r etc. in C++. If we ever need the fixed point mangling, I think it can be readded but better with a different prefix so that it doesn't conflict with the published standard manglings. So, this patch also kills the fixed point mangling and implements the DF <number> _ demangling. The patch predefines __STDCPP_FLOAT{16,32,64,128}_T__ macros when those types are available, but only for C++23, while the underlying types are available in C++98 and later including the {f,F}{16,32,64,128} literal suffixes (but those with a pedwarn for C++20 and earlier). My understanding is that it needs to be predefined by the compiler, on the other side predefining even for older modes when <stdfloat> is a new C++23 header would be weird. One can find out if _Float{16,32,64,128,32x,64x,128x} is supported in C++ by __GNUC__ >= 13 && defined(__FLT{16,32,64,128,32X,64X,128X}_MANT_DIG__) (but that doesn't work well with older G++ 13 snapshots). As for std::bfloat16_t, three targets (aarch64, arm and x86) apparently "support" __bf16 type which has the bfloat16 format, but isn't really usable, e.g. {aarch64,arm,ix86}_invalid_conversion disallow any conversions from or to type with BFmode, {aarch64,arm,ix86}_invalid_unary_op disallows any unary operations on those except for ADDR_EXPR and {aarch64,arm,ix86}_invalid_binary_op disallows any binary operation on those. So, I think we satisfy: "If the implementation supports an extended floating-point type with the properties, as specified by ISO/IEC/IEEE 60559, of radix (b) of 2, storage width in bits (k) of 16, precision in bits (p) of 8, maximum exponent (emax) of 127, and exponent field width in bits (w) of 8, then the typedef-name std::bfloat16_t is defined in the header <stdfloat> and names such a type, the macro __STDCPP_BFLOAT16_T__ is defined, and the floating-point literal suffixes bf16 and BF16 are supported." because we don't really support those right now. 2022-09-27 Jakub Jelinek <jakub@redhat.com> PR c++/106652 PR c++/85518 gcc/ * tree-core.h (enum tree_index): Add TI_FLOAT128T_TYPE enumerator. * tree.h (float128t_type_node): Define. * tree.cc (build_common_tree_nodes): Initialize float128t_type_node. * builtins.def (DEF_FLOATN_BUILTIN): Adjust comment now that _Float<N> is supported in C++ too. * config/i386/i386.cc (ix86_mangle_type): Only mangle as "g" float128t_type_node. * config/i386/i386-builtins.cc (ix86_init_builtin_types): Use float128t_type_node for __float128 instead of float128_type_node and create it if NULL. * config/i386/avx512fp16intrin.h (_mm_setzero_ph, _mm256_setzero_ph, _mm512_setzero_ph, _mm_set_sh, _mm_load_sh): Use 0.0f16 instead of 0.0f. * config/ia64/ia64.cc (ia64_init_builtins): Use float128t_type_node for __float128 instead of float128_type_node and create it if NULL. * config/rs6000/rs6000-c.cc (is_float128_p): Also return true for float128t_type_node if non-NULL. * config/rs6000/rs6000.cc (rs6000_mangle_type): Don't mangle float128_type_node as "u9__ieee128". * config/rs6000/rs6000-builtin.cc (rs6000_init_builtins): Use float128t_type_node for __float128 instead of float128_type_node and create it if NULL. gcc/c-family/ * c-common.cc (c_common_reswords): Change _Float{16,32,64,128} and _Float{32,64,128}x flags from D_CONLY to 0. (shorten_binary_op): Punt if common_type returns error_mark_node. (shorten_compare): Likewise. (c_common_nodes_and_builtins): For C++ record _Float{16,32,64,128} and _Float{32,64,128}x builtin types if available. For C++ clear float128t_type_node. * c-cppbuiltin.cc (c_cpp_builtins): Predefine __STDCPP_FLOAT{16,32,64,128}_T__ for C++23 if supported. * c-lex.cc (interpret_float): For q/Q suffixes prefer float128t_type_node over float128_type_node. Allow {f,F}{16,32,64,128} suffixes for C++ if supported with pedwarn for C++20 and older. Allow {f,F}{32,64,128}x suffixes for C++ with pedwarn. Don't call excess_precision_type for C++. gcc/cp/ * cp-tree.h (cp_compare_floating_point_conversion_ranks): Implement P1467R9 - Extended floating-point types and standard names except for std::bfloat16_t for now. Declare. (extended_float_type_p): New inline function. * mangle.cc (write_builtin_type): Mangle float{16,32,64,128}_type_node as DF{16,32,64,128}_. Mangle float{32,64,128}x_type_node as DF{32,64,128}x. Remove FIXED_POINT_TYPE mangling that conflicts with that. * typeck2.cc (check_narrowing): If one of ftype or type is extended floating-point type, compare floating-point conversion ranks. * parser.cc (cp_keyword_starts_decl_specifier_p): Handle CASE_RID_FLOATN_NX. (cp_parser_simple_type_specifier): Likewise and diagnose missing _Float<N> or _Float<N>x support if not supported by target. * typeck.cc (cp_compare_floating_point_conversion_ranks): New function. (cp_common_type): If both types are REAL_TYPE and one or both are extended floating-point types, select common type based on comparison of floating-point conversion ranks and subranks. (cp_build_binary_op): Diagnose operation with floating point arguments with unordered conversion ranks. * call.cc (standard_conversion): For floating-point conversion, if either from or to are extended floating-point types, set conv->bad_p for implicit conversion from larger to smaller conversion rank or with unordered conversion ranks. (convert_like_internal): Emit a pedwarn on such conversions. (build_conditional_expr): Diagnose operation with floating point arguments with unordered conversion ranks. (convert_arg_to_ellipsis): Don't promote extended floating-point types narrower than double to double. (compare_ics): Implement P1467R9 [over.ics.rank]/4 changes. gcc/testsuite/ * g++.dg/cpp23/ext-floating1.C: New test. * g++.dg/cpp23/ext-floating2.C: New test. * g++.dg/cpp23/ext-floating3.C: New test. * g++.dg/cpp23/ext-floating4.C: New test. * g++.dg/cpp23/ext-floating5.C: New test. * g++.dg/cpp23/ext-floating6.C: New test. * g++.dg/cpp23/ext-floating7.C: New test. * g++.dg/cpp23/ext-floating8.C: New test. * g++.dg/cpp23/ext-floating9.C: New test. * g++.dg/cpp23/ext-floating10.C: New test. * g++.dg/cpp23/ext-floating.h: New file. * g++.target/i386/float16-1.C: Adjust expected diagnostics. libcpp/ * expr.cc (interpret_float_suffix): Allow {f,F}{16,32,64,128} and {f,F}{32,64,128}x suffixes for C++. include/ * demangle.h (enum demangle_component_type): Add DEMANGLE_COMPONENT_EXTENDED_BUILTIN_TYPE. (struct demangle_component): Add u.s_extended_builtin member. libiberty/ * cp-demangle.c (d_dump): Handle DEMANGLE_COMPONENT_EXTENDED_BUILTIN_TYPE. Don't handle DEMANGLE_COMPONENT_FIXED_TYPE. (d_make_extended_builtin_type): New function. (cplus_demangle_builtin_types): Add _Float entry. (cplus_demangle_type): For DF demangle it as _Float<N> or _Float<N>x rather than fixed point which conflicts with it. (d_count_templates_scopes): Handle DEMANGLE_COMPONENT_EXTENDED_BUILTIN_TYPE. Just break; for DEMANGLE_COMPONENT_FIXED_TYPE. (d_find_pack): Handle DEMANGLE_COMPONENT_EXTENDED_BUILTIN_TYPE. Don't handle DEMANGLE_COMPONENT_FIXED_TYPE. (d_print_comp_inner): Likewise. * cp-demangle.h (D_BUILTIN_TYPE_COUNT): Bump. * testsuite/demangle-expected: Replace _Z3xxxDFyuVb test with _Z3xxxDF16_DF32_DF64_DF128_CDF16_Vb. Add _Z3xxxDF32xDF64xDF128xCDF32xVb test. fixincludes/ * inclhack.def (glibc_cxx_floatn_1, glibc_cxx_floatn_2, glibc_cxx_floatn_3): New fixes. * tests/base/bits/floatn.h: New file. * fixincl.x: Regenerated. |
||
---|---|---|
.. | ||
tests/base | ||
aclocal.m4 | ||
ChangeLog | ||
check.tpl | ||
config.h.in | ||
configure | ||
configure.ac | ||
fixfixes.c | ||
fixinc.in | ||
fixincl.c | ||
fixincl.tpl | ||
fixincl.x | ||
fixlib.c | ||
fixlib.h | ||
fixopts.c | ||
fixtests.c | ||
genfixes | ||
inclhack.def | ||
Makefile.in | ||
mkfixinc.sh | ||
mkheaders.in | ||
procopen.c | ||
README | ||
README-fixinc | ||
server.c | ||
server.h | ||
system.h |
GCC MAINTAINER INFORMATION ========================== If you are having some problem with a system header that is either broken by the manufacturer, or is broken by the fixinclude process, then you will need to alter or add information to the include fix definitions file, ``inclhack.def''. Please also send relevant information to gcc-bugs@gcc.gnu.org, gcc-patches@gcc.gnu.org and, please, to me: bkorb@gnu.org. To make your fix, you will need to do several things: 1. Obtain access to the AutoGen program on some platform. It does not have to be your build platform, but it is more convenient. 2. Edit "inclhack.def" to reflect the changes you need to make. See below for information on how to make those changes. 3. Run the "genfixes" shell script to produce a new copy of the "fixincl.x" file. 4. Rebuild the compiler and check the header causing the issue. Make sure it is now properly handled. Add tests to the "test_text" entry(ies) that validate your fix. This will help ensure that future fixes won't negate your work. Do *NOT* specify test text for "wrap" or "replacement" fixes. There is no real possibility that these fixes will fail. If they do, you will surely know straight away. NOTE: "test_text" is interpreted by the shell as it gets copied into the test header. THEREFORE you must quote dollar sign characters and back quotes -- unless you mean for them to be interpreted by the shell. e.g. the math_huge_val_from_dbl_max test_text needs to put text into both float.h and math.h, so it includes a back-quoted script to add text to float.h. 5. Go into the fixincludes build directory and type, "make check". You are guaranteed to have issues printed out as a result. Look at the diffs produced. Make sure you have not clobbered the proper functioning of a different fix. Make sure your fix is properly tested and it does what it is supposed to do. 6. Now that you have the right things happening, synchronize the $(srcdir)/tests/base directory with the $(builddir)/tests/res directory. The output of "make check" will be some diffs that should give you some hints about what to do. 7. Rerun "make check" and verify that there are no issues left. MAKING CHANGES TO INCLHACK.DEF ============================== 0. If you are not the fixincludes maintainer, please send that person email about any changes you may want to make. Thanks! 1. Every fix must have a "hackname" that is compatible with C syntax for variable names and is unique without regard to alphabetic case. Please keep them alphabetical by this name. :-) 2. If the problem is known to exist only in certain files, then identify the files with "files = " entries. If you use fnmatch(3C) wild card characters in a "files" entry, be certain that the first "files" entry has no such character. Otherwise, the "make check" machinery will attempt to create files with those characters in the name. That is inconvenient. 3. It is relatively expensive to fire off a process to fix a source file, therefore write apply tests to avoid unnecessary fix processes. The preferred apply tests are "select", "bypass", "mach" "sum", and "c-test" because they are performed internally: * select - Run a regex on the contents of the file being considered. All such regex-es must match. Matching is done with extended regular expressions. * bypass - Run a regex on the contents of the file being considered. No such regex may match. * sum - Select a specific version of a file that has a matching check sum. The BSD version of checksum ["sum(1BSD)"] is used. Each "sum" entry should contain exactly three space separated tokens: the sum, some number and the basename of the file. The "some number" is ignored. If there are multiple "sum" entries, only one needs to match in order to pass. For example: sum = '1234 3 foobar.h'; specifies that the "foobar.h" header in any directory will match if it has the checksum 1234. * c_test - call a function in fixtests.c. See that file. * files - the "fnmatch" pattern of the file(s) to examine for the issue. There may be several copies of this attribute. If the header lives in a /usr/include subdirectory, be sure to include that subdirectory in the name. e.g. net/if.h * mach - Match the output of config.guess against a series of fnmatch patterns. It must match at least one of the patterns, unless "not-machine" has also been specified. In that case, the config.guess output must not match any of the patterns. The next test is relatively slow because it must be handled in a separate shell process. Some platforms do not support server shells, so the whole process is even slower and more cumbersome there. * test - These should be arguments to the program, "/bin/test". You may perform multiple commands, if you enclose them in backquotes and echo out valid test arguments. For example, you might echo out '0 -eq 1' if you want a false result, or '0 -eq 0' for a true result. These tests are required to: 1. Be positive for all header files that require the fix. It is desirable to: 2. Be negative as often as possible whenever the fix is not required, avoiding the process overhead. It is nice if: 3. The expression is as simple as possible to both process and understand by people. :-) Please take advantage of the fact AutoGen will glue together string fragments. It helps. Also take note that double quote strings and single quote strings have different formation rules. Double quote strings are a tiny superset of ANSI-C string syntax. Single quote strings follow shell single quote string formation rules, except that the backslash is processed before '\\', '\'' and '#' characters (using C character syntax). Each test must pass or the fix is not applied. For example, all "select" expressions must be found and not one "bypass" selection may be found. Examples of test specifications: hackname = broken_assert_stdio; files = assert.h; select = stderr; bypass = "include.*stdio.h"; The ``broken_assert_stdio'' fix will be applied only to a file named "assert.h" if it contains the string "stderr" _and_ it does _not_ contain the expression "include.*stdio.h". hackname = no_double_slash; c_test = "double_slash"; The ``no_double_slash'' fix will be applied if the ``double_slash_test()'' function says to. See ``fixtests.c'' for documentation on how to include new functions into that module. 4. There are currently four methods of fixing a file: 1. a series of sed expressions. Each will be an individual "-e" argument to a single invocation of sed. Unless you need to use multiple or complex sed expressions, please use the "replacement text" method instead. 2. a shell script. These scripts are _required_ to read all of stdin in order to avoid pipe stalls. They may choose to discard the input. 3. Replacement text. If the replacement is empty, then no fix is applied. Otherwise, the replacement text is written to the output file and no further fixes are applied. If you really want a no-op file, replace the file with a comment. Replacement text "fixes" must be first in this file!! 4. A C language subroutine method for both tests and fixes. See ``fixtests.c'' for instructions on writing C-language applicability tests and ``fixfixes.c'' for C-language fixing. These files also contain tables that describe the currently implemented fixes and tests. If at all possible, you should try to use one of the C language fixes as it is far more efficient. There are currently five such fixes, three of which are very special purpose: i) char_macro_def - This function repairs the definition of an ioctl macro that presumes CPP macro substitution within pairs of single quote characters. ii) char_macro_use - This function repairs the usage of ioctl macros that no longer can wrap an argument with single quotes. iii) machine_name - This function will look at "#if", "#ifdef", "#ifndef" and "#elif" directive lines and replace the first occurrence of a non-reserved name that is traditionally pre-defined by the native compiler. The next two are for general use: iv) wrap - wraps the entire file with "#ifndef", "#define" and "#endif" self-exclusionary text. It also, optionally, inserts a prolog after the "#define" and an epilog just before the "#endif". You can use this for a fix as follows: c_fix = wrap; c_fix_arg = "/* prolog text */"; c_fix_arg = "/* epilog text */"; If you want an epilog without a prolog, set the first "c_fix_arg" to the empty string. Both or the second "c_fix_arg"s may be omitted and the file will still be wrapped. THERE IS A SPECIAL EXCEPTION TO THIS, HOWEVER: If the regular expression '#if.*__need' is found, then it is assumed that the file needs to be read and interpreted more than once. However, the prolog and epilog text (if any) will be inserted. v) format - Replaces text selected with a regular expression with a specialized formating string. The formatting works as follows: The format text is copied to the output until a '%' character is found. If the character after the '%' is another '%', then one '%' is output and processing continues. If the following character is not a digit, then the '%' and that character are copied and processing continues. Finally, if the '%' *is* followed by a digit, that digit is used as an index into the regmatch_t array to replace the two characters with the matched text. i.e.: "%0" is replaced by the full matching text, "%1" is the first matching sub-expression, etc. This is used as follows: c_fix = format; c_fix_arg = "#ifndef %1\n%0\n#endif"; c_fix_arg = "#define[ \t]+([A-Z][A-Z0-9a-z_]*).*"; This would wrap a one line #define inside of a "#ifndef"/"#endif" pair. The second "c_fix_arg" may be omitted *IF* there is at least one select clause and the first one identifies the text you wish to reformat. It will then be used as the second "c_fix_arg". You may delete the selected text by supplying an empty string for the replacement format (the first "c_fix_arg"). Note: In general, a format c_fix may be used in place of one sed expression. However, it will need to be rewritten by hand. For example: sed = 's@^#if __GNUC__ == 2 && __GNUC_MINOR__ >= 7$' '@& || __GNUC__ >= 3@'; may be rewritten using a format c_fix as: c_fix = format; c_fix_arg = '%0 || __GNUC__ >= 3'; c_fix_arg = '^#if __GNUC__ == 2 && __GNUC_MINOR__ >= 7$'; Multiple sed substitution expressions probably ought to remain sed expressions in order to maintain clarity. Also note that if the second sed expression is the same as the first select expression, then you may omit the second c_fix_arg. The select expression will be picked up and used in its absence. EXAMPLES OF FIXES: ================== hackname = AAA_ki_iface; replace; /* empty replacement -> no fixing the file */ When this ``fix'' is invoked, it will prevent any fixes from being applied. ------------------ hackname = AAB_svr4_no_varargs; replace = "/* This file was generated by fixincludes. */\n" "#ifndef _SYS_VARARGS_H\n" "#define _SYS_VARARGS_H\n\n" "#ifdef __STDC__\n" "#include <stdarg.h>\n" "#else\n" "#include <varargs.h>\n" "#endif\n\n" "#endif /* _SYS_VARARGS_H */\n"; When this ``fix'' is invoked, the replacement text will be emitted into the replacement include file. No further fixes will be applied. ------------------ hackname = hpux11_fabsf; files = math.h; select = "^[ \t]*#[ \t]*define[ \t]+fabsf\\(.*"; bypass = "__cplusplus"; c_fix = format; c_fix_arg = "#ifndef __cplusplus\n%0\n#endif"; test_text = "# define fabsf(x) ((float)fabs((double)(float)(x)))\n"; This fix will ensure that the #define for fabs is wrapped with C++ protection, providing the header is not already C++ aware. ------------------ 5. Testing fixes. The brute force method is, of course, to configure and build GCC. But you can also: cd ${top_builddir}/gcc rm -rf include-fixed/ stmp-fixinc make stmp-fixinc I would really recommend, however: cd ${top_builddir}/fixincludes make check To do this, you *must* have autogen installed on your system. The "check" step will proceed to construct a shell script that will exercise all the fixes, using the sample test_text provided with each fix. Once done, the changes made will be compared against the changes saved in the source directory. If you are changing the tests or fixes, the change will likely be highlighted.