GCC 10 Release Series
Changes, New Features, and Fixes

This page is a "brief" summary of some of the huge number of improvements in GCC 10. You may also want to check out our Porting to GCC 10 page and the full GCC documentation.

Caveats

• An ABI incompatibility between C++14 and C++17 has been fixed. On some targets a class with a zero-sized subobject would be passed incorrectly when compiled as C++17 or C++20. See the C++ notes below for more details.
• The deprecated Profile Mode and array_allocator extensions have been removed from libstdc++.
• The non-standard std::__is_nullptr_t type trait is deprecated and will be removed from libstdc++ in a future release. The standard trait std::is_null_pointer should be instead.
• The minimum version of the MPFR library required for building GCC has been increased to version 3.1.0 (released 2011-10-03).
• The automatic template instantiation at link time (-frepo) has been removed.
• The --param allow-store-data-races internal parameter has been removed in favor of a new official option -fallow-store-data-races. While default behavior is unchanged and the new option allows to correctly maintain a per compilation unit setting across link-time optimization, alteration of the default via --param allow-store-data-races will now be diagnosed and build systems have to be adjusted accordingly.
• Offloading to Heterogeneous System Architecture Intermediate Language (HSAIL) has been deprecated and will likely be removed in a future release.

General Improvements

• New built-in functions:
  • The __has_builtin built-in preprocessor operator can be used to query support for built-in functions provided by GCC and other compilers that support it.
  • __builtin_roundeven for the corresponding function from ISO/IEC TS 18661.
• New command-line options:
  • -fallocation-dce removes unneeded pairs of new and delete operators.
  • -fprofile-partial-training can now be used to inform the compiler that code paths not covered by the training run should not be optimized for size.
  • -fprofile-reproducible controls level of reproducibility of profile gathered by -fprofile-generate. This makes it possible to rebuild program with same outcome which is useful, for example, for distribution packages.
  • -fprofile-prefix-path can be used in combination with -fprofile-generate=profile_dir and -fprofile-use=profile_dir to inform GCC where the base directory of build source tree is in case it differs between instrumentation and optimized builds.
  • -fanalyzer enables a new static analysis pass and associated warnings. This pass performs a time-consuming exploration of paths through the code in the hope of detecting various common errors, such as double-free bugs. This option should be regarded as experimental in this release. In particular, analysis of non-C code is unlikely to work.
• Inter-procedural optimization improvements:
  • The inter-procedural scalar replacement of aggregates (IPA-SRA) pass was re-implemented to work at link-time and can now also remove computing and returning unused return values.
  • -finline-functions is now enabled at -O2 and was retuned for better code size versus runtime performance trade-offs. Inliner heuristics was also significantly sped up to avoid negative impact to -flto -O2 compile times.
  • Inliner heuristics and function cloning can now use value-range information to predict effectivity of individual transformations.
  • During link-time optimization the C++ One Definition Rule is used to increase precision of type based alias analysis.
• Link-time optimization improvements:
  • A new binary lto-dump has been added. The program can dump various information about a LTO bytecode object file.
  • Parallel phase of the LTO can automatically detect a running make's jobserver or can fall back to number of available cores.
  • The LTO bytecode can be compressed with zstd algorithm. Configure script can automatically detect the zstd support.
  • Most --param values can now be specified at translation unit granularity. This includes all parameters controlling the inliner and other inter-procedural optimizations. Unlike earlier releases, GCC 10 will ignore parameters controlling optimizations specified at link-time and apply parameters specified at compile-time in the same manner as done for optimization flags.
• Profile driven optimization improvements:
  • Profile maintenance during compilation and hot/cold code partitioning have been improved.
  • Using -fprofile-values, an instrumented binary can track multiple values (up to 4) for e.g. indirect calls and provide more precise profile information.

New Languages and Language-Specific Improvements

• Version 2.6 of the OpenACC specification is now supported in the C, C++ and Fortran compilers. See the implementation status section on the OpenACC wiki page and the run-time library documentation for further information.
• GCC 10 adds a number of newly implemented OpenMP 5.0 features on top of the GCC 9 release such as conditional lastprivate clause, scan and loop directives, order(concurrent) and use_device_addr clauses support, if clause on simd construct or partial support for the declare variant directive, getting closer to full support of the OpenMP 5.0 standard.
• OpenMP and OpenACC now support offloading to AMD Radeon (GCN) GPUs; supported are the third-generation Fiji (fiji) and the fifth-generation VEGA 10/VEGA 20 (gfx900 or gfx906).

C family

• New attributes:
  • The access function and type attribute has been added to describe how a function accesses objects passed to it by pointer or reference, and to associate such arguments with integer arguments denoting the objects' sizes. The attribute is used to enable the detection of invalid accesses by user-defined functions, such as those diagnosed by -Wstringop-overflow.
  • The symver attribute can be used to bind symbols to specific version nodes on ELF platforms. This is preferred to using inline assembly with GNU as symver directive because the latter is not compatible with link-time optimizations.
• New warnings:
  • -Wstring-compare, enabled by -Wextra, warns about equality and inequality expressions between zero and the result of a call to either strcmp and strncmp that evaluate to a constant as a result of the length of one argument being greater than the size of the array pointed to by the other.
  • -Wzero-length-bounds, enabled by -Warray-bounds, warns about accesses to elements of zero-length arrays that might overlap other members of the same object.
• Enhancements to existing warnings:
  • -Warray-bounds detects more out-of-bounds accesses to member arrays as well as accesses to elements of zero-length arrays.
  • -Wformat-overflow makes full use of string length information computed by the strlen optimization pass.
  • -Wrestrict detects overlapping accesses to dynamically allocated objects.
  • -Wreturn-local-addr diagnoses more instances of return statements returning addresses of automatic variables.
  • -Wstringop-overflow detects more out-of-bounds stores to member arrays including zero-length arrays, dynamically allocated objects and variable length arrays, as well as more instances of reads of unterminated character arrays by string built-in functions. The warning also detects out-of-bounds accesses by calls to user-defined functions declared with the new attribute access.
• Extended characters in identifiers may now be specified directly in the input encoding (UTF-8, by default), in addition to the UCN syntax (\uNNNN or \UNNNNNNNN) that is already supported:

  static const int π = 3;
  int get_naïve_pi() {
    return π;
  }

C

• Several new features from the upcoming C2X revision of the ISO C standard are supported with -std=c2x and -std=gnu2x. Some of these features are also supported as extensions when compiling for older language versions. In addition to the features listed, some features previously supported as extensions and now added to the C standard are enabled by default in C2X mode and not diagnosed with -std=c2x -Wpedantic.
  • The [[]] attribute syntax is supported, as in C++. Existing attributes can be used with this syntax in forms such as [[gnu::const]]. The standard attributes [[deprecated]], [[fallthrough]] and [[maybe_unused]] are supported.
  • UTF-8 character constants using the u8'' syntax are supported.
  • <float.h> defines macros FLT_NORM_MAX, DBL_NORM_MAX and LDBL_NORM_MAX.
  • When decimal floating-point arithmetic is supported, <float.h> defines macros DEC32_TRUE_MIN, DEC64_TRUE_MIN and DEC128_TRUE_MIN, in addition to the macros that were previously only defined if __STDC_WANT_DEC_FP__ was defined before including <float.h>.
  • In C2X mode, empty parentheses in a function definition give that function a type with a prototype for subsequent calls; other old-style function definitions are diagnosed by default in C2X mode.
  • The strftime format checking supports the %OB and %Ob formats.
  • In C2X mode, -fno-fp-int-builtin-inexact is enabled by default.
• GCC now defaults to -fno-common. As a result, global variable accesses are more efficient on various targets. In C, global variables with multiple tentative definitions now result in linker errors. With -fcommon such definitions are silently merged during linking.

C++

• Several C++20 features have been implemented:
  • P1668R1, Permitting Unevaluated inline-assembly in constexpr Functions
  • P1161R3, Deprecate a[b,c]
  • P0848R3, Conditionally Trivial Special Member Functions
  • P1091R3, Extending structured bindings
  • P1143R2, Adding the constinit keyword
  • P1152R4, Deprecating volatile
  • P0388R4, Permit conversions to arrays of unknown bound
  • P0784R7, constexpr new
  • Concepts, including P0734R0, P0857R0, P1084R2, P1141R2, P0848R3, P1616R1, P1452R2
  • P1301R4, [[nodiscard("with reason")]]
  • P1814R0, class template argument deduction for alias templates
  • P1816R0, class template argument deduction for aggregates
  • P0960R3, Parenthesized initialization of aggregates
  • P1331R2, Allow trivial default initialization in constexpr contexts
  • P1327R1, Allowing dynamic_cast and polymorphic typeid in constexpr contexts
  • P0912R5, Coroutines (requires -fcoroutines)
• Several C++ Defect Reports have been resolved, e.g.:
  • DR 1560, lvalue-to-rvalue conversion in ?:
  • DR 1813, __is_standard_layout for a class with repeated bases
  • DR 2094, volatile scalars are trivially copyable,
  • DR 2096, constraints on literal unions
  • DR 2413, typename in conversion-function-ids
  • DR 2352, Similar types and reference binding
  • DR 1601, Promotion of enumeration with fixed underlying type
  • DR 330, Qualification conversions and pointers to arrays of pointers
  • DR 1307, Overload resolution based on size of array initializer-list
  • DR 1710, Missing template keyword in class-or-decltype
• New warnings:
  • -Wmismatched-tags, disabled by default, warns about declarations of structs, classes, and class templates and their specializations with a class-key that does not match either the definition or the first declaration if no definition is provided. The option is provided to ease portability to Windows-based compilers.
  • -Wredundant-tags, disabled by default, warns about redundant class-key and enum-key in contexts where the key can be eliminated without causing an syntactic ambiguity.
• G++ can now detect modifying constant objects in constexpr evaluation (which is undefined behavior).
• G++ no longer emits bogus -Wsign-conversion warnings with explicit casts.
• Narrowing is now detected in more contexts (e.g., case values).
• Memory consumption of the compiler has been reduced in constexpr evaluation.
• The noexcept-specifier is now properly treated as a complete-class context as per [class.mem].
• The attribute deprecated can now be used on namespaces too.
• The ABI of passing and returning certain C++ classes by value changed on several targets in GCC 10, including AArch64, ARM, PowerPC ELFv2, S/390 and Itanium. These changes affect classes with a zero-sized subobject (an empty base class, or data member with the [[no_unique_address]] attribute) where all other non-static data members have the same type (this is called a "homogeneous aggregate" in some ABI specifications, or if there is only one such member, a "single element"). In -std=c++17 and -std=c++20 modes, classes with an empty base class were not considered to have a single element or to be a homogeneous aggregate, and so could be passed differently (in the wrong registers or at the wrong stack address). This could make code compiled with -std=c++17 and -std=c++14 ABI incompatible. This has been corrected and the empty bases are ignored in those ABI decisions, so functions compiled with -std=c++14 and -std=c++17 are now ABI compatible again. Example: struct empty {}; struct S : empty { float f; }; void f(S);. Similarly, in classes containing non-static data members with empty class types using the C++20 [[no_unique_address]] attribute, those members weren't ignored in the ABI argument passing decisions as they should be. Both of these ABI changes are now diagnosed with -Wpsabi.

Runtime Library (libstdc++)

• Improved experimental C++2a support, including:
  • Library concepts in <concepts> and <iterator>.
  • Constrained algorithms in <ranges>, <algorithm>, and <memory> (thanks to Patrick Palka).
  • New algorithms shift_left and shift_right (thanks to Patrick Palka).
  • std::span (thanks to JeanHeyd Meneide).
  • Three-way comparisons in <compare> and throughout the library.
  • Constexpr support in <algorithm> and elsewhere (thanks to Edward Smith-Rowland).
  • <stop_token> and std::jthread (thanks to Thomas Rodgers).
  • std::atomic_ref and std::atomic<floating point>.
  • Integer comparison functions (cmp_equal, cmp_less etc.).
  • std::ssize, std::to_array.
  • std::construct_at, std::destroy, constexpr std::allocator.
  • Mathematical constants in <numbers>.
• Support for RDSEED in std::random_device.
• Reduced header dependencies, leading to faster compilation for some code.

Fortran

• use_device_addr of version 5.0 of the OpenMP specification is now supported. Note that otherwise OpenMP 4.5 is partially supported in the Fortran compiler; the largest missing item is structure element mapping.
• The default buffer size for I/O using unformatted files has been increased to 1048576. The buffer size for can now be set at runtime via the environment variables GFORTRAN_FORMATTED_BUFFER_SIZE and GFORTRAN_UNFORMATTED_BUFFER_SIZE for formatted and unformatted files, respectively.
• Mismatches between actual and dummy argument lists in a single file are now rejected with an error. Use the new option -fallow-argument-mismatch to turn these errors into warnings; this option is implied with -std=legacy. -Wargument-mismatch has been removed.
• The handling of a BOZ literal constant has been reworked to provide better conformance to the Fortran 2008 and 2018 standards. In these Fortran standards, a BOZ literal constant is a typeless and kindless entity. As a part of the rework, documented and undocumented extensions to the Fortran standard now emit errors during compilation. Some of these extensions are permitted with the -fallow-invalid-boz, where the error is degraded to a warning and the code is compiled as with older gfortran.
• At any optimization level except-Os, gfortran now uses inline packing for arguments instead of calling a library routine. If the source contains a large number of arguments that need to be repacked, code size or time for compilation can become excessive. If that is the case, -fno-inline-arg-packing can be used to disable inline argument packing.
• Legacy extensions:
  • For formatted input/output, if the explicit widths after the data-edit descriptors I, F and G have been omitted, default widths are used.
  • A blank format item at the end of a format specification, i.e. nothing following the final comma, is allowed. Use the option -fdec-blank-format-item; this option is implied with -fdec.
  • The existing support for AUTOMATIC and STATIC attributes has been extended to allow variables with the AUTOMATIC attribute to be used in EQUIVALENCE statements. Use -fdec-static; this option is implied by -fdec.
  • Allow character literals in assignments and DATA statements for numeric (INTEGER, REAL, or COMPLEX) or LOGICAL variables. Use the option -fdec-char-conversions; this option is implied with -fdec.
  • DEC comparisons, i.e. allow Hollerith constants to be used in comparisons with INTEGER, REAL, COMPLEX and CHARACTER expressions. Use the option -fdec.
• Character type names in errors and warnings now include len in addition to kind; * is used for assumed length. The kind is omitted if it is the default kind. Examples: CHARACTER(12), CHARACTER(6,4).
• CO_BROADCAST now supports derived type variables including objects with allocatable components. In this case, the optional arguments STAT= and ERRMSG= are currently ignored.
• The handling of module and submodule names has been reworked to allow the full 63-character length mandated by the standard. Previously symbol names were truncated if the combined length of module, submodule, and function name exceeded 126 characters. This change therefore breaks the ABI, but only for cases where this 126 character limit was exceeded.

New Targets and Target Specific Improvements

AArch64 & arm

• The AArch64 and arm ports now support condition flag output constraints in inline assembly, as indicated by the __GCC_ASM_FLAG_OUTPUTS__. On arm this feature is only available for A32 and T32 targets. Please refer to the documentation for more details.

AArch64

• There have been several improvements related to the Scalable Vector Extension (SVE):
  • The SVE ACLE types and intrinsics are now supported. They can be accessed using the header file arm_sve.h.
  • It is now possible to create fixed-length SVE types using the arm_sve_vector_bits attribute. For example:

    #if __ARM_FEATURE_SVE_BITS==512
    typedef svint32_t vec512 __attribute__((arm_sve_vector_bits(512)));
    typedef svbool_t pred512 __attribute__((arm_sve_vector_bits(512)));
    #endif

  • -mlow-precision-div, -mlow-precision-sqrt and -mlow-precision-recip-sqrt now work for SVE.
  • -msve-vector-bits=128 now generates vector-length-specific code for little-endian targets. It continues to generate vector-length-agnostic code for big-endian targets, just as previous releases did for all targets.
  • The vectorizer is now able to use extending loads and truncating stores, including gather loads and scatter stores.
  • The vectorizer now compares the cost of vectorizing with SVE and vectorizing with Advanced SIMD and tries to pick the best one. Previously it would always use SVE if possible.
  • If a vector loop uses Advanced SIMD rather than SVE, the vectorizer now considers using SVE to vectorize the left-over elements (the “scalar tail” or “epilog”).
  • Besides these specific points, there have been many general improvements to the way that the vectorizer uses SVE.
• The -mbranch-protection=pac-ret option now accepts the optional argument +b-key extension to perform return address signing with the B-key instead of the A-key.
• The option -moutline-atomics has been added to aid deployment of the Large System Extensions (LSE) on GNU/Linux systems built with a baseline architecture targeting Armv8-A. When the option is specified code is emitted to detect the presence of LSE instructions at runtime and use them for standard atomic operations. For more information please refer to the documentation.
• The Transactional Memory Extension is now supported through ACLE intrinsics. It can be enabled through the +tme option extension (for example, -march=armv8.5-a+tme).
• A number of features from Armv8.5-A are now supported through ACLE intrinsics. These include:
  • The random number instructions that can be enabled through the (already present in GCC 9.1) +rng option extension.
  • Floating-point intrinsics to round to integer instructions from Armv8.5-A when targeting -march=armv8.5-a or later.
  • Memory Tagging Extension intrinsics enabled through the +memtag option extension.
• Similarly, the following Armv8.6-A features are now supported through ACLE intrinsics:
  • The bfloat16 extension. This extension is enabled automatically when Armv8.6-A is selected (such as by -march=armv8.6-a). It can also be enabled for Armv8.2-A and later using the +bf16 option extension.
  • The Matrix Multiply extension. This extension is split into three parts, one for each supported data type:
    • Support for 8-bit integer matrix multiply instructions. This extension is enabled automatically when Armv8.6-A is selected. It can also be enabled for Armv8.2-A and later using the +i8mm option extension.
    • Support for 32-bit floating-point matrix multiply instructions. This extension can be enabled using the +f32mm option extension, which also has the effect of enabling SVE.
    • Support for 64-bit floating-point matrix multiply instructions. This extension can be enabled using the +f64mm option extension, which likewise has the effect of enabling SVE.
• SVE2 is now supported through ACLE intrinsics and (to a limited extent) through autovectorization. It can be enabled through the +sve2 option extension (for example, -march=armv8.5-a+sve2). Additional extensions can be enabled through +sve2-sm4, +sve2-aes, +sve2-sha3 and +sve2-bitperm.
• Support has been added for the following processors (GCC identifiers in parentheses):
  • Arm Cortex-A77 (cortex-a77).
  • Arm Cortex-A76AE (cortex-a76ae).
  • Arm Cortex-A65 (cortex-a65).
  • Arm Cortex-A65AE (cortex-a65ae).
  • Arm Cortex-A34 (cortex-a34).
  • Marvell ThunderX3 (thunderx3t110).
  The GCC identifiers can be used as arguments to the -mcpu or -mtune options, for example: -mcpu=cortex-a77 or -mtune=cortex-a65ae or as arguments to the equivalent target attributes and pragmas.

arm

• Support for the FDPIC ABI has been added. It uses 64-bit function descriptors to represent pointers to functions, and enables code sharing on MMU-less systems. The corresponding target triple is arm-uclinuxfdpiceabi, and the C library is uclibc-ng.
• Support has been added for the Arm EABI on NetBSD through the arm*-*-netbsdelf-*eabi* triplet.
• The handling of 64-bit integer operations has been significantly reworked and improved leading to improved performance and reduced stack usage when using 64-bit integral data types. The option -mneon-for-64bits is now deprecated and will be removed in a future release.
• Support has been added for the following processors (GCC identifiers in parentheses):
  • Arm Cortex-A77 (cortex-a77).
  • Arm Cortex-A76AE (cortex-a76ae).
  • Arm Cortex-M35P (cortex-m35p).
  The GCC identifiers can be used as arguments to the -mcpu or -mtune options, for example: -mcpu=cortex-a77 or -mtune=cortex-m35p.
• Support has been extended for the ACLE data-processing intrinsics to include 32-bit SIMD, saturating arithmetic, 16-bit multiplication and other related intrinsics aimed at DSP algorithm optimization.
• Support for -mpure-code in Thumb-1 (v6m) has been added: this M-profile feature is no longer restricted to targets with MOVT. For example, -mcpu=cortex-m0 now supports this option.
• Support for the Custom Datapath Extension beta ACLE intrinsics has been added.

AMD Radeon (GCN)

• The code generation and in particular the vectorization support has been much improved.

ARC

• The interrupt service routine functions save all used registers, including extension registers and auxiliary registers used by Zero Overhead Loops.
• Improve code size by using multiple short instructions instead of a single long mov or ior instruction when its long immediate constant is known.
• Fix usage of the accumulator register for ARC600.
• Fix issues with uncached attribute.
• Remove -mq-class option.
• Improve 64-bit integer addition and subtraction operations.

AVR

• Support for the XMEGA-like devices

  ATtiny202, ATtiny204, ATtiny402, ATtiny404, ATtiny406, ATtiny804, ATtiny806, ATtiny807, ATtiny1604, ATtiny1606, ATtiny1607, ATmega808, ATmega809, ATmega1608, ATmega1609, ATmega3208, ATmega3209, ATmega4808, ATmega4809

  has been added.
• A new command-line option -nodevicespecs has been added. It allows to provide a custom device-specs file by means of

  avr-gcc -nodevicespecs -specs=my-spec-file <options>

  and without the need to provide options -B and -mmcu=. See AVR command-line options for details. This feature is also available in GCC 9.3+ and GCC 8.4+.
• New command-line options -mdouble=[32,64] and -mlong-double=[32,64] have been added. They allow to choose the size (in bits) of the double and long double types, respectively. Whether or not the mentioned layouts are available, whether the options act as a multilib option, and the default for either option are controlled by the new AVR configure options --with-double= and --with-long-double=.
• A new configure option --with-libf7= has been added. It controls to which level avr-libgcc provides 64-bit floating point support by means of Libf7.
• A new configure option --with-double-comparison= has been added. It's unlikely you need to set this option by hand.

IA-32/x86-64

• Support to expand __builtin_roundeven into the appropriate SSE 4.1 instruction has been added.
• New ISA extension support for Intel ENQCMD was added to GCC. ENQCMD intrinsics are available via the -menqcmd compiler switch.
• GCC now supports the Intel CPU named Cooperlake through -march=cooperlake. The switch enables the AVX512BF16 ISA extensions.
• GCC now supports the Intel CPU named Tigerlake through -march=tigerlake. The switch enables the MOVDIRI MOVDIR64B AVX512VP2INTERSECT ISA extensions.

MIPS

• The mips*-*-linux* targets now mark object files with appropriate GNU-stack note, facilitating use of non-executable stack hardening on GNU/Linux. The soft-float targets have this feature enabled by default, while for hard-float targets it is required for GCC to be configured with --with-glibc-version=2.31 against glibc 2.31 or later.

PowerPC / PowerPC64 / RS6000

• Many vector builtins have been listed as deprecated in the 64-Bit ELF V2 ABI Specification for quite a number of years. The vector builtins listed in Tables A.8 through A.10 are now deprecated for GCC 10, and will likely be removed from support in GCC 11.

PRU

• A new back end targeting TI PRU I/O processors has been contributed to GCC.

RISC-V

• The riscv*-*-* targets now require GNU binutils version 2.30 or later, to support new assembly instructions produced by GCC.

V850

• The ABI for V850 nested functions has been changed. Previously the V850 port used %r20 for the static chain pointer, now the port uses %r19. This corrects a long standing latent bug in the v850 port where a call to a nested function would unexpectedly change the value in %r20.

Operating Systems

Improvements for plugin authors

• GCC diagnostics can now have a chain of events associated with them, describing a path through the code that triggers the problem. These can be printed by the diagnostics subsystem in various ways, controlled by the -fdiagnostics-path-format option, or captured in JSON form via -fdiagnostics-format=json.
• GCC diagnostics can now be associated with CWE weakness identifiers, which will appear on the standard error stream, and in the JSON output from -fdiagnostics-format=json.

Other significant improvements

• To allow inline expansion of both memcpy and memmove, the existing movmem instruction patterns used for non-overlapping memory copies have been renamed to cpymem. The movmem name is now used for overlapping memory moves, consistent with the library functions memcpy and memmove.
• For many releases, when GCC emits a warning it prints the option controlling that warning. As of GCC 10, that option text is now a clickable hyperlink for the documentation of that option (assuming a sufficiently capable terminal). This behavior can be controlled via a new -fdiagnostics-urls option (along with various environment variables and heuristics documented with that option).

GCC 10.1

This is the list of problem reports (PRs) from GCC's bug tracking system that are known to be fixed in the 10.1 release. This list might not be complete (that is, it is possible that some PRs that have been fixed are not listed here).