28 Document Revision History

Revision A (June 2019)

Initial revision of the document.

Revision B (October 2021)

Added code coverage information
Added stack guidance
Reformatted and renumbered

Revision C (April 2023)

Updated library information to reflect recent changes, including the Microchip Unified Standard Library
Corrected examples of using the __pack specifier
Removed 'How To' on the cause or reset, which was not appropriate for PIC32C/SAM devices
Updated the size of floating-point types to reflect recent compiler changes
Updated information on how to ensure a function is not removed
Updated information relating to using a debugger
Updated information on the compiler option controls available in the MPLAB X IDE
Updated information on the runtime startup code, including information on the data initialization template
Provided dedicated sections on all compiler options in addition to a summary table
Added new information on -mdfp, -mpure-code, -feliminate-unused-debug-symbols, -Og, and --dinit-compress options
Some options and description were not relevant for PIC32/SAM devices and were removed
Added divergences from the C99 language standard
Expanded configuration bit access information
Improved information on tightly coupled memories
Added used variable attribute
Added externally_visible and nopa function attributes
Added information on placing code in eXecute Only Memory (XOM)
Updated information on interrupt context switch
Improved information on smart IO features
Many general corrections and improvements

Revision D (September 2023)

Updated references to the C++ standard to reflect that the ISO/IEC 14882:2014 C++ program language is now supported
Defined the compiler's implementation-defined behavior for the C++14 standard in a new section
Included additional information on interrupt operation
Removed the CCI chapter and references to this, as CCI is not supported for C/SAM devices and the feature is disabled in the MPLAB X IDE

Revision E (June 2024)

Added new How To explaining how the stack usage for each function in a module can be displayed
Added new How To showing how to prevent a program's checksum to change
Added summary of supported Arm architectures and cores
Removed reference to the unsupported -mreserve option from a How To
Updated compiler options used with the Linktime Optimizations setting in XC32 Global options tab of the MPLAB X IDE Project Properties
Added option present in the xc32-ar (Library Archive) tab of the MPLAB X IDE Project Properties
Added a diagram indicating the internal compiler applications

Added the following options to the guide:


`-marm`	`-mcmse`	`-mlibc-variant`
`-mmalloc-variant`	`--version`	`-std`
`-Wformat-smart-io`	`-dA`	`-fdiagnostics-show-caret`
`-fframe-base-loclist`	`-mno-info-linker`	`-madj-exiting-edge-prob`
`-mauto-indexing`	`-ffat-lto-objects`	`-finline-small-functions`
`-mloop-rename-registers`	`-flto-partition`	`-fpeel-loops`
`-mthumb2-align-loop`	`-ftracer`	`-fuse-linker-plugin`
`-T`

Corrected typo in the name of what should have been the --dint-compression option and specified this with the -Wl, prefix, since it is a linker option
Added tcm variable attribute
Removed erroneous text and corrected the example in the Variables in Register section
Updated the Dynamic Memory Allocation section to include the new allocation schemes
Updated the runtime startup code section to include the on-exit feature

Revision F (June 2025)

Modified command-line examples to now use the -mdfp option
Adjusted the default setting for the feature controlled by -fcommon, which has recently changed in the compiler
Indicated that improved object file compatibility might mean that some Microchip-specific features may not work with third party tools
Clarified that the -msmart-io option does not work when using the -mchp-stdlib=picolibc option.
Removed the section for -aux-info, as this option is not fully supported
Corrected information in the -D and -U preprocessor macro option sections relating to when this option is processed
Corrected the name give to sections used to hold objects or functions that use the addressattribute
Added more information to the -mdfp option regarding the use of DFPs
Added description of the -msemihost and -mgeneral-regs-only options
Added description of the -mchp-stdlibc and -nostdlib++ options
Added description of the -ftree-vrp option
Corrected typo in -fcall-used-reg option description
Clarified that objects and library archives built with LTO enabled are platform dependent
Added new options: -Warray-compare, -Warray-parameter, -Wenum-int-mismatch, -Wmismatched-dealloc, -Wmismatched-new-delete, -Wself-move, -Wsizeof-array-div, -Wvla-parameter, -Wzero-length-bounds, -Wabsolute-value, -Waddress-of-packed-member, -Wbidi-chars, -Wc++11-extensions, -Wc++14-extensions, -Wc++17-extensions, -Wc++20-extensions, -Wc++23-extensions, -Wctad-maybe-unsupported, -Wdangling-reference, -Wdeprecated-enum-enum-conversion, -Wdeprecated-enum-float-conversion, -Wmissing-requires, -Wrange-loop-construct, -Wstring-compare, -Wstringop-overread, -Wvexing-parse, and -Wxor-used-as-pow
Added arguments to the -falign-functions, -falign-labels, -falign-loopsnow supported by the compiler
Removed floating-point macro definitions that are now documented in the Microchip Unified Standard Library Reference Guide
Added section describing noload variable attribute
Added description of the no_stack_protector attribute to discussions on stack protection feature
Elaborated the text in the section on inline assembly code to indicate how the present of assembly labels can lead to build failures
Provided code example that can save the FPU registers for interrupt handlers using floating-point code on Cortex-A devices
Indicated in the descriptions of the -o and -save-temps options that they additionally control the name and location of aux and dump files
Added new section relating to the thread-local storage (TLS) memory management feature relevant for programs using threads
Corrected floating-point options used for multilib generation
Added description of new __thread object qualifier
Added description of new __has_builtin operator
Added information relating to the picolibc library, with support for thread-local storage
Added new section on the __thread qualifier
Added description of new access and stack_protect attributes
Expanded information on interrupt context switching relating to FPU registers
Updated the example that shows how to set canary values with the stack smashing protection feature
Indicated that section names associated with persistent objects can vary if other attributes are used with the object
Indicated that section names associated with absolute objects and functions have arbitrary names
Added description of new __builtin_has_attributebuilt-in function
Added descriptions for new __builtin_bswap16, __builtin_bswap32, and __builtin_bswap64 built-in functions
Provided more examples of function attributes
Added section listing new ACLE built-in functions
Included xc32-objdump command to generate an assembly listing in the How Do I Know What Code the Compiler Is Producing? How To
Clarified the difference between Free and PRO compilers in the What's the Difference Between a Free and PRO Compiler? How To.

Revision G (January 2026)

Added information relating to the new Picolibc library, an alternative to MUSL, selectable using the --mchp-stdlibc option
Indicated that both Picolibc and MUSL libraries now provide support functions for thread local storage of objects
Added descriptions of the -Wextra option and expanded the description of the -Werror option
Added sections for the -fzero-call-used-regs and -ftrivial-auto-var-init options, which are new to the compiler
Added section on the new --fill option, which fills unused memory locations
Added descriptions of the -Wtrivial-auto-var-init, -Wuninitialized, and -Wno-analyzer-use-of-uninitialized-value warning options
Corrected information in the -D and -U preprocessor macro option sections relating to when this option is processed
Added section for the copy variable and function attributes, which has been recently added to the compiler
Added description of the uninitialized and variable attribute
Added sections for the alias function and variable attributes
Added descriptions of graphical controls for the compiler options available when using the MPLAB Extensions for VS Code
Removed reference to "return address," which was incorrectly stated as being part of the call frame
Removed ye olde ASCII table

Revision H (April 2026)

Removed PRO licensing information relating to activation and use of some optimizations
Added information relating to the new word-oriented data initialization feature and associated --mchp-dinit option