Re: [RFC] Introduce -finline-memset-loops

[Fangrui Song <maskray@google.com>]

On Fri, Jun 2, 2023 at 3:11 AM Alexandre Oliva via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> On Jan 19, 2023, Alexandre Oliva <oliva@adacore.com> wrote:
>
> > Would it make more sense to extend it, even constrained by the
> > limitations mentioned above, or handle memset only?  In the latter case,
> > would it still make sense to adopt a command-line option that suggests a
> > broader effect than it already has, even if it's only a hopeful future
> > extension?  -finline-all-stringops[={memset,memcpy,...}], that you
> > suggested, seems to be a reasonable and extensible one to adopt.
>
> I ended up implementing all of memset, memcpy, memmove, and memcmp:
>
> Introduce -finline-stringops
>
> try_store_by_multiple_pieces was added not long ago, enabling
> variable-sized memset to be expanded inline when the worst-case
> in-range constant length would, using conditional blocks with powers
> of two to cover all possibilities of length and alignment.
>
> This patch introduces -finline-stringops[=fn] to request expansions to
> start with a loop, so as to still take advantage of known alignment
> even with long lengths, but without necessarily adding store blocks
> for every power of two.
>
> This makes it possible for the supported stringops (memset, memcpy,
> memmove, memset) to be expanded, even if storing a single byte per
> iteration.  Surely efficient implementations can run faster, with a
> pre-loop to increase alignment, but that would likely be excessive for
> inline expansions.
>
> Still, in some cases, such as in freestanding environments, users
> prefer to inline such stringops, especially those that the compiler
> may introduce itself, even if the expansion is not as performant as a
> highly optimized C library implementation could be, to avoid
> depending on a C runtime library.
>
> Regstrapped on x86_64-linux-gnu, also bootstrapped with
> -finline-stringops enabled by default, and tested with arm, aarch, 32-
> and 64-bit riscv with gcc-12.  Ok to install?
>[...]

This seems to be related to Clang's __builtin_mem{set,cpy}_inline . I
just created
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110094 ("Support
__builtin_mem*_inline").