About

Often the compiler has enough information to efficiently expand mem* or str* routines inline.  Say for example a 2 word copy or a string comparison against a constant.

By expanding these sequences inline, we can avoid the overhead of a function call and  expose more of the underlying semantics of the call to the optimizers, thus potentially allowing further optimization.  It also creates larger scheduling blocks and fewer optimization barriersThis covers multiple efforts.

1. Scalar inline expansion of various mem* and str* routines in GCC.  Essentially when some of the arguments are known the compiler can emit efficient versions of those routines while at the same time exposing the semantics to the optimizers allowing further optimization.  The scalar inline expansion work in GCC is largely complete at this point.  This work was largely done by VRULL under contract to Ventana.
2. Vector inline expansion.  Similar to scalar expansion, but using vector instructions when possible.  The bulk of this work has already been integrated for gcc-14 and is the combination of work from Ventana (Robin Dapp) and EMBECOSM (Joern Rennecke).
3. Additional vector inline expansion.  memmove, memset and memcmp can also be expanded inline by the compiler using vector instructions.  Sergei from Rivos has submitted an implementation of these routines, but the submission missed the gcc-14 deadline.  The implementation looks pretty reasonable and is expected to integrate into GCC in the late spring. It is expected this work will provide another percent or so improvement on the GCC workload within spec2017.
4. VRULL's engineers are also working on improving scalar codegen for inlined cases in gcc.  For example using cboz for clearing cache lines, zbb for str[n]cmp, etc.

Stakeholders/Partners

RISE:

Ventana: Jeff Law& Robin

Rivos: Palmer & Sergei

External:

...

                    VRULL: Christoph Muller

                    EMBECOSM: Joern Rennecke

Dependencies

Status

Updates

29 Jun 2024

• Sergei's last patch has been integrated upstream.
• VRULL's patches have been integrated upstream.
• We'll break out the glibc work into a 2H2024 item 

23 Jun 2024 

• Sergei's 2nd patch (setmem) has been updated for the current trunk and re-submitted for inclusion.

09 May 2024 

• VRULL is currently upstreaming various bits to improve some of the scalar inline expansion of these routines
• Ventana has upstreamed VRULL's patch that allows overlapping memory references when inlining memcpy and similar routines
• We're hoping to move on Sergei's vector versions shortly

30 Jan 2024 

• Added some additional text for the cases covered by this change as well as some performance expectations.
• Jeff L has done some review work on Sergei's work and is planning to incorporate that work into his upstream GCC tester ASAP.

29 Dec 2023 

• Robin Dapp from Ventana has submitted & integrated vector versions of str[n]cmp, strlen
• Christoph has submitted and integrated scalar versions of str[n]cmp, strlen, memcpy
• Sergei Lewis has submitted vector version of  memset, memmove and memcmp.

04 Oct 2023 

• Joern has submitted Embecosm's work to inline vectorized memcpy.

27 Sep 2023  

• Project reported as a priority for 1H2024.
• Christoph has submitted VRULL's work for str(n)cmp inline expansion and it has been integrated upstream

...