On Sun, Jan 1, 2023 at 3:55 PM Roger Sayle <ro...@nextmovesoftware.com> wrote:
>
>
> Hi Uros,
> Happy New Year.
> As requested here's a revised version of my patch to introduce a pattern for
> extendditi2, but implementing your suggestion to re-use the existing
> extendsidi2_1 splitters and peephole2 optimizations by using DWI/DWIH
> mode iterators.
>
> This patch has been tested on x86_64-pc-linux-gnu with make bootstrap
> and make -k check, both with and without --target_board=unix{-m32},
> with no new failures. Ok for mainline?
>
> 2023-01-01 Roger Sayle <ro...@nextmovesoftware.com>
> Uroš Bizjak <ubiz...@gmail.com>
>
> gcc/ChangeLog
> * config/i386/i386.md (extendditi2): New define_insn.
> (define_split): Use DWIH mode iterator to treat new extendditi2
> identically to existing extendsidi2_1.
> (define_peephole2): Likewise.
> (define_peephole2): Likewise.
> (define_split): Likewise.
>
> gcc/testsuite/ChangeLog
> * gcc.target/i386/extendditi2-1.c: New test case.
> * gcc.target/i386/extendditi2-2.c: Likewise.
OK.
Thanks - and HNY,
Uros.
>
>
> Thanks in advance,
> Roger
> --
>
> > -----Original Message-----
> > From: Uros Bizjak <ubiz...@gmail.com>
> > Sent: 28 December 2022 09:28
> > To: Roger Sayle <ro...@nextmovesoftware.com>
> > Cc: GCC Patches <gcc-patches@gcc.gnu.org>
> > Subject: Re: [x86_64 PATCH] Add post-reload splitter for extendditi2.
> >
> > On Wed, Dec 28, 2022 at 1:32 AM Roger Sayle
> > <ro...@nextmovesoftware.com> wrote:
> > >
> > >
> > > This is another step towards a possible solution for PR 105137.
> > > This patch introduces a define_insn_and_split for extendditi2, that
> > > allows DImode to TImode sign-extension to be represented in the early
> > > RTL optimizers, before being split post-reload into the exact same
> > > idiom as currently produced by RTL expansion.
> >
> > Please see extendsidi2_1 insn pattern and follow-up splitters and
> > peephole2 patterns that do exactly what you want to achieve, but they are
> > currently handling only SImode to DImode on 32-bit targets. OTOH, these
> > patterns handle several more cases (e.g. split to the memory
> > output) and just have to be macroized with DWIH mode iterator to also handle
> > DImode to TImode on 64-bit targets. Probably, an extendsidi expander will
> > have
> > to be slightly adjusted when macroized to signal middle end the
> > availability of
> > extendditi pattern.
> >
> > Following macroization, any possible follow-up optimizations and
> > improvements
> > will then be automatically applied also to 32-bit targets.
> >
> > Uros.
> >
> > >
> > > Typically this produces the identical code, so the first new test
> > > case:
> > > __int128 foo(long long x) { return (__int128)x; }
> > >
> > > continues to generate:
> > > foo: movq %rdi, %rax
> > > cqto
> > > ret
> > >
> > > The "magic" is that this representation allows combine and the other
> > > RTL optimizers to do a better job. Hence, the second test case:
> > >
> > > __int128 foo(__int128 a, long long b) {
> > > a += ((__int128)b) << 70;
> > > return a;
> > > }
> > >
> > > which mainline with -O2 currently generates as:
> > >
> > > foo: movq %rsi, %rax
> > > movq %rdx, %rcx
> > > movq %rdi, %rsi
> > > salq $6, %rcx
> > > movq %rax, %rdi
> > > xorl %eax, %eax
> > > movq %rcx, %rdx
> > > addq %rsi, %rax
> > > adcq %rdi, %rdx
> > > ret
> > >
> > > with this patch now becomes:
> > > foo: movl $0, %eax
> > > salq $6, %rdx
> > > addq %rdi, %rax
> > > adcq %rsi, %rdx
> > > ret
> > >
> > > i.e. the same code for the signed and unsigned extension variants.
> > >
> > > This patch has been tested on x86_64-pc-linux-gnu with make bootstrap
> > > and make -k check, both with and without --target_board=unix{-m32},
> > > with no new failures. Ok for mainline?
> > >
> > > 2022-12-28 Roger Sayle <ro...@nextmovesoftware.com>
> > >
> > > gcc/ChangeLog
> > > * config/i386/i386.md (extendditi2): New define_insn_and_split
> > > to split DImode to TImode sign-extension after reload.
> > >
> > > gcc/testsuite/ChangeLog
> > > * gcc.target/i386/extendditi2-1.c: New test case.
> > > * gcc.target/i386/extendditi2-2.c: Likewise.
> > >
> > >
> > > Thanks in advance,
> > > Roger
> > > --
> > >
