* [PATCH] RISC-V: Change VSETVL PASS always call split_all_insns
@ 2023-01-18 2:50 juzhe.zhong
2023-01-22 19:39 ` Jeff Law
0 siblings, 1 reply; 3+ messages in thread
From: juzhe.zhong @ 2023-01-18 2:50 UTC (permalink / raw)
To: gcc-patches; +Cc: kito.cheng, palmer, Ju-Zhe Zhong
From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
Since LCM will destroy CFG, we are going to reorder the location of VSETVL PASS
at least before bbro (block-reorder PASS) which is before split3 PASS. We need
to call it in VSETVL PASS to get final RVV instructions patterns.
gcc/ChangeLog:
* config/riscv/riscv-vsetvl.cc (pass_vsetvl::execute): Always call split_all_insns.
---
gcc/config/riscv/riscv-vsetvl.cc | 10 ++++------
1 file changed, 4 insertions(+), 6 deletions(-)
diff --git a/gcc/config/riscv/riscv-vsetvl.cc b/gcc/config/riscv/riscv-vsetvl.cc
index 0245124e28f..d494369a603 100644
--- a/gcc/config/riscv/riscv-vsetvl.cc
+++ b/gcc/config/riscv/riscv-vsetvl.cc
@@ -3092,12 +3092,10 @@ pass_vsetvl::execute (function *)
if (n_basic_blocks_for_fn (cfun) <= 0)
return 0;
- /* The reason we have this since we didn't finish splitting yet
- when optimize == 0. In this case, we should conservatively
- split all instructions here to make sure we don't miss any
- RVV instruction. */
- if (!optimize)
- split_all_insns ();
+ /* The RVV instruction may change after split which is not a stable
+ instruction. We need to split it here to avoid potential issue
+ since the VSETVL PASS is insert before split PASS. */
+ split_all_insns ();
/* Early return for there is no vector instructions. */
if (!has_vector_insn (cfun))
--
2.36.3
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] RISC-V: Change VSETVL PASS always call split_all_insns
2023-01-18 2:50 [PATCH] RISC-V: Change VSETVL PASS always call split_all_insns juzhe.zhong
@ 2023-01-22 19:39 ` Jeff Law
2023-01-27 9:36 ` Kito Cheng
0 siblings, 1 reply; 3+ messages in thread
From: Jeff Law @ 2023-01-22 19:39 UTC (permalink / raw)
To: juzhe.zhong, gcc-patches; +Cc: kito.cheng, palmer
On 1/17/23 19:50, juzhe.zhong@rivai.ai wrote:
> From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
>
> Since LCM will destroy CFG, we are going to reorder the location of VSETVL PASS
> at least before bbro (block-reorder PASS) which is before split3 PASS. We need
> to call it in VSETVL PASS to get final RVV instructions patterns.
Just for the record. LCM does not destroy the CFG, it just splits
critical edges.
>
> gcc/ChangeLog:
>
> * config/riscv/riscv-vsetvl.cc (pass_vsetvl::execute): Always call split_all_insns.
OK.
jeff
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] RISC-V: Change VSETVL PASS always call split_all_insns
2023-01-22 19:39 ` Jeff Law
@ 2023-01-27 9:36 ` Kito Cheng
0 siblings, 0 replies; 3+ messages in thread
From: Kito Cheng @ 2023-01-27 9:36 UTC (permalink / raw)
To: Jeff Law; +Cc: juzhe.zhong, gcc-patches, palmer
[-- Attachment #1: Type: text/plain, Size: 684 bytes --]
committed, thanks
On Mon, Jan 23, 2023 at 3:39 AM Jeff Law via Gcc-patches <
gcc-patches@gcc.gnu.org> wrote:
>
>
> On 1/17/23 19:50, juzhe.zhong@rivai.ai wrote:
> > From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
> >
> > Since LCM will destroy CFG, we are going to reorder the location of
> VSETVL PASS
> > at least before bbro (block-reorder PASS) which is before split3 PASS.
> We need
> > to call it in VSETVL PASS to get final RVV instructions patterns.
> Just for the record. LCM does not destroy the CFG, it just splits
> critical edges.
>
> >
> > gcc/ChangeLog:
> >
> > * config/riscv/riscv-vsetvl.cc (pass_vsetvl::execute): Always
> call split_all_insns.
> OK.
> jeff
>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2023-01-27 9:36 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-01-18 2:50 [PATCH] RISC-V: Change VSETVL PASS always call split_all_insns juzhe.zhong
2023-01-22 19:39 ` Jeff Law
2023-01-27 9:36 ` Kito Cheng
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).