From: "Maciej W. Rozycki" <macro@embecosm.com>
To: Jeff Law <jeffreyalaw@gmail.com>
Cc: gcc-patches@gcc.gnu.org, Andrew Waterman <andrew@sifive.com>,
Jim Wilson <jim.wilson.gcc@gmail.com>,
Kito Cheng <kito.cheng@gmail.com>,
Palmer Dabbelt <palmer@dabbelt.com>
Subject: Re: [PATCH 09/44] RISC-V: Rework branch costing model for if-conversion
Date: Wed, 29 Nov 2023 12:01:39 +0000 (GMT) [thread overview]
Message-ID: <alpine.DEB.2.20.2311291115010.5892@tpp.orcam.me.uk> (raw)
In-Reply-To: <86917a66-8442-49d5-a2f2-fb8b991ccaf3@gmail.com>
On Tue, 28 Nov 2023, Jeff Law wrote:
> FWIW, I was looking at a regression with our internal tests after your
> changes. It was quite nice to see how well twiddling -mbranch-cost
> correlated to how many instructions we would allow in a conditional move
> sequence.
I'm a bit concerned though that our interpretation of `-mbranch-cost=0'
is different from the middle end's, such as in `emit_store_flag':
/* If we reached here, we can't do this with a scc insn, however there
are some comparisons that can be done in other ways. Don't do any
of these cases if branches are very cheap. */
if (BRANCH_COST (optimize_insn_for_speed_p (), false) == 0)
return 0;
> The downside is it highlighted the gimple vs RTL use issue. I'm confident
> that we would like to see a higher branch cost in the RTL phases for our
> uarch, but I'm much less comfortable with how that's going to change the
> decisions made in trees/gimple. We'll have to investigate that at some depth.
Ack.
> > I've looked at it already and it's the middle end that ends up with the
> > zero-extension, specifically `convert_move' invoked from `emit_cstore'
> > down the call to `noce_try_store_flag_mask', to widen the output from
> > `cstoredi4', so I don't think we can do anything in the backend to prevent
> > it from happening. And neither I think we can do anything useful about
> > `cstoredi4' having a SImode output, as it's a pattern matched by name
> > rather than RTX, so we can't provide variants having a SImode and a DImode
> > output each both at a time, as that would cause a name clash.
> We're actually tracking some of these extraneous extensions. Do you happen to
> know if the zero-extended object happens to be (subreg:SI (reg:DI)) kind of
> construct? That's the kind of thing we're chasing down right now from various
> points. Vineet has already fixed one class of them. Jivan and I are looking
> at others.
Under GDB it's a plain move from (reg:SI 140) to (reg:DI 139), as in the
FROM and TO arguments to `convert_move' respectively. This makes it call
`convert_mode_scalar', which then chooses between `zext_optab' and
`sext_optab' as appropriate, under:
/* If the target has a converter from FROM_MODE to TO_MODE, use it. */
to produce:
(set (reg:DI 139)
(zero_extend:DI (reg:SI 140)))
ending up with this complete sequence:
(insn 27 0 28 (set (reg:SI 140)
(eq:SI (reg/v:DI 137 [ c ])
(const_int 0 [0]))) -1
(nil))
(insn 28 27 29 (set (reg:DI 139)
(zero_extend:DI (reg:SI 140))) -1
(nil))
(insn 29 28 30 (set (reg:DI 141)
(neg:DI (reg:DI 139))) -1
(nil))
(insn 30 29 0 (set (reg/v:DI 134 [ <retval> ])
(and:DI (reg/v:DI 135 [ a ])
(reg:DI 141))) -1
(nil))
passed to `targetm.noce_conversion_profitable_p' right away. Maybe you
can teach `emit_cstore' or `convert_move' to use a subreg when it is known
for the particular target that the value produced by the conditional-set
machine instruction emitted by `cstoreMODE4' is valid unchanged in both
modes.
You can fiddle with it by trying:
$ gcc -march=rv64gc -mbranch-cost=3 -O2 -S gcc/testsuite/gcc.target/riscv/pr105314.c
Set a breakpoint at `noce_try_store_flag_mask' and then single-step to see
how things proceed.
Maciej
next prev parent reply other threads:[~2023-11-29 12:01 UTC|newest]
Thread overview: 101+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-11-19 5:35 [PATCH 00/44] RISC-V: Various if-conversion fixes and improvements Maciej W. Rozycki
2023-11-18 16:50 ` [PATCH 13/44] RISC-V/testsuite: Add branchless cases for FP cond-move operations Maciej W. Rozycki
2023-11-18 18:03 ` Jeff Law
2023-11-19 6:27 ` Maciej W. Rozycki
2023-11-19 5:37 ` Maciej W. Rozycki
2023-11-19 5:35 ` [PATCH 01/44] testsuite: Add cases for conditional-move and conditional-add operations Maciej W. Rozycki
2023-11-19 5:52 ` Kito Cheng
2023-11-20 10:16 ` Maciej W. Rozycki
2023-11-20 12:57 ` Richard Biener
2023-11-22 1:33 ` Maciej W. Rozycki
2023-11-19 5:35 ` [PATCH 02/44] RISC-V/testsuite: Add cases for integer SFB cond-move operations Maciej W. Rozycki
2023-11-19 5:53 ` Kito Cheng
2023-11-19 5:35 ` [PATCH 03/44] RISC-V: Reorder comment on SFB patterns Maciej W. Rozycki
2023-11-19 5:53 ` Kito Cheng
2023-11-19 5:36 ` [PATCH 04/44] RISC-V: Sanitise NEED_EQ_NE_P case with `riscv_emit_int_compare' Maciej W. Rozycki
2023-11-19 5:53 ` Kito Cheng
2023-11-19 5:36 ` [PATCH 05/44] RISC-V: Fix `mode' usage in `riscv_expand_conditional_move' Maciej W. Rozycki
2023-11-19 5:54 ` Kito Cheng
2023-11-19 5:36 ` [PATCH 06/44] RISC-V: Avoid repeated GET_MODE calls " Maciej W. Rozycki
2023-11-19 5:55 ` Kito Cheng
2023-11-19 5:36 ` [PATCH 07/44] RISC-V: Use `nullptr' " Maciej W. Rozycki
2023-11-19 5:53 ` Kito Cheng
2023-11-19 5:36 ` [PATCH 08/44] RISC-V: Simplify EQ vs NE selection " Maciej W. Rozycki
2023-11-19 5:56 ` Kito Cheng
2023-11-19 5:36 ` [PATCH 09/44] RISC-V: Rework branch costing model for if-conversion Maciej W. Rozycki
2023-11-19 18:52 ` Jeff Law
2023-11-23 18:34 ` Maciej W. Rozycki
2023-11-29 1:19 ` Jeff Law
2023-11-29 12:01 ` Maciej W. Rozycki [this message]
2023-11-19 5:37 ` [PATCH 10/44] RISC-V/testsuite: Add branched cases for integer cond-move operations Maciej W. Rozycki
2023-11-19 6:44 ` Kito Cheng
2023-11-19 5:37 ` [PATCH 11/44] RISC-V/testsuite: Add branchless " Maciej W. Rozycki
2023-11-19 6:47 ` Kito Cheng
2023-11-23 19:18 ` Maciej W. Rozycki
2023-11-19 5:37 ` [PATCH 12/44] RISC-V/testsuite: Add branched cases for FP " Maciej W. Rozycki
2023-11-19 6:48 ` Kito Cheng
2023-11-19 5:38 ` [PATCH 14/44] RISC-V: Also invert the cond-move condition for GEU and LEU Maciej W. Rozycki
2023-11-19 6:50 ` Kito Cheng
2023-11-19 5:38 ` [PATCH 15/44] RISC-V/testsuite: Add branched cases for GEU and LEU cond-move operations Maciej W. Rozycki
2023-11-19 17:42 ` Jeff Law
2023-11-19 5:38 ` [PATCH 16/44] RISC-V/testsuite: Add branchless " Maciej W. Rozycki
2023-11-19 7:22 ` Kito Cheng
2023-11-19 5:38 ` [PATCH 17/44] RISC-V: Avoid extraneous EQ or NE operation in cond-move expansion Maciej W. Rozycki
2023-11-19 17:45 ` Jeff Law
2023-11-19 5:38 ` [PATCH 18/44] RISC-V/testsuite: Add branched cases for equality cond-move operations Maciej W. Rozycki
2023-11-19 17:45 ` Jeff Law
2023-11-19 5:39 ` [PATCH 19/44] RISC-V/testsuite: Add branchless " Maciej W. Rozycki
2023-11-19 17:46 ` Jeff Law
2023-11-19 5:39 ` [PATCH 20/44] RISC-V: Also accept constants for T-Head cond-move comparison operands Maciej W. Rozycki
2023-11-19 17:48 ` Jeff Law
2023-11-19 5:39 ` [PATCH 21/44] RISC-V: Also accept constants for T-Head cond-move data input operands Maciej W. Rozycki
2023-11-19 17:50 ` Jeff Law
2023-11-19 5:40 ` [PATCH 22/44] RISC-V: Fold all the cond-move variants together Maciej W. Rozycki
2023-11-19 18:35 ` Jeff Law
2023-11-19 5:40 ` [PATCH 23/44] RISC-V/testsuite: Add branched cases for T-Head non-equality cond moves Maciej W. Rozycki
2023-11-19 17:54 ` Jeff Law
2023-11-19 5:40 ` [PATCH 24/44] RISC-V/testsuite: Add branchless " Maciej W. Rozycki
2023-11-19 17:54 ` Jeff Law
2023-11-19 5:40 ` [PATCH 25/44] RISC-V: Implement `riscv_emit_unary' helper Maciej W. Rozycki
2023-11-19 17:54 ` Jeff Law
2023-11-19 5:40 ` [PATCH 26/44] RISC-V: Add `movMODEcc' implementation for generic targets Maciej W. Rozycki
2023-11-19 18:18 ` Jeff Law
2023-11-23 22:16 ` Maciej W. Rozycki
2023-11-19 5:40 ` [PATCH 27/44] RISC-V/testsuite: Add branched cases for generic integer cond moves Maciej W. Rozycki
2023-11-19 18:18 ` Jeff Law
2023-11-19 5:41 ` [PATCH 28/44] RISC-V/testsuite: Add branchless " Maciej W. Rozycki
2023-11-19 18:19 ` Jeff Law
2023-11-19 5:41 ` [PATCH 29/44] RISC-V: Add `addMODEcc' implementation for generic targets Maciej W. Rozycki
2023-11-19 18:23 ` Jeff Law
2023-11-23 22:36 ` Maciej W. Rozycki
2023-11-19 5:41 ` [PATCH 30/44] RISC-V/testsuite: Add branched cases for generic integer cond adds Maciej W. Rozycki
2023-11-19 18:23 ` Jeff Law
2023-11-19 5:41 ` [PATCH 31/44] RISC-V/testsuite: Add branchless " Maciej W. Rozycki
2023-11-19 18:25 ` Jeff Law
2023-11-23 22:48 ` Maciej W. Rozycki
2023-11-19 5:42 ` [PATCH 32/44] RISC-V: Only use SUBREG if applicable in `riscv_expand_float_scc' Maciej W. Rozycki
2023-11-19 18:26 ` Jeff Law
2023-11-19 5:42 ` [PATCH 33/44] RISC-V: Also allow FP conditions in `riscv_expand_conditional_move' Maciej W. Rozycki
2023-11-19 18:30 ` Jeff Law
2023-11-23 22:55 ` Maciej W. Rozycki
2023-11-19 5:42 ` [PATCH 34/44] RISC-V: Provide FP conditional-branch instructions for if-conversion Maciej W. Rozycki
2023-11-19 19:42 ` Jeff Law
2023-11-23 23:26 ` Maciej W. Rozycki
2023-11-19 5:42 ` [PATCH 35/44] RISC-V: Avoid extraneous integer comparison for FP comparisons Maciej W. Rozycki
2023-11-19 19:44 ` Jeff Law
2023-11-19 5:42 ` [PATCH 36/44] RISC-V/testsuite: Add branched cases for generic FP cond moves Maciej W. Rozycki
2023-11-19 19:45 ` Jeff Law
2023-11-19 5:43 ` [PATCH 37/44] RISC-V/testsuite: Add branchless " Maciej W. Rozycki
2023-11-19 19:46 ` Jeff Law
2023-11-19 5:43 ` [PATCH 38/44] RISC-V/testsuite: Add branched cases for generic FP cond adds Maciej W. Rozycki
2023-11-19 19:46 ` Jeff Law
2023-11-19 5:43 ` [PATCH 39/44] RISC-V/testsuite: Add branchless " Maciej W. Rozycki
2023-11-19 19:47 ` Jeff Law
2023-11-19 5:43 ` [PATCH 40/44] RISC-V: Handle FP NE operator via inversion in cond-operation expansion Maciej W. Rozycki
2023-11-19 19:51 ` Jeff Law
2023-11-22 1:37 ` Maciej W. Rozycki
2023-11-19 5:43 ` [PATCH 41/44] RISC-V/testsuite: Add branched cases for FP NE cond-move operations Maciej W. Rozycki
2023-11-19 5:43 ` [PATCH 42/44] " Maciej W. Rozycki
2023-11-19 5:44 ` [PATCH 43/44] RISC-V/testsuite: Add branched cases for FP NE cond-add operation Maciej W. Rozycki
2023-11-19 5:44 ` [PATCH 44/44] RISC-V/testsuite: Add branchless " Maciej W. Rozycki
2023-11-19 5:52 ` [PATCH 00/44] RISC-V: Various if-conversion fixes and improvements Kito Cheng
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=alpine.DEB.2.20.2311291115010.5892@tpp.orcam.me.uk \
--to=macro@embecosm.com \
--cc=andrew@sifive.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=jeffreyalaw@gmail.com \
--cc=jim.wilson.gcc@gmail.com \
--cc=kito.cheng@gmail.com \
--cc=palmer@dabbelt.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).