* [PATCH] rtl-optimization/109237 - quadraticness in delete_trivially_dead_insns
@ 2023-03-22 10:01 Richard Biener
0 siblings, 0 replies; 2+ messages in thread
From: Richard Biener @ 2023-03-22 10:01 UTC (permalink / raw)
To: gcc-patches; +Cc: Jakub Jelinek
The following addresses quadraticness in processing debug insns
in delete_trivially_dead_insns and insn_live_p by using TREE_VISITED
on the INSN_VAR_LOCATION_DECL to indicate a later debug bind
with the same decl and no intervening real insn or debug marker.
That gets rid of the NEXT_INSN walk in insn_live_p in favor of
first clearing TREE_VISITED in the first loop over insn and
the book-keeping of decls we set the bit since we need to clear
them when visiting a real or debug marker insn.
That improves the time spent in delete_trivially_dead_insns from
10.6s to 2.2s for the testcase.
Bootstrapped and tested on x86_64-unknown-linux-gnu.
OK?
Thanks,
Richard.
PR rtl-optimization/109237
* cse.cc (insn_live_p): Remove NEXT_INSN walk, instead check
TREE_VISITED on INSN_VAR_LOCATION_DECL.
(delete_trivially_dead_insns): Maintain TREE_VISITED on
active debug bind INSN_VAR_LOCATION_DECL.
---
gcc/cse.cc | 39 ++++++++++++++++++++++++---------------
1 file changed, 24 insertions(+), 15 deletions(-)
diff --git a/gcc/cse.cc b/gcc/cse.cc
index 8fbda4ecc86..204047b0e0b 100644
--- a/gcc/cse.cc
+++ b/gcc/cse.cc
@@ -6906,22 +6906,12 @@ insn_live_p (rtx_insn *insn, int *counts)
}
else if (DEBUG_INSN_P (insn))
{
- rtx_insn *next;
-
if (DEBUG_MARKER_INSN_P (insn))
return true;
- for (next = NEXT_INSN (insn); next; next = NEXT_INSN (next))
- if (NOTE_P (next))
- continue;
- else if (!DEBUG_INSN_P (next))
- return true;
- /* If we find an inspection point, such as a debug begin stmt,
- we want to keep the earlier debug insn. */
- else if (DEBUG_MARKER_INSN_P (next))
- return true;
- else if (INSN_VAR_LOCATION_DECL (insn) == INSN_VAR_LOCATION_DECL (next))
- return false;
+ if (DEBUG_BIND_INSN_P (insn)
+ && TREE_VISITED (INSN_VAR_LOCATION_DECL (insn)))
+ return false;
return true;
}
@@ -7007,8 +6997,11 @@ delete_trivially_dead_insns (rtx_insn *insns, int nreg)
counts = XCNEWVEC (int, nreg * 3);
for (insn = insns; insn; insn = NEXT_INSN (insn))
if (DEBUG_BIND_INSN_P (insn))
- count_reg_usage (INSN_VAR_LOCATION_LOC (insn), counts + nreg,
- NULL_RTX, 1);
+ {
+ count_reg_usage (INSN_VAR_LOCATION_LOC (insn), counts + nreg,
+ NULL_RTX, 1);
+ TREE_VISITED (INSN_VAR_LOCATION_DECL (insn)) = 0;
+ }
else if (INSN_P (insn))
{
count_reg_usage (insn, counts, NULL_RTX, 1);
@@ -7048,6 +7041,7 @@ delete_trivially_dead_insns (rtx_insn *insns, int nreg)
the setter. Then go through DEBUG_INSNs and if a DEBUG_EXPR
has been created for the unused register, replace it with
the DEBUG_EXPR, otherwise reset the DEBUG_INSN. */
+ auto_vec<tree, 32> later_debug_set_vars;
for (insn = get_last_insn (); insn; insn = prev)
{
int live_insn = 0;
@@ -7110,6 +7104,21 @@ delete_trivially_dead_insns (rtx_insn *insns, int nreg)
}
cse_cfg_altered |= delete_insn_and_edges (insn);
}
+ else
+ {
+ if (!DEBUG_INSN_P (insn) || DEBUG_MARKER_INSN_P (insn))
+ {
+ for (tree var : later_debug_set_vars)
+ TREE_VISITED (var) = 0;
+ later_debug_set_vars.truncate (0);
+ }
+ else if (DEBUG_BIND_INSN_P (insn)
+ && !TREE_VISITED (INSN_VAR_LOCATION_DECL (insn)))
+ {
+ later_debug_set_vars.safe_push (INSN_VAR_LOCATION_DECL (insn));
+ TREE_VISITED (INSN_VAR_LOCATION_DECL (insn)) = 1;
+ }
+ }
}
if (MAY_HAVE_DEBUG_BIND_INSNS)
--
2.35.3
^ permalink raw reply [flat|nested] 2+ messages in thread
[parent not found: <20230322100210.AE2793857B9B@sourceware.org>]
* Re: [PATCH] rtl-optimization/109237 - quadraticness in delete_trivially_dead_insns
[not found] <20230322100210.AE2793857B9B@sourceware.org>
@ 2023-03-31 16:30 ` Jeff Law
0 siblings, 0 replies; 2+ messages in thread
From: Jeff Law @ 2023-03-31 16:30 UTC (permalink / raw)
To: Richard Biener, gcc-patches; +Cc: Jakub Jelinek
On 3/22/23 04:01, Richard Biener via Gcc-patches wrote:
> The following addresses quadraticness in processing debug insns
> in delete_trivially_dead_insns and insn_live_p by using TREE_VISITED
> on the INSN_VAR_LOCATION_DECL to indicate a later debug bind
> with the same decl and no intervening real insn or debug marker.
> That gets rid of the NEXT_INSN walk in insn_live_p in favor of
> first clearing TREE_VISITED in the first loop over insn and
> the book-keeping of decls we set the bit since we need to clear
> them when visiting a real or debug marker insn.
>
> That improves the time spent in delete_trivially_dead_insns from
> 10.6s to 2.2s for the testcase.
>
> Bootstrapped and tested on x86_64-unknown-linux-gnu.
>
> OK?
>
> Thanks,
> Richard.
>
> PR rtl-optimization/109237
> * cse.cc (insn_live_p): Remove NEXT_INSN walk, instead check
> TREE_VISITED on INSN_VAR_LOCATION_DECL.
> (delete_trivially_dead_insns): Maintain TREE_VISITED on
> active debug bind INSN_VAR_LOCATION_DECL.
OK
jeff
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2023-03-31 16:30 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-03-22 10:01 [PATCH] rtl-optimization/109237 - quadraticness in delete_trivially_dead_insns Richard Biener
[not found] <20230322100210.AE2793857B9B@sourceware.org>
2023-03-31 16:30 ` Jeff Law
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).