[Bug middle-end/31531] A microoptimization of isnegative of signed integer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 Andrew Pinski changed: What|Removed |Added Status|ASSIGNED|RESOLVED Resolution|--- |FIXED Target Milestone|--- |14.0 --- Comment #22 from Andrew Pinski --- Fixed.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 --- Comment #21 from CVS Commits --- The trunk branch has been updated by Andrew Pinski : https://gcc.gnu.org/g:29a4453c7b8a86d242dab89b9e4d222749fd911e commit r14-4661-g29a4453c7b8a86d242dab89b9e4d222749fd911e Author: Andrew Pinski Date: Sun Oct 15 15:18:42 2023 -0700 [PR31531] MATCH: Improve ~a < ~b and ~a < CST, allow a nop cast inbetween ~ and a/b Currently we able to simplify `~a CMP ~b` to `b CMP a` but we should allow a nop conversion in between the `~` and the `a` which can show up. A similarly thing should be done for `~a CMP CST`. I had originally submitted the `~a CMP CST` case as https://gcc.gnu.org/pipermail/gcc-patches/2021-November/585088.html; I noticed we should do the same thing for the `~a CMP ~b` case and combined it with that one here. OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions. PR tree-optimization/31531 gcc/ChangeLog: * match.pd (~X op ~Y): Allow for an optional nop convert. (~X op C): Likewise. gcc/testsuite/ChangeLog: * gcc.dg/tree-ssa/pr31531-1.c: New test. * gcc.dg/tree-ssa/pr31531-2.c: New test.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 Andrew Pinski changed: What|Removed |Added URL|https://gcc.gnu.org/piperma |https://gcc.gnu.org/piperma |il/gcc-patches/2021-Novembe |il/gcc-patches/2023-October |r/585088.html |/633090.html --- Comment #20 from Andrew Pinski --- Updated patch: https://gcc.gnu.org/pipermail/gcc-patches/2023-October/633090.html
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 --- Comment #19 from Andrew Pinski --- Created attachment 56119 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=56119=edit testcase from the original attachment For easier access, I posting this here so don't need to untar the file.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 Andrew Pinski changed: What|Removed |Added URL||https://gcc.gnu.org/piperma ||il/gcc-patches/2021-Novembe ||r/585088.html Keywords||patch --- Comment #18 from Andrew Pinski --- Patch posted: https://gcc.gnu.org/pipermail/gcc-patches/2021-November/585088.html
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 Andrew Pinski changed: What|Removed |Added Attachment #26645|0 |1 is obsolete|| --- Comment #17 from Andrew Pinski --- Created attachment 51841 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51841=edit Patch which I am testing Attached patch which I am testing. should resolve this fully.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531
--- Comment #16 from Andrew Pinski ---
The only patch which is needed now:
diff --git a/gcc/match.pd b/gcc/match.pd
index 37c5be9e5f4..ca6c9eff624 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -4729,10 +4729,11 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
(for cmp (simple_comparison)
scmp (swapped_simple_comparison)
(simplify
- (cmp (bit_not@2 @0) CONSTANT_CLASS_P@1)
+ (cmp (nop_convert?:s (bit_not@2 @0)) CONSTANT_CLASS_P@1)
(if (single_use (@2)
&& (TREE_CODE (@1) == INTEGER_CST || TREE_CODE (@1) == VECTOR_CST))
- (scmp @0 (bit_not @1)
+ (with { tree type1 = TREE_TYPE (@1); }
+(scmp (convert:type1 @0) (bit_not @1))
(for cmp (simple_comparison)
/* Fold (double)float1 CMP (double)float2 into float1 CMP float2. */
I will add a testcase and submit it in a few.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 Andrew Pinski changed: What|Removed |Added Status|NEW |ASSIGNED --- Comment #15 from Andrew Pinski --- (In reply to Andrew Pinski from comment #14) > My patch needs to be re-written to use match.pd instead. I am going to update it.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 Andrew Pinski changed: What|Removed |Added Status|ASSIGNED|NEW --- Comment #14 from Andrew Pinski --- My patch needs to be re-written to use match.pd instead.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 Bug 31531 depends on bug 15459, which changed state. Bug 15459 Summary: [meta-bug] there should be a tree combiner like the rtl one https://gcc.gnu.org/bugzilla/show_bug.cgi?id=15459 What|Removed |Added Status|ASSIGNED|RESOLVED Resolution|--- |FIXED
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 --- Comment #13 from Andrew Pinski pinskia at gcc dot gnu.org 2012-02-14 03:59:12 UTC --- I committed the branch to the git only branch pinskia/treecombine. The only regression is gcc.target/i386/umod-3.c. This is not really a regression but rather a testcase issue as we produce better code and not produce a division at all.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531
--- Comment #9 from Andrew Pinski pinskia at gcc dot gnu.org 2012-02-12
22:58:41 UTC ---
(In reply to comment #8)
forwprop already handles:
int f(int a)
{
int b = ~a;
return b0;
}
It just needs to handle:
int f(unsigned a)
{
int b = ~a;
return b0;
}
forward_propagate_into_comparison only handles combing of two SSA_NAMEs, it
could handle dealing with a conversion also.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 --- Comment #10 from Andrew Pinski pinskia at gcc dot gnu.org 2012-02-13 00:44:11 UTC --- I have a patch which adds this optimization to fold. And a partial patch which adds some of it to forwprop but that fails because we have to create a temp variable.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 --- Comment #11 from Andrew Pinski pinskia at gcc dot gnu.org 2012-02-13 06:30:29 UTC --- I have a full patch now which also handles PR 14792 once that folding is included in fold.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 --- Comment #12 from Andrew Pinski pinskia at gcc dot gnu.org 2012-02-13 07:46:45 UTC --- Created attachment 26645 -- http://gcc.gnu.org/bugzilla/attachment.cgi?id=26645 Patch which fixes the problem (well the fold-const.c is really only needed for the original testcase) ChangeLog: * fold-const.c (fold_comparison): Handle ((CAST)~X). * tree-ssa-forwprop.c (combine_cond_expr_cond): Swap operands so that fold_binary_loc would not return a non-folded tree. Don't call canonicalize_cond_expr_cond. (forward_propagate_into_comparison_1): Also combine casts into the comparisons. (expand_possible_comparison): New function. (forward_propagate_into_comparison): Call expand_possible_comparison. (forward_propagate_into_gimple_cond): Take also a gsi. Call expand_possible_comparison. (forward_propagate_into_cond): Call expand_possible_comparison. (ssa_forward_propagate_and_combine): Update call to forward_propagate_into_gimple_cond.
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531
--- Comment #6 from Andrew Pinski pinskia at gcc dot gnu.org 2012-02-12
04:35:53 UTC ---
The shortest testcase for the problem function:
int isnegative_optimized_4(unsigned int X) {
int result; // Y is the conditional expression of if-else.
if ((~X) 31) result = 0;
elseresult = 1;
return result;
}
We need to combine the following gimple:
D.2293_3 = ~X_2(D);
D.2294_4 = (int) D.2293_3;
D.2424_9 = D.2294_4 = 0;
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 Andrew Pinski pinskia at gcc dot gnu.org changed: What|Removed |Added Status|NEW |ASSIGNED AssignedTo|unassigned at gcc dot |pinskia at gcc dot gnu.org |gnu.org | --- Comment #7 from Andrew Pinski pinskia at gcc dot gnu.org 2012-02-12 04:37:08 UTC --- Note: D.2424_9 = D.2294_4 = 0; is the same as: D.2424_9 = D.2294_4 0;
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531
--- Comment #8 from Andrew Pinski pinskia at gcc dot gnu.org 2012-02-12
04:39:01 UTC ---
forwprop already handles:
int f(int a)
{
int b = ~a;
return b0;
}
It just needs to handle:
int f(unsigned a)
{
int b = ~a;
return b0;
}
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531 Andrew Pinski pinskia at gcc dot gnu.org changed: What|Removed |Added Keywords||TREE --- Comment #5 from Andrew Pinski pinskia at gcc dot gnu.org 2011-07-18 22:46:55 UTC --- All the functions produce the same assembler output in 4.6.1 on x86. Only isnegative_optimized_4 produces a difference on some targets (and a difference IR coming into expand).
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
--- Comment #4 from pinskia at gcc dot gnu dot org 2007-04-15 06:43 --- There are a couple of ways of fixing this bug, folding while going out of ssa. Doing a tree combined which is PR 15459. Or the patch which I attached. -- pinskia at gcc dot gnu dot org changed: What|Removed |Added BugsThisDependsOn||15459 Status|UNCONFIRMED |NEW Ever Confirmed|0 |1 Last reconfirmed|-00-00 00:00:00 |2007-04-15 06:43:24 date|| http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
--- Comment #2 from pinskia at gcc dot gnu dot org 2007-04-10 19:17 --- What the real issue is that we don't fold !(x = 0) into x 0 before expanding. -- pinskia at gcc dot gnu dot org changed: What|Removed |Added Component|c |middle-end Keywords||missed-optimization Summary|A microoptimization of |A microoptimization of |isnegative or |isnegative of signed integer |greaterthan2millions. | http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531
[Bug middle-end/31531] A microoptimization of isnegative of signed integer
--- Comment #3 from pinskia at gcc dot gnu dot org 2007-04-10 19:20 ---
A quick patch like:
Index: expr.c
===
--- expr.c (revision 123691)
+++ expr.c (working copy)
@@ -6828,7 +6828,7 @@
tree type;
int unsignedp;
enum machine_mode mode;
- enum tree_code code = TREE_CODE (exp);
+ enum tree_code code;
optab this_optab;
rtx subtarget, original_target;
int ignore;
@@ -6840,6 +6840,9 @@
type) \
: (expr))
+ if (TREE_CODE (exp) != COND_EXPR !GIMPLE_TUPLE_P (exp))
+exp = fold (exp);
+ code = TREE_CODE(exp);
if (GIMPLE_STMT_P (exp))
{
type = void_type_node;
Makes all functions emit the same code.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31531
