diff options
author | Jakub Jelinek <jakub@redhat.com> | 2022-01-29 17:54:43 +0100 |
---|---|---|
committer | Jakub Jelinek <jakub@redhat.com> | 2022-01-29 17:54:43 +0100 |
commit | a1544878966020d1f7a640b35d1f7a5f0e055624 (patch) | |
tree | 6578967d246f4f0b2147cb07cc2915b4772b552a | |
parent | f6f2d6cfec1c2fe9570b98211be58329d8d7749b (diff) | |
download | gcc-a1544878966020d1f7a640b35d1f7a5f0e055624.zip gcc-a1544878966020d1f7a640b35d1f7a5f0e055624.tar.gz gcc-a1544878966020d1f7a640b35d1f7a5f0e055624.tar.bz2 |
match.pd: Fix up 1 / X for unsigned X optimization [PR104280]
On Fri, Jan 28, 2022 at 11:38:23AM -0700, Jeff Law wrote:
> Thanks. Given the original submission and most of the review work was done
> prior to stage3 closing, I went ahead and installed this on the trunk.
Unfortunately this breaks quite a lot of things.
The main problem is that GIMPLE allows EQ_EXPR etc. only with BOOLEAN_TYPE
or with TYPE_PRECISION == 1 integral type (or vector boolean).
Violating this causes verification failures in tree-cfg.cc in some cases,
in other cases wrong-code issues because before it is verified we e.g.
transform
1U / x
into
x == 1U
and later into
x (because we assume that == type must be one of the above cases and
when it is the same type as the type of the first operand, for boolean-ish
cases it should be equivalent).
Fixed by changing that
(eq @1 { build_one_cst (type); })
into
(convert (eq:boolean_type_node @1 { build_one_cst (type); }))
Note, I'm not 100% sure if :boolean_type_node is required in that case,
I see some spots in match.pd that look exactly like this, while there is
e.g. (convert (le ...)) that supposedly does the right thing too.
The signed integer 1/X case doesn't need changes changes, for
(cond (le ...) ...)
le gets correctly boolean_type_node and cond should use type.
I've also reformatted it, some lines were too long, match.pd uses
indentation by 1 column instead of 2 etc.
2022-01-29 Jakub Jelinek <jakub@redhat.com>
Andrew Pinski <apinski@marvell.com>
PR tree-optimization/104279
PR tree-optimization/104280
PR tree-optimization/104281
* match.pd (1 / X -> X == 1 for unsigned X): Build eq with
boolean_type_node and convert to type. Formatting fixes.
* gcc.dg/torture/pr104279.c: New test.
* gcc.dg/torture/pr104280.c: New test.
* gcc.dg/torture/pr104281.c: New test.
-rw-r--r-- | gcc/match.pd | 28 | ||||
-rw-r--r-- | gcc/testsuite/gcc.dg/torture/pr104279.c | 12 | ||||
-rw-r--r-- | gcc/testsuite/gcc.dg/torture/pr104280.c | 16 | ||||
-rw-r--r-- | gcc/testsuite/gcc.dg/torture/pr104281.c | 22 |
4 files changed, 66 insertions, 12 deletions
diff --git a/gcc/match.pd b/gcc/match.pd index 0544ddd..b942cb2 100644 --- a/gcc/match.pd +++ b/gcc/match.pd @@ -435,18 +435,22 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT) && TYPE_UNSIGNED (type)) (trunc_divmod @0 @1)))) - /* 1 / X -> X == 1 for unsigned integer X. - 1 / X -> X >= -1 && X <= 1 ? X : 0 for signed integer X. - But not for 1 / 0 so that we can get proper warnings and errors, - and not for 1-bit integers as they are edge cases better handled elsewhere. */ -(simplify - (trunc_div integer_onep@0 @1) - (if (INTEGRAL_TYPE_P (type) && !integer_zerop (@1) && TYPE_PRECISION (type) > 1) - (if (TYPE_UNSIGNED (type)) - (eq @1 { build_one_cst (type); }) - (with { tree utype = unsigned_type_for (type); } - (cond (le (plus (convert:utype @1) { build_one_cst (utype); }) { build_int_cst (utype, 2); }) - @1 { build_zero_cst (type); }))))) +/* 1 / X -> X == 1 for unsigned integer X. + 1 / X -> X >= -1 && X <= 1 ? X : 0 for signed integer X. + But not for 1 / 0 so that we can get proper warnings and errors, + and not for 1-bit integers as they are edge cases better handled + elsewhere. */ +(simplify + (trunc_div integer_onep@0 @1) + (if (INTEGRAL_TYPE_P (type) + && !integer_zerop (@1) + && TYPE_PRECISION (type) > 1) + (if (TYPE_UNSIGNED (type)) + (convert (eq:boolean_type_node @1 { build_one_cst (type); })) + (with { tree utype = unsigned_type_for (type); } + (cond (le (plus (convert:utype @1) { build_one_cst (utype); }) + { build_int_cst (utype, 2); }) + @1 { build_zero_cst (type); }))))) /* Combine two successive divisions. Note that combining ceil_div and floor_div is trickier and combining round_div even more so. */ diff --git a/gcc/testsuite/gcc.dg/torture/pr104279.c b/gcc/testsuite/gcc.dg/torture/pr104279.c new file mode 100644 index 0000000..7c4fc20 --- /dev/null +++ b/gcc/testsuite/gcc.dg/torture/pr104279.c @@ -0,0 +1,12 @@ +/* PR tree-optimization/104279 */ +/* { dg-do compile } */ + +unsigned a, b; + +int +main () +{ + b = ~(0 || ~0); + a = ~b / ~a; + return 0; +} diff --git a/gcc/testsuite/gcc.dg/torture/pr104280.c b/gcc/testsuite/gcc.dg/torture/pr104280.c new file mode 100644 index 0000000..c5a1d9b --- /dev/null +++ b/gcc/testsuite/gcc.dg/torture/pr104280.c @@ -0,0 +1,16 @@ +/* PR tree-optimization/104280 */ +/* { dg-do run } */ + +int +foo (unsigned b, int c) +{ + return b / c; +} + +int +main () +{ + if (foo (1, 2) != 0) + __builtin_abort (); + return 0; +} diff --git a/gcc/testsuite/gcc.dg/torture/pr104281.c b/gcc/testsuite/gcc.dg/torture/pr104281.c new file mode 100644 index 0000000..35c8e20 --- /dev/null +++ b/gcc/testsuite/gcc.dg/torture/pr104281.c @@ -0,0 +1,22 @@ +/* PR tree-optimization/104281 */ +/* { dg-do run } */ + +unsigned a = 1; +int b, c = 2; +long d; + +int +main () +{ + while (1) + { + int m = a; + L: + a = ~(-(m || b & d)); + b = ((1 ^ a) / c); + if (b) + goto L; + break; + } + return 0; +} |