| Message ID | 1636834427-22589-1-git-send-email-apinski@marvell.com |
|---|---|
| State | New |
| Headers | show |
| Series | tree-optimization: [PR103218] Fold ((type)(a<0)) << SIGNBITOFA into ((type)a) & signbit | expand |
On Sat, Nov 13, 2021 at 9:14 PM apinski--- via Gcc-patches <gcc-patches@gcc.gnu.org> wrote: > > From: Andrew Pinski <apinski@marvell.com> > > This folds Fold ((type)(a<0)) << SIGNBITOFA into ((type)a) & signbit inside match.pd. > This was already handled in fold-cost by: > /* A < 0 ? <sign bit of A> : 0 is simply (A & <sign bit of A>). */ > I have not removed as we only simplify "a ? POW2 : 0" at the gimple level to "a << CST1" > and fold actually does the reverse of folding "(a<0)<<CST" into "(a<0) ? 1<<CST : 0". > OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions. OK. Thanks, Richard. > PR tree-optimization/103218 > > gcc/ChangeLog: > > * match.pd: New pattern for "((type)(a<0)) << SIGNBITOFA". > > gcc/testsuite/ChangeLog: > > * gcc.dg/tree-ssa/pr103218-1.c: New test. > --- > gcc/match.pd | 10 ++++++++ > gcc/testsuite/gcc.dg/tree-ssa/pr103218-1.c | 28 ++++++++++++++++++++++ > 2 files changed, 38 insertions(+) > create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr103218-1.c > > diff --git a/gcc/match.pd b/gcc/match.pd > index a319aefa808..df31964e02f 100644 > --- a/gcc/match.pd > +++ b/gcc/match.pd > @@ -865,6 +865,16 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT) > { tree utype = unsigned_type_for (type); } > (convert (rshift (lshift (convert:utype @0) @2) @3)))))) > > +/* Fold ((type)(a<0)) << SIGNBITOFA into ((type)a) & signbit. */ > +(simplify > + (lshift (convert (lt @0 integer_zerop@1)) INTEGER_CST@2) > + (if (TYPE_SIGN (TREE_TYPE (@0)) == SIGNED > + && wi::eq_p (wi::to_wide (@2), TYPE_PRECISION (TREE_TYPE (@0)) - 1)) > + (with { wide_int wone = wi::one (TYPE_PRECISION (type)); } > + (bit_and (convert @0) > + { wide_int_to_tree (type, > + wi::lshift (wone, wi::to_wide (@2))); })))) > + > /* Fold (-x >> C) into -(x > 0) where C = precision(type) - 1. */ > (for cst (INTEGER_CST VECTOR_CST) > (simplify > diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr103218-1.c b/gcc/testsuite/gcc.dg/tree-ssa/pr103218-1.c > new file mode 100644 > index 00000000000..f086f073b38 > --- /dev/null > +++ b/gcc/testsuite/gcc.dg/tree-ssa/pr103218-1.c > @@ -0,0 +1,28 @@ > +/* { dg-do compile } */ > +/* { dg-options "-O2 -fdump-tree-optimized" } */ > +/* PR tree-optimization/103218 */ > + > +/* These first two are removed during forwprop1 */ > +signed char f(signed char a) > +{ > + signed char t = a < 0; > + int tt = (unsigned char)(t << 7); > + return tt; > +} > +signed char f0(signed char a) > +{ > + unsigned char t = a < 0; > + int tt = (unsigned char)(t << 7); > + return tt; > +} > + > +/* This one is removed during phiopt. */ > +signed char f1(signed char a) > +{ > + if (a < 0) > + return 1u<<7; > + return 0; > +} > + > +/* These three examples should remove "a < 0" by optimized. */ > +/* { dg-final { scan-tree-dump-times "< 0" 0 "optimized"} } */ > -- > 2.17.1 >
diff --git a/gcc/match.pd b/gcc/match.pd index a319aefa808..df31964e02f 100644 --- a/gcc/match.pd +++ b/gcc/match.pd @@ -865,6 +865,16 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT) { tree utype = unsigned_type_for (type); } (convert (rshift (lshift (convert:utype @0) @2) @3)))))) +/* Fold ((type)(a<0)) << SIGNBITOFA into ((type)a) & signbit. */ +(simplify + (lshift (convert (lt @0 integer_zerop@1)) INTEGER_CST@2) + (if (TYPE_SIGN (TREE_TYPE (@0)) == SIGNED + && wi::eq_p (wi::to_wide (@2), TYPE_PRECISION (TREE_TYPE (@0)) - 1)) + (with { wide_int wone = wi::one (TYPE_PRECISION (type)); } + (bit_and (convert @0) + { wide_int_to_tree (type, + wi::lshift (wone, wi::to_wide (@2))); })))) + /* Fold (-x >> C) into -(x > 0) where C = precision(type) - 1. */ (for cst (INTEGER_CST VECTOR_CST) (simplify diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr103218-1.c b/gcc/testsuite/gcc.dg/tree-ssa/pr103218-1.c new file mode 100644 index 00000000000..f086f073b38 --- /dev/null +++ b/gcc/testsuite/gcc.dg/tree-ssa/pr103218-1.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-O2 -fdump-tree-optimized" } */ +/* PR tree-optimization/103218 */ + +/* These first two are removed during forwprop1 */ +signed char f(signed char a) +{ + signed char t = a < 0; + int tt = (unsigned char)(t << 7); + return tt; +} +signed char f0(signed char a) +{ + unsigned char t = a < 0; + int tt = (unsigned char)(t << 7); + return tt; +} + +/* This one is removed during phiopt. */ +signed char f1(signed char a) +{ + if (a < 0) + return 1u<<7; + return 0; +} + +/* These three examples should remove "a < 0" by optimized. */ +/* { dg-final { scan-tree-dump-times "< 0" 0 "optimized"} } */
From: Andrew Pinski <apinski@marvell.com> This folds Fold ((type)(a<0)) << SIGNBITOFA into ((type)a) & signbit inside match.pd. This was already handled in fold-cost by: /* A < 0 ? <sign bit of A> : 0 is simply (A & <sign bit of A>). */ I have not removed as we only simplify "a ? POW2 : 0" at the gimple level to "a << CST1" and fold actually does the reverse of folding "(a<0)<<CST" into "(a<0) ? 1<<CST : 0". OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions. PR tree-optimization/103218 gcc/ChangeLog: * match.pd: New pattern for "((type)(a<0)) << SIGNBITOFA". gcc/testsuite/ChangeLog: * gcc.dg/tree-ssa/pr103218-1.c: New test. --- gcc/match.pd | 10 ++++++++ gcc/testsuite/gcc.dg/tree-ssa/pr103218-1.c | 28 ++++++++++++++++++++++ 2 files changed, 38 insertions(+) create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr103218-1.c