From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 2136) id 8087F3857B89; Wed, 14 Sep 2022 15:08:07 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 8087F3857B89 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org; s=default; t=1663168087; bh=6WpJwTnY6T4N9Ce0uQSyaEzcZX8YUjQsAcQBu6/7SzU=; h=From:To:Subject:Date:From; b=HVzd4a6J/e5bEZ2f/VEefEhd4VznfVWcEVA11X7q0WnATODBrNem135vvxy/H1D7t 2QAwJwcbyKKIqQMTxzdr4TpYGgYDjFVstnGzzDWAyd/TXQQjq7qdamYqc6dBBaC53u /gogJTCs/xi5POyQh8g12hvAk8HIlI/pI0S7Ixjw= MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset="utf-8" From: Aldy Hernandez To: gcc-cvs@gcc.gnu.org Subject: [gcc r13-2671] frange: add both zeros to ranges when there's the possiblity of equality. X-Act-Checkin: gcc X-Git-Author: Aldy Hernandez X-Git-Refname: refs/heads/master X-Git-Oldrev: 7d647a8d49d49dfbb82061e6c66a19c142331169 X-Git-Newrev: 3c85c1188579bee9c6f8b8f5223daf49b772e594 Message-Id: <20220914150807.8087F3857B89@sourceware.org> Date: Wed, 14 Sep 2022 15:08:07 +0000 (GMT) List-Id: https://gcc.gnu.org/g:3c85c1188579bee9c6f8b8f5223daf49b772e594 commit r13-2671-g3c85c1188579bee9c6f8b8f5223daf49b772e594 Author: Aldy Hernandez Date: Wed Sep 14 08:11:54 2022 +0200 frange: add both zeros to ranges when there's the possiblity of equality. Every time there's equality at play, we must be careful that any equality with zero matches both -0.0 and +0.0 when honoring signed zeros. We were doing this correctly for the == and != op1_range operators (albeit inefficiently), but aren't doing it at all when building >= and <=. This fixes the oversight. There is change in functionality here for the build_* functions. This is the last "simple" patch I submit before overhauling NAN and sign tracking. And that will likely be after Cauldron because it will need further testing (lapack, ppc64le, -ffinite-math-only, etc). Regstrapped on x86-64 Linux, plus I ran selftests for -ffinite-math-only. gcc/ChangeLog: * range-op-float.cc (frange_add_zeros): New. (build_le): Call frange_add_zeros. (build_ge): Same. (foperator_equal::op1_range): Same. (foperator_not_equal::op1_range): Same. Diff: --- gcc/range-op-float.cc | 31 +++++++++++++++++++++++++------ 1 file changed, 25 insertions(+), 6 deletions(-) diff --git a/gcc/range-op-float.cc b/gcc/range-op-float.cc index 8f3e5241313..fbc14a730ad 100644 --- a/gcc/range-op-float.cc +++ b/gcc/range-op-float.cc @@ -208,6 +208,19 @@ frange_drop_ninf (frange &r, tree type) r.intersect (tmp); } +// If zero is in R, make sure both -0.0 and +0.0 are in the range. + +static inline void +frange_add_zeros (frange &r, tree type) +{ + if (r.undefined_p () || r.known_nan ()) + return; + + if (HONOR_SIGNED_ZEROS (type) + && (real_iszero (&r.lower_bound ()) || real_iszero (&r.upper_bound ()))) + r.set_signbit (fp_prop::VARYING); +} + // Build a range that is <= VAL and store it in R. static bool @@ -219,6 +232,10 @@ build_le (frange &r, tree type, const frange &val) return false; } r.set (type, dconstninf, val.upper_bound ()); + + // Add both zeros if there's the possibility of zero equality. + frange_add_zeros (r, type); + return true; } @@ -257,6 +274,10 @@ build_ge (frange &r, tree type, const frange &val) return false; } r.set (type, val.lower_bound (), dconstinf); + + // Add both zeros if there's the possibility of zero equality. + frange_add_zeros (r, type); + return true; } @@ -376,9 +397,8 @@ foperator_equal::op1_range (frange &r, tree type, case BRS_TRUE: // If it's true, the result is the same as OP2. r = op2; - // Make sure we don't copy the sign bit if we may have a zero. - if (HONOR_SIGNED_ZEROS (type) && r.contains_p (build_zero_cst (type))) - r.set_signbit (fp_prop::VARYING); + // Add both zeros if there's the possibility of zero equality. + frange_add_zeros (r, type); // The TRUE side of op1 == op2 implies op1 is !NAN. r.clear_nan (); break; @@ -480,9 +500,8 @@ foperator_not_equal::op1_range (frange &r, tree type, case BRS_FALSE: // If it's false, the result is the same as OP2. r = op2; - // Make sure we don't copy the sign bit if we may have a zero. - if (HONOR_SIGNED_ZEROS (type) && r.contains_p (build_zero_cst (type))) - r.set_signbit (fp_prop::VARYING); + // Add both zeros if there's the possibility of zero equality. + frange_add_zeros (r, type); // The FALSE side of op1 != op2 implies op1 is !NAN. r.clear_nan (); break;