August 2011 - PPL-devel - BUGSENG Mailman

[clattner@apple.com: Re: [cfe-dev] -frounding-math and llvm?]
by Jack Howarth 03 Aug '11

03 Aug '11

----- Forwarded message from Chris Lattner <clattner(a)apple.com> ----- X-Spam-Checker-Version: SpamAssassin 3.2.5 (2008-06-10) on bromo.med.uc.edu X-Spam-Level: X-Spam-Status: No, score=-0.9 required=4.4 tests=AWL,BAYES_00, DNS_FROM_OPENWHOIS,FH_DATE_PAST_20XX,RCVD_IN_DNSWL_MED autolearn=no version=3.2.5 X-Original-To: howarth(a)bromo.med.uc.edu Delivered-To: howarth(a)bromo.med.uc.edu X-AuditID: 11807134-b7c71ae0000014d0-a0-4e397ac8109b Subject: Re: [cfe-dev] -frounding-math and llvm? From: Chris Lattner <clattner(a)apple.com> In-reply-to: <20110803164538.GA27855(a)bromo.med.uc.edu> Date: Wed, 03 Aug 2011 09:48:43 -0700 Cc: Eli Friedman <eli.friedman(a)gmail.com>, cfe-dev(a)cs.uiuc.edu To: Jack Howarth <howarth(a)bromo.med.uc.edu> X-Mailer: Apple Mail (2.1084) X-Brightmail-Tracker: H4sIAAAAAAAAA+NgFlrDLMWRmVeSWpSXmKPExsUi+EjZRPdElaWfwcuFahbrln5mtjh07hWb xcOd2xkdmD2WHj3H4jHr6yQmj52z7rIHMEdx2aSk5mSWpRbp2yVwZbRsn81YcJyl4ureJ4wN jCeYuxg5OSQETCTOL1/GAmGLSVy4t56ti5GLQ0hgOZPEuzlngBwODmEBA4lL02VBangFDCWW bmpnB7GZBbQkbvx7yQRSwiagIfFjqQpImFPATOLd1CZWEJtFQEXiedsqVohyB4nLM25B2doS yxa+ZoYYaSNxcm8nE8TaJYwSd1ZfYQNJiAAVTdm8G+wECQFZiaZlGRMY+WchuWIWkitmIRm7 gJF5FaNgUWpOYqWhiV5iQUFOql5yfu4mRlAYNhSa7GA8+JP/EKMAB6MSD69kgqWfEGtiWXFl 7iFGCQ5mJRHeeZVAId6UxMqq1KL8+KLSnNTiQ4zSHCxK4ry335v5CQmkJ5akZqemFqQWwWSZ ODilGhhnt9Z5BjLw3Fiq499VI5qT965uzd64J+w/zQ9abZj2TnO2yN1Lzc8mb0z8keZ22iiR 8bz7zR06ju+lStbOD0vOvWtzYVLciaz7pb/cAp6W2gokqfhnL55QOXfpSTVNqYNfPitpbpjV 3PI0hWFaxpmri33CBGVtCr9WxK5ak+VVaWtv3Pnp8TwlluKMREMt5qLiRADA4eqjPwIAAA== On Aug 3, 2011, at 9:45 AM, Jack Howarth wrote: >>> Thanks in advance for any clarifications on the status of -frounding-math in the llvm >>> based compilers. >> >> Not supported, and I don't know of anyone who plans on working on it >> (or FENV_ACCESS) in the near future. > > Eli, > Is this the same situation with llvm-gcc? I don't see the warnings from llvm-gcc > but we do see the same failures in ppl's testsuite. Is this a case of llvm-gcc silently not supporting > -frounding-math? Yep, llvm-gcc just accepts and ignores the flag. -Chris ----- End forwarded message -----

1 0

[eli.friedman@gmail.com: Re: [cfe-dev] -frounding-math and llvm?]
by Jack Howarth 03 Aug '11

03 Aug '11

----- Forwarded message from Eli Friedman <eli.friedman(a)gmail.com> ----- X-Spam-Checker-Version: SpamAssassin 3.2.5 (2008-06-10) on bromo.med.uc.edu X-Spam-Level: X-Spam-Status: No, score=-0.3 required=4.4 tests=AWL,BAYES_00, DNS_FROM_OPENWHOIS,FH_DATE_PAST_20XX,RCVD_IN_DNSWL_LOW autolearn=no version=3.2.5 X-Original-To: howarth(a)bromo.med.uc.edu Delivered-To: howarth(a)bromo.med.uc.edu DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type:content-transfer-encoding; bh=+clIIC+b2uBpRTrZPIiFF4+CSc18YiLxGJDVuGChFk0=; b=AeEhd421aEAQjlZKu51PpDoxmuWbZxku6i16QBjTs6uwWgM8WIFz8kyVf5RNUxtKK2 JP19CcVMm6+RggxRUoNlDKlbzB+AHSPb04M7uREm7zawkrF0oTXBbmGahaYem84FAFMs rGljg0ZaAb5MwJhfULZZ0uhqRGRd6PifFzN9k= In-Reply-To: <20110803154556.GA26318(a)bromo.med.uc.edu> Date: Wed, 3 Aug 2011 09:10:42 -0700 Subject: Re: [cfe-dev] -frounding-math and llvm? From: Eli Friedman <eli.friedman(a)gmail.com> To: Jack Howarth <howarth(a)bromo.med.uc.edu> Cc: cfe-dev(a)cs.uiuc.edu On Wed, Aug 3, 2011 at 8:45 AM, Jack Howarth <howarth(a)bromo.med.uc.edu> wrote: > What exactly is the situation with support for -frounding-math in llvm-gcc/clang? I ask > because we are seeing regressions in the ppl 0.11.2 testsuite when compiling with either llvm-gcc > or clang from Xcode 4.1 as well as clang svn. The ppl developers seem to be seeing warnings to > the effect... > > clang: warning: argument unused during compilation: '-frounding-math' > > in their builds. > > http://www.cs.unipr.it/pipermail/ppl-devel/2011-August/017559.html > > Thanks in advance for any clarifications on the status of -frounding-math in the llvm > based compilers. Not supported, and I don't know of anyone who plans on working on it (or FENV_ACCESS) in the near future. -Eli ----- End forwarded message -----

1 0

[GIT] ppl/ppl(master): Silenced a few warnings.
by Roberto Bagnara 02 Aug '11

02 Aug '11

Module: ppl/ppl Branch: master Commit: 95879d0d2423f52e49571520780e919c6b142fa3 URL: http://www.cs.unipr.it/git/gitweb.cgi?p=ppl/ppl.git;a=commit;h=95879d0d2423… Author: Roberto Bagnara <bagnara(a)cs.unipr.it> Date: Tue Aug 2 21:55:22 2011 +0200 Silenced a few warnings. --- src/Interval.inlines.hh | 10 ++++++++-- src/Interval_Info.inlines.hh | 4 ++-- src/Interval_Restriction.defs.hh | 3 +++ 3 files changed, 13 insertions(+), 4 deletions(-) diff --git a/src/Interval.inlines.hh b/src/Interval.inlines.hh index f0e860a..f8329e0 100644 --- a/src/Interval.inlines.hh +++ b/src/Interval.inlines.hh @@ -518,7 +518,9 @@ Interval<To_Boundary, To_Info>::refine_universal(Relation_Symbol rel, return combine(V_EQ, V_EQ); info().clear_boundary_properties(UPPER); Result ru = Boundary_NS::assign(UPPER, upper(), info(), - LOWER, f_lower(x), SCALAR_INFO, !is_open(LOWER, f_lower(x), f_info(x))); + LOWER, f_lower(x), SCALAR_INFO, + !is_open(LOWER, f_lower(x), f_info(x))); + used(ru); normalize(); return I_ANY; } @@ -529,6 +531,7 @@ Interval<To_Boundary, To_Info>::refine_universal(Relation_Symbol rel, info().clear_boundary_properties(UPPER); Result ru = Boundary_NS::assign(UPPER, upper(), info(), LOWER, f_lower(x), SCALAR_INFO); + used(ru); normalize(); return I_ANY; } @@ -538,7 +541,9 @@ Interval<To_Boundary, To_Info>::refine_universal(Relation_Symbol rel, return combine(V_EQ, V_EQ); info().clear_boundary_properties(LOWER); Result rl = Boundary_NS::assign(LOWER, lower(), info(), - UPPER, f_upper(x), SCALAR_INFO, !is_open(UPPER, f_upper(x), f_info(x))); + UPPER, f_upper(x), SCALAR_INFO, + !is_open(UPPER, f_upper(x), f_info(x))); + used(rl); normalize(); return I_ANY; } @@ -549,6 +554,7 @@ Interval<To_Boundary, To_Info>::refine_universal(Relation_Symbol rel, info().clear_boundary_properties(LOWER); Result rl = Boundary_NS::assign(LOWER, lower(), info(), UPPER, f_upper(x), SCALAR_INFO); + used(rl); normalize(); return I_ANY; } diff --git a/src/Interval_Info.inlines.hh b/src/Interval_Info.inlines.hh index 9cb27a5..d22f458 100644 --- a/src/Interval_Info.inlines.hh +++ b/src/Interval_Info.inlines.hh @@ -35,12 +35,12 @@ Interval_Info_Null<Policy>::swap(Interval_Info_Null<Policy>&) { template <typename Policy> inline void -Interval_Info_Null<Policy>::ascii_dump(std::ostream& s) const { +Interval_Info_Null<Policy>::ascii_dump(std::ostream&) const { } template <typename Policy> inline bool -Interval_Info_Null<Policy>::ascii_load(std::istream& s) { +Interval_Info_Null<Policy>::ascii_load(std::istream&) { return true; } diff --git a/src/Interval_Restriction.defs.hh b/src/Interval_Restriction.defs.hh index 51c6e05..dd28edf 100644 --- a/src/Interval_Restriction.defs.hh +++ b/src/Interval_Restriction.defs.hh @@ -618,6 +618,9 @@ inline bool diff_restriction(Interval_Restriction_Integer_Modulo<T, Base>& to, const From1& x, const From2& y) { // FIXME: to be written + used(to); + used(x); + used(y); return true; }

1 0

[GIT] ppl/ppl(master): RAII_Temporary_Real_Relaxation is a struct.
by Roberto Bagnara 02 Aug '11

02 Aug '11

Module: ppl/ppl Branch: master Commit: aee79293c2b6bbce1f7087ea9276301aed8c760c URL: http://www.cs.unipr.it/git/gitweb.cgi?p=ppl/ppl.git;a=commit;h=aee79293c2b6… Author: Roberto Bagnara <bagnara(a)cs.unipr.it> Date: Tue Aug 2 21:49:45 2011 +0200 RAII_Temporary_Real_Relaxation is a struct. --- src/MIP_Problem.defs.hh | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/MIP_Problem.defs.hh b/src/MIP_Problem.defs.hh index c99cf80..492d85e 100644 --- a/src/MIP_Problem.defs.hh +++ b/src/MIP_Problem.defs.hh @@ -603,7 +603,7 @@ private: std::swap(i_vars, lp.i_variables); } }; - friend class RAII_Temporary_Real_Relaxation; + friend struct RAII_Temporary_Real_Relaxation; //! A tag type to distinguish normal vs. inheriting copy constructor. struct Inherit_Constraints {};

1 0

[GIT] ppl/ppl(master): Do not mix `int' and `dimension_type'.
by Roberto Bagnara 02 Aug '11

02 Aug '11

Module: ppl/ppl Branch: master Commit: d35568b97460439faf1f3f03aede2a5826038f1d URL: http://www.cs.unipr.it/git/gitweb.cgi?p=ppl/ppl.git;a=commit;h=d35568b97460… Author: Roberto Bagnara <bagnara(a)cs.unipr.it> Date: Tue Aug 2 21:25:01 2011 +0200 Do not mix `int' and `dimension_type'. --- src/Linear_Form.inlines.hh | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-) diff --git a/src/Linear_Form.inlines.hh b/src/Linear_Form.inlines.hh index 741d9bd..f2c3a33 100644 --- a/src/Linear_Form.inlines.hh +++ b/src/Linear_Form.inlines.hh @@ -200,7 +200,7 @@ Linear_Form<C>::ascii_dump(std::ostream& s) const { const char separator = ' '; dimension_type space = space_dimension(); s << space << "\n"; - for (int i = 0; i <= space; ++i) + for (dimension_type i = 0; i <= space; ++i) s << vec[i] << separator; s << "\n"; } @@ -214,7 +214,7 @@ Linear_Form<C>::ascii_load(std::istream& s) { return false; vec.resize(new_dim + 1, zero); - for (int i = 0; i <= new_dim; ++i) { + for (dimension_type i = 0; i <= new_dim; ++i) { if (!(s >> vec[i])) return false; }

1 0

Fwd: ppl 0.11.2 vs llvm
by Roberto Bagnara 02 Aug '11

02 Aug '11

-------- Original Message -------- Subject: ppl 0.11.2 vs llvm Date: Tue, 2 Aug 2011 11:19:11 -0400 From: Jack Howarth <howarth(a)bromo.med.uc.edu> To: bagnara(a)cs.unipr.it Roberto, Have you done any build of ppl 0.11.2 against the llvm-gcc or clang compilers? Mac OS X 10.7 (Lion) defaults to llvm-gcc and the fink project has switch to clang for their 10.7 builds. Both llvm based compilers produce a ppl 0.11.2 build which fails a subset of tests. For example the second build of affineimage1 in Box fails with... [MacPro:build/tests/Box] root# ./affineimage1 tests failed: test08 The problem occurs with both llvm-gcc-4.2 and clang3.0svn from Xcode 4.1 as well as current clang svn. We should try to find out if this is a bug in ppl or in the llvm backend of the llvm-gcc and clang compilers. Thanks in advance for any help in pinning this down. Jack ps The problem isn't system related since the Xcode gcc-4.2 compiler still produces a ppl build with no testsuite regressions. Apple will be depeciating the gcc-4.2 compiler out of Xcode in favor of llvm-gcc-4.2 at some point so we do need to address these ppl failures under llvm-based compilers.

1 0