[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
--- Comment #13 from spop at gcc dot gnu dot org 2009-08-28 20:39 --- Subject: Bug 40965 Author: spop Date: Fri Aug 28 20:39:43 2009 New Revision: 151190 URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=151190 Log: 2009-08-28 Sebastian Pop sebastian@amd.com PR middle-end/40965 * graphite-poly.c (apply_poly_transforms): Remove legality test before any transform. Modified: trunk/gcc/ChangeLog trunk/gcc/ChangeLog.graphite trunk/gcc/graphite-poly.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
--- Comment #14 from spop at gcc dot gnu dot org 2009-08-28 20:44 --- Fixed. -- spop at gcc dot gnu dot org changed: What|Removed |Added Status|ASSIGNED|RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
--- Comment #12 from spop at gcc dot gnu dot org 2009-08-27 19:13 --- Subject: Re: [4.5 Regression][graphite] slow compilation This fix also decreases the compile time of graphite in general. Â I will send a pointer to the automatic tester output that shows this on the SPEC benchmarks. Before data dep compile time fixes: http://gcc-graphite-test.googlegroups.com/attach/2944d6b8db76028b/cpu2006.compile.png?view=1part=4 http://gcc-graphite-test.googlegroups.com/attach/2944d6b8db76028b/cpu2006?view=1part=2 After: http://gcc-graphite-test.googlegroups.com/attach/e5e1c39697fdc3e6/cpu2006.compile.png?view=1part=3 http://gcc-graphite-test.googlegroups.com/attach/e5e1c39697fdc3e6/cpu2006?view=1part=2 Sebastian -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
--- Comment #11 from spop at gcc dot gnu dot org 2009-08-26 14:42 --- This is a fix for this particular bug: I am considering this bug closed, and I will close it when the patch will be in trunk. This fix also decreases the compile time of graphite in general. I will send a pointer to the automatic tester output that shows this on the SPEC benchmarks. Sebastian -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
--- Comment #8 from spop at gcc dot gnu dot org 2009-08-25 21:03 --- Disabling one of the useless gcc_asserts that verifies the consistency of the original representation before any transform, I am down to 0 seconds for the data dependence test, and the following compile times for graphite: Graphite loop transforms: 2.60 (29%) usr 0.12 (60%) sys 2.69 (29%) wall 2365 kB ( 7%) ggc Graphite code generation: 0.14 ( 2%) usr 0.00 ( 0%) sys 0.21 ( 2%) wall 157 kB ( 0%) ggc TOTAL : 9.09 0.20 9.37 34582 kB This is in the graphite branch, and so -O3 contains -floop-strip-mine and -fgraphite-identity. I also have tested with -floop-interchange and there is no more overhead. I will test the patch in the graphite branch and then I will commit it to trunk. Sebastian -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
--- Comment #9 from spop at gcc dot gnu dot org 2009-08-25 21:58 --- Subject: Bug 40965 Author: spop Date: Tue Aug 25 21:58:20 2009 New Revision: 151102 URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=151102 Log: Remove legality test before any transform. 2009-08-25 Sebastian Pop sebastian@amd.com PR middle-end/40965 * graphite-poly.c (apply_poly_transforms): Remove legality test before any transform. Modified: branches/graphite/gcc/ChangeLog.graphite branches/graphite/gcc/graphite-poly.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
--- Comment #10 from howarth at nitro dot med dot uc dot edu 2009-08-26 03:49 --- Does r151102 address the exponential code generation in graphite generally or just that for this particular test case? -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
-- mmitchel at gcc dot gnu dot org changed: What|Removed |Added Priority|P3 |P2 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
--- Comment #7 from spop at gcc dot gnu dot org 2009-08-12 15:03 --- This still fails on rev150694. The problem is in the data dependence analyzer: Graphite loop transforms: 0.34 ( 0%) usr 0.00 ( 0%) sys 0.29 (0%) wall 2105 kB ( 7%) ggc Graphite data dep analysis: 107.68 (95%) usr 0.67 (92%) sys 109.43 (95%) wall 0 kB ( 0%) ggc Graphite code generation: 0.21 ( 0%) usr 0.00 ( 0%) sys 0.15 (0%) wall 157 kB ( 1%) ggc -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
--- Comment #5 from jv244 at cam dot ac dot uk 2009-08-05 10:59 --- Checked against the 4.4 branch, with the following timings: 4.4 w/o1.79 w 1.80 4.5 w/o4.23 w144.30 -- jv244 at cam dot ac dot uk changed: What|Removed |Added Known to fail||4.5.0 Known to work||4.4.2 Summary|[graphite] slow compilation |[4.5 Regression][graphite] ||slow compilation Target Milestone|--- |4.5.0 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
Re: [Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
What changed from 4.4 to 4.5 is that we now get to compile larger SCoPs with Graphite. In 4.5, Graphite can deal with reductions and other unhandled constructs like the NE_EXPR that Fortran is frequently using for representing the exit condition of DO loops.
[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
--- Comment #6 from sebpop at gmail dot com 2009-08-05 14:04 --- Subject: Re: [4.5 Regression][graphite] slow compilation What changed from 4.4 to 4.5 is that we now get to compile larger SCoPs with Graphite. In 4.5, Graphite can deal with reductions and other unhandled constructs like the NE_EXPR that Fortran is frequently using for representing the exit condition of DO loops. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
[Bug middle-end/40965] [4.5 Regression][graphite] slow compilation
-- spop at gcc dot gnu dot org changed: What|Removed |Added AssignedTo|unassigned at gcc dot gnu |spop at gcc dot gnu dot org |dot org | Status|UNCONFIRMED |ASSIGNED Ever Confirmed|0 |1 Last reconfirmed|-00-00 00:00:00 |2009-08-05 14:05:14 date|| http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40965
