[Bug middle-end/47048] misc vect.exp failures with -fgraphite-identity enabled at -O2.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=47048 Andrew Pinski changed: What|Removed |Added Status|ASSIGNED|NEW Keywords||missed-optimization Assignee|spop at gcc dot gnu.org|unassigned at gcc dot gnu.org CC||pinskia at gcc dot gnu.org
[Bug middle-end/47048] misc vect.exp failures with -fgraphite-identity enabled at -O2.
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47048 --- Comment #7 from Jack Howarth howarth at nitro dot med.uc.edu 2011-07-29 01:54:25 UTC --- Checking the failures in vect.exp at r176905 with -fgraphite-identity enabled at -O2, we seem to have picked up a bunch for -flto at -m32 on x86_64-apple-darwin11... FAIL: gcc.dg/vect/pr36493.c -flto scan-tree-dump-times vect vectorized 1 loops 1 FAIL: gcc.dg/vect/pr38529.c -flto scan-tree-dump-times vect OUTER LOOP VECTORIZED 1 FAIL: gcc.dg/vect/vect-iv-8.c -flto scan-tree-dump-times vect vectorized 1 loops 1 FAIL: gcc.dg/vect/vect-outer-2.c -flto scan-tree-dump-times vect OUTER LOOP VECTORIZED 1 FAIL: gcc.dg/vect/vect-outer-2a.c -flto scan-tree-dump-times vect OUTER LOOP VECTORIZED 1 FAIL: gcc.dg/vect/vect-outer-2c.c -flto scan-tree-dump-times vect OUTER LOOP VECTORIZED 1 FAIL: gcc.dg/vect/slp-13.c -flto scan-tree-dump-times vect vectorized 2 loops 1 FAIL: gcc.dg/vect/slp-widen-mult-half.c -flto scan-tree-dump-times vect vect_recog_widen_mult_pattern: detected 2 FAIL: gcc.dg/vect/bb-slp-16.c -flto scan-tree-dump-times slp basic block vectorized using SLP 1
[Bug middle-end/47048] misc vect.exp failures with -fgraphite-identity enabled at -O2.
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47048 --- Comment #6 from Jack Howarth howarth at nitro dot med.uc.edu 2011-02-07 12:29:10 UTC --- (In reply to comment #4) More specifically, the code that is now confused by the (int) cast is: dr_analyze_innermost (dr); dr_analyze_indices (dr, nest, loop); in create_data_ref. Is this fixable for the 4.6.0 release or would that be to invasive for stage 4?
[Bug middle-end/47048] misc vect.exp failures with -fgraphite-identity enabled at -O2.
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47048 --- Comment #1 from Jack Howarth howarth at nitro dot med.uc.edu 2011-02-01 20:20:28 UTC --- With the patches proposed in http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40979#c14 and http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40979#c17 and -fgraphite-identity enabled at -O2, the -m32 vect.exp failures are reduced to just 24... FAIL: gcc.dg/vect/pr36493.c scan-tree-dump-times vect vectorized 1 loops 1 FAIL: gcc.dg/vect/pr36630.c scan-tree-dump-times vect vectorized 1 loops 1 XPASS: gcc.dg/vect/vect-iv-7.c scan-tree-dump-times vect vectorized 1 loops 1 FAIL: gcc.dg/vect/vect-iv-8.c scan-tree-dump-times vect vectorized 1 loops 1 FAIL: gcc.dg/vect/vect-outer-2.c scan-tree-dump-times vect OUTER LOOP VECTORIZED 1 FAIL: gcc.dg/vect/vect-outer-2a.c scan-tree-dump-times vect OUTER LOOP VECTORIZED 1 FAIL: gcc.dg/vect/vect-outer-2c.c scan-tree-dump-times vect OUTER LOOP VECTORIZED 1 FAIL: gcc.dg/vect/slp-13.c scan-tree-dump-times vect vectorized 2 loops 1 FAIL: gcc.dg/vect/bb-slp-16.c scan-tree-dump-times slp basic block vectorized using SLP 1 FAIL: gcc.dg/vect/wrapv-vect-7.c scan-tree-dump-times vect vectorized 2 loops 1 FAIL: gcc.dg/vect/no-scevccp-vect-iv-3.c scan-tree-dump-times vect vect_recog_widen_sum_pattern: detected 1 XPASS: gcc.dg/vect/no-scevccp-outer-12.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-outer-15.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-outer-16.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-outer-17.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-outer-18.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-outer-2.c scan-tree-dump-times vect OUTER LOOP VECTORIZED 1 FAIL: gcc.dg/vect/no-scevccp-outer-20.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-outer-21.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-outer-3.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-outer-5.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-outer-6.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-outer-8.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-noreassoc-outer-3.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-noreassoc-outer-4.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1 FAIL: gcc.dg/vect/no-scevccp-noreassoc-outer-5.c scan-tree-dump-times vect OUTER LOOP VECTORIZED. 1
[Bug middle-end/47048] misc vect.exp failures with -fgraphite-identity enabled at -O2.
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47048
Sebastian Pop spop at gcc dot gnu.org changed:
What|Removed |Added
Status|UNCONFIRMED |ASSIGNED
Last reconfirmed||2011.02.01 21:15:06
AssignedTo|unassigned at gcc dot |spop at gcc dot gnu.org
|gnu.org |
Ever Confirmed|0 |1
--- Comment #2 from Sebastian Pop spop at gcc dot gnu.org 2011-02-01 21:15:06
UTC ---
Looks like the classic data dependence test can be improved:
A[0] does not overlap with A[1], and so there is no dependence:
Data ref a:
#(Data Ref:
# bb: 6
# stmt: points[D.2720_36][0] = start$0_8(D);
# ref: points[D.2720_36][0];
# base_object: points
# Access function 0: 0
# Access function 1: (int) {0, +, 1}_2;
#)
Data ref b:
#(Data Ref:
# bb: 6
# stmt: points[D.2720_36][1] = D.2723_40;
# ref: points[D.2720_36][1];
# base_object: points
# Access function 0: 1
# Access function 1: (int) {0, +, 1}_2;
#)
affine dependence test not usable: access function not affine or constant.
(dependence classified: scev_not_known)
)
/home/seb/gcc/trunk/gcc/testsuite/gcc.dg/vect/pr33866.c:13: note: not
vectorized: data ref analysis failed points[D.2720_36][0] = start$0_8(D);
[Bug middle-end/47048] misc vect.exp failures with -fgraphite-identity enabled at -O2.
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47048 --- Comment #3 from Sebastian Pop spop at gcc dot gnu.org 2011-02-01 23:28:33 UTC --- I think that the problem is (again) the type of the induction variables that we create in graphite: in pr33866.c we have D.2720_37 = (int) graphite_IV.5_34; and this (int) confuses the data reference analysis.
[Bug middle-end/47048] misc vect.exp failures with -fgraphite-identity enabled at -O2.
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47048 --- Comment #4 from Sebastian Pop spop at gcc dot gnu.org 2011-02-01 23:32:20 UTC --- More specifically, the code that is now confused by the (int) cast is: dr_analyze_innermost (dr); dr_analyze_indices (dr, nest, loop); in create_data_ref.
[Bug middle-end/47048] misc vect.exp failures with -fgraphite-identity enabled at -O2.
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47048 --- Comment #5 from Joost VandeVondele Joost.VandeVondele at pci dot uzh.ch 2011-02-02 07:42:22 UTC --- (In reply to comment #4) More specifically, the code that is now confused by the (int) cast There might be overlap with PR47341 (also casts that confuse dependency info)?
