On Mon, Aug 15, 2011 at 07:40, Joseph S. Myers <[email protected]> wrote: > I'd also like to point out that this is simply not a proper patch > submission at all, independent of the specific issues about changes to > prerequisites that require things to be raised and explained in a > self-contained way on the gcc list. Every patch submission needs an > explanation of the purpose of the patch - not just the ChangeLog entry > saying at a low level what changed, but the higher-level motivation for > why the patch was implemented and why it was implemented the way it was. > It also needs to say how the patch was tested. And a multi-patch series > should contain such an overview for the whole series (which might also > make clear whether the patches were tested individually or only in > combination), typically as patch 00/nn. Note that as stated in > contribute.html this all applies even when the patch is being sent by a > maintainer of the relevant part of the compiler.
Sorry I haven't had time to write all that is required. All the patches that I submitted are completely untested, except the make -k check RUNTESTFLAGS=graphite.exp The following fails are to be expected on amd64-linux, after the change to the data dependence using ISL: FAIL: gcc.dg/graphite/block-0.c execution test FAIL: gcc.dg/graphite/block-0.c scan-tree-dump-not graphite "will be loop blocked" FAIL: gfortran.dg/graphite/pr29581.f90 -O0 execution test FAIL: gfortran.dg/graphite/pr29581.f90 -O1 execution test FAIL: gfortran.dg/graphite/pr29581.f90 -O2 execution test FAIL: gfortran.dg/graphite/pr29581.f90 -O3 -fomit-frame-pointer execution test FAIL: gfortran.dg/graphite/pr29581.f90 -O3 -fomit-frame-pointer -funroll-loops execution test FAIL: gfortran.dg/graphite/pr29581.f90 -O3 -fomit-frame-pointer -funroll-all-loops -finline-functions execution test FAIL: gfortran.dg/graphite/pr29581.f90 -O3 -g execution test FAIL: gfortran.dg/graphite/pr29581.f90 -Os execution test FAIL: gfortran.dg/graphite/pr29832.f90 -O0 execution test FAIL: gfortran.dg/graphite/pr29832.f90 -O1 execution test FAIL: gfortran.dg/graphite/pr29832.f90 -O2 execution test FAIL: gfortran.dg/graphite/pr29832.f90 -O3 -fomit-frame-pointer execution test FAIL: gfortran.dg/graphite/pr29832.f90 -O3 -fomit-frame-pointer -funroll-loops execution test FAIL: gfortran.dg/graphite/pr29832.f90 -O3 -fomit-frame-pointer -funroll-all-loops -finline-functions execution test FAIL: gfortran.dg/graphite/pr29832.f90 -O3 -g execution test FAIL: gfortran.dg/graphite/pr29832.f90 -Os execution test I suspect that these are problems in the code generation part, now that the dependence test is validating more transforms. Sebastian
