| Message ID | CAAe5K+U1P8kS5=0Bmb3Zgy1JeqDw-ZCoeQvCa_6dtG3h+7i59w@mail.gmail.com |
|---|---|
| State | New |
| Headers | show |
Probably need to file a bug to track the copyrename2 problem. David On Wed, Oct 1, 2014 at 9:59 AM, Teresa Johnson <tejohnson@google.com> wrote: > This patch removes some asserts my jump threading patch r215739 added. > An upstream pass (copyrename2) is introducing some bogus profile > counts, so we can't assert that counts are 0 when there is no profile > data for the function. > > Tested on testcase attached to PR63422. Currently running gcc > regression tests for x86_64-unknown-linux-gnu. Ok for trunk if that > passes? > > Thanks, > Teresa > > 2014-10-01 Teresa Johnson <tejohnson@google.com> > > PR middle-end/63422 > * tree-ssa-threadupdate.c (freqs_to_counts_path): Remove > asserts to handle incoming insanities. > > Index: tree-ssa-threadupdate.c > =================================================================== > --- tree-ssa-threadupdate.c (revision 215739) > +++ tree-ssa-threadupdate.c (working copy) > @@ -977,20 +977,14 @@ freqs_to_counts_path (struct redirection_data *rd) > edge ein; > edge_iterator ei; > FOR_EACH_EDGE (ein, ei, e->dest->preds) > - { > - gcc_assert (!ein->count); > - ein->count = EDGE_FREQUENCY (ein); > - } > + ein->count = EDGE_FREQUENCY (ein); > > for (unsigned int i = 1; i < path->length (); i++) > { > edge epath = (*path)[i]->e; > - gcc_assert (!epath->count); > edge esucc; > FOR_EACH_EDGE (esucc, ei, epath->src->succs) > - { > - esucc->count = EDGE_FREQUENCY (esucc); > - } > + esucc->count = EDGE_FREQUENCY (esucc); > epath->src->count = epath->src->frequency; > } > } > > > -- > Teresa Johnson | Software Engineer | tejohnson@google.com | 408-460-2413
On Wed, Oct 1, 2014 at 10:03 AM, Xinliang David Li <davidxl@google.com> wrote: > Probably need to file a bug to track the copyrename2 problem. I have filed https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63443 to track this. > > David > > On Wed, Oct 1, 2014 at 9:59 AM, Teresa Johnson <tejohnson@google.com> wrote: >> This patch removes some asserts my jump threading patch r215739 added. >> An upstream pass (copyrename2) is introducing some bogus profile >> counts, so we can't assert that counts are 0 when there is no profile >> data for the function. >> >> Tested on testcase attached to PR63422. Currently running gcc >> regression tests for x86_64-unknown-linux-gnu. Ok for trunk if that >> passes? Regression tests pass, ok for trunk? Thanks, Teresa --- Teresa Johnson | Software Engineer | tejohnson@google.com | 408-460-2413
On 10/01/14 10:59, Teresa Johnson wrote: > This patch removes some asserts my jump threading patch r215739 added. > An upstream pass (copyrename2) is introducing some bogus profile > counts, so we can't assert that counts are 0 when there is no profile > data for the function. > > Tested on testcase attached to PR63422. Currently running gcc > regression tests for x86_64-unknown-linux-gnu. Ok for trunk if that > passes? > > Thanks, > Teresa > > 2014-10-01 Teresa Johnson <tejohnson@google.com> > > PR middle-end/63422 > * tree-ssa-threadupdate.c (freqs_to_counts_path): Remove > asserts to handle incoming insanities. OK. Jeff
On 10/01/14 11:03, Xinliang David Li wrote:
> Probably need to file a bug to track the copyrename2 problem.
Yea, that would definitely be appreciated.
Jeff
Index: tree-ssa-threadupdate.c =================================================================== --- tree-ssa-threadupdate.c (revision 215739) +++ tree-ssa-threadupdate.c (working copy) @@ -977,20 +977,14 @@ freqs_to_counts_path (struct redirection_data *rd) edge ein; edge_iterator ei; FOR_EACH_EDGE (ein, ei, e->dest->preds) - { - gcc_assert (!ein->count); - ein->count = EDGE_FREQUENCY (ein); - } + ein->count = EDGE_FREQUENCY (ein); for (unsigned int i = 1; i < path->length (); i++) { edge epath = (*path)[i]->e; - gcc_assert (!epath->count); edge esucc; FOR_EACH_EDGE (esucc, ei, epath->src->succs) - { - esucc->count = EDGE_FREQUENCY (esucc); - } + esucc->count = EDGE_FREQUENCY (esucc); epath->src->count = epath->src->frequency; } }
This patch removes some asserts my jump threading patch r215739 added. An upstream pass (copyrename2) is introducing some bogus profile counts, so we can't assert that counts are 0 when there is no profile data for the function. Tested on testcase attached to PR63422. Currently running gcc regression tests for x86_64-unknown-linux-gnu. Ok for trunk if that passes? Thanks, Teresa 2014-10-01 Teresa Johnson <tejohnson@google.com> PR middle-end/63422 * tree-ssa-threadupdate.c (freqs_to_counts_path): Remove asserts to handle incoming insanities.