GitLab

Line 61 in pdep_kernel.cpp to set the stride in StreamExpandKernel was previously commented out.

        //setStride(4 * b->getBitBlockWidth());

However, on cs-osl-11 this leads to premature zeroing of output for some reason.

This was observed with the following command which should produce colorized output on all lines:

bin/icgrep '^([0-9]+[.]){3}' ../QA/testfiles/ips

Reinstating the line explicitly setting the stride has resolved the issue for this case, but this may just be hiding a bug.

This is on the current master branch which has merged recent changes/fixes from memory-aware-scheduling.