https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120805
Avinash Jayakar <avinashd at linux dot ibm.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |avinashd at linux dot ibm.com
--- Comment #4 from Avinash Jayakar <avinashd at linux dot ibm.com> ---
I am fairly new to the gcc repository and I was looking at into this issue.
I had a few doubts on these changes that were done from using log2 to using
division. So the difference we are seeing is mainly due to the if condition
if (stmts != NULL && const_vf > 0)
being always active. Previously this if condition was
if (stmts != NULL && log_vf)
the condition inside would be false, when log_vf was null, this would happen
when LOOP_VINFO_USING_PARTIAL_VECTORS_P (loop_vinfo) was true, which I think is
for the epilogue of the loop which has to use partial vectors.
But with the new change, the const_vf variable will always be 4 since in the
condition
if (vf.is_constant (&const_vf)
&& !LOOP_VINFO_USING_PARTIAL_VECTORS_P (loop_vinfo))
we assign its value to the vector length.
I believe that the previous behaviour was correct, where you do not need to
assign the range info to the epilogue. Adding the range info has different
lowering path within the rtl for powerpc, which I have not yet investigated
yet.
So my main doubt here is const_vf, is supposed to be 0 for the epilogue block
right, just like log_vf was null for the epilogue. If so, this is a simple fix,
by assigning a temporary_const_vf, and assigning the actual value inside the
latter mentioned if block. Please do let me know, in this case I can create a
patch.
