[Bug tree-optimization/89176] Vectorizer fails to consider narrower vector width for res[i] = v1[i] < v2[i] ? v2[i] : v1[i]
rguenth at gcc dot gnu.org
gcc-bugzilla@gcc.gnu.org
Mon Feb 4 09:51:00 GMT 2019
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89176
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2019-02-04
Blocks| |53947
Ever confirmed|0 |1
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Oddly enough with --param vect-epilogues-nomask=1 we vectorize the epilogue but
immediately throw it away afterwards.
The reason it works with multiplication is that basic-block vectorization
works there but in the condition case is presented with non-ifconverted
form in multiple BBs which it doesn't handle.
I think this one is then a duplicate of a bug that mentions that phiopt
doesn't use .FMIN/MAX internal functions if available.
The epilogue vectorization issue also needs investigation.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53947
[Bug 53947] [meta-bug] vectorizer missed-optimizations
More information about the Gcc-bugs
mailing list