llvm dev - Feb 2012 - [LLVMdev] SelectionDAG scalarizes vector operations.

On Feb 8, 2012, at 9:01 AM, David A. Greene wrote:
> "Rotem, Nadav" <nadav.rotem at intel.com> writes:
> 
>> Hi David!
>> 
>> I'd be interested in hearing about the places that you had to fix. 
It
>> seems like there is a number of people who are starting to look at the
>> quality of the generated vector code.  Maybe we should report our
>> findings in bug reports, so that we could share the work and discuss
>> possible findings.  I also plan to fill a few bug reports with
>> suboptimal code.
> 
> That's a good idea.  I confess that I don't have a laundry list
at-hand.
> These were fixes I made two years ago.  I'll have to do a giant diff
> against the upstream LLVM.  :)
> 
> But yes, I fully intend to send these fixes up.
Would it make sense to start something like a lib/Target/README-Vectors.txt file
with a collection of small examples of pessimizations?  We've successfully
used text files like this to track "laundry lists" like this.  High
priority problems (e.g. the sorts that Duncan is pointing out) makes sense to
track in bugzilla, of course.

-Chris

Chris Lattner <clattner at apple.com> writes:
> Would it make sense to start something like a
> lib/Target/README-Vectors.txt file with a collection of small examples
> of pessimizations?  We've successfully used text files like this to
> track "laundry lists" like this.  High priority problems (e.g.
the
> sorts that Duncan is pointing out) makes sense to track in bugzilla,
> of course.
Sounds fine to me, but is there any reason *not* to file bugs other than
increasing the number of bugs in Bugzilla?  Bugzilla is always where I
go first to see if a problem has already been identified.

                          -Dave

On Feb 8, 2012, at 12:42 PM, David A. Greene wrote:
> Chris Lattner <clattner at apple.com> writes:
> 
>> Would it make sense to start something like a
>> lib/Target/README-Vectors.txt file with a collection of small examples
>> of pessimizations?  We've successfully used text files like this to
>> track "laundry lists" like this.  High priority problems
(e.g. the
>> sorts that Duncan is pointing out) makes sense to track in bugzilla,
>> of course.
> 
> Sounds fine to me, but is there any reason *not* to file bugs other than
> increasing the number of bugs in Bugzilla?  Bugzilla is always where I
> go first to see if a problem has already been identified.
Personally, I think that bugzilla is overkill (i.e. the wrong answer) for a long
tail of random micro-optimizations.  Much of the stuff in X86/README has been in
there for years.  Having it in bugzilla would just clutter things up.

-Chris