[PATCH] Global merge on constants

Bill Wendling wendling at apple.com
Fri Mar 15 10:49:04 PDT 2013


On Mar 15, 2013, at 9:51 AM, Quentin Colombet <qcolombet at apple.com> wrote:

> Hi Bill,
> 
> I’ve attached the measurement made on compile time for O3.
> In that file, the Reference column is llvm r177120 without the proposed patch and Test is  the same compiler with the proposed patch and constant merging enabled.
> Both compilers have been generated without assertions and with optimizations (i.e., release mode).
> 
> The speedup column reports a ratio: Test/Reference, thus the bigger the better.
> 
> On average, the impact is limited (0.99 speedup, thus 1% slow-down). The maximum and minimum reported numbers are, in my opinion, not much relevant as the related tests (SingleSource/UnitTests/Vector/build and MultiSource/Benchmarks/MiBench/network-dijkstra/network-dijkstr) are compiled very quickly and the measurement method is not that accurate.
> Moreover, we have to keep in mind that this patch adds a check for a currently bogus behavior with global variables maked as “used”.
> 
> Feel free to comment!
> 
Hrm. I was hoping that the results would be no change. The slowdowns are small, but then so are the test cases. Would you be able to run some of the SPEC tests to see if it shows up there?

Sorry to ask you to do this. I know it's a pain. You may want to talk with Nadav to see what he thinks. If he isn't concerned, then I will back off and you can commit the new patch. :)

-bw





More information about the llvm-commits mailing list