Mahesh Madhav

If the work is repetitive (the same code is being executed over and over with same/similar data), then it is not interesting. If a script exercises different parts of the...

Thank you for a much better description. Yes, different types of engines and crafts, and timed events.

In SPEC CPUv8 we started converting std::partition to std::stable_partition, and also investigating converting std::unordered_map to std::map and std::unordered_set to std::set. Even after all that there is still variation between different...

Sure thing. Since we are running within the SPEC CPU harness, all files are built with the same compiler switches (that is one of the requirements for the benchmark). We...

No, we replaced our flags with your flags. I confirmed that the gains are not coming from LTO by removing -flto and rerunning; the asm produced is isolated to one...

That's a cool graph! Can you explain the axes? The title says "CPU-time vs compression", so I thought that means "Y-axis vs X-axis", but the axes are labeled differently (Compression...

Hi @ktmf01! I took your changes from !700 and !702, applied them on top of the changes here, and observed some more performance on my dev machine (Ampere Altra). I...

I want to share that the SPEC CPU development committee has been testing with this edit for a month, without any issues across a variety of hardware and compilers (AMD...