[llvm] [MachineCopyPropagation, Scheduler] Detect and fix suboptimal instruction order to enable optimizations (PR #98087)

Michael Maitland via llvm-commits llvm-commits at lists.llvm.org
Mon Aug 5 07:34:37 PDT 2024


michaelmaitland wrote:

> So what do you think of an implementation in MCP using the existing ScheduleDAG?

I think this makes sense. Once you have an implementation, I am happy to take a look into compile time vs performance impact on spec using SiFive hardware.

> The impact needs to be negligible unless we show significant performance boost across the board (which I doubt we'll achieve)

@qcolombet is your main concern the compile time impact or that there will be a lack of performance impact?

https://github.com/llvm/llvm-project/pull/98087


More information about the llvm-commits mailing list