[fpc-devel] Debugging Loop Unroll Optimization

[Florian Klämpfl]

Am 16.05.2018 um 14:57 schrieb Martok:
> Hi all,
> 
> as we have discovered in 0033576 and 0033614, there is a bug somewhere in the
> loop unroll optimization pass that only appears in complex enough code. The
> problem is, it doesn't happen in the single testsuite test, and the observable
> crashes in Lazarus are caused by memory corruption at some point unrelated to
> the crash, so it's hard to debug (at least on Windows, without rr...).
> 
> I have one other project that has sporadic crashes with -OoLOOPUNROLL (I wish I
> had figured that out back then), but that is about as difficult as Lazarus,
> where it's at least 100% reproducible.
> 
> Does anyone have more complete test cases, or maybe smaller affected projects?
> 

How big is the project? Normally, the number of unrolled loops is reasonable small so comparing the
generated assembler with and without unrolling should be feasible.