Re: [hatari-devel] Hatari data cache tests

[ Thread Index | Date Index | More Archives ]

Hi Eero,

> Can you provide test-case triggering it?

Douglas, are you seeing these still?

TBH I haven't been doing much of that kind of testing recently but it can be re-tested against the original code. Unfortunately no tidy/small test case yet :-/

I plan to use the profiler again soon so I'll see if it recurs. I'm pretty sure it occured a few times recently and I just began to ignore it....
> So, problem is probably with manually written assembly code.
> You'll want to grep out loop addresses from the symbol list.

Did that fix the slowdown?

I very rarely use global symbols at loop addresses, and I do grep out locals - mainly because it skews profiler call counts and makes symbol tracing impossible. But for a certain answer I'd need to look again. Some of the past events here are quite fuzzy now in my head - it decays pretty fast :D

Part of my problem here is being short of time and as soon as I get unblocked I jump onto the next thing - I don't always get to the bottom of such issues. I'll try to capture a bit more context in future...

Mail converted by MHonArc 2.6.19+