Hi folks, primarily ARM folks. I have a series of CLs ( https://gem5-review.googlesource.com/c/public/gem5/+/38721/2) which implement a series of optimizations for the HDLCD which make a significant (20%) performance improvement booting linux when using a Fast Model CPU.
With Fast Model, and probably also with KVM, the HDLCD takes up a much bigger portion of execution time, and optimizations there make a much bigger difference there than they would under, say, the atomic CPU. My best estimate is that there is still maybe 2-3% improvement in total run time possible if fetching a frame and rendering it was infinitely fast (measured by commenting out the code that does that work), so I think with these changes we get to the point of diminishing returns. Gabe
_______________________________________________ gem5-dev mailing list -- [email protected] To unsubscribe send an email to [email protected] %(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
