Date: | Fri, 30 Jan 2009 07:55:54 -0600 |
---|---|
From: | Dan Gibson <degibson@xxxxxxxx> |
Subject: | Re: [Gems-users] GEMS performance and hardware configuration |
Thomas, The short answer is: We all pull our hair out waiting for simulations to finish. Others have reported relative slowdowns in the neighborhood of 30,000x. The longer answer is: * We do everything possible to start Ruby at the last possible moment. I.e. run through thread startup, warmups, etc., in 'Simics-only', then load Ruby once meaningful computation actually begins. * We run a lot of simulations concurrently, aiming for throughput * We don't try to do I/O operations under timing mode (at least not I/O with realistic timing) Regards, Dan
On Fri, Jan 30, 2009 at 5:27 AM, Thomas Canhao Xu <canxu@xxxxxx> wrote: Hi all! -- http://www.cs.wisc.edu/~gibson [esc]:wq! |
[← Prev in Thread] | Current Thread | [Next in Thread→] |
---|---|---|
|
Previous by Date: | [Gems-users] GEMS performance and hardware configuration, Thomas Canhao Xu |
---|---|
Next by Date: | , (nil) |
Previous by Thread: | [Gems-users] GEMS performance and hardware configuration, Thomas Canhao Xu |
Next by Thread: | [Gems-users] GEMS TLB Operation, Abhishek Bhattacharjee |
Indexes: | [Date] [Thread] |