|
From: | Joel Sherrill |
Subject: | Re: [Simulavr-devel] Uniform the code |
Date: | Sun, 12 Apr 2009 20:50:08 -0500 |
User-agent: | Thunderbird 2.0.0.21 (X11/20090320) |
Weddington, Eric wrote:
-----Original Message-----From: address@hidden [mailto:address@hidden Sent: Sunday, April 12, 2009 10:39 AM To: address@hidden; '' simulavr-devel @ nongnu . org ''; 'Michael N . Moran'; address@hidden; address@hidden; Weddington, EricSubject: RE: [Simulavr-devel] Uniform the code A major increase in the number of items to allocate would be a major increase in the allocation time.My recollection is that simulated RAM is allocated a byteat a time.can't the RAM be allocated as a single object? Would that make it faster?I don't know if it is correct or not. If it is, then whyIs that correct? If so, can it be changed?I will wade back through all this tomorrow but it would be interesting to baseline the time required to run the tests now on avrtest or simulavr and go ahead and try to run them on simulavrxx. It should have all the key features for automated testing now and the issue will be tests that pass on one of the other simulator and fail on simulavrxx. Eric are you the one who runs it?I used to run it on a daily basis. Now it's on an "occasional" basis. Mike Stein has set up some automatic testing on the GCC Compile Farm which tests the avr on a daily basis.
I have emailed him a few times. I don't know which targets he knows about in detail but he runs a lot of them.
Once the context switch code is working on rtems/avr, I should be able to turn on avr-rtems testing as well. We have a couple of quad-core machines we use for gcc testing.After I wade through some higher priority stuff, I'll see about getting a baseline to work from, with avrtest, and simulavrxx HEAD.
--joel
[Prev in Thread] | Current Thread | [Next in Thread] |