Weddington, Eric wrote: > Hi All, Hi, Eric
> There's a lot of new work being done at the SimulAVR project: > <http://savannah.nongnu.org/projects/simulavr> > > Specifically the "old" simulavr project hasn't had much maintenance > done on it in years. There is an "avrtest" project (hosted at WinAVR) > that is used exclusively as a simulator for the GCC Regression Test > Suite. The newer, C++ based simulavrxx is being worked on heavily, with > the intent that it will supersede both the old simulavr and avrtest and > include as much functionality from both as possible. > > We would like to turn the simulavrxx branch into a best-of-class AVR > simulator that will work for all of the use cases, on all platforms that > the AVR toolchain is used on, so we don't have various project forks > that don't work effectively for all. At some point we would like for the > old simulavr and avrtest to go away and to be able to use the C++ based > simulavr for everything. I understand the reasons to want this, but I believe the initial premises for the avrtest development still apply: it is very hard to do a full-featured simulator (with complete peripheral emulation, gdb-server, etc.) that runs gcc regression tests as fast as possible. IIRC, in the latest benchmarks I did with avrtest, it performed as fast as a 100Mhz AVR on a 2.8GHz pentium 4. The problem with the regression tests, is that when a test fails and is stuck in a infinite loop (or similar) it takes a long time to execute enough instructions to be confident that the test is actually stuck and not just taking a long time to complete. Most tests ran in fractions of a second, but when this happened it could take a few minutes to fail the test, and this represented a big slice of the total time to run the regression suite. I think that the effort of making simulavrxx configurable enough to work as a full-featured simulator or a "regression test fast executer" is actually larger than the maintenance required to keep avrtest working with new architectures. Although I do realize that my opinion might be a little biased here ;) -- Paulo Marques Software Development Department - Grupo PIE, S.A. Phone: +351 252 290600, Fax: +351 252 290601 Web: www.grupopie.com "Who is general Failure and why is he reading my disk?" _______________________________________________ AVR-GCC-list mailing list AVR-GCC-list@nongnu.org http://lists.nongnu.org/mailman/listinfo/avr-gcc-list
