On Sun, 20 Jun 2010, Eliot Miranda wrote:
> Hi All, > > > it gives me great pleasure to announce that the Teleplace Cog VMs are now > available. Huge thanks to all at Teleplace who have given me the This is fantastic, thanks. <snip> > attempt to port it, e.g. to ARM, PowerPC and x86-64. The Cogit (excuse the > pun) is so named because it is both an interpreter and a JIT, choosing not > to generate machine code for large methods, interpreting them instead, the > default policy being not to JIT methods with more than 60 literals. I think it would be better to use an adaptive policy later. Something like: all methods will be inlined if they are evaluated enough times. Say a method with 10 literals will be inlined if it's evaluated 100 times. A method with 100 literals will need 10000 runs, etc. (note: i just made up the numbers). Levente _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
On Jun 20, 2010, at 11:47 45PM, Eliot Miranda wrote: > > I expect the next few weeks to involve some bug fixing as it is used in a much broader context (no pun intended). Y'all can help by giving me reproducible test cases. I made the Float/LargeInt and special object array changes, not sure if it 's just my derived images missing more or an actual bug, but... [|b| 5] value, and [: a | |b| 5 ] value: 1 crashes in both Squeak and Pharo. Compiler evaluate: '[ :a | | t | ] value: 1' crashes the VM in Pharo 1.1, but not in Squeak. It returns 1 rather than nil though, as is expected in Pharo at least. Cheers, Henry _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
Hi Henry,
On Mon, Jun 21, 2010 at 10:22 AM, Henrik Johansen <[hidden email]> wrote:
Crashes my VM too :) Thanks!! Compiler evaluate: '[ :a | | t | ] value: 1' crashes the VM in Pharo 1.1, but not in Squeak. This doesn't crash for me. So this is likely a closure compiler issue. Someone might want to compare the closure compiler in Squeak trunk against Pharo and move the updates across. I'll happily review.
It returns 1 rather than nil though, as is expected in Pharo at least.
_______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project |
In reply to this post by Levente Uzonyi-2
Hi Eliot,
Eliot Miranda wrote: > > Hi All, > > > it gives me great pleasure to announce that the Teleplace Cog VMs are > now available. Huge thanks to all at Teleplace who have given me the > opportunity to build Cog and release it as open source, been willing > guinea pigs braving its bugs, and providing indispensable > participation in getting Cog to its current state. Huge thanks are > also due to the original Back To The Future team whose VMMaker Cog > extends to write the VM, and to Peter Deutsch from whom I've taken > many ideas. > This is wonderful news! Thank you, Andreas, and all the people at Teleplace who made this possible! I tried to run Cuis on the Windows VM Andreas did, and I'm having trouble. I did as follows: 1) Start Cuis 2.4 with the Closure VM 2) Install the attached changeset (made after your post and the tweaks 3) Exit saving the image 4) Start it with the Cog VM When starting, it does an invalid memory access, and terminates without any error message or log. I guess I'd need to setup a VM building and debugging environment, etc, and learn quite a bit to fix this myself. I guess it would take you minutes to spot this, compared with the days I'd need to spend... I'd very much appreciate your help. Thank you. Cheers, Juan Vuletich _______________________________________________ Pharo-project mailing list [hidden email] http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project CogCuis.2.zip (4K) Download Attachment |
Free forum by Nabble | Edit this page |