|
|
On Thu, 22 Nov 2018 at 01:32, Todd Blanchard via Pharo-users < [hidden email]> wrote: OK, sorry. I missed that.
Now I've read it. And it requires a customized VM that can count bytecodes executed against a budget and terminate execution if the budget is exceeded.
For the developer it requires an awareness of the relationship between bytecodes and source code. I, personally, have virtually no idea what that relationship is. I used Smalltalk as a high level language and try not to worry about how expensive things are until it actually matters. I think that is most developers.
So I'm not really a fan of this contest approach. The implementation is certainly beyond my skills. I have been dabbling with Squeak since its release and only recently trying to get into Pharo "in anger" because I have a couple projects I personally want to realize. I'm finding the entire thing to be really fast moving and hard to get my head around and the VM is pretty much still impenetrable to me after a couple months trying to build one.
It was a real nice feeling when I first managed to build the VM.
Buzz me on Discord next week and I can help you out with this.
So I'm definitely not your guy and the number of people who are on top of the VM enough to pull this off is likely on the order of ten.
If its okay to execute entries on a StackInterpreter VM, then I've dabbled enough that I may know "just enough" to achieve this. The added complexity of the Cog JIT would put it beyond my skills since the interpreter loop is cycled for every bytecode. In concept, add an instance variable to Process and increment that at the start of every interpreter loop. Can someone comment on the validity of that concept?
cheers -ben
|
Locked
