cfbolz changed the topic of #pypy to: PyPy, the flexible snake (IRC logs: https://botbot.me/freenode/pypy/ ) | use cffi for calling C | the secret reason for us trying to get PyPy users: to test the JIT well enough that we're somewhat confident about it
forgottenone has quit [Quit: Konversation terminated!]
stduolc has quit [Quit: Connection closed for inactivity]
nimaje has quit [Remote host closed the connection]
MerlinTheWizard has joined #pypy
<MerlinTheWizard>
Hi all.
<MerlinTheWizard>
Can anyone tell me what architectural barriers there might be to making pypy as fast as node.js?
nimaje has joined #pypy
<simpson>
MerlinTheWizard: I suppose that you'd have to first find a way to compare the speed of PyPy and Node meaningfully. Do you have a way of measuring this?
nimaje has quit [Remote host closed the connection]
<simpson>
It'd be nice to not guess. Unfortunately, the owners of the benchmark game are kind of silly about their rules, so it could be hard to make a reasonable comparison.
<simpson>
The problem is that speed is often something that PyPy could improve for a particular application, but not necessarily for everybody.
<MerlinTheWizard>
simpson, I do feel that doing something like running a braindead simple recursive function can be pretty indicative of overall performance.
<simpson>
Okay. We don't, though. PyPy speed bugs are example-driven. There's not much that can be done for a bug that just says 'make this synthetic benchmark faster'
<MerlinTheWizard>
simpson, so it sounds like the pypy optimization approach isn't really comparing with other languages at all?
<simpson>
MerlinTheWizard: I'm saying that for *any* two interpreters which interpret totally different languages, comparing them is ultimately a matter of picking the programs to compare.
<simpson>
I'm also saying that the PyPy team cares more about whether your *actual* Python program is slow.
<MerlinTheWizard>
simpson, yes, I gather that python has always relied on C backend code when serious speed is needed. I was kind of wondering if Python would ever get beyond this paradigm.
<MerlinTheWizard>
It would seem like that would be beneficial, since there is such a huge volume of code written in python.
<cfbolz>
MerlinTheWizard: to make pypy as fast as node, the thing that is needed is money
<cfbolz>
PyPy runs on a shoestring budget, compared to what Google invests into V8
<MerlinTheWizard>
cfbolz, that's what I would imagine. It seems to me that if you've got a compiler under the hood, as pypy does, then assuming there are no architectural issues standing in the way, it should be possible to do the same thing with PyPy that is done with V8.
<cfbolz>
well, the languages are very different
<cfbolz>
their details differ
<MerlinTheWizard>
cfbolz, yes. And I'm sort of new to Python. So I'm wondering if any of those details might impede optimization efforts. From what I know of the language so far, it seems like the answer should be 'no'. But of course I know very little about it right now.
<cfbolz>
MerlinTheWizard: well, javascript has more annoying properties, really