I am working on it.
Vladimir
On Oct 20, 2012, at 9:39 AM, Florian Weimer wrote:
> * Remi Forax:
>
>> Even a simple code like the one below, there is no scalar replacement
>> (OSR or not),
>> Float f = new Float(0);
>> for(int i=0; i> f = new Float(f + 1.0f);
>> }
>> floa
* Remi Forax:
> Even a simple code like the one below, there is no scalar replacement
> (OSR or not),
> Float f = new Float(0);
> for(int i=0; if = new Float(f + 1.0f);
> }
> float result = f;
> System.out.println(result);
Last time I looked at this, scalar repl
On Wed, Oct 17, 2012 at 3:48 PM, Remi Forax wrote:
> Graal also does partial EA, the code is available and readable.
I will try to have a look. With JRuby 1.7.0 out soon, I'm going to do
a bit more JVM hacking than in the past :)
- Charlie
___
mlvm-dev
On Wed, Oct 17, 2012 at 7:58 PM, Philip Jenvey wrote:
> PyPy will allocate them on the stack. I'm not sure what you mean by having
> "unoptimized, unemitted operations in the IR". It optimizes the operations
> around them too.
>
> This is a pretty good description of it all if you haven't seen i
On Oct 17, 2012, at 12:59 PM, Charles Oliver Nutter wrote:
> On Wed, Oct 17, 2012 at 2:07 PM, Christian Thalinger
> wrote:
>> On Oct 17, 2012, at 8:33 AM, David Chase wrote:
>>>
>>> I'm very new to this (have not even looked at the source code to Hotspot
>>> yet), but is it possible
>>> to pu
On 10/17/2012 09:07 PM, Christian Thalinger wrote:
> On Oct 17, 2012, at 8:33 AM, David Chase wrote:
>
>> On 2012-10-16, at 8:53 PM, Charles Oliver Nutter wrote:
>>> So *almost* everything is inlining, but one path (I believe it's the
>>> failure path from GWT after talking with Christian) is not
On Wed, Oct 17, 2012 at 2:07 PM, Christian Thalinger
wrote:
> On Oct 17, 2012, at 8:33 AM, David Chase wrote:
>>
>> I'm very new to this (have not even looked at the source code to Hotspot
>> yet), but is it possible
>> to push the allocation/boxing to paths that are believed to be rarely taken?
On Oct 17, 2012, at 8:33 AM, David Chase wrote:
>
> On 2012-10-16, at 8:53 PM, Charles Oliver Nutter wrote:
>>
>> So *almost* everything is inlining, but one path (I believe it's the
>> failure path from GWT after talking with Christian) is not reached.
>> Because Hotspot's EA can't do partia
On 2012-10-16, at 8:53 PM, Charles Oliver Nutter wrote:
>
> So *almost* everything is inlining, but one path (I believe it's the
> failure path from GWT after talking with Christian) is not reached.
> Because Hotspot's EA can't do partial EA, any unfollowed paths that
> would receive the allocat
On Wed, Oct 17, 2012 at 2:54 PM, Charles Oliver Nutter
wrote:
> I will indeed! Just preparing ahead of time for the hype machine to go
> into overdrive. Regardless of initial speed, there's an incredibly
> long tail to any Ruby implementation, and new ones won't be useful
> until months or years a
I will indeed! Just preparing ahead of time for the hype machine to go
into overdrive. Regardless of initial speed, there's an incredibly
long tail to any Ruby implementation, and new ones won't be useful
until months or years after they're first released.
- Charlie
On Wed, Oct 17, 2012 at 3:03 A
On 10/17/2012 02:53 AM, Charles Oliver Nutter wrote:
> Hello all!
>
> I've recently been informed that a new Ruby implementation is about to
> be announced that puts JRuby's numeric perf to shame. Boo hoo.
>
> It's not like I expected us to retain the numeric crown since we're
> still allocating ob
Hi Charlie,
Can you send us a decent link or two once it actually does drop. I'm
not much of a Ruby head generally, but would like to see the numbers
(and, of course, take a quick look at their testing / benching
methodology).
Thanks,
Ben
On Wed, Oct 17, 2012 at 1:53 AM, Charles Oliver Nutter
Hello all!
I've recently been informed that a new Ruby implementation is about to
be announced that puts JRuby's numeric perf to shame. Boo hoo.
It's not like I expected us to retain the numeric crown since we're
still allocating objects for every number in the system, but hopefully
we can get th
14 matches
Mail list logo
