Dan Sugalski [EMAIL PROTECTED] writes:
At 11:39 AM 11/12/2001 -0500, Ken Fox wrote:
Simon Cozens wrote:
You save one level of indirection, at a large complexity
cost.
A lot less complexity than a JIT though. 100% portable
code too.
It's got the same sort of issue that a lot of other
I few days ago I suggested inlining some PMC methods
would be good for performance. It turns out that this
question has been heavily studied by the OO community
for at least 10 years. A common solution is to
dynamically replace a method call with the body of the
method wrapped in an if statement.
On Mon, Nov 12, 2001 at 10:15:19AM -0500, Ken Fox wrote:
Where this fits into Parrot's interpreter is that
languages could pre-generate ops corresponding to
dynamically generated inlined caches. All we need is a
way to replace the simple method call op with the
inlined one.
You save one
Simon Cozens wrote:
You save one level of indirection, at a large complexity
cost.
A lot less complexity than a JIT though. 100% portable
code too.
It's also something that can be bolted on later, so there's
no reason to reject it now. I'm just throwing it out to the
list because I know other
At 11:39 AM 11/12/2001 -0500, Ken Fox wrote:
Simon Cozens wrote:
You save one level of indirection, at a large complexity
cost.
A lot less complexity than a JIT though. 100% portable
code too.
It's got the same sort of issue that a lot of other inlining's got, but...
In those cases where
On Mon, Nov 12, 2001 at 11:39:34AM -0500, Ken Fox wrote:
It's also something that can be bolted on later, so there's
no reason to reject it now.
I'm not *reject*ing it now. I'm rejecting it *now*. :)
--
The trouble with computers is that they do what you tell them, not what
you want.