[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: threaded code in bytecode.c

From: Jaeyoun Chung
Subject: Re: threaded code in bytecode.c
Date: Mon, 03 May 2004 13:15:38 +0900
User-agent: Gnus/5.110002 (No Gnus v0.2) Emacs/21.3.50 (gnu/linux)

* Miles Bader <address@hidden> at 2004-05-02 21:35-0400
| On Mon, May 03, 2004 at 10:10:11AM +0900, address@hidden wrote:
>> I've just added gcc's computed goto based bytecode interpretation in
>> bytecode.c without affecting any other parts. 
| Does using this gcc extension make any noticeable speed difference?

Well, I didn't checked it out but I'm sure it will. at least the dispatching
overhead has been half-ed. After all, reducing the bytecode interpretation
overhead worth trying.

I really wanted to specify some register dedication for 'op' and `top' and a
few other fields but didn't apply them yet since I don't know how to make it
portable except for sparc and x86 machines.

| BTW, your patch contains a lot of unnecessary reformatting &c of the
| byte-code #defines (white space changes, replacing `foo + N' with `fooN'),
| which makes the patch harder to read and will likely cause merge conflicts.

well, i think realigning the number would be better for the readability.
I'll not change the spaces so that spurious diffs does not bother developers.

And converting 'foo + N' to fooN is a necessity to give it a label in case
THREADED_CODE is defined. And also, I think it is more explicit than using the
implicit '+ N' rule for those operator.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]