[Fwd: Re: Another task for the interested - t/pmc/pmc.t]
[ shoud have gone to the list probaby ] Original Message Subject: Re: Another task for the interested - t/pmc/pmc.t Date: Sat, 12 Mar 2005 07:50:04 +0100 (CET) From: Steven Schubiger [EMAIL PROTECTED] To: [EMAIL PROTECTED] On 9 Mar, Leopold Toetsch wrote: The test file t/pmc/pmc.t is full of Perl* tests. These should be factored out into t/pmc/perl*.t. The file itself should contain just common PMC functionality tests, like the range or type checks. Thanks, leo I'll have a try at this one (for the moment), if no one else bothers.
RE: A task for the interested
Dan Sugalski:
# One of the things that the configure script needs to do is
# generate the
# opcode dispatch macro to either be a giant switch statement (with a
# fallthrough default to handle cases we don't know about) or
# the function
# table dispatch we have now.
If somebody codes up the alternate dispatch, I can easily modify
Configure.pl, config_h.in and the hints files to handle it. Something
like this, perhaps:
#define USE_${dispatch}_DISPATCH
#ifdef USE_switch_DISPATCH
#define DISPATCHER
switch-based dispatcher code here
#else
#define DISPATCHER
function table-based dispatcher code here
#endif
--Brent Dax
[EMAIL PROTECTED]
They *will* pay for what they've done.
Re: A task for the interested
On Tue, Sep 18, 2001 at 02:32:32PM -0700, Brent Dax wrote: If somebody codes up the alternate dispatch, I can easily modify Configure.pl, config_h.in and the hints files to handle it. Something like this, perhaps: Something like that, but the Right Way would be to rewrite DO_OP in interp_guts.h Simon -- } /* the next line is indented funny to preserve old indentation */ - plan9 has a bad day
Re: A task for the interested
On Tue, Sep 18, 2001 at 03:55:23PM -0400, Dan Sugalski wrote: Anyone care to take a shot at it? Having an extra overridable column in the opcode_table file (so we know which opcodes are overridable, and thus can't be in the switch) would be a good thing while you were at it... I will do this tonight, if nobody else gets to it first. - Damien
RE: A task for the interested
Simon Cozens:
# On Tue, Sep 18, 2001 at 02:32:32PM -0700, Brent Dax wrote:
# If somebody codes up the alternate dispatch, I can easily modify
# Configure.pl, config_h.in and the hints files to handle it.
# Something
# like this, perhaps:
#
# Something like that, but the Right Way would be to rewrite DO_OP in
# interp_guts.h
So, something more like this?
~~
--- ..\..\parrot-cvs\parrot\config_h.in Fri Sep 14 09:40:36 2001
+++ config_h.in Tue Sep 18 14:44:48 2001
@@ -24,7 +24,14 @@
#define MASK_CHUNK_LOW_BITS 0xf000
+#define USE_DISPATCH_${dispatch}
+#ifdef USE_DISPATCH_SWITCH
+# undef DO_OP
+# define DO_OP code here
+#endif
+
+
${headers}
--- ..\..\parrot-cvs\parrot\build_interp_starter.pl Tue Sep 18
10:54:52 2001
+++ build_interp_starter.pl Tue Sep 18 14:48:26 2001
@@ -59,12 +59,13 @@
#
print INTERP EOI;
-
-#define DO_OP(w,x,y,z) do { \\
-x = (void *)z-opcode_funcs; \\
-(void *)y = x[*w]; \\
-w = (y)(w,z); \\
- } while (0);
+#ifndef DO_OP
+# define DO_OP(w,x,y,z) do { \\
+ x = (void *)z-opcode_funcs; \\
+ (void *)y = x[*w]; \\
+ w = (y)(w,z); \\
+} while (0);
+#endif
EOI
# Spit out the OPCODE_FINGERPRINT macro
~~~
(With, of course, the correct code put in config_h.in, and various
changes and additions to the hints/ folder to automagically suggest the
right dispatch to use.)
--Brent Dax
[EMAIL PROTECTED]
They *will* pay for what they've done.
RE: A task for the interested
Simon Cozens:
# On Tue, Sep 18, 2001 at 02:51:35PM -0700, Brent Dax wrote:
# So, something more like this?
#
# Urh, how can I put this? No.
#
# I *really* want to avoid macro hackery - undef'ing this and
# then testing if it's defined and then redefining it, and
# urgh, urgh, urgh. No.
#
# I was thinking more like:
#
# +++ build_interp_starter.pl Tue Sep 18 14:48:26 2001
# @@ -59,12 +59,13 @@
# #
#
# if ($Config{dispatch} eq fpointer) {
# print INTERP EOI;
# -#define DO_OP(w,x,y,z) do { \\
# -x = (void *)z-opcode_funcs; \\
# -(void *)y = x[*w]; \\
# -w = (y)(w,z); \\
# - } while (0);
# EOI
# } elsif ($config{dispatch} eq switch) {
#print INTERP EOI;
# #define DO_OP(w,x,y,z) do { \\
# switch(*w) {...\\
# ...
# EOF
# } elsif ( ...
Okay, that's actually not too hard--just add a 'use Parrot::Config;' to
the top and check something in %PConfig. Alright.
--Brent Dax
[EMAIL PROTECTED]
They *will* pay for what they've done.
Re: A task for the interested
On Tuesday 18 September 2001 03:55 pm, Dan Sugalski wrote: One of the things that the configure script needs to do is generate the opcode dispatch macro to either be a giant switch statement (with a fallthrough default to handle cases we don't know about) or the function table dispatch we have now. FWIW, having the function dispatch table as the default case in a switch wasn't a big win in my initial testing. The savings came from an initial test to decide to switch or lookup, but that requires the non-overrideable to be blocked so that they're easily tested (ie, opcodes 0-n). Parrot's performance does seem to differ from my initial testing, though, so you may want to benchmark it every which way. -- Bryan C. Warnock [EMAIL PROTECTED]
