Change 33355 by [EMAIL PROTECTED] on 2008/02/22 22:30:05
As best as I (and my minion, gcc -Os) can tell, PL_curcop and
PL_in_eval do not need to be volatile. This improves the generated code
measurably - for example toke.o is 1.5% smaller. Every little helps.
Affected files ...
... //depot/perl/intrpvar.h#234 edit
Differences ...
==== //depot/perl/intrpvar.h#234 (text) ====
Index: perl/intrpvar.h
--- perl/intrpvar.h#233~32959~ 2008-01-11 09:00:59.000000000 -0800
+++ perl/intrpvar.h 2008-02-22 14:30:05.000000000 -0800
@@ -124,7 +124,7 @@
PERLVAR(Icurstash, HV *) /* symbol table for current package */
PERLVAR(Irestartop, OP *) /* propagating an error from croak? */
-PERLVAR(Icurcop, COP * VOL)
+PERLVAR(Icurcop, COP *)
PERLVAR(Icurstack, AV *) /* THE STACK */
PERLVAR(Icurstackinfo, PERL_SI *) /* current stack + context */
PERLVAR(Imainstack, AV *) /* the stack when nothing funny is
@@ -189,7 +189,7 @@
PERLVAR(Icolorset, bool) /* from regcomp.c */
PERLVARI(Idirty, bool, FALSE) /* in the middle of tearing things
down? */
-PERLVAR(Iin_eval, VOL U8) /* trap "fatal" errors? */
+PERLVAR(Iin_eval, U8) /* trap "fatal" errors? */
PERLVAR(Itainted, bool) /* using variables controlled by $< */
/* This value may be set when embedding for full cleanup */
End of Patch.
