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.

p4raw-id: //depot/perl@33355

diff --git a/intrpvar.h b/intrpvar.h
index 06014ef..4a62779 100644 (file)
--- a/intrpvar.h
+++ b/intrpvar.h
@@ -124,7 +124,7 @@ PERLVAR(Idefstash,  HV *)           /* main symbol table */
 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(Ilocalizing,        U8)             /* are we processing a local() list? */
 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  */