This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: genattrtab speedup 1/4


On Sun, 2005-07-31 at 23:44 +0200, Michael Matz wrote:
> Hi,
> 
> I seem to be the master of bad timing, posting this long standing thing 
> again during stage 3 :-|  But IMHO it could still be accepted as the scope 
> is relatively limited (and there are some real bugs fixed too).
Sigh.  Definitely the master of bad timing.  Mark (of course) has the
final call, but I'm only going to look at the bugfixes and removal
of the dead code you identified.


[ ... ]

> This first patch is merely a cleanup, which is fairly obvious.  
> 'current_alternative_string' is not set anywhere, hence all code depending 
> on it is dead.  Second there is a small type in simplify_test_exp in the 
> AND case, where it tests 'left' twice for falseness and misses testing 
> 'right'.  Fixing this obvious type btw. makes the resulting insn-attrtab.c 
> even larger and slower on x86-64 :-)
> 
> Okay?
> 
> 
> Ciao,
> Michael.
> -- 
> 	* genattrtab.c (current_alternative_string): Remove.
> 	(SIMPLIFY_ALTERNATIVE): Ditto.
> 	(attr_alt_bit_p): Ditto.
> 	(alternative_name): Make const char *.
> 	(evaluate_eq_attr): Remove use of above things.
> 	(simplify_test_exp): Ditto.
> 	(simplify_test_exp <EQ_ATTR>): Guard for insn_code < 0 .
> 	(simplify_test_exp <AND>): Correct typo (test 'right' not 'left').
This is OK.

jeff



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]