Email: Password: Remember Me | Create Account (Free)

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
Jan Waclawek
02/16/07 02:45
Read: 502 times
Bratislava
Slovakia


 
#133114 - SDCC peephole optimiser conflict
Responding to: Jan Waclawek's previous message
I had a closer look at that last example, as I think this is quite a simple case (djnz).

I found out that this is exactly the case, when one of the strengths of SDCC - the peephole optimizer - turned into its weakness. The trouble is, that the peephole optimisations do interact.

In this case, there is a peephole optimisation (#257.b) which is supposed to turn a dec/jnz pair into djnz; but it is "spoiled" by peephole #160.c, which turns jnz (and subsequent jmp) into jz...

JW


List of 68 messages in thread
TopicAuthorDate
Speed Recommendation      David Lewis      02/14/07 08:00      
   Time for processing      Lynn Reed      02/14/07 08:11      
   Get a 1X Chip      Michael Karas      02/14/07 08:25      
      another 485 issue      Erik Malund      02/14/07 08:36      
         slightly less      Jan Waclawek      02/14/07 08:47      
            who gives a hoot      Erik Malund      02/14/07 09:04      
               maybe yes, maybe not...      Jan Waclawek      02/14/07 09:30      
   Choose the appropriate tool!      Andy Neil      02/14/07 08:26      
      Don't have much of a choice!      David Lewis      02/15/07 00:46      
         drop the "coding standards"      Jan Waclawek      02/15/07 00:53      
         Yikes.      Christoph Franck      02/15/07 01:00      
            down memory lane      Erik Malund      02/15/07 06:10      
               a possible compromise...      Jan Waclawek      02/15/07 06:29      
                  Well...        Christoph Franck      02/15/07 06:54      
               The other end of the spectrum ...      Christoph Franck      02/15/07 07:10      
                  afer quite a few iterations....      Jan Waclawek      02/15/07 07:37      
                     The guide to writing unmaintainable code says:      Christoph Franck      02/15/07 08:28      
         That's the kind of nonsense...      Andy Neil      02/15/07 12:30      
   put some load on the hardware      Jan Waclawek      02/14/07 08:33      
      Jan, oh Jan      Erik Malund      02/14/07 08:43      
      9-bit - more work for the hardware!      Andy Neil      02/14/07 08:43      
         true, but...      Jan Waclawek      02/14/07 08:49      
            It is...      David Lewis      02/15/07 00:55      
               calculate the ISR after making the ISR skip      Erik Malund      02/15/07 06:18      
   Optimizing C        Christoph Franck      02/14/07 08:54      
      just one example      Erik Malund      02/14/07 09:12      
         The advantage is...      Andy Neil      02/14/07 10:01      
            If the thinking had preceded the coding ...      Richard Erlacher      02/14/07 12:45      
               you miss the point      Erik Malund      02/14/07 13:04      
                  then what's the chance of having him to C-craft?      Jan Waclawek      02/14/07 15:26      
                     zero      Erik Malund      02/14/07 15:41      
                     Job security, of course.      Christoph Franck      02/15/07 01:05      
      Examples are not equivalent        Chris Kelly      02/15/07 06:41      
         Wasn\'t saying that they were.      Christoph Franck      02/15/07 07:02      
            I just tried SDCC...      Jan Waclawek      02/15/07 08:23      
               Hm, can you post the assembly output ?      Christoph Franck      02/15/07 08:27      
                  Keil does this      Chris Kelly      02/15/07 16:41      
                     Interesting.      Christoph Franck      02/16/07 02:06      
                     Optimizing and Keil/SDCC/Raisonance      Jon Ward      02/19/07 09:33      
                        Retort      Chris Kelly      02/19/07 18:14      
                           Register Banks...      Jon Ward      02/20/07 07:25      
                              benchmarks      Jan Waclawek      02/20/07 07:40      
                              tools differ      Erik Malund      02/20/07 08:12      
                              Clarification      Chris Kelly      02/20/07 15:52      
                                 comments      Erik Malund      02/21/07 06:04      
                                    not to write things twice      Jan Waclawek      02/21/07 06:25      
                                       no need      Erik Malund      02/21/07 06:42      
                                          And how does this produce two incarnations?      Jan Waclawek      02/21/07 06:59      
                                             like this      Erik Malund      02/21/07 07:06      
                                                hehe      Jan Waclawek      02/21/07 07:30      
                                                   this is no better confusion-wise      Erik Malund      02/21/07 07:48      
                                                      what is "efficient"      Jan Waclawek      02/21/07 08:11      
                                                         ok is \'safe\' better?      Erik Malund      02/21/07 08:36      
                                                            How to convey our ideas to the mcu?      Jan Waclawek      02/21/07 08:49      
                                                               Carefull now, you could end with Prolog :)      Caslav Nedeljkovic      02/21/07 10:33      
                                                                  my kids      Jan Waclawek      02/21/07 15:07      
                                                            Huge code != huge time      Chris Kelly      02/21/07 16:13      
                                                               This comes under the heading of "famous last words      Erik Malund      02/22/07 06:20      
                        the question I always raise on this      Erik Malund      02/20/07 05:56      
                           Here's Why...      Jon Ward      02/20/07 18:25      
                  SDCC      Jan Waclawek      02/16/07 02:09      
                     Thanks. Also interesting.      Christoph Franck      02/16/07 02:23      
                     SDCC peephole optimiser conflict      Jan Waclawek      02/16/07 02:45      
                  Raisonance      Rob Klein      02/16/07 04:34      
   help me ,I NEED SOME HELP ON MY PROJECT      Rahul Bisht      02/19/07 10:23      
      that is NOT 'help'      Erik Malund      02/19/07 10:52      
      Well that is not the way to get it!      Andy Neil      02/19/07 14:16      
      start a new thread!      Jan Waclawek      02/19/07 17:08      

Back to Subject List