pic c

Didi wrote:

Walter,

thanks for taking the time to do it. I am impressed by the fact that a compiler goes to such optimization depth, I did not expect it. This does not make me more likely to give HLLs yet another chance as my main tool, but I can sense you have done some good work allright.

Dimiter

Please do not use html in newsgroups.

Please do not use SIG's of more than 4 lines..... We can all nit pick but it is not productive.

Actually, in this case it is quite relevant - many newsservers, including mine, cut out html posts. So I can't see Walter's recent html post (though I see his other plain text posts - this must have been a one-off mistake).

I have to say - that's really impressive! Are SKP1 and SKP2 discovered by the peephole optimiser?

cheers, Rich.

Html to get fixed spacing in the code I included in the message

w..

#pragma memory ROM [2000] @ 0x2000 registera AC;

void main (void) { ENTRY1: AC = 1; // prepare value in B-acc. goto RealEntry; ENTRY2: AC = 2; // prepare another value in b goto RealEntry; ENTRY3: AC = 0; RealEntry: NOP(); }

======== Listing File =========================

void main (void) {

ENTRY1:

No they are part of a comprehensive branch jump call optimizer. We do full application optimization in our compilers / linkers. There is actually a translation error from Didi's code that I missed until after I posted the code fragment and listing yesterday.

The first "goto RealEntry" should have been goto l2; to chain the goto's. I looked at the generated code was happy with it and posted it. The optimizer chained the skips. The following is the corrected source which generates the same output.

#pragma memory ROM [200] @ 0x200; registera AC;

void main (void) {

ENTRY1: AC = 1; // prepare value in B-acc. goto l2; ENTRY2: AC = 2; // prepare another value in b l2: goto RealEntry; ENTRY3: AC = 0; RealEntry: NOP(); }

Regards,

Walter Banks

-- Byte Craft Limited Tel. (519) 888-6911

Thanks. For good HLLs think data base of programmers tricks

w..