Cyprian Posted May 26, 2022 Share Posted May 26, 2022 interesting. will "or y2,y2" fix that score-boarding problem? Quote Link to comment Share on other sites More sharing options...
ggn Posted May 26, 2022 Share Posted May 26, 2022 This could potentially be something we can catch in rmac while assembling the code and issue a warning. Of course I'm not sure that this won't also give false positives: for example it's not easy to predict how many cycles a "load" can take so it's unclear how many instructions after the load the scoreboarding bug is not applicable. Nevertheless, is this something that people would find beneficial or confusing? Quote Link to comment Share on other sites More sharing options...
42bs Posted May 26, 2022 Share Posted May 26, 2022 1 hour ago, Cyprian said: interesting. will "or y2,y2" fix that score-boarding problem? Yes, but only if placed directly after the load. Quote Link to comment Share on other sites More sharing options...
42bs Posted May 26, 2022 Share Posted May 26, 2022 21 minutes ago, ggn said: This could potentially be something we can catch in rmac while assembling the code and issue a warning. Of course I'm not sure that this won't also give false positives: for example it's not easy to predict how many cycles a "load" can take so it's unclear how many instructions after the load the scoreboarding bug is not applicable. Nevertheless, is this something that people would find beneficial or confusing? The problem is, rmac needs to know if it is a load from within the GPU or from outside. Matthias sent me jwarn which (after a quick look) offers such warnings. Quote Link to comment Share on other sites More sharing options...
Cyprian Posted May 26, 2022 Share Posted May 26, 2022 5 hours ago, 42bs said: Yes, but only if placed directly after the load. does it mean that "or y2,y2" in that case will not prevent score-boarding bug? load (points_ptr),y2 jr nz,.not_last addq #4,points_ptr or y2,y2 movefa xy0.a,y2 ; get back first point Quote Link to comment Share on other sites More sharing options...
42bs Posted May 27, 2022 Share Posted May 27, 2022 I really had to place the OR directly after the LOAD. Else it did not work. Rearranged it now, but it is again a good hint to test regularly on the real hardware. 1 Quote Link to comment Share on other sites More sharing options...
ggn Posted May 27, 2022 Share Posted May 27, 2022 22 hours ago, 42bs said: The problem is, rmac needs to know if it is a load from within the GPU or from outside. Matthias sent me jwarn which (after a quick look) offers such warnings. This? https://github.com/djipi/Jwarn From a quick glance this doesn't seem too bad to implement inside rmac so people won't need a post processor (or pre processor) to scan the file for issues. I've added an issue for this in the rmac bug tracker - it's lucky ticket #200! 2 Quote Link to comment Share on other sites More sharing options...
+CyranoJ Posted May 27, 2022 Share Posted May 27, 2022 33 minutes ago, ggn said: From a quick glance this doesn't seem too bad to implement inside rmac so people won't need a post processor (or pre processor) to scan the file for issues. I've added an issue for this in the rmac bug tracker - it's lucky ticket #200! Please give it an on/off toggle Wouldn't want RMAC to start slipping the way of SMAC! Quote Link to comment Share on other sites More sharing options...
Cyprian Posted May 27, 2022 Share Posted May 27, 2022 1 hour ago, ggn said: This? https://github.com/djipi/Jwarn nice tool, I didn't know about it Quote Link to comment Share on other sites More sharing options...
ggn Posted May 27, 2022 Share Posted May 27, 2022 2 hours ago, CyranoJ said: Please give it an on/off toggle Wouldn't want RMAC to start slipping the way of SMAC! Yeah yeah, don't worry about it. I know the score, 30+ year old tool, 30+ year old codebases. This will probably go under -s or something similar. This will be a while until it's implemented because there's loads of unmerged code lurking in the issue tracker and I just don't want to pile MOAR in there. Also: features get added in order of maintainers' preference And with that: apologies for the diversion, please continue talking about triangles! 1 Quote Link to comment Share on other sites More sharing options...
dilinger Posted May 27, 2022 Share Posted May 27, 2022 4 hours ago, Cyprian said: nice tool, I didn't know about it Due to the rmac one pass assembly, Jwarn may not be very useful because it needs complete opcodes from output listing assembly. A discussion about the subject http://jlhconsulting.gotdns.com/bugs/show_bug.cgi?id=192 Only gasm and vasm output listing can be used with Jwarn at the moment. Quote Link to comment Share on other sites More sharing options...
JagChris Posted May 28, 2022 Share Posted May 28, 2022 SMAC started life as a dual pass assembler. It's a shame that's all been lost now. Or maybe today rmac or SMAC would be able to use things like this. Quote Link to comment Share on other sites More sharing options...
+CyranoJ Posted May 28, 2022 Share Posted May 28, 2022 41 minutes ago, JagChris said: SMAC started life as a dual pass assembler. It's a shame that's all been lost now. Or maybe today rmac or SMAC would be able to use things like this. As a non coder you have no idea how annoying SMAC was altering your code and causing sections to not work because it thought it knew better, wasting literally hours of your time. We're all better off without its crap. 1 1 Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.