Dolphin, the GameCube and Wii emulator - Forums

(03-06-2016, 04:28 AM)retroben Wrote: [ -> ]Just informing this even though I can always use 3689 since it runs a bit faster from the optimizations.

That's not true. Master now has more optimizations, including the ones in PR 3689. Either just use master builds, or follow all new PRs. When they're merged through, then master should have the same effect.

Thanks for telling me,I could have sworn Brawl was still a tiny bit slower,but that might have been shader generating again when in the challenges menu,other than that,it no longer slows down in there.
I guess I could always stick with a slightly earlier master build if I needed cheats to work for whatever specific reason.

(03-06-2016, 05:18 AM)retroben Wrote: [ -> ]Cheats work on 9061 but hold on and wait for the edit.

Working: 9061,9079,9081,9083,9085,9087

Unsure: 9100 (arm64 build warning,and afraid to try it)

No more waiting.

Edit: Only build 9102 it seems.

To be honest, I have no idea how 4.0-9100 may have broken cheats. But as there were lots of arm changes, it's not that unlikely. May you please double check that 4.0-9087 did work fine, and 4.0-9100 is broken?

If you want to help me debugging, you can check which of those 12 commits in there is the broken one. Hint: git bisect start; git checkout 1b65adc; git bisect good; git checkout 9bf50fb; git bisect bad. Now repeat: build, test, git bisect good or git bisect bad. After 4 iterations, it will show you a commit.

Edit: There is no need to skip 4.0-9100 is your first run. The "warning" is not a warning, it's the title of the change. As all changes are build on the last one, 9102 also includes this changes.

I found the text info,and the warnings are only for VC games it seems,so I tried 9100 and the codes don't work.

I am positively sure the codes work on 9087 since I could mash buttons to repeatedly cancel into other attacks and fly with up specials using Bowser.

The issue could be in 9088-9099 for all we know,just bad luck to split the builds right at that point instead of being every other build if the issue is in one of those instead of just 9100.

As many people can't do for various reasons,I can't compile,and I bet the requirements are too steep for a weak single-core PC to compile it being unable to handle dependencies,and even then,it would take ages to compile on it,and my limited access to it.
Sorry that I can't provide that high of a level for testing.

I'm on a dual core, and it takes about 5 minutes to compile here. Please be honest and say that you just don't want to. That's fine, but me neither. So this bug will stay as it is until it is fixed by accident (very unlikely, maybe in years) or someone else appear and want to fix this bug

To be completely honest,I REALLY don't know how to do anything with compiling,and I was and still am uneasy about how large some dependencies can be,some are 1GB and larger for downloading which would take ages if the server is not fast.
To you it may sound stupid,but I was always too scared in trying to compile anything because of the many posts of people getting errors when trying to compile stuff,meaning I would wait for the compile only to get a whole bunch of errors and never getting a successful result.

I am sure either someone who loves cheats for extra fun even more than I do,or someone who wants them for use with 60fps/widescreen hacks might try to figure something out.

I belive that you have no clue how to do it. But it doesn't sound like you want to learn it. For me, the later matters more.

It's really actually not that hard, and if you are on a fast connection, 1GB is nothing. But with Windows and a weak CPU, that may indeed not be the most enjoyable thing…

Here are the relevant commits:

9bf50fb - JitArm64: Fix mulli. (3 days ago) <degasus>
9ed465f - JitArm64: Implement mulhwx (3 days ago) <degasus>
889a0d3 - JitArm64: Align farcode entry points to 16 bytes (3 days ago) <degasus>
eaa0e27 - JitArm64: Use immediate WriteExceptionExit more often. (3 days ago) <degasus>
689205f - JitArm64: small cleanup, no changes (3 days ago) <degasus>
7b017c6 - JitArm64: Use a register as PC argument for the dispatcher. (3 days ago) <degasus>
050932e - JitArm64: Rename global registers. (3 days ago) <degasus>
57d76ce - JitArm64: Merge WriteExceptionExit. (3 days ago) <degasus>
55b9ce9 - JitArm64: Store memory pointer once, not per block. (3 days ago) <degasus>
56c8f65 - JitArm64: Precheck for exceptions before calling PowerPC::CheckExceptions. (3 days ago) <degasus>
aea48c4 - JitArm64: Drop unused WriteExceptionExit. (3 days ago) <degasus>
c5b0dc9 - JitArm64: Fix LWZ idle skipping. (3 days ago) <degasus>

55b9ce9 is 4.0-9087, and 4.0-9095 is 9bf50fb, and since you said 4.0-9087 works, I'm going to skip that and only build the commits in between those two (can't bisect as I don't have an ARMv8 device and the games to try it.) Will update this post with the links when I have built them.

edit: here is the link to the APKs.

(03-08-2016, 06:52 AM)leolam Wrote: [ -> ]It's really actually not that hard, and if you are on a fast connection, 1GB is nothing. But with Windows and a weak CPU, that may indeed not be the most enjoyable thing…

Here are the relevant commits:

9bf50fb - JitArm64: Fix mulli. (3 days ago) <degasus>
9ed465f - JitArm64: Implement mulhwx (3 days ago) <degasus>
889a0d3 - JitArm64: Align farcode entry points to 16 bytes (3 days ago) <degasus>
eaa0e27 - JitArm64: Use immediate WriteExceptionExit more often. (3 days ago) <degasus>
689205f - JitArm64: small cleanup, no changes (3 days ago) <degasus>
7b017c6 - JitArm64: Use a register as PC argument for the dispatcher. (3 days ago) <degasus>
050932e - JitArm64: Rename global registers. (3 days ago) <degasus>
57d76ce - JitArm64: Merge WriteExceptionExit. (3 days ago) <degasus>
55b9ce9 - JitArm64: Store memory pointer once, not per block. (3 days ago) <degasus>
56c8f65 - JitArm64: Precheck for exceptions before calling PowerPC::CheckExceptions. (3 days ago) <degasus>
aea48c4 - JitArm64: Drop unused WriteExceptionExit. (3 days ago) <degasus>
c5b0dc9 - JitArm64: Fix LWZ idle skipping. (3 days ago) <degasus>

55b9ce9 is 4.0-9087, and 4.0-9095 is 9bf50fb, and since you said 4.0-9087 works, I'm going to skip that and only build the commits in between those two (can't bisect as I don't have an ARMv8 device and the games to try it.) Will update this post with the links when I have built them.

edit: here is the link to the APKs.

I can help you but what problem I have to look?

(03-06-2016, 06:31 PM)degasus Wrote: [ -> ]To be honest, I have no idea how 4.0-9100 may have broken cheats. But as there were lots of arm changes, it's not that unlikely. May you please double check that 4.0-9087 did work fine, and 4.0-9100 is broken?

If you want to help me debugging, you can check which of those 12 commits in there is the broken one. Hint: git bisect start; git checkout 1b65adc; git bisect good; git checkout 9bf50fb; git bisect bad. Now repeat: build, test, git bisect good or git bisect bad. After 4 iterations, it will show you a commit.

I actually don't need help I just read that cheats may have been broken with the recent JIT changes, so I built the relevant commits to try to help debugging this issue by letting retroben figure out which change broke it more easily.

Of course, if you want to try and see if cheats work, you're more than welcome doing that.