- User Since
- Dec 21 2016, 1:38 PM (190 w, 2 d)
Sat, Aug 8
Fri, Aug 7
Cap the instancing uniform at 64 following Optimus' testing.
Actually do something... Managed to break it yesterday.
gpuskinning is an option in default.cfg that you have to activate using the config file. You could try (de)/activating that
Fixes a remaining issue with GPUs skinning.
Fixed the pgl stuff locally, so enabling it.
Fix GPU skinning. In general don't expect to really benefit from instancing, but I think this might actually instance trireme oars somewhat (at least testing locally), so that might be positive.
Switch back to 32 to get more pessimistic improvements (better for testing imo).
Also fix the slight ARB pessimisation.
Updated version with less terrible code, and also enables using "preferglsl=no" so you can compare with a baseline.
On my intel card, the GLSL version with instancing is always faster, while also being better looking.
Thu, Aug 6
Extreme rebase (aka complete rewrite but the idea is there).
@Freagarach all yours
Implement unordered_map conversion.
I quite like how this looks.
Fixed and umlauted. Committing when lights are green.
This might want to compile.
Still working, overall still looks fine to me. I agree that it's better that way.
- Show hotkey conflicts in the hotkey picker window (nothing on the main window - the data structures aren't good for it and I don't think it's too bad).
- Adds a simple text filter for hotkeys.
- Clean up the "left/right" meta keys situation.
- Include the hotkeys in the session menu.
TODO post SM52
TODO post SM52
Following discussion on IRC -> it seems we should instead delete the oldest corpse, as this looks rather ugly.
Update comment. Ready for commit.
Re-checked my maths, and I had a mistake for the min-range case.
@StefanBruens Let me know if the credits are OK for you or if you want to appear differently. I'll commit then.
Years, programming credits.
This looks fine now :)
Tested on Dodecanese which uses a DiskPlacer with a small radius -> does seem faster. Code seems obviously faster, though nothing is obvious in JS-land.
Thanks for the comments. Overall I don't think they warrant a lint-commit on this soonish, I'll keep it in mind for my next atlas diff I think.
I feel like skirmish maps make a better first impression as they are generally better maps and load much faster, so I would edge towards rejecting this.
Alright, thanks for reporting.
cc @borg- you want this I believe.
I feel like something like this (order random)
gather_metal_0_serfs.json gather_metal_1_servants.json gather_metal_2_shaftmining.json gather_stone_0_silvermining.json gather_stone_1_slaves.json gather_stone_2_wedgemallet.json
might be reasonably efficient at conveying order, but also nicer than hardcoding "city"
Wed, Aug 5
I've kept the settings for DisplaYFrustrum and such since those aren't config options. I'm not sure we want that or not.
I think we could make them SYSTEM-level configs.
Erf, I'd suggest to revert that then...
I mean I'd argue that "user XP" changes basically none between "x" and "×"
See above graph.
Ready for commit.
Bumped, fix the comments in IcmpRangeManager.h, spaces for consistency.
@borg- update on that?
@borg- what do you think? Do you prefer Restricted classes ? Or making them "not" preferred classes?
I believe this might no longer need the GUI patches since we removed ballistic tower?
I think this is acceptable as a first diff - I'll clarify in the commit message that this is merely one step forward.
I meant the old code logic :p
Add a new script that ought to do something similar to what we're using poediff form.
I'd like to have @vladislavbelov's take on this before merging
Upload with context.
Tue, Aug 4
It's in the template itself.
(needs a rebase now)
Fix small mistake in boudicca footprint (Circle conflicting with Square)
Ready for commit.
This makes the Hero Elephant an absurdly good siege unit.
That seems objectively superior to our current organisation to me.
I'm kind of surprised that this change alone is sufficient to rebalance ranged vs melee units?