Test results: Snapshot (r4227) on Ubuntu 13.04 with Intel/ATI OSS drivers

[tomreyn]

I just tried the latest MegaGlest snapshot (r4227) on my intelbox (Intel GMA X4500MHD, more specific specs below), running (and booting) off a USB 2.0 key on Ubuntu 13.04 - and what can I say:

MegaGlest worked well, right out of the box!

Results:

• GLSelectBuf still works fine there - no selection issues
• The red custom mouse pointer works well
• 30-40 FPS while playing against a CPU on Conflict with Desert2 tileset, with a minimum of 23 FPS with debug view on, all values based on 800x600, with just a ~5 FPS decrease when switching to 1280x800
• No visible problems after activating texture compression (but also no noticeably increased FPS)
• CPU load was low or not elevated (the fan remained silent, and it would not when under load)

Minor issues:

• It feels like units move a bit slow, but I'm not sure why this is, can't put it into words
• The right-click-walking-instead-of-attacking issue is still present
• My locale was auto-detected as LC, a language file for which was not found, so MG defaulted to english.lng

Not or not sufficiently tested:

• Multi-player
• CPU (I did not actually inspect how load spread across this systems' two cores)
• Memory allocation (I did not check at all)

Summing it up: I'm excited.

[tomreyn]

Test results of MegaGlest r4227 running on Ubuntu 13.04 on my atibox (hardware specs in forums), using the r300g driver:

• GLSelectBuf still doesn't work well: With one or more units selected, right-clicking on a location to move  to (or attack) is often registered at a different location. It's with r300 / ati only, though, this didn't happen with i915. no such problems with wz2100.
• The red triangle mouse cursor moves too slowly on large screens / resolutions (1680x1050 here), lacks acceleration and an ability to configure it.
• With an 8 player map, 7 Mega Persian CPUs and a human observer, desert tileset, and 2x speed, the red triangle mouse cursor becomes unusable.
• With the same game setup, the load is not evenly enough spread out across all processes. one process causes (maximum) 150% load (i.e. the kernel needs to spread it out across two cores, but none of the cores is fully loaded), another at 20%, and the other two are between 1% and 10%. on the end of this game, the total memory allocation of megaglest processes was 650 MB. FPS was much decreased by then (averaging at about 20-25 FPS), but that's an old effect. Here's the debug view at the time (may be interesting in terms of caches):

• With frustrum calc ordering units around works, with the known limitations: while drawing large drawboxes FPS decreases by 50% and more, selection of units in front or behind trees or other objects is impossible.
• Standard FPS on conflict map with desert2 tileset, playing indian against a single CPU easy opponent (same setup as yesterday on i915) is 60, whether or not debug info is on hardly makes a difference.
• There is no selection box painted / drawn in first person view (but you can still select multiple units this way)
• Texture compression works without errors, again no noticeable difference.

G3DViewer

• transparency issues remain:

This archer is actually looking towards the camera and has his arm in front of his head while placing an arrow on the bow.

Map Editor

• looks good

WZ2100 side note

On a side note, here's a WZ2100 debug log (game started using: warzone2100 --texturecompression --shadows --debug=all), they do some nice logging about 3D capabilities http://paste.ubuntu.com/5595903/ - if you just want the 3d events logged then it's --debug=3d

[titi]

hm, maybe the zbuffer is disabled while rendering the archer, I will check for it