dsda-doom

dsda-doom 0.30.0 (the PrBoom+ successor, with the same GL renderer) plays Freedoom’s DEMO1 on its GL renderer at 55.9 steady fps with no stalls, after big-endian and GL fixes of ours. Its OpenGL played session (2026-10-25) had a 58 fps median and was smooth; the software launcher 44. What is left is the MIDI music, which still chops: FluidSynth’s voice DSP on the SPEs halved the underruns, but hundreds a minute remain.

Packages and settings

Our build of Debian’s dsda-doom (0.30.0+dfsg-1+ps3.N, packages/dsda-doom), replacing Debian’s, with Debian’s freedoom 0.13.0 (/usr/share/games/doom/freedoom2.wad). The release candidate is ps3.361.

• Two menu entries of ours (desktop/launchers) replace Debian’s (diverted): “DSDA-Doom (OpenGL)” at 1920x1080 and “DSDA-Doom (software)” at 960x540 scaled (at 1920x1080 the PPE manages about 5 fps in software).
• Music is MIDI through SDL_mixer and our FluidSynth (libfluidsynth3 2.6.1+dfsg-1+ps3.N, its voice DSP on the SPEs through libspeq, SPEQ_SYNTH=auto|spu|ppe|off; SPU-AUDIO.md).

How it performs

The batching gave +3% and better lows (CPU 118 against 161 s). The gameplay profile (batch 2026-10-19-ssdroot2-gameplay-profiles): Mesa 22% of the samples, the game 20.5%, the music about 31% on the audio thread (FluidSynth 14.2%, libspeq 11.9%: speq_synth_add_v2 7.4%, speq_synth_render_v2 2.0%; ALSA 3.4%), kernel 12.7%, ps3rsx 3.3%, Xorg 8.7% of the game’s; 2793 underruns in the run. The music is the largest share of the seven profiles (HOTPATHS.md).

The benchmark (recipe): -timedemo demo1 (Freedoom’s own DEMO1 lump, 2664 gametics, 76 s at 35 Hz), 6 passes (DSDA_BENCH_LOOPS=6, PASSES=1). The load ends at Playing demo: (the recipe’s LOAD_MARK; 0.30 never prints the R_Init lines on the PS3), where the window, perf and the memory snapshots start. Base is the software renderer at 960x540 scaled by SDL; the bench row of the suite is gl. Freedoom 0.13’s own demos desync (freedoom issue #1532), which doesn’t matter for a timedemo: the same tics play every run. Variants include play (the demo at normal speed with GL, 60 s, for the sound), playnomusic, tuned (OPL music, 48 kHz, a 2048 buffer, 60 fps cap), playopl, playspu, playppe, playoff, the batching switches and owncfg (base on a config of the run’s own, -config in the run directory: dsda-doom’s defaults, the player’s file untouched).

Known problems

• The MIDI music chops: about 465 sound underruns a minute with the synthesizer on the SPEs; the OpenGL session counted 7.5 audible dropouts a minute, the software one 35.9 (the CPU renders there). The decision on the default (FluidSynth on the SPEs, OPL, or no music) waits on the cause (the investigation).

Investigations

• 2026-10-16 the big-endian and GL fixes (0001-0004).
• 2026-10-17 the music on the SPEs.

Research

What the surveys and plans said about it, newest first; each line names its source, the section and the date.

• 2026-10-16, game-candidates.md, “The recommended suite and why”: PrBoom+ GL (on the user’s list) is covered by dsda-doom, its successor with the same GL renderer, packaged for ppc64.
• 2026-10-15, the user’s first wish list (backlog.txt, commit 7d58e3d), “2. GPU driver and game support”: “PrBoom+, with its GL renderer” a stretch goal: easy, a light test of the 2D/texture paths more than real 3D.