Warning
0.9.5 raises the chip supply on every FM-1 (core 1.38 V, DC-DC 1.50 V), above what JieLi uses (1.35 / 1.45 V). It runs well on our unit, but the chip's absolute limits are not public. Use 0.9.6 instead: there the raised supply is an opt-in setting, off by default.
Install: ./instalar-linux.sh jangada-0.9.5.fwsc, or the web installer. Same sound as 0.9.4.
Both cores (#4, the pointer from @gospodindilly-ops; the second core found by @keremimo in Melodee and shipped in X0X by @charlesvestal)
The FM-1's chip has two cores, and Jangada now uses both. The second core renders part of the synth parts every block, so the same patch costs about a third less CPU (31% → 21% on our test load), leaving more room for heavy engines (FM6, PHYS) and more voices. The sound is bit-for-bit the same as on one core.
The supply fix. On some FM-1s, at the supply the boot loader leaves, the second core misreads RAM (single bit flips) and crashes within seconds. That's what made 0.9.3 and X0X click and freeze on our unit. Before starting the second core, Jangada now raises the DC-DC rail to 1.50 V and the core rail to 1.38 V, a step at a time. On our unit that took it from crashing in 6 s to 15 minutes under load with no errors, then a 33-minute real session with no errors. The measurements are in BUILDING.md.
If the second core ever faults anyway, it is stopped and the FM-1 carries on with one core. A crash early in a boot keeps the next boot on one core.
New on the USB console (tools/fm1_console.py): status shows sysvdd, vdc14 and the cpu2_* counters. vdd S [D] and p33 are for measuring.
Full Changelog: v0.9.4...v0.9.5