Glasgow Embedded / Glasgow - History

applet.memory.onfi: explicitly flush IN FIFO.

Improves throughput.

support.logging: dump_hex: show total bytes, not elided bytes.

It's not actually clear how many bytes are shown, so printing how
many bytes are elided isn't helpful at all. Instead, print the total
count, which is handy e.g. to see if a complete packet was received.

revC1: add io port pinout to back silkscreen

Shortened version with just numbers, V instead of Vio, aligned numbers

fx2.FX2Arbiter→fx2_crossbar.FX2Crossbar

It's not so much an arbiter (there's only one bus master, the FPGA)
as a crossbar switch with an integrated control signal sequencer.

applet.internal.benchmark: add a counter, for debugging.

If the transfer hangs or errors out, the counter will show where
that happened.

gateware.i2c: fix fencepost error in reads.

The ACK period would be interpreted as starting one bit time too
early, and SDA would be released. Not only that is a fencepost error,
but releasing SDA during the sample half-period is illegal; it would
be correctly detected as a stop condition, terminating trasnaction.

This commit reworks the read part of the FSM to only change SDA level
during setup half-period, and fixes the fencepost error.

gateware.i2c: load new data after each byte is read.

Also, fix testbench to release SDA after writing a bit.

applet.audio.yamaha_opl: mark as broken.

Looks like the changes in e59730fb disturbed a fragile balance
between coroutine scheduling and buffering in the hardware.
To fix this, the applet likely requires an explicit feedback path
between the playback and record coroutines, and/or further rework
of pipelining.