<omnitechnomancer>
if you find valgrind too slow and aren't after uninit memory try -fsanitize=address
FabM has joined #apicula
<omnitechnomancer>
problems going away when you shuffle things is usually a sign of out of bounds writes that have happened to hit benign spots now
<pepijndevos>
ohhhh I'm actually getting to the bits where I need to express the constraints on the slices, exciting stuff :))
<pepijndevos>
ERROR: Found two arcs with same sink wire R5C23_CE1: cpu.latched_rd[2]_LUT2_I0_F_LUT3_I0_F_LUT3_I1_F (3) vs cpu.latched_rd[2]_LUT2_I1_I0_LUT2_I0_F_LUT2_I1_F (3)
<pepijndevos>
there is a constraint that makes all the clocks the same per tile, but not for enables and resets yet :)))
<omnitechnomancer>
\o/
<pepijndevos>
And Gowin actually has more fine-grained clock/reset/enable than per tile, so yay
<omnitechnomancer>
can you have two of each per tile?
<pepijndevos>
Each tile has for pairs that each have their own
<pepijndevos>
*four
<omnitechnomancer>
ah
FabM has quit [Quit: Leaving]
<pepijndevos>
blegh
<pepijndevos>
routing makes it a long way and then
<pepijndevos>
Warning: Failed to find a route for arc 0 of net cpu.mem_wdata[0]_DFFE_D_Q[7].
<pepijndevos>
ERROR: Routing design failed.
<pepijndevos>
let's try a simpler design...
<pepijndevos>
Weird that it'd fail on some completely normal-looking DFF after routing a few thousand others
<pepijndevos>
Ah, I think some routing is missing from the IOB to the fabric