<litghost>
hackerfoo: I think when picking sample points, we should be maxizing the number of nodes at the location
<litghost>
hackerfoo: I added a print for each expanded RR node, and only 10 NR1BEG nodes were expands
<litghost>
hackerfoo: I would've expected 3 * 3 * 4 = 36, nodes to be expanded
<litghost>
hackerfoo: Because there are 4 NR1BEG wires per interconnect switch box
<litghost>
hackerfoo: This might explain why there is more data sparcity, but I'm not confident yet
<hackerfoo>
litghost: I thought of just expanding a window at each point until there are enough samples, but I don't know how many is enough.
<hackerfoo>
Also that's probably expensive to compute without some spatial data structure.
freemint has quit [Ping timeout: 245 seconds]
<hackerfoo>
Okay, I'm going to build a grid per segment type with counts of that segment type, and then pick the maximum count in each grid cell, nearest the center to break ties.
<hackerfoo>
litghost: Does that sound reasonable?
<litghost>
That was my thinking
<litghost>
One detail that may or may not matter is whether to use vpr or canonical point for centering
<hackerfoo>
The canonical point is probably better, right?
<hackerfoo>
Maybe I shouldn't use VPR coordinates at all.
freemint has joined #symbiflow
freemint has quit [Ping timeout: 245 seconds]
stzsch has quit [Ping timeout: 252 seconds]
stzsch has joined #symbiflow
stzsch has quit [Ping timeout: 276 seconds]
stzsch has joined #symbiflow
stzsch has quit [Ping timeout: 276 seconds]
stzsch_ has joined #symbiflow
stzsch_ has quit [Ping timeout: 276 seconds]
stzsch has joined #symbiflow
rvalles has quit [Ping timeout: 246 seconds]
rvalles has joined #symbiflow
_whitelogger has joined #symbiflow
Vonter has quit [Ping timeout: 250 seconds]
craigo has joined #symbiflow
Vonter has joined #symbiflow
kraiskil has joined #symbiflow
OmniMancer has joined #symbiflow
citypw has joined #symbiflow
_whitelogger has joined #symbiflow
craigo has quit [Quit: leaving]
Bertl_zZ is now known as Bertl
freemint has joined #symbiflow
_whitelogger has joined #symbiflow
freemint has quit [Remote host closed the connection]
freemint has joined #symbiflow
kuldeep has joined #symbiflow
kraiskil has quit [Ping timeout: 240 seconds]
Bertl is now known as Bertl_oO
proteus-guy has joined #symbiflow
kraiskil has joined #symbiflow
davidthings has joined #symbiflow
davidthings has quit [Read error: Connection reset by peer]
davidw has joined #symbiflow
davidw is now known as Guest42602
Guest42602 has quit [Read error: Connection reset by peer]
<litghost>
Yosys new inserts BUFG's automatically if "-noclkbuf" is not supplied. In the ROI graph, the BUFG is outside of the ROI. In the ROI-less graph the BUFG is required.
<litghost>
As we are finalizing the ROI breakout work, we have intentionally disabled the yosys BUFG insertion behavior to accomidate both ROI and ROI-less graphs
<litghost>
Once all xc7 graphs are ROI-less, we can re-enable Yosys's BUFG insertion
Vonter has joined #symbiflow
<lromor[m]>
I see. Thanks for clarifying!
citypw has quit [Ping timeout: 276 seconds]
OmniMancer has quit [Quit: Leaving.]
proteus-guy has quit [Ping timeout: 245 seconds]
ZipCPU has quit [Ping timeout: 240 seconds]
<hackerfoo>
litghost: Is it worth changing the `set_X` methods in vtr::Rect to the builder pattern so you can write e.g. `rect.set_ymin(1).set_ymax(2).set_xmin(3).set_xmax(4)`?
<litghost>
hackerfoo: I'm relatively ambivalent on it, but you are welcome to make that change upstream and incorporate downstream
<hackerfoo>
Okay. I don't like methods to return `void` if they can return something useful.
ZipCPU has joined #symbiflow
proteus-guy has joined #symbiflow
kraiskil has quit [Ping timeout: 276 seconds]
proteus-guy has quit [Ping timeout: 265 seconds]
freemint has quit [Remote host closed the connection]