_whitelogger has joined #yosys
perryprog has joined #yosys
citypw has joined #yosys
nrossi has joined #yosys
proteusguy has quit [Ping timeout: 268 seconds]
proteusguy has joined #yosys
svenn4 has quit [Ping timeout: 268 seconds]
_whitelogger has joined #yosys
dys has joined #yosys
Jybz has joined #yosys
_whitelogger has joined #yosys
lutsabound has joined #yosys
gorbak25 has quit [Quit: gorbak25]
X-Scale has quit [Ping timeout: 276 seconds]
X-Scale has joined #yosys
X-Scale has quit [Ping timeout: 265 seconds]
X-Scale` has joined #yosys
X-Scale` is now known as X-Scale
Boko_dabce has joined #yosys
Boko_dabce has quit [Read error: Connection reset by peer]
Boko_debac has joined #yosys
Boko_debac has quit [Read error: Connection reset by peer]
citypw has quit [Ping timeout: 250 seconds]
Boko_edeea has joined #yosys
<
ZirconiumX>
So, for the $alu cell, what would be a cell that would change BI at runtime as opposed to it being constant?
Boko_edeea has quit [Read error: Connection reset by peer]
Jybz has quit [Ping timeout: 276 seconds]
Jybz has joined #yosys
Boko_bbbde has joined #yosys
Boko_bbbde has quit [Read error: Connection reset by peer]
<
ZirconiumX>
daveshah, maybe?
<
daveshah>
Hypothetically an adder/subtractor
<
daveshah>
I don't know if Yosys maps these yet (from add sub and mux)
<
daveshah>
But it's something I've wanted to look at at some point
<
ZirconiumX>
I'm wondering if trying to specialise based on CI and/or BI is worth it or not
Boko_aeadc has joined #yosys
<
daveshah>
BI=1 is a common case (subtractor)
<
daveshah>
Fixed BI and CI is much more common than variable
<
ZirconiumX>
I've been inspecting the output of Quartus and it's interesting (to me) that comparisons are not done with carry chains
<
ZirconiumX>
Which is actually probably cmp2lut
Boko_aeadc has quit [Read error: Connection reset by peer]
rombik_su has joined #yosys
<
daveshah>
Not just cmp2lut, we should probably have a mode in Yosys not to convert any compare to soft logic
<
mwk>
tbh I think we shouldcompletely dump convering compares to sub
<
mwk>
it's completely bonkers
develonepi3 has quit [Remote host closed the connection]
<
daveshah>
If nothing else because it prevents any serious downstream optimisation
<
mwk>
well it's also horribly inefficient
<
mwk>
at least on xilinx
<
mwk>
you could reduce area 2× by using the carry chain in a better way for the variable inputs case (two pairs of input bits per LUT)
<
mwk>
and at least 5× when one of the inputs is const
Jybz has quit [Quit: Konversation terminated!]
nrossi has quit [Quit: Connection closed for inactivity]
lutsabound has quit [Quit: Connection closed for inactivity]
lutsabound has joined #yosys
rombik_su has quit [Read error: Connection reset by peer]
tpb has quit [Remote host closed the connection]
tpb has joined #yosys