Twitter | Search | |
David Shah
Open Source Electronics, FPGA and Software developer. Contributor to Project Icestorm, currently building open source tools for the ECP5 and other FPGAs. he/him
998
Tweets
694
Following
3,252
Followers
Tweets
David Shah retweeted
Al Jun 17
Today I am mostly baking, baking carriers that is...
Reply Retweet Like
David Shah Jun 16
Reply Retweet Like
David Shah Jun 13
Replying to @whitequark
Another of their parts is OpenRISC based (yes, really, not just or32 extension)! - mirror:
Reply Retweet Like
David Shah Jun 13
Replying to @whitequark
Sorry, don't have a csdn account, only pudn
Reply Retweet Like
David Shah Jun 13
Replying to @whitequark
Reply Retweet Like
David Shah Jun 13
Replying to @whitequark
Reply Retweet Like
David Shah Jun 8
I'm pretty sure that Lattice Diamond LSE has the same behaviour
Reply Retweet Like
David Shah Jun 8
Muxes would usually come from if statements where you deliberately writing a set/reset; gates might be general logic that would cause packing problems if mapped (only one SR signal per tile of 8 flops)
Reply Retweet Like
David Shah Jun 8
This is definitely deliberate in how the pass that extracts iCE40 (and the equivalent pass I wrote for ECP5) sync SRs. It maps muxes selecting between a signal and 0/1 to reset or set; but not ANDNOT or OR gates.
Reply Retweet Like
David Shah Jun 8
Replacing the ?: with out <= in & ~{WORD_WIDTH{annul}}; means that Yosys no longer infers a sync reset, but Vivado still does infer the reset (don't have Quartus installed)
Reply Retweet Like
David Shah Jun 8
Using & instead of ?: should prevent this
Reply Retweet Like
David Shah Jun 8
FYI, Yosys synth_ice40 with default settings will infer a sync reset from
Reply Retweet Like
David Shah Jun 5
The limit for ECP5 IOLOGIC is 800Mbps (400MHz DDR clock), enough for 760Mbps 1080p30 HDMI
Reply Retweet Like
David Shah Jun 4
Replying to @SingularitySur1
It would certainly be a challenge, but it might be doable. I suspect the problem wouldn't be any particular part on its own, more the sheer number of parts in aggregate (you might be able to cut this down if you don't need all the features)
Reply Retweet Like
David Shah Jun 3
Replying to @CRvanRensburg
The chasing LEDs (the contrast in the video isn't great unfortunately, but they are about right irl)
Reply Retweet Like
David Shah Jun 3
Replying to @gojimmypi @ico_TC
Probably we will do a small-medium batch, might have some plans too
Reply Retweet Like
David Shah Jun 3
Replying to @gojimmypi
Reply Retweet Like
David Shah Jun 3
Reply Retweet Like
David Shah May 24
Replying to @sahajsarup
Ah, I see. Are you using the "Assign Budget" button in the GUI?
Reply Retweet Like
David Shah May 24
Replying to @sahajsarup
Compared to what? It is quite common for Yosys/nextpnr to be about 20-30% less than Diamond at the moment due to a various small issues combined.
Reply Retweet Like