Replies: 1 comment
-
Woo! Makes a lot of sense and this makes a strong case for deprecating std_mem completely in my opinion |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Just a quick note about resource sharing in Vivado.
If you have a large Calyx design, you should map memories to map to either URAM or BRAM. This requires using
seq_mem
s in Calyx instead ofstd_mem
s. Also, the device you are targeting should support URAM/BRAM. Since havarti gets alveo boards, I used the specific part namexcu250-figd2104-2-e
.The reason why you want to map memories to URAM/BRAM is because it significantly speeds up how long resource estimates take. LeNet w/ memories mapped to URAM and targeting alveo boards takes about 10 minutes to get resource estimates, whereas LeNet w/ out memories mapped to URAM and targeting a smaller board takes > 24 hours to get resource estimates.
Beta Was this translation helpful? Give feedback.
All reactions