Open xfzhou01 opened 1 year ago
Hi Xiaofeng,
Thank you for your interest in our project and feel free to ask more questions if the following answers still don't make sense.
Dear Saranyu,
Thank you very much for your timely reply! I have understood these points. I may have another small question, in Table I, why benchmarks including keypair, gsm, HLSCNN, FlexNLP, Dataflow and Opticalflow are not listed, is it affected by the batch size of sub-accelerators?
Best Wishes, Xiaofeng ZHOU
Hi, thanks for your query. For sub-accelerators with batch size of one, strong FC (intra-batch FC check requires at least a batch size of two) will catch bugs like initialization and Out-of-Bound bugs which can also be caught by running off-the-shelf checkers like Frama-c and Infer so we didn't run FC for sub-accelerators with batch size of one. Let me know.
Hi Saranyu, so sorry to bother you again, I am recently collecting some benchmarks for conducting model checking on HLS designs, if possible, could you provide the source code of grayscale[32|64|128] and mean[32|64|128]? Thank you so much ~
Hi Xiaofeng,
Can you share your email id? I'll email them to you directly.
Hi Saranyu, thank you so much for your timely response. My email address is [xzhoubu@connect.ust.hk]. Thanks ^^
Dear authors,
I read your work "Scaling Up Hardware Accelerator Verification using A-QED with Functional Decomposition" published in FMCAD, which is a very impressive work and and I am very interested in it. I may have some questions about the experiment part of your work.
About the dSAC verification, how do you get the Spec function from different kinds of representations?
It would be very appreciated if you could take a look at my questions above. So sorry for bothering you.
Best Wishes, Xiaofeng ZHOU