diff options
author | Yann Herklotz <git@yannherklotz.com> | 2021-04-07 15:17:33 +0100 |
---|---|---|
committer | Yann Herklotz <git@yannherklotz.com> | 2021-04-07 15:17:33 +0100 |
commit | e45902cf8e19ea8ee7573f5d7629fa33ce47a470 (patch) | |
tree | 8b646b89359713f48ccf4b89874a32b1f5edaef2 /reviews | |
parent | b4273c305ec683aac168a08969ad956b12e460ba (diff) | |
download | oopsla21_fvhls-e45902cf8e19ea8ee7573f5d7629fa33ce47a470.tar.gz oopsla21_fvhls-e45902cf8e19ea8ee7573f5d7629fa33ce47a470.zip |
Add changes
Diffstat (limited to 'reviews')
-rw-r--r-- | reviews/pldi.org | 24 |
1 files changed, 10 insertions, 14 deletions
diff --git a/reviews/pldi.org b/reviews/pldi.org index 765e7e6..2f14d3e 100644 --- a/reviews/pldi.org +++ b/reviews/pldi.org @@ -71,7 +71,7 @@ sound and provide a novel and proper foundation for extending to optimizations, so I'm giving the paper a small benefit of the doubt in my rating. -* Other apparent weaknesses in the design +** Other apparent weaknesses in the design :PROPERTIES: :CUSTOM_ID: other-apparent-weaknesses-in-the-design :END: @@ -87,7 +87,7 @@ I can sympathize with the difficulty of getting Verilog compilers to identify RAMs. IMO, this is not a significant demerit of the current approach as part of a research prototype. -* Evaluation +** Evaluation :PROPERTIES: :CUSTOM_ID: evaluation :END: @@ -103,7 +103,7 @@ coaxed into doing that kind of partial evaluation! What would be different in that approach, considering implementation and proof? It would make for another worthwhile comparison as evaluation. -* Size of contribution +** Size of contribution :PROPERTIES: :CUSTOM_ID: size-of-contribution :END: @@ -119,7 +119,7 @@ It is not clear that there are strong research contributions in the proofs, just engineering, which seems worth doing even if there are not surprises in the results. -* Small stuff +** Small stuff :PROPERTIES: :CUSTOM_ID: small-stuff :END: @@ -206,11 +206,7 @@ mechanically verified HLS. The tools is (somewhat) comparable to its (unverified-) competitor based on the aforementioned matrices, with room for significant improvements. -** Comments for author - :PROPERTIES: - :CUSTOM_ID: comments-for-author-1 - :END: -* Strengths +** Strengths :PROPERTIES: :CUSTOM_ID: strengths :END: @@ -225,7 +221,7 @@ for significant improvements. the test candidates by preparing an alternative version of the benchmarks that uses repeated division and multiplications by 2. -* A few Questions/Concerns +** A few Questions/Concerns :PROPERTIES: :CUSTOM_ID: a-few-questionsconcerns :END: @@ -264,7 +260,7 @@ when it is trivial to achieve that. should be a as challenging as certifying the instruction translation from C to Verilog. -* nits +** nits :PROPERTIES: :CUSTOM_ID: nits :END: @@ -329,7 +325,7 @@ experience study on what was required to plug a different (formally verified) backend to the Compiler toolchain, a backend that caters to Verilog as the target language. -* Strengths +** Strengths :PROPERTIES: :CUSTOM_ID: strengths-1 :END: @@ -339,7 +335,7 @@ Verilog as the target language. studies 3. I enjoyed studying Figure 9 -* Weaknesses +** Weaknesses :PROPERTIES: :CUSTOM_ID: weaknesses :END: @@ -545,7 +541,7 @@ into account new results, which this will have to count as. I would encourage the authors to take the feedback gathered here into account and resubmit with the new results to a different venue. -** Response by Yann Herklotz +* Response by Yann Herklotz [[mailto:yann.herklotz15@imperial.ac.uk][yann.herklotz15@imperial.ac.uk]] :PROPERTIES: :CUSTOM_ID: response-by-yann-herklotz-yann.herklotz15imperial.ac.uk |