Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
fn decode_coefs
is slow #1180.This makes
TxfmSize
andRectTxfmSize
into a realenum
, which eliminates many bounds checks in hotfn
s likefn decode_ceofs
. It also lets us further simplify theitx
macros and eliminate the distinction between theR
ect ones.Note that
TxfmSize
defined the square sizes; these ones are named with anS
prefix, while theR
-prefixedRectTxfmSize
stayR
-prefixed.There are a few new places where we have to create
TxfmSize
s usingfrom_repr
and thus do a bounds check. These should be pretty cheap, but there are optimizable out, as they ultimately come fromstatic
/const
data whos bounds are known but LLVM is not able to deduce them. This is why I added 7f0cdc2, starting to try to fix this, though it's a little complicated, so I didn't want to finish it in this PR.