Make the slice iterator optimise better. #21448
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
r? @brson (rust_highfive has picked a reviewer for you, use r? to override) |
Due to the null pointer optimisation, LLVM was leaving totally useless
null pointer checks in tight loops using this iterator. We compile
`Some(a_reference)` to a no-op, and matching against None in an
`Option<&T>` is just comparing against 0, this means that the `None`
branch in `match Some(a_reference) { None => ... }` (as appears
effectively in a loop using `slice::Iter`) can't be eliminated unless
`a_reference` is known to be non-null.
LLVM currently doesn't seem to understand this in every situation (often
due to inlining, I am told), particularly painfully in ones involving
this iterator, but using `assume` means it understands almost always.
for i in x {
test::black_box(i);
}
Before:
.LBB0_4:
leaq 4(%rcx), %rsi
testq %rcx, %rcx
je .LBB0_5
movq %rcx, (%rsp)
#APP
#NO_APP
cmpq %rsi, %rax
movq %rsi, %rcx
jne .LBB0_4
After:
.LBB0_4:
movq %rcx, (%rsp)
addq $4, %rcx
#APP
#NO_APP
cmpq %rcx, %rax
jne .LBB0_4
This leads to significantly better unrolling and vectorisation too.
|
@bors r+ a62c rollup |
|
@bors r- Hm, this appears to be aborting in the bootstrap. Debugging now. |
|
We may want to hold off on |
|
Woah, I noticed my builds were taking a while but I didn't connect that it was caused by |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Due to the null pointer optimisation, LLVM was leaving totally useless
null pointer checks in tight loops using this iterator. We compile
Some(a_reference)to a no-op, and matching against None in anOption<&T>is just comparing against 0, this means that theNonebranch in
match Some(a_reference) { None => ... }(as appearseffectively in a loop using
slice::Iter) can't be eliminated unlessa_referenceis known to be non-null.LLVM currently doesn't seem to understand this in every situation (often
due to inlining, I am told), particularly painfully in ones involving
this iterator, but using
assumemeans it understands almost always.Before:
After:
This leads to significantly better unrolling and vectorisation too.