Align boost::uuid to 16 bytes
#148
Open
+12
−2
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
Lastique
force-pushed
the
feature/alignas16
branch
3 times, most recently
from
a21d9bc
to
e8b1914
Compare
|
I've updated the PR to only require alignment of 4 on 32-bit platforms. At least, on Windows, 32-bit processes only get 4-byte aligned stack. |
jeking3
reviewed
May 14, 2024
Lastique
force-pushed
the
feature/alignas16
branch
from
e8b1914
to
6ac3681
Compare
|
@pdimov What's the purpose of |
|
It's something that works today. |
|
Do you mean that In any case, my question was what was the intended tested feature in this test. IOW, are you testing that |
On 64-bit targets, this improves performance of operations on UUIDs as the UUID object now does not cross page boundary. It may also improve code generation by compilers, which can now assume that loads and stores of UUIDs are not as expensive and therefore it may be more preferable to use vector instructions for that. On 32-bit targets, we cannot guarantee that boost::uuid is always aligned to 16 bytes because stack and dynamic memory allocations typically have lower default alignment (typically 4 or 8). For this reason, we only require alignment of 4, and also do not change SIMD intrinsics used in boost::uuid operations to their aligned equivalents. Using unaligned load/store instructions on actually aligned memory location on modern CPUs has the same performance as using aligned instructions, so the performance benefit is still there on 64-bit targets. Closes boostorg#139.
Lastique
force-pushed
the
feature/alignas16
branch
from
6ac3681
to
9dcdf15
Compare
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Outdated suggestions cannot be applied.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.
This improves performance of operations on UUIDs as the UUID object now does not(*) cross page boundary. It may also improve code generation by compilers, which can now assume that loads and stores of UUIDs are not as expensive and therefore it may be more preferable to use vector instructions for that.
(*) It is technically possible that the
boost::uuidobject is still not properly aligned under some conditions. For example, on some 32-bit targets the stack is not aligned to 16 bytes, so placing the UUID object on the unaligned stack may result in unaligned UUID. For this reason, this commit does not change the SIMD intrinsics used inboost::uuidoperations to their aligned equivalents. Using unaligned load/store instructions on actually aligned memory location on modern CPUs has the same performance as using aligned instructions, so the performance benefit is still there.Closes #139.