AArch64 Instruction support & minor bug fixes #273
Merged
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
FinnWilkinson
added
bug
…the mbind syscall.
… index) NEON instructions with tests.
FinnWilkinson
force-pushed
the
RIKEN-code-support
branch
from
08b5165 to
67f2781
Compare
|
#rerun tests |
…f SME src/dest operands.
jj16791
requested changes
Jan 30, 2023
The instructions in ROB are sorted by their insnID, hence we can use binary search to efficiently find them.
This avoids having to make a new vector for every tick of DispatchIssueUnit
Replace std::vector with std::list as it supports O(1) deletion.
dANW34V3R
reviewed
Jan 30, 2023
Closed
dANW34V3R
reviewed
Jan 30, 2023
jj16791
requested changes
Jan 30, 2023
dANW34V3R
reviewed
Jan 30, 2023
jj16791
approved these changes
Jan 30, 2023
dANW34V3R
approved these changes
Jan 30, 2023
jj16791
approved these changes
Jan 30, 2023
jj16791
pushed a commit
that referenced
this pull request
May 19, 2023
As well as some AArch64 instruction support, the following fixes and optimisations are implemented : - Added example SME core (A64FX with in-core SME support) - Fixes MIPS calculation in CoreWrapper.cc - Fixed store instruction group allocation logic for non-SVE instructions - Corrected SME instruction group allocation in - Reassign LdAddr iterator after erasing element in LSQ load conflict logic (originally in PR Reassign iterator after erasing element in std::vector #272) - Moved optinisations of PR Minor optimizations #274 into this PR : - Refactored the ReorderBuffer::commitMicroOps method by using binary search to find the list of uops with the same instruction id. This improves the complexity from O(n) to O(logn) - Changed the DispatchIssueUnit::tick method to use a dynamic array instead of initializing a vector for each method call. This eliminates ~ 14m memory allocations for stream-triad. - Replaced std::vector to std::list in LoadStoreQueue::startLoad as it provides O(1) deletion compared to O(n) (worst-case). - Removed the explicit copy constructor from the AArch64 Instruction class - Changed MatrixRow-Count config option to Matrix-Count - Makes it easier for the user to understand how many physcial ZA registers they are defining - Hides the internal ZA implementation, again making it easier for users to understand and use
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.
As well as some AArch64 instruction support, the following fixes are implemented :
ReorderBuffer::commitMicroOpsmethod by using binary search to find the list of uops with the same instruction id. This improves the complexity from O(n) to O(logn)DispatchIssueUnit::tickmethod to use a dynamic array instead of initializing a vector for each method call. This eliminates ~ 14m memory allocations for stream-triad.std::vectortostd::listinLoadStoreQueue::startLoadas it provides O(1) deletion compared to O(n) (worst-case).MatrixRow-Countconfig option toMatrix-Count