JIT ARM64-SVE: Add BitwiseClear and BooleanNot#101853
JIT ARM64-SVE: Add BitwiseClear and BooleanNot#101853kunalspathak merged 1 commit intodotnet:mainfrom
Conversation
|
Note regarding the |
|
Tagging subscribers to this area: @dotnet/area-system-runtime-intrinsics |
| fmt = IF_SVE_GW_3A; | ||
| break; | ||
|
|
||
| case INS_sve_clz: |
There was a problem hiding this comment.
The asserts for these instructions were wrong - it allowed cases that don't exist.
|
@dotnet/arm64-contrib @kunalspathak |
|
/azp run runtime-coreclr superpmi-replay, runtime-coreclr superpmi-diffs |
|
Azure Pipelines successfully started running 2 pipeline(s). |
kunalspathak
left a comment
There was a problem hiding this comment.
nit: Suggested change to remove instruction from summary docs.
| /// BIC Ztied1.B, Pg/M, Ztied1.B, Zop2.B | ||
| /// BIC Zresult.D, Zop1.D, Zop2.D | ||
| /// svbool_t svbic[_b]_z(svbool_t pg, svbool_t op1, svbool_t op2) | ||
| /// BIC Presult.B, Pg/Z, Pop1.B, Pop2.B |
There was a problem hiding this comment.
I don't think we are supporting the BIC Presult.B, Pg/Z, Pop1.B, Pop2.B. Can we remove it from the summary docs in this and other APIs?
There was a problem hiding this comment.
Sve. BitwiseClear(mask) will convert the mask to a normal vector and use the vector BIC instruction. That's not great. We should be using the mask variant where possible. We need something in lowering? hwinstrinsics? to spot this.
I've raised an issue to cover it: #101970
Is it ok to leave the summary as is, because we'll have to restore them when fixing the issue. Plus it makes it easier to spot them when we fix it.
Test results: