[SPIRVIntrinsics] Subgroup shuffle and barrier intrinsics

[christiangnrd]

Easier to review than #413 and then we can release SPIRVIntrinsics.jl 0.5.7 and remove the temporary compat workarounds in the oneAPI, OpenCL, and KA POCL backend branches.

[github-actions]

Your PR requires formatting changes to meet the project's style guidelines.
Please consider running Runic (git runic master) to apply these changes.

Click here to view the suggested changes.

diff --git a/lib/intrinsics/src/synchronization.jl b/lib/intrinsics/src/synchronization.jl
index 7d88b69..c8abe6d 100644
--- a/lib/intrinsics/src/synchronization.jl
+++ b/lib/intrinsics/src/synchronization.jl
@@ -156,7 +156,9 @@ export barrier, work_group_barrier, sub_group_barrier
                     MemorySemantics.SequentiallyConsistent | mem_fence_flags_to_semantics(flags))
 
 @inline sub_group_barrier(flags, scope = memory_scope_sub_group) =
-    control_barrier(Scope.Subgroup, cl_scope_to_spirv(scope),
-                    MemorySemantics.SequentiallyConsistent | mem_fence_flags_to_semantics(flags))
+    control_barrier(
+    Scope.Subgroup, cl_scope_to_spirv(scope),
+    MemorySemantics.SequentiallyConsistent | mem_fence_flags_to_semantics(flags)
+)
 
 barrier(flags) = work_group_barrier(flags)

[simeonschaub]

What's the reason we can't just merge this together with the OpenCL.jl changes in #413? I am a little weary of merging this without its tests

[christiangnrd]

What's the reason we can't just merge this together with the OpenCL.jl changes in #413? I am a little weary of merging this without its tests

The main reason was that this is useful on its own for other packages, so I was hoping that splitting it up would make it easier for a reviewer to find the time.

I’ll address the comments in #413 directly and close this once #413 is merged.