KernelInterface

.buildkite/pipeline.yml

@@ -74,7 +74,6 @@ steps:
     matrix:
       setup:
         version:
-          - "1.10"
           - "1.11"
           - "1.12"
         test_args:
@@ -84,6 +83,32 @@ steps:
       (build.message =~ /\[only [^\]]*(tests|julia)/ ||
        build.message !~ /\[only / && !build.pull_request.draft)
 
+  # XXX: KernelInterface 0.3 isn't registered yet, and Julia 1.10 ignores [sources]: develop
+  #      it from a clone before instantiating, which the julia-test plugin would do first
+  - <<: *julia
+    matrix:
+      setup:
+        version:
+          - "1.10"
+        test_args:
+          - "--verbose"
+    plugins:
+      - JuliaCI/julia#v1:
+          version: "{{matrix.version}}"
+      - JuliaCI/julia-coverage#v1:
+          codecov: true
+    command: |
+      git clone --depth 1 --branch tb/ki-0.3 https://github.com/JuliaGPU/KernelAbstractions.jl ka
+      julia --project -e '
+        using Pkg
+        Pkg.develop(path="ka/lib/KernelInterface")
+        Pkg.instantiate()
+        Pkg.test(; coverage=true, test_args=`{{matrix.test_args}}`)'
+    if: |
+      build.message !~ /\[skip [^\]]*(tests|julia)/ &&
+      (build.message =~ /\[only [^\]]*(tests|julia)/ ||
+       build.message !~ /\[only / && !build.pull_request.draft)
+
   - <<: *julia
     matrix:
       setup:
@@ -127,10 +152,14 @@ steps:
     plugins:
       - JuliaCI/julia#v1:
           version: "1.10"
-      - JuliaCI/julia-test#v1:
-          run_tests: false
+    # XXX: KernelInterface 0.3 isn't registered yet, and Julia 1.10 ignores [sources]
     command: |
+      git clone --depth 1 --branch tb/ki-0.3 https://github.com/JuliaGPU/KernelAbstractions.jl ka
       julia --project -e '
+        using Pkg
+        Pkg.develop(path="ka/lib/KernelInterface")
+        Pkg.instantiate()
+
         using AMDGPU
         @assert !AMDGPU.functional()'
     agents:

Project.toml

@@ -19,6 +19,7 @@ GPUArrays = "0c68f7d7-f131-5f86-a1c3-88cf8149b2d7"
 GPUCompiler = "61eb1bfa-7361-4325-ad38-22787b887f55"
 GPUToolbox = "096a3bc2-3ced-46d0-87f4-dd12716f4bfc"
 KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
+KernelInterface = "4ee993da-d684-4d17-a7dd-4e58e78d92bf"
 LLVM = "929cbde3-209d-540e-8aea-75f648917ca0"
 LLVMDowngrader_jll = "f52de702-fb25-5922-94ba-81dd59b07444"
 Libdl = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
@@ -48,6 +49,9 @@ AMDGPUEnzymeCoreExt = "EnzymeCore"
 AMDGPUSparseMatricesCSRExt = "SparseMatricesCSR"
 AMDGPUSpecialFunctionsExt = "SpecialFunctions"
 
+[sources]
+KernelInterface = {url = "https://github.com/JuliaGPU/KernelAbstractions.jl", rev = "tb/ki-0.3", subdir = "lib/KernelInterface"}
+
 [compat]
 AMDGPU_LLVM_Backend_jll = "23"
 AbstractFFTs = "1.0"
@@ -63,6 +67,7 @@ GPUArrays = "11.5.14"
 GPUCompiler = "2.9"
 GPUToolbox = "3"
 KernelAbstractions = "0.9.2"
+KernelInterface = "0.3"
 LLVM = "9"
 LLVMDowngrader_jll = "0.11"
 Libdl = "1"

docs/src/api/devices.md

@@ -42,6 +42,7 @@ AMDGPU.device_id!
 AMDGPU.HIP.name
 AMDGPU.HIP.wavefrontsize
 AMDGPU.HIP.gcn_arch
+AMDGPU.HIP.max_workgroup_dims
 AMDGPU.HIP.device_id
 AMDGPU.HIP.properties
 ```

docs/src/api/intrinsics.md

@@ -23,6 +23,7 @@ AMDGPU.Device.blockDim
 
 ```@docs
 AMDGPU.sync_workgroup
+AMDGPU.sync_wavefront
 AMDGPU.sync_workgroup_count
 AMDGPU.sync_workgroup_and
 AMDGPU.sync_workgroup_or

src/AMDGPU.jl

@@ -60,6 +60,7 @@ include("discovery/discovery.jl")
 using .ROCmDiscovery
 using .ROCmDiscovery: AMDGPU_LLVM_Backend_jll, LLVMDowngrader_jll
 
+using KernelInterface
 include("utils.jl")
 
 include("hsa/HSA.jl")
@@ -82,6 +83,7 @@ Base.Experimental.@MethodTable(method_table)
 #needs to be before Device since sync uses this
 const syncscope_agent = UnsafeAtomics.Internal.LLVMSyncScope{:agent}()
 const syncscope_workgroup = UnsafeAtomics.Internal.LLVMSyncScope{:workgroup}()
+const syncscope_wavefront = UnsafeAtomics.Internal.LLVMSyncScope{:wavefront}()
 
 # Referenced by the generated `kernel_state()`, and generators run in the world they are
 # defined in, so this has to precede the device code.
@@ -102,7 +104,7 @@ import .Device: ROCDeviceArray, AS, HostCall, HostCallHolder, hostcall!
 import .Device: @ROCDynamicLocalArray, @ROCStaticLocalArray
 import .Device: workitemIdx, workgroupIdx, workgroupDim, gridItemDim, gridGroupDim
 import .Device: threadIdx, blockIdx, blockDim
-import .Device: sync_workgroup, sync_workgroup_count, sync_workgroup_and, sync_workgroup_or
+import .Device: sync_workgroup, sync_wavefront, sync_workgroup_count, sync_workgroup_and, sync_workgroup_or
 import .Device: @rocprint, @rocprintln, @rocprintf
 
 export ROCDeviceArray, @ROCDynamicLocalArray, @ROCStaticLocalArray
@@ -147,10 +149,15 @@ function Atomix.modify!(ref::ROCIndexableRef, op::OP, x, ord) where OP <: Union{
     GC.@preserve root UnsafeAtomics.modify!(ptr, op, x, ord, syncscope_agent)
 end
 
-include("ROCKernels.jl")
+# KernelAbstractions
+include("ROCKernelsOld.jl")
 import .ROCKernels: ROCBackend
 export ROCBackend
 
+# KernelInterface
+include("ROCKernels.jl")
+import .ROCInterface
+
 include("precompile.jl")
 
 function __init__()