Update for PolyesterWeave 0.1.1

.github/workflows/ci.yml

@@ -11,27 +11,26 @@ defaults:
   run:
     shell: bash
 jobs:
-  test:
-    name: Julia ${{ matrix.version }}/${{ matrix.threads }} threads/${{ matrix.os }}/${{ matrix.arch }}/${{ github.event_name }}
+  coverage:
+    name: coverage=true/Julia ${{ matrix.version }}/${{ matrix.threads }} threads/${{ matrix.os }}/${{ matrix.arch }}/${{ github.event_name }}
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
       matrix:
         arch:
           - x64
-          - x86
+        eltype:
+          - Float64
+          - Float32
+          - Int64
+          - Int32	  
         os:
           - ubuntu-latest
-          - windows-latest
-          - macOS-latest
         threads:
           - '1'
           - '3' # GitHub runners have 2 cores, so `NUM_CORES+1` is 3
         version:
           - '1' # automatically expands to the latest stable 1.x release of Julia
-        exclude:
-          - os: macOS-latest
-            arch: x86 # 32-bit Julia binaries are not available on macOS
     steps:
       - uses: actions/checkout@v2
       - uses: julia-actions/setup-julia@v1
@@ -51,24 +50,35 @@ jobs:
       - uses: julia-actions/julia-buildpkg@v1
       - uses: julia-actions/julia-runtest@v1
         with:
-          coverage: false
+          coverage: true
         env:
           JULIA_NUM_THREADS: ${{ matrix.threads }}
-  coverage:
-    name: coverage=true/Julia ${{ matrix.version }}/${{ matrix.threads }} threads/${{ matrix.os }}/${{ matrix.arch }}/${{ github.event_name }}
+          JULIA_TEST_ELTYPE: ${{ matrix.eltype }}
+      - uses: julia-actions/julia-processcoverage@v1
+      - uses: codecov/codecov-action@v1
+        with:
+          file: lcov.info
+  test:
+    name: Julia ${{ matrix.version }}/${{ matrix.threads }} threads/${{ matrix.os }}/${{ matrix.arch }}/${{ github.event_name }}
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
       matrix:
         arch:
           - x64
+          - x86
         os:
           - ubuntu-latest
+          - windows-latest
+          - macOS-latest
         threads:
           - '1'
           - '3' # GitHub runners have 2 cores, so `NUM_CORES+1` is 3
         version:
           - '1' # automatically expands to the latest stable 1.x release of Julia
+        exclude:
+          - os: macOS-latest
+            arch: x86 # 32-bit Julia binaries are not available on macOS
     steps:
       - uses: actions/checkout@v2
       - uses: julia-actions/setup-julia@v1
@@ -88,13 +98,9 @@ jobs:
       - uses: julia-actions/julia-buildpkg@v1
       - uses: julia-actions/julia-runtest@v1
         with:
-          coverage: true
+          coverage: false
         env:
           JULIA_NUM_THREADS: ${{ matrix.threads }}
-      - uses: julia-actions/julia-processcoverage@v1
-      - uses: codecov/codecov-action@v1
-        with:
-          file: lcov.info
   docs:
     name: Documentation
     runs-on: ubuntu-latest

Project.toml

@@ -1,7 +1,7 @@
 name = "Octavian"
 uuid = "6fd5a793-0b7e-452c-907f-f8bfe9c57db4"
 authors = ["Mason Protter", "Chris Elrod", "Dilum Aluthge", "contributors"]
-version = "0.3.3"
+version = "0.3.4"
 
 [deps]
 ArrayInterface = "4fba245c-0d91-5ea0-9b3e-6abc04ee57a9"
@@ -16,12 +16,12 @@ VectorizationBase = "3d5dd08c-fd9d-11e8-17fa-ed2836048c2f"
 [compat]
 ArrayInterface = "3.1.14"
 IfElse = "0.1"
-LoopVectorization = "0.12.34"
+LoopVectorization = "0.12.86"
 ManualMemory = "0.1.1"
-PolyesterWeave = "0.1"
+PolyesterWeave = "0.1.1"
 Static = "0.2, 0.3"
 ThreadingUtilities = "0.4.6"
-VectorizationBase = "0.21.5"
+VectorizationBase = "0.21.15"
 julia = "1.6"
 
 [extras]

src/complex_matmul.jl

@@ -52,7 +52,8 @@ for AT in [:AbstractVector, :AbstractMatrix]  # to avoid ambiguity error
             η = ifelse(ArrayInterface.is_lazy_conjugate(_A), StaticInt(-1), StaticInt(1))
             (+ᶻ, -ᶻ) = ifelse(ArrayInterface.is_lazy_conjugate(_C), (-, +), (+, -))
 
-            @tturbo for n ∈ indices((C, B), (3, 2)), m ∈ indices((C, A), 2)
+            # @tturbo for n ∈ indices((C, B), (3, 2)), m ∈ indices((C, A), 2)
+            @turbo for n ∈ indices((C, B), (3, 2)), m ∈ indices((C, A), 2)
                 Cmn_re = zero(T)
                 Cmn_im = zero(T)
                 for k ∈ indices((A, B), (3, 1))
@@ -130,4 +131,4 @@ for AT in [:AbstractVector, :AbstractMatrix]  # to avoid ambiguity error
             _C
         end
     end
-end
+end

src/matmul.jl

@@ -360,7 +360,7 @@ function __matmul!(
     clamp(div_fast(M * N, StaticInt{256}() * W), 0, _nthread-1)
   end
   # nkern = cld_fast(M * N,  MᵣW * Nᵣ)
-  threads, torelease = PolyesterWeave.__request_threads(_nrequest % UInt32, PolyesterWeave.worker_pointer())
+  threads, torelease = PolyesterWeave.__request_threads(_nrequest % UInt32, PolyesterWeave.worker_pointer(), nothing)
   # _threads, _torelease = PolyesterWeave.request_threads(Threads.threadid()%UInt32, _nrequest)
 
   nrequest = threads.i