test fixes

Author: chriselrod

src/LoopVectorization.jl

@@ -115,6 +115,7 @@ include("precompile.jl")
 _precompile_()
 
 _vreduce(+, Float64[1.0])
+# matmul_params(64, 32, 64)
 
 # import ChainRulesCore, ForwardDiff
 # include("vmap_grad.jl")

test/gemm.jl

@@ -633,17 +633,17 @@
     C[m,n] = ΔCₘₙ
     end) |> LoopVectorization.loopset;
     if LoopVectorization.register_count() == 32
-        if LoopVectorization.register_size() == 64
-        # @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :m, :n, :m, 1, 8)
-            @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :k, :n, :m, 1, 8)
-        elseif LoopVectorization.register_size() == 16
-            @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :m, :n, :m, 4, 4)
-        end            
+      if LoopVectorization.register_size() == 64
+        @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :m, :n, :m, 1, 8)
+        # @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :k, :n, :m, 1, 8)
+      elseif LoopVectorization.register_size() == 16
+        @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :m, :n, :m, 4, 4)
+      end            
     elseif LoopVectorization.register_count() == 16
-        # @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :m, :n, :m, 2, 4)
-        @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :n, :m, :n, 2, 4)
+      @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :m, :n, :m, 2, 4)
+      # @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :n, :m, :n, 2, 4)
     elseif LoopVectorization.register_count() == 8
-        @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :m, :n, :m, 1, 4)
+      @test LoopVectorization.choose_order(lsAtmulBt8) == ([:n, :m, :k], :m, :n, :m, 1, 4)
     end
 
     struct SizedMatrix{M,N,T} <: DenseMatrix{T}

test/runtests.jl

@@ -19,7 +19,7 @@ if LOOPVECTORIZATION_TEST == "all"
     env = copy(ENV)
     env["LOOPVECTORIZATION_TEST"] = "part$i"
     env["JULIA_NUM_THREADS"] = string(Threads.nthreads())
-    processes[i] = run(pipeline(setenv(`$(Base.julia_cmd()) $(@__FILE__)`, env), stderr = io, stdout = io), wait=false)
+    processes[i] = run(pipeline(setenv(`$(Base.julia_cmd()) $(@__FILE__) --project=$(Base.active_project())`, env), stderr = io, stdout = io), wait=false)
   end
   completed = fill(false, NUMGROUPS)
   while true