Merge branch 'dev' into dev

Author: kirillgarbar

benchmarks/GraphBLAS-sharp.Benchmarks/Matrix/SpGeMM/Expand.fs

@@ -15,7 +15,7 @@ open GraphBLAS.FSharp.Backend
 [<AbstractClass>]
 [<IterationCount(100)>]
 [<WarmupCount(10)>]
-[<Config(typeof<Configs.Matrix2>)>]
+[<Config(typeof<Configs.Matrix>)>]
 type Benchmarks<'elem when 'elem : struct>(
         buildFunToBenchmark,
         converter: string -> 'elem,
@@ -24,11 +24,9 @@ type Benchmarks<'elem when 'elem : struct>(
 
     let mutable funToBenchmark = None
 
-    let mutable firstMatrix = Unchecked.defaultof<ClMatrix<'elem>>
-    let mutable secondMatrix = Unchecked.defaultof<ClMatrix<'elem>>
+    let mutable matrix = Unchecked.defaultof<ClMatrix<'elem>>
 
-    let mutable firstMatrixHost = Unchecked.defaultof<_>
-    let mutable secondMatrixHost = Unchecked.defaultof<_>
+    let mutable matrixHost = Unchecked.defaultof<_>
 
     member val ResultMatrix = Unchecked.defaultof<ClMatrix.COO<'elem> option> with get, set
 
@@ -38,7 +36,7 @@ type Benchmarks<'elem when 'elem : struct>(
     [<ParamsSource("InputMatrixProvider")>]
     member val InputMatrixReader = Unchecked.defaultof<MtxReader> with get, set
 
-    member this.OclContext:ClContext = (fst this.OclContextInfo).ClContext
+    member this.OclContext: ClContext = (fst this.OclContextInfo).ClContext
     member this.WorkGroupSize = snd this.OclContextInfo
 
     member this.Processor =
@@ -78,24 +76,21 @@ type Benchmarks<'elem when 'elem : struct>(
         reader.ReadMatrix converter
 
     member this.Mxm() =
-        this.ResultMatrix <- this.FunToBenchmark this.Processor DeviceOnly firstMatrix secondMatrix
+        this.ResultMatrix <- this.FunToBenchmark this.Processor DeviceOnly matrix matrix
 
     member this.ClearInputMatrices() =
-        firstMatrix.Dispose this.Processor
-        secondMatrix.Dispose this.Processor
+        matrix.Dispose this.Processor
 
     member this.ClearResult() =
         match this.ResultMatrix with
         | Some matrix -> matrix.Dispose this.Processor
         | None -> ()
 
     member this.ReadMatrices() =
-        firstMatrixHost <- this.ReadMatrix this.InputMatrixReader
-        secondMatrixHost <- this.ReadMatrix this.InputMatrixReader
+        matrixHost <- this.ReadMatrix this.InputMatrixReader
 
     member this.LoadMatricesToGPU () =
-        firstMatrix <- buildMatrix this.OclContext firstMatrixHost
-        secondMatrix <- buildMatrix this.OclContext secondMatrixHost
+        matrix <- buildMatrix this.OclContext matrixHost
 
     abstract member GlobalSetup : unit -> unit
 

paket.lock

@@ -761,11 +761,11 @@ NUGET
     BinaryDefense.FSharp.Analyzers.Hashing (0.2.2)
       FSharp.Analyzers.SDK (>= 0.8) - restriction: >= net5.0
       FSharp.Core (>= 5.0.1) - restriction: >= net5.0
-    FSharp.Analyzers.SDK (0.12) - restriction: >= net5.0
-      FSharp.Compiler.Service (>= 43.7.200) - restriction: >= net6.0
-      FSharp.Core (>= 7.0.200) - restriction: >= net6.0
-      McMaster.NETCore.Plugins (>= 1.4) - restriction: >= net6.0
-    FSharp.Compiler.Service (43.7.300) - restriction: >= net6.0
+    FSharp.Analyzers.SDK (0.11) - restriction: >= net5.0
+      FSharp.Compiler.Service (>= 41.0.1) - restriction: >= net5.0
+      FSharp.Core (>= 6.0.1) - restriction: >= net5.0
+      McMaster.NETCore.Plugins (>= 1.4) - restriction: >= net5.0
+    FSharp.Compiler.Service (43.7.300) - restriction: >= net5.0
       FSharp.Core (7.0.300) - restriction: >= netstandard2.0
       System.Buffers (>= 4.5.1) - restriction: >= netstandard2.0
       System.Collections.Immutable (>= 6.0) - restriction: >= netstandard2.0
@@ -775,7 +775,7 @@ NUGET
       System.Reflection.Metadata (>= 6.0.1) - restriction: >= netstandard2.0
       System.Runtime.CompilerServices.Unsafe (>= 6.0) - restriction: >= netstandard2.0
     FSharp.Core (7.0.300) - restriction: >= net5.0
-    McMaster.NETCore.Plugins (1.4) - restriction: >= net6.0
+    McMaster.NETCore.Plugins (1.4) - restriction: >= net5.0
       Microsoft.DotNet.PlatformAbstractions (>= 3.1.6) - restriction: >= netcoreapp2.1
       Microsoft.Extensions.DependencyModel (>= 5.0) - restriction: >= netcoreapp2.1
     Microsoft.DotNet.PlatformAbstractions (3.1.6) - restriction: >= net6.0
@@ -785,13 +785,17 @@ NUGET
     System.Buffers (4.5.1) - restriction: >= net6.0
     System.Collections.Immutable (7.0) - restriction: >= net6.0
       System.Runtime.CompilerServices.Unsafe (>= 6.0) - restriction: || (>= net462) (&& (>= net6.0) (< net7.0)) (&& (< net6.0) (>= netstandard2.0))
-    System.Diagnostics.DiagnosticSource (7.0.2) - restriction: >= net6.0
+    System.Diagnostics.DiagnosticSource (7.0.2) - restriction: >= net5.0
+      System.Memory (>= 4.5.5) - restriction: || (>= net462) (&& (< net6.0) (>= netstandard2.0))
       System.Runtime.CompilerServices.Unsafe (>= 6.0) - restriction: || (>= net462) (&& (>= net6.0) (< net7.0)) (&& (< net6.0) (>= netstandard2.0))
     System.Memory (4.5.5) - restriction: >= net6.0
       System.Buffers (>= 4.5.1) - restriction: || (&& (>= monoandroid) (< netstandard1.1)) (&& (< monoandroid) (< net45) (>= netstandard1.1) (< netstandard2.0) (< win8) (< wpa81)) (&& (< monoandroid) (< netstandard1.1) (>= portable-net45+win8+wpa81) (< win8)) (>= monotouch) (&& (>= net45) (< netstandard2.0)) (&& (< net45) (< netcoreapp2.0) (>= netstandard2.0)) (>= net461) (&& (< netstandard1.1) (>= win8)) (&& (< netstandard2.0) (< uap10.1) (>= wpa81)) (>= xamarinios) (>= xamarinmac) (>= xamarintvos) (>= xamarinwatchos)
       System.Runtime.CompilerServices.Unsafe (>= 4.5.3) - restriction: || (&& (>= monoandroid) (< netstandard1.1)) (&& (< monoandroid) (< net45) (>= netstandard1.1) (< netstandard2.0) (< win8) (< wpa81)) (&& (< monoandroid) (>= netcoreapp2.0) (< netcoreapp2.1)) (&& (< monoandroid) (< netstandard1.1) (>= portable-net45+win8+wpa81) (< win8)) (>= monotouch) (&& (>= net45) (< netstandard2.0)) (&& (< net45) (< netcoreapp2.0) (>= netstandard2.0)) (>= net461) (&& (< netstandard1.1) (>= win8)) (&& (< netstandard2.0) (>= wpa81)) (>= uap10.1) (>= xamarinios) (>= xamarinmac) (>= xamarintvos) (>= xamarinwatchos)
     System.Reflection.Emit (4.7) - restriction: >= net6.0
-    System.Reflection.Metadata (7.0.1) - restriction: >= net6.0
+    System.Reflection.Metadata (7.0.1) - restriction: >= net6.
+    System.Numerics.Vectors (4.5) - restriction: || (&& (>= net462) (>= net5.0)) (&& (>= net5.0) (< net6.0))
+    System.Reflection.Emit (4.7) - restriction: >= net5.0
+    System.Reflection.Metadata (7.0.1) - restriction: >= net5.0
       System.Collections.Immutable (>= 7.0) - restriction: || (>= net462) (>= netstandard2.0)
     System.Runtime.CompilerServices.Unsafe (6.0) - restriction: >= net6.0
     System.Text.Encodings.Web (7.0) - restriction: >= net6.0

src/GraphBLAS-sharp.Backend/Common/Sort/Radix.fs

@@ -318,6 +318,7 @@ module Radix =
                     keysPair <- swap keysPair
                     valuesPair <- swap valuesPair
 
+                    globalOffset.Free processor
                     localOffset.Free processor
                     shift.Free processor
 

src/GraphBLAS-sharp.Backend/Common/Sum.fs

@@ -704,6 +704,7 @@ module Reduce =
                         reducedValues.Free processor
                         reducedKeys.Free processor
                         resultPositions.Free processor
+
                         None
                     else
                         // write values
@@ -988,6 +989,7 @@ module Reduce =
                         firstReducedKeys.Free processor
                         secondReducedKeys.Free processor
                         resultPositions.Free processor
+
                         None
                     else
                         // write value

src/GraphBLAS-sharp.Backend/Matrix/SpGeMM/Expand.fs

@@ -401,6 +401,10 @@ module Expand =
 
                     // compute sub result
                     let length, result = runCOO subMatrix
+
+                    // release subMatrix (TODO(non blocking Free)
+                    subMatrix.Dispose processor
+
                     // increase workOffset according to previous expand
                     let workOffset = workOffset + length
 
@@ -447,7 +451,12 @@ module Expand =
             if generalLength < maxAllocSize then
                 segmentLengths.Free processor
 
-                runOneStep processor allocationMode leftMatrix rightMatrixRowsNNZ rightMatrix
+                let result =
+                    runOneStep processor allocationMode leftMatrix rightMatrixRowsNNZ rightMatrix
+
+                rightMatrixRowsNNZ.Free processor
+
+                result
             else
                 let result =
                     runManySteps