youmuliu/stgsharp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

rewrite of f32 fma, and other optmization of simd kernel ops

Browse Source
This commit is contained in:
Youmu Liu
2025-11-10 20:40:19 +08:00
parent 1dee467220
commit 678a846e54
18 changed files with 302 additions and 156 deletions
Download Patch File Download Diff File Expand all files Collapse all files
StgSharp.Mathematics/Numeric/MatrixParallel.Await.cs
+23 -5
View File
@@ -2,8 +2,8 @@
// -----------------------------------------------------------------------
// file="MatrixParallel.Await"
// Project: StgSharp
// AuthorGroup: Nitload Space
// Copyright (c) Nitload Space. All rights reserved.
// AuthorGroup: Nitload
// Copyright (c) Nitload. All rights reserved.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
@@ -56,7 +56,16 @@ namespace StgSharp.Mathematics.Numeric
min,
recommendCount
);
_awaitingLeaderHandle.Enqueue(handle);
do
{
if (Interlocked.Exchange(ref _sortState, 1) == 0)
{
_awaitingLeaderHandle.Enqueue(handle);
_ = Interlocked.Exchange(ref _sortState, 0);
Interlocked.Increment(ref _createGeneration);
break;
}
} while (true);
int cycleGeneration = _createGeneration;
do
{
@@ -66,7 +75,7 @@ namespace StgSharp.Mathematics.Numeric
{
if (Interlocked.Exchange(ref _sortState, 1) == 0)
{
if (_awaitingLeaderHandle.TryPopIfRefEqual(cycleGeneration, ref handle))
if (_awaitingLeaderHandle.TryPopIfRefEqual(cycleGeneration, in handle))
{
// Sort is implemented in TryPopIfRefEqual
ret = _threads.PopRange(handle.BestCount);
@@ -101,7 +110,16 @@ namespace StgSharp.Mathematics.Numeric
public static void ReturnParallelResources(ref MatrixParallelThread[] pool)
{
_threads.PushRange(pool);
while (true)
{
if (Interlocked.Exchange(ref _sortState, 1) == 0)
{
_threads.PushRange(pool);
Interlocked.Exchange(ref _sortState, 0);
break;
}
}
pool = null!;
}