Misaki/Misaki.HighPerformance
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add AllBitsSet, refactor WideLane, improve math paths

Browse Source 
- Add static AllBitsSet property to ISPMDLane and implement in ScalarLane and WideLane
- Refactor WideLane shuffle table pointers and update usages
- Improve pointer safety and mask handling in CompressStore, Gather, and MaskLoad
- Enhance Sin, Cos, SinCos with fast-math and hardware fallback
- Add Newton-Raphson refinement for reciprocal/sqrt when not fast-math
- Optimize MathV.Vector vector loading (struct init, pointer ops)
- Update project file: version 1.3.4, content packaging, AOT settings
- Minor code cleanup and naming consistency fixes
This commit is contained in:
Misaki
2026-05-01 12:19:58 +09:00
parent 5b4832a886
commit 18a181f57a
7 changed files with 209 additions and 124 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
Misaki.HighPerformance.Mathematics.SPMD/Templates/IJobSPMD.tt
View File

@@ -10,7 +10,6 @@ using System.Numerics;
namespace Misaki.HighPerformance.Mathematics.SPMD;
<#
const string TLane = "TLane";
const string TNumber = "TNumber";
const string GenericParameters = $"{TLane}, {TNumber}";

75
Misaki.HighPerformance.Mathematics.SPMD/Templates/MathV.Vector.gen.cs
View File

@@ -44,21 +44,21 @@ public static unsafe partial class MathV
where TLane : unmanaged, ISPMDLane<TLane, TNumber>
where TNumber : unmanaged, INumber<TNumber>, IBinaryNumber<TNumber>, IMinMaxValue<TNumber>, IBitwiseOperators<TNumber, TNumber, TNumber>
{
var width = TLane.LaneWidth;
Unsafe.SkipInit(out TLane x);
var px = (TNumber*)&x;
Unsafe.SkipInit(out TLane y);
var py = (TNumber*)&y;
var x = stackalloc TNumber[width];
var y = stackalloc TNumber[width];
for (var i = 0; i < width; i++)
for (var i = 0; i < TLane.LaneWidth; i++)
{
x[i] = pSrc[i * 2 + 0];
y[i] = pSrc[i * 2 + 1];
px[i] = pSrc[i * 2 + 0];
py[i] = pSrc[i * 2 + 1];
}
return new Vector2<TLane, TNumber>
{
x = TLane.Load(x),
y = TLane.Load(y),
x = x,
y = y,
};
}
@@ -515,24 +515,25 @@ public static unsafe partial class MathV
where TLane : unmanaged, ISPMDLane<TLane, TNumber>
where TNumber : unmanaged, INumber<TNumber>, IBinaryNumber<TNumber>, IMinMaxValue<TNumber>, IBitwiseOperators<TNumber, TNumber, TNumber>
{
var width = TLane.LaneWidth;
Unsafe.SkipInit(out TLane x);
var px = (TNumber*)&x;
Unsafe.SkipInit(out TLane y);
var py = (TNumber*)&y;
Unsafe.SkipInit(out TLane z);
var pz = (TNumber*)&z;
var x = stackalloc TNumber[width];
var y = stackalloc TNumber[width];
var z = stackalloc TNumber[width];
for (var i = 0; i < width; i++)
for (var i = 0; i < TLane.LaneWidth; i++)
{
x[i] = pSrc[i * 3 + 0];
y[i] = pSrc[i * 3 + 1];
z[i] = pSrc[i * 3 + 2];
px[i] = pSrc[i * 3 + 0];
py[i] = pSrc[i * 3 + 1];
pz[i] = pSrc[i * 3 + 2];
}
return new Vector3<TLane, TNumber>
{
x = TLane.Load(x),
y = TLane.Load(y),
z = TLane.Load(z),
x = x,
y = y,
z = z,
};
}
@@ -1024,27 +1025,29 @@ public static unsafe partial class MathV
where TLane : unmanaged, ISPMDLane<TLane, TNumber>
where TNumber : unmanaged, INumber<TNumber>, IBinaryNumber<TNumber>, IMinMaxValue<TNumber>, IBitwiseOperators<TNumber, TNumber, TNumber>
{
var width = TLane.LaneWidth;
Unsafe.SkipInit(out TLane x);
var px = (TNumber*)&x;
Unsafe.SkipInit(out TLane y);
var py = (TNumber*)&y;
Unsafe.SkipInit(out TLane z);
var pz = (TNumber*)&z;
Unsafe.SkipInit(out TLane w);
var pw = (TNumber*)&w;
var x = stackalloc TNumber[width];
var y = stackalloc TNumber[width];
var z = stackalloc TNumber[width];
var w = stackalloc TNumber[width];
for (var i = 0; i < width; i++)
for (var i = 0; i < TLane.LaneWidth; i++)
{
x[i] = pSrc[i * 4 + 0];
y[i] = pSrc[i * 4 + 1];
z[i] = pSrc[i * 4 + 2];
w[i] = pSrc[i * 4 + 3];
px[i] = pSrc[i * 4 + 0];
py[i] = pSrc[i * 4 + 1];
pz[i] = pSrc[i * 4 + 2];
pw[i] = pSrc[i * 4 + 3];
}
return new Vector4<TLane, TNumber>
{
x = TLane.Load(x),
y = TLane.Load(y),
z = TLane.Load(z),
w = TLane.Load(w),
x = x,
y = y,
z = z,
w = w,
};
}

11
Misaki.HighPerformance.Mathematics.SPMD/Templates/MathV.Vector.tt
View File

@@ -65,23 +65,22 @@ public static unsafe partial class MathV
<#= TLaneRestrictions #>
<#= TNumberRestrictions #>
{
var width = TLane.LaneWidth;
<# for (int i = 0; i < dimension; i++) { #>
var <#= components[i] #> = stackalloc <#= TNumber #>[width];
Unsafe.SkipInit(out TLane <#= components[i] #>);
var p<#= components[i] #> = (<#= TNumber #>*)&<#= components[i] #>;
<# } #>
for (var i = 0; i < width; i++)
for (var i = 0; i < TLane.LaneWidth; i++)
{
<# for (int i = 0; i < dimension; i++) { #>
<#= components[i] #>[i] = pSrc[i * <#= dimension #> + <#= i #>];
p<#= components[i] #>[i] = pSrc[i * <#= dimension #> + <#= i #>];
<# } #>
}
return new <#= vectorType #>
{
<# for (int i = 0; i < dimension; i++) { #>
<#= components[i] #> = <#= TLane #>.Load(<#= components[i] #>),
<#= components[i] #> = <#= components[i] #>,
<# } #>
};
}