Skip to content

Common Implemenatation for MatMul and MatMulTran for both aligned and unaligned arrays #1218

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 8 commits into from
Oct 16, 2018
Merged

Common Implemenatation for MatMul and MatMulTran for both aligned and unaligned arrays #1218

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
6221e1b
implemenatation and unitTests added
Anipik Oct 10, 2018
9515ebe
added performance test for matmul and matmulTrans
Anipik Oct 10, 2018
9c34ff2
load combined with math operation
Anipik Oct 11, 2018
1793109
add flag removed
Anipik Oct 13, 2018
421e33d
TransPA removed as nobody uses this combination of flags
Anipik Oct 13, 2018
348f329
removed firstTime and corrected nativePerformanceTests
Anipik Oct 15, 2018
a19069d
Merge branch 'master' into matMul
Anipik Oct 15, 2018
aa805f0
removed branch from hot path sseintrinsics
Anipik Oct 16, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 7 additions & 16 deletions src/Microsoft.ML.CpuMath/Avx.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,7 @@ public static bool CheckAvx()
return Thunk.ChkAvx();
}

public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, AlignedArray src, AlignedArray dst, int crun)
public static void MatTimesSrc(bool tran, AlignedArray mat, AlignedArray src, AlignedArray dst, int crun)
{
Contracts.Assert(Compat(mat));
Contracts.Assert(Compat(src));
Expand All @@ -50,18 +50,18 @@ public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, AlignedArr
if (!tran)
{
Contracts.Assert(0 <= crun && crun <= dst.Size);
Thunk.MatMulX(add, Ptr(mat, pmat), Ptr(src, psrc), Ptr(dst, pdst), crun, src.Size);
Thunk.MatMulX(Ptr(mat, pmat), Ptr(src, psrc), Ptr(dst, pdst), crun, src.Size);
}
else
{
Contracts.Assert(0 <= crun && crun <= src.Size);
Thunk.MatMulTranX(add, Ptr(mat, pmat), Ptr(src, psrc), Ptr(dst, pdst), dst.Size, crun);
Thunk.MatMulTranX(Ptr(mat, pmat), Ptr(src, psrc), Ptr(dst, pdst), dst.Size, crun);
}
}
}
}

public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, int[] rgposSrc, AlignedArray srcValues,
public static void MatTimesSrc(AlignedArray mat, int[] rgposSrc, AlignedArray srcValues,
int posMin, int iposMin, int iposLim, AlignedArray dst, int crun)
{
Contracts.Assert(Compat(mat));
Expand All @@ -73,8 +73,7 @@ public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, int[] rgpo

if (iposMin >= iposLim)
{
if (!add)
dst.ZeroItems();
dst.ZeroItems();
return;
}
Contracts.AssertNonEmpty(rgposSrc);
Expand All @@ -85,16 +84,8 @@ public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, int[] rgpo
fixed (float* psrc = &srcValues.Items[0])
fixed (int* ppossrc = &rgposSrc[0])
{
if (!tran)
{
Contracts.Assert(0 <= crun && crun <= dst.Size);
Thunk.MatMulPX(add, Ptr(mat, pmat), ppossrc, Ptr(srcValues, psrc), posMin, iposMin, iposLim, Ptr(dst, pdst), crun, srcValues.Size);
}
else
{
Contracts.Assert(0 <= crun && crun <= srcValues.Size);
Thunk.MatMulTranPX(add, Ptr(mat, pmat), ppossrc, Ptr(srcValues, psrc), posMin, iposMin, iposLim, Ptr(dst, pdst), dst.Size);
}
Contracts.Assert(0 <= crun && crun <= dst.Size);
Thunk.MatMulPX(Ptr(mat, pmat), ppossrc, Ptr(srcValues, psrc), posMin, iposMin, iposLim, Ptr(dst, pdst), crun, srcValues.Size);
}
}
}
Expand Down
550 changes: 380 additions & 170 deletions src/Microsoft.ML.CpuMath/AvxIntrinsics.cs
View file
Open in desktop

Large diffs are not rendered by default.

20 changes: 6 additions & 14 deletions src/Microsoft.ML.CpuMath/CpuAligenedMathUtils.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,40 +75,32 @@ public static void AssertCompatible(ICpuFullMatrix mat, ICpuVector src, ICpuVect

/// <summary>
/// Matrix multiplication:
/// if (add)
/// dst = mat * src
/// else
/// dest += mat * src
/// dst = mat * src
/// </summary>
/// <param name="add">The addition flag</param>
/// <param name="mat">The multiplier matrix</param>
/// <param name="src">The source vector</param>
/// <param name="dst">The destination vector</param>
public static void MatTimesSrc(bool add, ICpuFullMatrix mat, ICpuVector src, ICpuVector dst)
public static void MatTimesSrc(ICpuFullMatrix mat, ICpuVector src, ICpuVector dst)
{
bool colMajor = typeof(TMatrix) == typeof(CpuAlignedMatrixCol);
AssertCompatible(mat, src, dst);
var m = A(mat);
CpuMathUtils.MatTimesSrc(colMajor, add, m.Items, A(src).Items, A(dst).Items, m.RunCnt);
CpuMathUtils.MatTimesSrc(colMajor, m.Items, A(src).Items, A(dst).Items, m.RunCnt);
}

/// <summary>
/// Matrix transpose multiplication:
/// if (add)
/// dst = mat' * src
/// else
/// dest += mat' * src
/// dst = mat' * src
/// </summary>
/// <param name="add">The addition flag</param>
/// <param name="mat">The multiplier matrix</param>
/// <param name="src">The source vector</param>
/// <param name="dst">The destination vector</param>
public static void MatTranTimesSrc(bool add, ICpuFullMatrix mat, ICpuVector src, ICpuVector dst)
public static void MatTranTimesSrc(ICpuFullMatrix mat, ICpuVector src, ICpuVector dst)
{
bool colMajor = typeof(TMatrix) == typeof(CpuAlignedMatrixCol);
AssertCompatible(mat, dst, src);
var m = A(mat);
CpuMathUtils.MatTimesSrc(!colMajor, add, m.Items, A(src).Items, A(dst).Items, m.RunCnt);
CpuMathUtils.MatTimesSrc(!colMajor, m.Items, A(src).Items, A(dst).Items, m.RunCnt);
}
}

Expand Down
105 changes: 20 additions & 85 deletions src/Microsoft.ML.CpuMath/CpuMathUtils.netcoreapp.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@ public static partial class CpuMathUtils
public static int GetVectorAlignment()
=> Avx.IsSupported ? Vector256Alignment : (Sse.IsSupported ? Vector128Alignment : FloatAlignment);

public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, AlignedArray src, AlignedArray dst, int crun)
public static void MatTimesSrc(bool tran, AlignedArray mat, AlignedArray src, AlignedArray dst, int crun)
{
Contracts.Assert(mat.Size == dst.Size * src.Size);
Contracts.Assert(crun >= 0);
Expand All @@ -34,25 +34,25 @@ public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, AlignedArr
if (!tran)
{
Contracts.Assert(crun <= dst.Size);
AvxIntrinsics.MatMulX(add, mat, src, dst, crun, src.Size);
AvxIntrinsics.MatMulX(mat, src, dst, crun, src.Size);
}
else
{
Contracts.Assert(crun <= src.Size);
AvxIntrinsics.MatMulTranX(add, mat, src, dst, dst.Size, crun);
AvxIntrinsics.MatMulTranX(mat, src, dst, dst.Size, crun);
}
}
else if (Sse.IsSupported)
{
if (!tran)
{
Contracts.Assert(crun <= dst.Size);
SseIntrinsics.MatMulA(add, mat, src, dst, crun, src.Size);
SseIntrinsics.MatMul(mat, src, dst, crun, src.Size);
}
else
{
Contracts.Assert(crun <= src.Size);
SseIntrinsics.MatMulTranA(add, mat, src, dst, dst.Size, crun);
SseIntrinsics.MatMulTran(mat, src, dst, dst.Size, crun);
}
}
else
Expand All @@ -68,14 +68,7 @@ public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, AlignedArr
dotProduct += mat[i * src.Size + j] * src[j];
}

if (add)
{
dst[i] += dotProduct;
}
else
{
dst[i] = dotProduct;
}
dst[i] = dotProduct;
}
}
else
Expand All @@ -89,20 +82,13 @@ public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, AlignedArr
dotProduct += mat[j * src.Size + i] * src[j];
}

if (add)
{
dst[i] += dotProduct;
}
else
{
dst[i] = dotProduct;
}
dst[i] = dotProduct;
}
}
}
}

public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, int[] rgposSrc, AlignedArray srcValues,
public static void MatTimesSrc(AlignedArray mat, int[] rgposSrc, AlignedArray srcValues,
int posMin, int iposMin, int iposLim, AlignedArray dst, int crun)
{
Contracts.AssertValue(rgposSrc);
Expand All @@ -113,8 +99,7 @@ public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, int[] rgpo

if (iposMin >= iposLim)
{
if (!add)
dst.ZeroItems();
dst.ZeroItems();
return;
}

Expand All @@ -123,76 +108,26 @@ public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, int[] rgpo

if (Avx.IsSupported)
{
if (!tran)
{
Contracts.Assert(crun <= dst.Size);
AvxIntrinsics.MatMulPX(add, mat, rgposSrc, srcValues, posMin, iposMin, iposLim, dst, crun, srcValues.Size);
}
else
{
Contracts.Assert(crun <= srcValues.Size);
AvxIntrinsics.MatMulTranPX(add, mat, rgposSrc, srcValues, posMin, iposMin, iposLim, dst, dst.Size);
}
Contracts.Assert(crun <= dst.Size);
AvxIntrinsics.MatMulPX(mat, rgposSrc, srcValues, posMin, iposMin, iposLim, dst, crun, srcValues.Size);
}
else if (Sse.IsSupported)
{
if (!tran)
{
Contracts.Assert(crun <= dst.Size);
SseIntrinsics.MatMulPA(add, mat, rgposSrc, srcValues, posMin, iposMin, iposLim, dst, crun, srcValues.Size);
}
else
{
Contracts.Assert(crun <= srcValues.Size);
SseIntrinsics.MatMulTranPA(add, mat, rgposSrc, srcValues, posMin, iposMin, iposLim, dst, dst.Size);
}
Contracts.Assert(crun <= dst.Size);
SseIntrinsics.MatMulPA(mat, rgposSrc, srcValues, posMin, iposMin, iposLim, dst, crun, srcValues.Size);
}
else
{
if (!tran)
{
Contracts.Assert(crun <= dst.Size);
for (int i = 0; i < crun; i++)
{
float dotProduct = 0;
for (int j = iposMin; j < iposLim; j++)
{
int col = rgposSrc[j] - posMin;
dotProduct += mat[i * srcValues.Size + col] * srcValues[col];
}

if (add)
{
dst[i] += dotProduct;
}
else
{
dst[i] = dotProduct;
}
}
}
else
Contracts.Assert(crun <= dst.Size);
for (int i = 0; i < crun; i++)
{
Contracts.Assert(crun <= srcValues.Size);
for (int i = 0; i < dst.Size; i++)
float dotProduct = 0;
for (int j = iposMin; j < iposLim; j++)
{
float dotProduct = 0;
for (int j = iposMin; j < iposLim; j++)
{
int col = rgposSrc[j] - posMin;
dotProduct += mat[col * dst.Size + i] * srcValues[col];
}

if (add)
{
dst[i] += dotProduct;
}
else
{
dst[i] = dotProduct;
}
int col = rgposSrc[j] - posMin;
dotProduct += mat[i * srcValues.Size + col] * srcValues[col];
}

dst[i] = dotProduct;
}
}
}
Expand Down
6 changes: 3 additions & 3 deletions src/Microsoft.ML.CpuMath/CpuMathUtils.netstandard.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,10 +16,10 @@ public static partial class CpuMathUtils
public static int GetVectorAlignment()
=> Vector128Alignment;

public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, AlignedArray src, AlignedArray dst, int crun) => SseUtils.MatTimesSrc(tran, add, mat, src, dst, crun);
public static void MatTimesSrc(bool tran, AlignedArray mat, AlignedArray src, AlignedArray dst, int crun) => SseUtils.MatTimesSrc(tran, mat, src, dst, crun);

public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, int[] rgposSrc, AlignedArray srcValues,
int posMin, int iposMin, int iposLim, AlignedArray dst, int crun) => SseUtils.MatTimesSrc(tran, add, mat, rgposSrc, srcValues, posMin, iposMin, iposLim, dst, crun);
public static void MatTimesSrc(AlignedArray mat, int[] rgposSrc, AlignedArray srcValues,
int posMin, int iposMin, int iposLim, AlignedArray dst, int crun) => SseUtils.MatTimesSrc(mat, rgposSrc, srcValues, posMin, iposMin, iposLim, dst, crun);

public static void Add(float a, Span<float> dst) => SseUtils.Add(a, dst);

Expand Down
23 changes: 7 additions & 16 deletions src/Microsoft.ML.CpuMath/Sse.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ private static bool Compat(AlignedArray a)
return q;
}

public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, AlignedArray src, AlignedArray dst, int crun)
public static void MatTimesSrc(bool tran, AlignedArray mat, AlignedArray src, AlignedArray dst, int crun)
{
Contracts.Assert(Compat(mat));
Contracts.Assert(Compat(src));
Expand All @@ -46,18 +46,18 @@ public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, AlignedArr
if (!tran)
{
Contracts.Assert(0 <= crun && crun <= dst.Size);
Thunk.MatMulA(add, Ptr(mat, pmat), Ptr(src, psrc), Ptr(dst, pdst), crun, src.Size);
Thunk.MatMul(Ptr(mat, pmat), Ptr(src, psrc), Ptr(dst, pdst), crun, src.Size);
}
else
{
Contracts.Assert(0 <= crun && crun <= src.Size);
Thunk.MatMulTranA(add, Ptr(mat, pmat), Ptr(src, psrc), Ptr(dst, pdst), dst.Size, crun);
Thunk.MatMulTran(Ptr(mat, pmat), Ptr(src, psrc), Ptr(dst, pdst), dst.Size, crun);
}
}
}
}

public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, int[] rgposSrc, AlignedArray srcValues,
public static void MatTimesSrc(AlignedArray mat, int[] rgposSrc, AlignedArray srcValues,
int posMin, int iposMin, int iposLim, AlignedArray dst, int crun)
{
Contracts.Assert(Compat(mat));
Expand All @@ -69,8 +69,7 @@ public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, int[] rgpo

if (iposMin >= iposLim)
{
if (!add)
dst.ZeroItems();
dst.ZeroItems();
return;
}
Contracts.AssertNonEmpty(rgposSrc);
Expand All @@ -81,16 +80,8 @@ public static void MatTimesSrc(bool tran, bool add, AlignedArray mat, int[] rgpo
fixed (float* psrc = &srcValues.Items[0])
fixed (int* ppossrc = &rgposSrc[0])
{
if (!tran)
{
Contracts.Assert(0 <= crun && crun <= dst.Size);
Thunk.MatMulPA(add, Ptr(mat, pmat), ppossrc, Ptr(srcValues, psrc), posMin, iposMin, iposLim, Ptr(dst, pdst), crun, srcValues.Size);
}
else
{
Contracts.Assert(0 <= crun && crun <= srcValues.Size);
Thunk.MatMulTranPA(add, Ptr(mat, pmat), ppossrc, Ptr(srcValues, psrc), posMin, iposMin, iposLim, Ptr(dst, pdst), dst.Size);
}
Contracts.Assert(0 <= crun && crun <= dst.Size);
Thunk.MatMulPA(Ptr(mat, pmat), ppossrc, Ptr(srcValues, psrc), posMin, iposMin, iposLim, Ptr(dst, pdst), crun, srcValues.Size);
}
}
}
Expand Down
Loading