public static Tensor Invoke(CudaCode kernels, string funcName, Tensor result, Tensor src)
{
try
{
TSCudaContext context = CudaHelpers.TSContextForTensor(src);
CudaContext cudaContext = context.CudaContextForTensor(src);
cudaContext.SetCurrent();
Tensor writeTarget = TensorResultBuilder.GetWriteTarget(result, src, false, src.Sizes);
long elementCount = writeTarget.ElementCount();
byte[] ptx = kernels.GetPtx(context.Compiler);
if (result == src)
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t1_" + funcName, writeTarget, elementCount);
}
else
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t2_" + funcName, writeTarget, src, elementCount);
}
return(writeTarget);
}
catch (Exception e)
{
Logger.WriteLine($"Error = '{e.Message}', Call stack = '{e.StackTrace}'");
throw;
}
}
public static Tensor Invoke(CudaCode kernels, Tensor result, Tensor srcY, Tensor srcX)
{
TSCudaContext context = CudaHelpers.TSContextForTensor(srcY);
CudaContext cudaContext = context.CudaContextForTensor(srcY);
cudaContext.SetCurrent();
Tensor writeTarget = TensorResultBuilder.GetWriteTarget(result, srcY, false, srcY.Sizes);
long elementCount = writeTarget.ElementCount();
byte[] ptx = kernels.GetPtx(context.Compiler);
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "atan2", writeTarget, srcY, srcX, elementCount);
return(writeTarget);
}
public static Tensor Invoke(CudaCode kernels, string funcName, Tensor result, Tensor src, float value)
{
TSCudaContext context = CudaHelpers.TSContextForTensor(src);
CudaContext cudaContext = context.CudaContextForTensor(src);
cudaContext.SetCurrent();
Tensor writeTarget = TensorResultBuilder.GetWriteTarget(result, src, false, src.Sizes);
long elementCount = writeTarget.ElementCount();
byte[] ptx = kernels.GetPtx(context.Compiler);
if (result == src)
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t1_" + funcName, writeTarget, value, elementCount);
}
else
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t2_" + funcName, writeTarget, src, value, elementCount);
}
return(writeTarget);
}
public static Tensor Invoke(CudaCode kernels, string funcName, Tensor result, Tensor lhs, Tensor rhs, Tensor rhs2)
{
var context = CudaHelpers.TSContextForTensor(lhs);
var cudaContext = context.CudaContextForTensor(lhs);
cudaContext.SetCurrent();
var writeTarget = TensorResultBuilder.GetWriteTarget(result, lhs, false, lhs.Sizes);
var elementCount = writeTarget.ElementCount();
var ptx = kernels.GetPtx(context.Compiler);
if (result == lhs)
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t1_" + funcName, writeTarget, rhs, rhs2, elementCount);
}
else
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t2_" + funcName, writeTarget, lhs, rhs, rhs2, elementCount);
}
return(writeTarget);
}
public static Tensor Invoke(CudaCode kernels, Tensor result, Tensor rhs)
{
try
{
TSCudaContext context = CudaHelpers.TSContextForTensor(rhs);
CudaContext cudaContext = context.CudaContextForTensor(rhs);
cudaContext.SetCurrent();
long elementCount = result.ElementCount();
byte[] ptx = kernels.GetPtx(context.Compiler);
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t1_atomicAdd", result, rhs, elementCount);
return(result);
}
catch (Exception e)
{
Logger.WriteLine($"Exception message = '{e.Message}', Call stack = '{e.StackTrace}'");
throw;
}
}
