devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merged in rmlarsen/eigen2 (pull request PR-292)

Browse Source 
Adds a fast memcpy function to Eigen.
This commit is contained in:
Benoit Steiner
2017-01-25 00:14:04 +00:00
parent 7b6aaa3440 3be5ee2352
commit e96c77668d
5 changed files with 54 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
unsupported/Eigen/CXX11/src/Tensor/TensorContraction.h
View File

@@ -56,7 +56,7 @@ void pack_simple(Scalar * dst, const Scalar * src, Index cols, Index rows, Index
} else {
// Naive memcpy calls
for (Index col = 0; col < cols; ++col) {
memcpy(dst + col*lddst, src + col*ldsrc, rows*sizeof(Scalar));
internal::fast_memcpy(dst + col*lddst, src + col*ldsrc, rows*sizeof(Scalar));
}
}
}

2
unsupported/Eigen/CXX11/src/Tensor/TensorDeviceDefault.h
View File

@@ -22,7 +22,7 @@ struct DefaultDevice {
internal::aligned_free(buffer);
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void memcpy(void* dst, const void* src, size_t n) const {
::memcpy(dst, src, n);
internal::fast_memcpy(dst, src, n);
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void memcpyHostToDevice(void* dst, const void* src, size_t n) const {
memcpy(dst, src, n);

2
unsupported/Eigen/CXX11/src/Tensor/TensorDeviceThreadPool.h
View File

@@ -106,7 +106,7 @@ struct ThreadPoolDevice {
}
EIGEN_STRONG_INLINE void memcpy(void* dst, const void* src, size_t n) const {
::memcpy(dst, src, n);
internal::fast_memcpy(dst, src, n);
}
EIGEN_STRONG_INLINE void memcpyHostToDevice(void* dst, const void* src, size_t n) const {
memcpy(dst, src, n);

4
unsupported/Eigen/CXX11/src/Tensor/TensorFFT.h
View File

@@ -253,7 +253,7 @@ struct TensorEvaluator<const TensorFFTOp<FFT, ArgType, FFTResultType, FFTDir>, D
// get data into line_buf
const Index stride = m_strides[dim];
if (stride == 1) {
memcpy(line_buf, &buf[base_offset], line_len*sizeof(ComplexScalar));
m_device.memcpy(line_buf, &buf[base_offset], line_len*sizeof(ComplexScalar));
} else {
Index offset = base_offset;
for (int j = 0; j < line_len; ++j, offset += stride) {
@@ -271,7 +271,7 @@ struct TensorEvaluator<const TensorFFTOp<FFT, ArgType, FFTResultType, FFTDir>, D
// write back
if (FFTDir == FFT_FORWARD && stride == 1) {
memcpy(&buf[base_offset], line_buf, line_len*sizeof(ComplexScalar));
m_device.memcpy(&buf[base_offset], line_buf, line_len*sizeof(ComplexScalar));
} else {
Index offset = base_offset;
const ComplexScalar div_factor = ComplexScalar(1.0 / line_len, 0);