mirror/eigen
2
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-01-24 14:45:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

Made the cost model cwiseMax and cwiseMin methods consts to help the PowerPC cuda compiler compile this code.

Browse Source
This commit is contained in:
Benoit Steiner 2016-08-18 13:46:36 -07:00
parent 647a51b426
commit 7944d4431f
1 changed files with 12 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
unsupported/Eigen/CXX11/src/Tensor/TensorCostModel.h
View File

@ -91,21 +91,21 @@ class TensorOpCost {
}
// TODO(rmlarsen): Define min in terms of total cost, not elementwise.
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost& cwiseMin(
const TensorOpCost& rhs) {
bytes_loaded_ = numext::mini(bytes_loaded_, rhs.bytes_loaded());
bytes_stored_ = numext::mini(bytes_stored_, rhs.bytes_stored());
compute_cycles_ = numext::mini(compute_cycles_, rhs.compute_cycles());
return *this;
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost cwiseMin(
const TensorOpCost& rhs) const {
double bytes_loaded = numext::mini(bytes_loaded_, rhs.bytes_loaded());
double bytes_stored = numext::mini(bytes_stored_, rhs.bytes_stored());
double compute_cycles = numext::mini(compute_cycles_, rhs.compute_cycles());
return TensorOpCost(bytes_loaded, bytes_stored, compute_cycles);
}
// TODO(rmlarsen): Define max in terms of total cost, not elementwise.
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost& cwiseMax(
const TensorOpCost& rhs) {
bytes_loaded_ = numext::maxi(bytes_loaded_, rhs.bytes_loaded());
bytes_stored_ = numext::maxi(bytes_stored_, rhs.bytes_stored());
compute_cycles_ = numext::maxi(compute_cycles_, rhs.compute_cycles());
return *this;
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost cwiseMax(
const TensorOpCost& rhs) const {
double bytes_loaded = numext::maxi(bytes_loaded_, rhs.bytes_loaded());
double bytes_stored = numext::maxi(bytes_stored_, rhs.bytes_stored());
double compute_cycles = numext::maxi(compute_cycles_, rhs.compute_cycles());
return TensorOpCost(bytes_loaded, bytes_stored, compute_cycles);
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost& operator+=(