Add tensor minimum/maximum ops

[yuecideng]

Add two operations Maximum, Minimum in tensor function.

This change is 

[update-docs]

Thanks for submitting this pull request! The maintainers of this repository would appreciate if you could update the CHANGELOG.md based on your changes.

[reyanshsolis]

at least for one case, we must also verify behavior over different dtypes such as int, uint, and bool. You may add this after the Special case.

[yuecideng]

Should we make bool dtype supported? I think LogicalAnd and LogicalOr are equivalent to Minimum and Maximum for bool respectively. @reyanshsolis

[yxlao]

Reviewed 5 of 7 files at r1, 1 of 2 files at r2, all commit messages.
Reviewable status: 6 of 7 files reviewed, 4 unresolved discussions (waiting on @reyanshsolis, @yuecideng, and @yxlao)

---

cpp/open3d/core/kernel/BinaryEW.h line 77 at r2 (raw file):

#endif

inline void Add(const Tensor& lhs, const Tensor& rhs, Tensor& dst) {

Clean up previous code, and remove all of these.

---

cpp/open3d/core/kernel/BinaryEWCUDA.cu line 61 at r2 (raw file):

    const scalar_t* rhs_ptr = static_cast<const scalar_t*>(rhs);
    if (*lhs_ptr > *rhs_ptr) {
        *static_cast<scalar_t*>(dst) = *lhs_ptr;

Can we try:
https://docs.nvidia.com/cuda/cuda-math-api/group__CUDA__MATH__INT.html#group__CUDA__MATH__INT

Code quote:

OPEN3D_HOST_DEVICE

---

cpp/pybind/core/tensor_function.cpp line 109 at r2 (raw file):

          R"(Computes the element-wise maximum of input and other. The tensors 
must have same data type and device.
          

nit: remove space

---

cpp/tests/core/TensorFunction.cpp line 268 at r2 (raw file):

Previously, yuecideng (Yueci Deng) wrote…

Should we make bool dtype supported? I think LogicalAnd and LogicalOr are equivalent to Minimum and Maximum for bool respectively. @reyanshsolis

Try directly use dispatch template with bool and see if result is correct

[yuecideng]

Reviewable status: 1 of 8 files reviewed, 3 unresolved discussions (waiting on @reyanshsolis and @yxlao)

---

cpp/open3d/core/kernel/BinaryEW.h line 77 at r2 (raw file):

Previously, yxlao (Yixing Lao) wrote…

Clean up previous code, and remove all of these.

Done.

---

cpp/open3d/core/kernel/BinaryEWCUDA.cu line 61 at r2 (raw file):

Previously, yxlao (Yixing Lao) wrote…

Can we try:
https://docs.nvidia.com/cuda/cuda-math-api/group__CUDA__MATH__INT.html#group__CUDA__MATH__INT

Done.

---

cpp/pybind/core/tensor_function.cpp line 109 at r2 (raw file):

Previously, yxlao (Yixing Lao) wrote…

nit: remove space

Done.

---

cpp/tests/core/TensorFunction.cpp line 268 at r2 (raw file):

Previously, yxlao (Yixing Lao) wrote…

Try directly use dispatch template with bool and see if result is correct

Use dispatch template with bool for Max and Min conditions. Also add more test cases for these two functions.

[yxlao]

LGTM after CI

Reviewed 5 of 7 files at r3, 5 of 5 files at r4, all commit messages.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @reyanshsolis)