Implementing hyperbolic trigonometric, exponential and rounding functions in NumPy

[czgdp1807]

Below I have provided some links to get started and then explained what you both are supposed to do,

Links to get started

1. https://github.com/lcompilers/lpython/blob/main/src/runtime/lpython_intrinsic_numpy.py - Contains definitions of sin, cos, tan, log, log2, log10, sqrt for f32 and f64 types. Some points to be noted - i) @vectorize decorator automatically vectorizes the function. Now by "vectorizes the function" I mean even if the functions are defined for single number input and single number outputs, you can call them upon are array inputs as well.
2. https://github.com/lcompilers/lpython/blob/main/integration_tests/elemental_01.py, https://github.com/lcompilers/lpython/blob/main/integration_tests/elemental_04.py - These are the test files for NumPy functions. You can see how sin, cos, log are called with array inputs even though they were defined for single input and single output. @vectorize does the job for you. You can also notice how the functions are tested. You can make the tests as natural/complicated as you want to. For example for testing sin, cos simultaneously, you can just verify sin(array)**2 + cos(array)**2 == 1. Similarly for trigonometric function, you can take a few trigonometric identities and try verifying them in the tests (with 1d, 2d, 5d, 6d, etc arrays as inputs). In fact you can also try finding the derivative of functions using arrays. For example for a function f, you can do (f(array + harray) - f(array))/harray and match with the formula you know for derivative of f you know. Bottomline is you can use your creativity to the full extent to write interesting test cases. In case you get errors don't work around the test case, instead let me know and I will fix the error for you.

Overall this task is easy-medium. Easy because extending src/runtime/lpython_intrinsic_numpy.py is fairly easy (just follow the already existing pattern) and medium because writing test cases requires some creativity and effort.

What you have to do?

• sinh - Sinh & cosh numpy_intrinsic functions implemented #1002
• cosh - Sinh & cosh numpy_intrinsic functions implemented #1002
• tanh - Implement tanh and exp in Numpy_intrinsic library #1064
• exp - Implement tanh and exp in Numpy_intrinsic library #1064
• arcsinh
• arccosh
• arctanh
• fix
• floor
• ceil
• trunc

If you complete the above tasks then please let me know. Feel free to reach out if you have any questions. :-)

[czgdp1807]

@Madhav2310 Assigning this to you (I cannot assign non-members of lcompilers Github organisation so consider this as the assignment)

[Madhav2310]

These

Below I have provided some links to get started and then explained what you both are supposed to do,

Links to get started

1. https://github.com/lcompilers/lpython/blob/main/src/runtime/lpython_intrinsic_numpy.py - Contains definitions of sin, cos, tan, log, log2, log10, sqrt for f32 and f64 types. Some points to be noted - i) @vectorize decorator automatically vectorizes the function. Now by "vectorizes the function" I mean even if the functions are defined for single number input and single number outputs, you can call them upon are array inputs as well.
2. https://github.com/lcompilers/lpython/blob/main/integration_tests/elemental_01.py, https://github.com/lcompilers/lpython/blob/main/integration_tests/elemental_04.py - These are the test files for NumPy functions. You can see how sin, cos, log are called with array inputs even though they were defined for single input and single output. @vectorize does the job for you. You can also notice how the functions are tested. You can make the tests as natural/complicated as you want to. For example for testing sin, cos simultaneously, you can just verify sin(array)**2 + cos(array)**2 == 1. Similarly for trigonometric function, you can take a few trigonometric identities and try verifying them in the tests (with 1d, 2d, 5d, 6d, etc arrays as inputs). In fact you can also try finding the derivative of functions using arrays. For example for a function f, you can do (f(array + harray) - f(array))/harray and match with the formula you know for derivative of f you know. Bottomline is you can use your creativity to the full extent to write interesting test cases. In case you get errors don't work around the test case, instead let me know and I will fix the error for you.

Overall this task is easy-medium. Easy because extending src/runtime/lpython_intrinsic_numpy.py is fairly easy (just follow the already existing pattern) and medium because writing test cases requires some creativity and effort.

What you have to do?

• sinh
• cosh
• tanh
• exp

If you complete the above tasks then please let me know. Feel free to reach out if you have any questions. :-)

These functions have been successfully implemented in #1064 and #1002

[khushi-411]

Hi, @czgdp1807! I'd like to work on these functions and get familiar with the codebase. The Lpython project looks interesting to me! Just wondering if I could pick the issue. Thanks :-)

[certik]

@khushi-411 you don't need to ask, just do it.

Many of these are already implemented, but I am sure many are still missing.

[khushi-411]

Hi, @certik @czgdp1807!
I think this issue could be closed as all the functions mentioned here are completed. Link
Just wanted to share. Thanks!

[certik]

Excellent, great job!

[khushi-411]

Thanks for all your helpful suggestions, @certik!