[pyc] use pyc file instead py file is not working #2217

WangWei90 · 2021-03-17T10:39:31Z

Describe the bug
Thank you for creating taichi language. I want to know how to use pyc files in computation.

To Reproduce
Consider 3 files below.

# hello.py
from core import *

print(add(1,2))

print(six_times_big(10))

# core.py
import taichi as ti


def add(a, b):
    return a+b


@ti.func
def triple(a):
    return a * 3


@ti.kernel
def six_times_big(a:float) -> float:
    temp = triple(a)
    return temp * 2

# compile.py
import py_compile
for i in ["hello.py", "core.py"]:
    py_compile.compile(i)

Use python3 compile.py and rename files in __pycache__ to hello.pyc and core.pyc then run python3 hello.pyc, this will work if using module numpy but not module taichi.

Log/Screenshots
Please post the full log of the program (instead of just a few lines around the error message, unless the log is > 1000 lines). This will help us diagnose what's happening. For example:

$ python my_sample_code.py
[Taichi] mode=release
[Taichi] preparing sandbox at /var/folders/25/2mxhd1b50sbc1_njhz_93jn80000gn/T/taichi-fcvyqhf8
[Taichi] version 0.7.15, llvm 10.0.0, commit cff542ce, osx, python 3.8.8
3
[Taichi] materializing...
Traceback (most recent call last):
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 150, in our_findsource
    return inspect._si_old_findsource(object)
  File "/usr/local/Cellar/python@3.8/3.8.8_1/Frameworks/Python.framework/Versions/3.8/lib/python3.8/inspect.py", line 798, in findsource
    raise OSError('could not get source code')
OSError: could not get source code

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 96, in blender_findsourcetext
    import bpy
ModuleNotFoundError: No module named 'bpy'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 153, in our_findsource
    return blender_findsource(object)
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 110, in blender_findsource
    lines, text_name = blender_findsourcetext(object)
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 98, in blender_findsourcetext
    raise IOError('Not in Blender environment!')
OSError: Not in Blender environment!

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 38, in remote_findsource
    assert module[0] == '<' and module[-1] == '>' or module in ['__console__', '__main__']
AssertionError

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 156, in our_findsource
    return remote_findsource(object)
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 40, in remote_findsource
    raise IOError(f'Object `{repr(object)}` does not come from console')
OSError: Object `<function six_times_big at 0x1329a3310>` does not come from console

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 159, in our_findsource
    return dill_findsource(object)
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 145, in dill_findsource
    return dill.source.findsource(object)
  File "/usr/local/lib/python3.8/site-packages/dill/source.py", line 154, in findsource
    raise IOError('could not extract source code')
OSError: could not extract source code

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "hello.py", line 5, in <module>
  File "/usr/local/lib/python3.8/site-packages/taichi/lang/kernel.py", line 571, in wrapped
    return primal(*args, **kwargs)
  File "/usr/local/lib/python3.8/site-packages/taichi/lang/kernel.py", line 499, in __call__
    self.materialize(key=key, args=args, arg_features=arg_features)
  File "/usr/local/lib/python3.8/site-packages/taichi/lang/kernel.py", line 311, in materialize
    src = remove_indent(oinspect.getsource(self.func))
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 185, in getsource
    return inspect.getsource(object)
  File "/usr/local/Cellar/python@3.8/3.8.8_1/Frameworks/Python.framework/Versions/3.8/lib/python3.8/inspect.py", line 997, in getsource
    lines, lnum = getsourcelines(object)
  File "/usr/local/Cellar/python@3.8/3.8.8_1/Frameworks/Python.framework/Versions/3.8/lib/python3.8/inspect.py", line 979, in getsourcelines
    lines, lnum = findsource(object)
  File "/usr/local/lib/python3.8/site-packages/sourceinspect/__init__.py", line 161, in our_findsource
    raise IOError(f'Could not get source code for object: {repr(object)}')
OSError: Could not get source code for object: <function six_times_big at 0x1329a3310>

Additional comments
If possible, please also consider attaching the output of command ti diagnose. This produces the detailed environment information and hopefully helps us diagnose faster.

[Taichi] mode=release
[Taichi] preparing sandbox at /var/folders/25/2mxhd1b50sbc1_njhz_93jn80000gn/T/taichi-0hhdclqp
[Taichi] version 0.7.15, llvm 10.0.0, commit cff542ce, osx, python 3.8.8

*******************************************
**      Taichi Programming Language      **
*******************************************

Docs:   https://taichi.rtfd.io/zh_CN/latest
GitHub: https://github.com/taichi-dev/taichi
Forum:  https://forum.taichi.graphics

Taichi system diagnose:

python: 3.8.8 (default, Feb 27 2021, 02:19:17) 
[Clang 12.0.0 (clang-1200.0.32.29)]
system: darwin
executable: /usr/local/opt/python@3.8/bin/python3.8
platform: macOS-11.2.3-x86_64-i386-64bit
architecture: 64bit 
uname: uname_result(system='Darwin', node='appledeMacBook-Pro.local', release='20.3.0', version='Darwin Kernel Version 20.3.0: Thu Jan 21 00:07:06 PST 2021; root:xnu-7195.81.3~1/RELEASE_X86_64', machine='x86_64', processor='i386')
locale: zh_CN.UTF-8
PATH: /Users/apple/Documents/1_gery/2_tool_by_gery:/usr/local/texlive/2020/bin/x86_64-darwin:/Users/apple/.nimble/bin:/Users/apple/.local/bin:/usr/local/opt/coreutils/libexec/gnubin:/Users/apple/bin:/usr/local/bin:/usr/local/bin:/usr/bin:/bin:/usr/sbin:/sbin
PYTHONPATH: ['/usr/local/bin', '/Users/apple/Documents/1_gery/2_tool_by_gery', '/usr/local/Cellar/python@3.8/3.8.8_1/Frameworks/Python.framework/Versions/3.8/lib/python38.zip', '/usr/local/Cellar/python@3.8/3.8.8_1/Frameworks/Python.framework/Versions/3.8/lib/python3.8', '/usr/local/Cellar/python@3.8/3.8.8_1/Frameworks/Python.framework/Versions/3.8/lib/python3.8/lib-dynload', '/usr/local/lib/python3.8/site-packages', '/Users/apple/Documents/1_gery/1_coding_zone/3_python_价值投资_极其仰赖/09stackprinter-master', '/usr/local/lib/python3.8/site-packages/taichi/core/../lib']

`lsb_release` not available: [Errno 2] No such file or directory: 'lsb_release'
TAICHI_REPO_DIR=


import: <module 'taichi' from '/usr/local/lib/python3.8/site-packages/taichi/__init__.py'>

cc: False
cpu: True
metal: True
opengl: False
cuda: False

`glewinfo` not available: [Errno 2] No such file or directory: 'glewinfo'

`nvidia-smi` not available: [Errno 2] No such file or directory: 'nvidia-smi'
[Taichi] mode=release
[Taichi] preparing sandbox at /var/folders/25/2mxhd1b50sbc1_njhz_93jn80000gn/T/taichi-6v_tgwo0
[Taichi] version 0.7.15, llvm 10.0.0, commit cff542ce, osx, python 3.8.8

[Taichi] mode=release
[Taichi] preparing sandbox at /var/folders/25/2mxhd1b50sbc1_njhz_93jn80000gn/T/taichi-3hhj9d5z
[Taichi] version 0.7.15, llvm 10.0.0, commit cff542ce, osx, python 3.8.8
[Taichi] Starting on arch=x64

[W 03/17/21 18:38:50.703] [__init__.py:adaptive_arch_select@589] Arch=[<Arch.opengl: 6>] is not supported, falling back to CPU
[Taichi] mode=release
[Taichi] preparing sandbox at /var/folders/25/2mxhd1b50sbc1_njhz_93jn80000gn/T/taichi-7p8g2sww
[Taichi] version 0.7.15, llvm 10.0.0, commit cff542ce, osx, python 3.8.8
[Taichi] Starting on arch=x64

[W 03/17/21 18:38:52.431] [__init__.py:adaptive_arch_select@589] Arch=[<Arch.cuda: 4>] is not supported, falling back to CPU
[Taichi] mode=release
[Taichi] preparing sandbox at /var/folders/25/2mxhd1b50sbc1_njhz_93jn80000gn/T/taichi-6u47384s
[Taichi] version 0.7.15, llvm 10.0.0, commit cff542ce, osx, python 3.8.8
[Taichi] Starting on arch=x64

[Taichi] mode=release
[Taichi] preparing sandbox at /var/folders/25/2mxhd1b50sbc1_njhz_93jn80000gn/T/taichi-f7hachrp
[Taichi] version 0.7.15, llvm 10.0.0, commit cff542ce, osx, python 3.8.8

*******************************************
**      Taichi Programming Language      **
*******************************************

Docs:   https://taichi.rtfd.io/zh_CN/latest
GitHub: https://github.com/taichi-dev/taichi
Forum:  https://forum.taichi.graphics

Running example minimal ...
[Taichi] Starting on arch=x64
[Taichi] materializing...
>>> Running time: 0.28s
42

Consider attaching this log when maintainers ask about system information.
>>> Running time: 19.70s

The text was updated successfully, but these errors were encountered:

WangWei90 · 2021-03-17T10:42:44Z

Maybe this issue is related, I do not know.

#1186

I just begin to use taichi, maybe there are some options to find out.

archibate · 2021-03-17T11:01:48Z

Hi, Wang: Taichi doesn't support to be run from .pyc at this moment due to the fact that it needs to get the source of functions to work properly. In fact, Taichi uses the `inspect` module to extract the source of function @ti.kernel is decorated, then modify the source using `ast` module in order to translate Python syntax into Taichi API calls, then re-compile the modifed source into binary code (pyc) and run it. NOTE: It might be possible to support Taichi in .pyc, that is to store the 'modifed' code object, but not sure it worth the complicity to do this change.. 无法顺畅的大口呼吸，是活着的最好证明

…

---Original--- From: "Walker ***@***.***> Date: Wed, Mar 17, 2021 18:43 PM To: ***@***.***>; Cc: ***@***.***>; Subject: Re: [taichi-dev/taichi] [pyc] use pyc file instead py file is not working (#2217) Maybe this issue is related, I do not know. #1186 I just begin to use taichi, maybe there are some options to find out. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.

WangWei90 · 2021-03-17T11:15:06Z

Thank you for the answer!

Can you show me a minimal process to do that? I want to evaluate the complexity.

WangWei90 · 2021-03-17T11:18:24Z

In fact, Taichi uses the inspect module to extract the source of function @ti.kernel is decorated, then modify the source using ast module in order to translate Python syntax into Taichi API calls, then re-compile the modifed source into binary code (pyc) and run it.

The print and range inside the taichi scope make me confusing at begin.
If it is possible to alias them to ti.print and ti.range (What should I do).
Maybe I will find more such needing in the future.

archibate · 2021-03-17T11:22:47Z

Thank for your passion! You may need to modify this file: https://github.com/taichi-dev/taichi/blob/python/taichi/lang/kernel.py Then save / restore the code object it generated to / from a .pyc file. See also https://taichi.rtfd.io/en/stable/dev_install.html for setting up a develop environment for Taichi. 无法顺畅的大口呼吸，是活着的最好证明

…

---Original--- From: "Walker ***@***.***> Date: Wed, Mar 17, 2021 19:15 PM To: ***@***.***>; Cc: ***@***.******@***.***>; Subject: Re: [taichi-dev/taichi] [pyc] use pyc file instead py file is not working (#2217) Thank you for the answer! Can you show me a minimal process to do that? I want to evaluate the complexity. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

k-ye · 2021-03-17T11:34:22Z

Hi,

The entire python AST transformation is done in https://github.com/taichi-dev/taichi/blob/master/python/taichi/lang/transformer.py. Specifically for your questions:

print:

taichi/python/taichi/lang/transformer.py

Lines 630 to 631 in 588fc8f

if func_name == 'print':

node.func = self.parse_expr('ti.ti_print')

range for:

taichi/python/taichi/lang/transformer.py

Lines 564 to 566 in 588fc8f

    
           elif isinstance(node.iter, ast.Call) and isinstance( 
        
                   node.iter.func, ast.Name) and node.iter.func.id == 'range': 
        
               return self.visit_range_for(node)

As @archibate has pointed out, things that are inside @ti.kernel is completely cut off from Python, and the complexity to make it work in .pyc could be very high. Instead, we should consider AOT compiling the kernels (the issue you have pointed out). Hopefully #2193 will get us closer to AOT a bit :) We are also finding ways to parallelize the JIT compilation to speed up the launch, if that's what motivated you in the first place.

archibate · 2021-03-17T12:04:45Z

> ti.print ti.range Exactly! That's how Taichi works. Please try appending this after `import taichi`: ```py ti.init(print_preprocessed=True) ``` and running it (as .py) will print the transformed source code (including the `ti.ti_print` you've pointed out). Now if we could store the transformed code into .pyc, Taichi would likely to work without requiring inspect the source at runtime! 无法顺畅的大口呼吸，是活着的最好证明

…

---Original--- From: "Walker ***@***.***> Date: Wed, Mar 17, 2021 19:18 PM To: ***@***.***>; Cc: ***@***.******@***.***>; Subject: Re: [taichi-dev/taichi] [pyc] use pyc file instead py file is not working (#2217) In fact, Taichi uses the inspect module to extract the source of function @ti.kernel is decorated, then modify the source using ast module in order to translate Python syntax into Taichi API calls, then re-compile the modifed source into binary code (pyc) and run it. The print and range inside the taichi scope make me confusing at begin. If it is possible to alias them to ti.print and ti.range (What should I do). Maybe I will find more such needing in the future. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

WangWei90 · 2021-03-17T12:25:03Z

Initial AST:
def six_times_big(a: float) ->float:
    temp = triple(a)
    return temp * 2

Preprocessed:
def six_times_big():
    import taichi as ti
    ti.decl_scalar_ret(float)
    a = ti.decl_scalar_arg(float)
    temp = ti.expr_init(triple(a))
    ti.core.create_kernel_return(ti.cast(ti.Expr(temp * 2), float).ptr)

Checked:
def six_times_big():
    import taichi as ti
    ti.decl_scalar_ret(float)
    a = ti.decl_scalar_arg(float)
    temp = ti.expr_init(triple(a))
    ti.core.create_kernel_return(ti.cast(ti.Expr(temp * 2), float).ptr)

Final AST:
def six_times_big():
    import taichi as ti
    ti.decl_scalar_ret(float)
    a = ti.decl_scalar_arg(float)
    temp = ti.expr_init(triple(a))
    ti.core.create_kernel_return(ti.cast(ti.Expr(temp * 2), float).ptr)

I think the transform has lost the argument?

k-ye · 2021-03-17T14:39:57Z

I think the transform has lost the argument?

Hi, this is actually intentional. The python source code inside @ti.kernel is only used to generate a series of Taichi's own instructions, which are then JIT compiled to backends like CPU/CUDA. When you call a @ti.kernel function, it's not the original Python function that's been executed. Instead, the decorator finds the JIT'ed function (in the C++ side) and forwards the args to that function. See

taichi/python/taichi/lang/kernel.py

Lines 494 to 500 in 588fc8f

    
           def __call__(self, *args, **kwargs): 
        
               _taichi_skip_traceback = 1 
        
               assert len(kwargs) == 0, 'kwargs not supported for Taichi kernels' 
        
               instance_id, arg_features = self.mapper.lookup(args) 
        
               key = (self.func, instance_id) 
        
               self.materialize(key=key, args=args, arg_features=arg_features) 
        
               return self.compiled_functions[key](*args)

WangWei90 added the potential bug Something that looks like a bug but not yet confirmed label Mar 17, 2021

k-ye mentioned this issue Mar 18, 2021

[stacktrace] with stackprinter #2216

Closed

bobcao3 closed this as completed Dec 13, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[pyc] use pyc file instead py file is not working #2217

[pyc] use pyc file instead py file is not working #2217

WangWei90 commented Mar 17, 2021

WangWei90 commented Mar 17, 2021

archibate commented Mar 17, 2021 via email

WangWei90 commented Mar 17, 2021

WangWei90 commented Mar 17, 2021

archibate commented Mar 17, 2021 via email

k-ye commented Mar 17, 2021

archibate commented Mar 17, 2021 via email

WangWei90 commented Mar 17, 2021

k-ye commented Mar 17, 2021

[pyc] use pyc file instead py file is not working #2217

[pyc] use pyc file instead py file is not working #2217

Comments

WangWei90 commented Mar 17, 2021

WangWei90 commented Mar 17, 2021

archibate commented Mar 17, 2021 via email

WangWei90 commented Mar 17, 2021

WangWei90 commented Mar 17, 2021

archibate commented Mar 17, 2021 via email

k-ye commented Mar 17, 2021

archibate commented Mar 17, 2021 via email

WangWei90 commented Mar 17, 2021

k-ye commented Mar 17, 2021