BUG: matrix multiplication causes compiler error on Windows

Describe the issue:

I'm getting a compiler error when trying to use a dot product on a Windows install of PyTensor (see version information). The issue goes away when either removing blas__ldflags=-lblas or setting optimizer=None in the config but I'd rather not do that if possible. From what I can understand the error message seems to be related specifically to the Windows compiler.

Some notes:

• I'm not using the latest version of PyTensor, but this is the latest version I could get on conda-forge with m2w64-toolchain
• This uses a very old version of gcc (5.3.0 from 2015)

Reproducable code example:

import pytensor.tensor as pt
import pytensor

# Same problem with pt.dmatrix
a = pt.dvector("a")
b = pt.dvector("b")
c = a @ b

fn = pytensor.function([a, b], c)

print(fn([1, 2], [3, 4]))

Error message:

You can find the C code in this temporary file: C:\Users\TOMHAM~1\AppData\Local\Temp\pytensor_compilation_error_aq1rh79d
library ine>:0:0: is not found.
library blas is not found.
library lapack is not found.
Traceback (most recent call last):
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\vm.py", line 1230, in make_all
    node.op.make_thunk(node, storage_map, compute_map, [], impl=impl)
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\op.py", line 119, in make_thunk
    return self.make_c_thunk(node, storage_map, compute_map, no_recycling)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\op.py", line 84, in make_c_thunk
    outputs = cl.make_thunk(
              ^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\basic.py", line 1186, in make_thunk
    cthunk, module, in_storage, out_storage, error_storage = self.__compile__(
                                                             ^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\basic.py", line 1103, in __compile__
    thunk, module = self.cthunk_factory(
                    ^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\basic.py", line 1631, in cthunk_factory
    module = cache.module_from_key(key=key, lnk=self)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\cmodule.py", line 1250, in module_from_key
    module = lnk.compile_cmodule(location)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\basic.py", line 1532, in compile_cmodule
    module = c_compiler.compile_str(
             ^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\cmodule.py", line 2676, in compile_str
    raise CompileError(
pytensor.link.c.exceptions.CompileError: Compilation failed (return status=1):
"C:\Users\TomHammond\miniconda3\envs\test_env_4\Library\mingw-w64\bin\g++.EXE" -shared -g -O3 -fno-math-errno -Wno-unused-label -Wno-unused-variable -Wno-write-strings -DNPY_NO_DEPRECATED_API=NPY_1_7_API_VERSION -m64 -DMS_WIN64 -I"C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\numpy\core\include" -I"C:\Users\TomHammond\miniconda3\envs\test_env_4\include" -I"C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\c_code" -L"C:\Users\TomHammond\miniconda3\envs\test_env_4\libs" -L"C:\Users\TomHammond\miniconda3\envs\test_env_4" -o "C:\Users\TomHammond\AppData\Local\PyTensor\compiledir_Windows-11-10.0.26200-SP0-Intel64_Family_6_Model_186_Stepping_2_GenuineIntel-3.12.13-64\tmp8d_ea1n_\me74e37e9c42976fda662dc2a17d4af4617dce1e2c92b41a904b6ced701075905.pyd" "C:\Users\TomHammond\AppData\Local\PyTensor\compiledir_Windows-11-10.0.26200-SP0-Intel64_Family_6_Model_186_Stepping_2_GenuineIntel-3.12.13-64\tmp8d_ea1n_\mod.cpp" -lblas -llapack "C:\Users\TomHammond\miniconda3\envs\test_env_4\python312.dll"
In file included from C:\Users\TomHammond\miniconda3\envs\test_env_4\include/Python.h:12:0,
                 from C:\Users\TomHammond\AppData\Local\PyTensor\compiledir_Windows-11-10.0.26200-SP0-Intel64_Family_6_Model_186_Stepping_2_GenuineIntel-3.12.13-64\tmp8d_ea1n_\mod.cpp:1:
C:\Users\TomHammond\miniconda3\envs\test_env_4\include/pyconfig.h:238:0: warning: "MS_WIN64" redefined
 #define MS_WIN64
 ^
<command-line>:0:0: note: this is the location of the previous definition
C:/Users/TomHammond/miniconda3/envs/test_env_4/Library/mingw-w64/bin/../lib/gcc/x86_64-w64-mingw32/5.3.0/../../../../x86_64-w64-mingw32/bin/ld.exe: cannot find -lblas
C:/Users/TomHammond/miniconda3/envs/test_env_4/Library/mingw-w64/bin/../lib/gcc/x86_64-w64-mingw32/5.3.0/../../../../x86_64-w64-mingw32/bin/ld.exe: cannot find -llapack
collect2.exe: error: ld returned 1 exit status


During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:\Users\TomHammond\Documents\PROTAC notebook v0.11.0\ternary_algos\test.py", line 8, in <module>
    fn = pytensor.function([a, b], c)
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\compile\function\__init__.py", line 318, in function
    fn = pfunc(
         ^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\compile\function\pfunc.py", line 465, in pfunc
    return orig_function(
           ^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\compile\function\types.py", line 1769, in orig_function
    fn = m.create(defaults)
         ^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\compile\function\types.py", line 1661, in create
    _fn, _i, _o = self.linker.make_thunk(
                  ^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\basic.py", line 245, in make_thunk
    return self.make_all(
           ^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\vm.py", line 1239, in make_all
    raise_with_op(fgraph, node)
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\utils.py", line 526, in raise_with_op
    raise exc_value.with_traceback(exc_trace)
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\vm.py", line 1230, in make_all
    node.op.make_thunk(node, storage_map, compute_map, [], impl=impl)
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\op.py", line 119, in make_thunk
    return self.make_c_thunk(node, storage_map, compute_map, no_recycling)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\op.py", line 84, in make_c_thunk
    outputs = cl.make_thunk(
              ^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\basic.py", line 1186, in make_thunk
    cthunk, module, in_storage, out_storage, error_storage = self.__compile__(
                                                             ^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\basic.py", line 1103, in __compile__
    thunk, module = self.cthunk_factory(
                    ^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\basic.py", line 1631, in cthunk_factory
    module = cache.module_from_key(key=key, lnk=self)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\cmodule.py", line 1250, in module_from_key
    module = lnk.compile_cmodule(location)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\basic.py", line 1532, in compile_cmodule
    module = c_compiler.compile_str(
             ^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\cmodule.py", line 2676, in compile_str
    raise CompileError(
pytensor.link.c.exceptions.CompileError: Compilation failed (return status=1):
"C:\Users\TomHammond\miniconda3\envs\test_env_4\Library\mingw-w64\bin\g++.EXE" -shared -g -O3 -fno-math-errno -Wno-unused-label -Wno-unused-variable -Wno-write-strings -DNPY_NO_DEPRECATED_API=NPY_1_7_API_VERSION -m64 -DMS_WIN64 -I"C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\numpy\core\include" -I"C:\Users\TomHammond\miniconda3\envs\test_env_4\include" -I"C:\Users\TomHammond\miniconda3\envs\test_env_4\Lib\site-packages\pytensor\link\c\c_code" -L"C:\Users\TomHammond\miniconda3\envs\test_env_4\libs" -L"C:\Users\TomHammond\miniconda3\envs\test_env_4" -o "C:\Users\TomHammond\AppData\Local\PyTensor\compiledir_Windows-11-10.0.26200-SP0-Intel64_Family_6_Model_186_Stepping_2_GenuineIntel-3.12.13-64\tmp8d_ea1n_\me74e37e9c42976fda662dc2a17d4af4617dce1e2c92b41a904b6ced701075905.pyd" "C:\Users\TomHammond\AppData\Local\PyTensor\compiledir_Windows-11-10.0.26200-SP0-Intel64_Family_6_Model_186_Stepping_2_GenuineIntel-3.12.13-64\tmp8d_ea1n_\mod.cpp" -lblas -llapack "C:\Users\TomHammond\miniconda3\envs\test_env_4\python312.dll"
In file included from C:\Users\TomHammond\miniconda3\envs\test_env_4\include/Python.h:12:0,
                 from C:\Users\TomHammond\AppData\Local\PyTensor\compiledir_Windows-11-10.0.26200-SP0-Intel64_Family_6_Model_186_Stepping_2_GenuineIntel-3.12.13-64\tmp8d_ea1n_\mod.cpp:1:
C:\Users\TomHammond\miniconda3\envs\test_env_4\include/pyconfig.h:238:0: warning: "MS_WIN64" redefined
 #define MS_WIN64
 ^
<command-line>:0:0: note: this is the location of the previous definition
C:/Users/TomHammond/miniconda3/envs/test_env_4/Library/mingw-w64/bin/../lib/gcc/x86_64-w64-mingw32/5.3.0/../../../../x86_64-w64-mingw32/bin/ld.exe: cannot find -lblas
C:/Users/TomHammond/miniconda3/envs/test_env_4/Library/mingw-w64/bin/../lib/gcc/x86_64-w64-mingw32/5.3.0/../../../../x86_64-w64-mingw32/bin/ld.exe: cannot find -llapack
collect2.exe: error: ld returned 1 exit status

Apply node that caused the error: CGemv{inplace}(AllocEmpty{dtype='float64'}.0, 1.0, ExpandDims{axis=0}.0, a, 0.0)
Toposort index: 2
Inputs types: [TensorType(float64, shape=(1,)), TensorType(float64, shape=()), TensorType(float64, shape=(1, None)), TensorType(float64, shape=(None,)), TensorType(float64, shape=())]

HINT: Use a linker other than the C linker to print the inputs' shapes and strides.
HINT: Re-running with most PyTensor optimizations disabled could provide a back-trace showing when this node was created. This can be done by setting the PyTensor flag 'optimizer=fast_compile'. If that does not work, PyTensor optimizations can be disabled with 'optimizer=None'.
HINT: Use the PyTensor flag `exception_verbosity=high` for a debug print-out and storage map footprint of this Apply node.

PyTensor version information:

On Windows 11 install using conda:
conda config --set channel_priority strict
conda config --prepend channels conda-forge
conda create -n test_env pytensor m2w64-toolchain

then fix a problem locating g++ by commenting out a line as described here: https://github.com/pymc-devs/pytensor/issues/234

conda list dump:

Context for the issue:

Found when rewriting/updating a program from PyMC3 with lots of other dependencies so it would be useful to have a fix that works on Windows/conda

Reproduce the dot-product example with the Windows conda environment described in the issue, then inspect pytensor/link/c/basic.py and pytensor/link/c/cmodule.py around C compilation and the reported -lblas/-llapack linker arguments. Compare the behavior with and without blas__ldflags; done means the example compiles and evaluates successfully without requiring those workarounds.