How to use slangtorch in a coda environment?

[oliver-batchelor]

I'm trying with python 3.11, torch 2.5.1 from conda/mamba. slangtorch 1.3.0 from pypi (otherwise I run into this as well #23)

If I use the system compilers I get this kind of error (both GCC and clang)

/local/slang-torch/examples/inline-mlp-example/.slangtorch_cache/image-model/39a1e66c93673ceb/image-model.cpp:493:9: error: unknown type name 'SlangUInt'
typedef SlangUInt UInt;
        ^
/local/slang-torch/examples/inline-mlp-example/.slangtorch_cache/image-model/39a1e66c93673ceb/image-model.cpp:494:9: error: unknown type name 'SlangInt'
typedef SlangInt Int;

If I use compilers installed from mamba (GCC or clang again) I get this error:

/local/slang-torch/examples/inline-mlp-example/.slangtorch_cache/image-model/39a1e66c93673ceb/image-model.cpp:434:9: warning: #pragma once in main file [-Wpragma-once-outside-header]
  434 | #pragma once
      |         ^
/local/slang-torch/examples/inline-mlp-example/.slangtorch_cache/image-model/39a1e66c93673ceb/image-model.cpp:476:10: fatal error: 'slang.h' file not found
  476 | #include "slang.h"

and in both cases a lot of warnings like this:

/local/slang-torch/examples/inline-mlp-example/.slangtorch_cache/image-model/39a1e66c93673ceb/image-model_cuda.cu(1203): warning #177-D: function "surfCubemapLayeredwrite(T, cudaSurfaceObject_t, int, int, int, cudaSurfaceBoundaryMode) [with T=__half4]" was declared but never referenced

[oliver-batchelor]

Basic workaround here: #10

[saipraveenb25]

@oliver-batchelor: thank you for raising this issue!

Could you try with a fresh environment that uses the latest slangtorch release (v1.3.2)?
We fixed the "slang.h" regression there.

[oliver-batchelor]

@oliver-batchelor: thank you for raising this issue!

Could you try with a fresh environment that uses the latest slangtorch release (v1.3.2)? We fixed the "slang.h" regression there.

Can confirm this works!
Thanks!