ENH: Add support for 16 and 32-bit radiograph images

[NicerNewerCar]

• Replace outdated texture references with texture objects (added in CUDA 5.0) for CUDA 12.0
• Closes issues
  • Cuda: Convert texture references to texture objects #42
  • Nvdia 12.0 Build error #72
  • SAM (cuda) crash with mayo shoulder data cfg load #241
• For future reference see https://developer.nvidia.com/blog/cuda-pro-tip-kepler-texture-objects-improve-performance-and-flexibility/

[amymmorton]

Successful build with 12.0, but when opening trial - autoscoper crashed:

C:\Users\amorton1\Documents\GitHub\Autoscoper\libautoscoper\src\View.cpp(430) : cudaSafeCall() Runtime API error : OS call failed or operation not supported on this OS.

[NicerNewerCar]

Successful build with 12.0, but when opening trial - autoscoper crashed:

@amymmorton what trial are you opening? Did you make sure to build with a clean build tree? I have opened and tracked WN00106 and SOL001A successfully on this branch.

[amymmorton]

Yes. tried again - changed to main then back to PR 76:
Release configuration
same error on trial open:

Filename: P:/iTWA_Instrumented_Total_Wrist/Subjects/NORMAL/WN00106/Autoscoper/flx_ext.cfg
Volume Name: P:\iTWA_Instrumented_Total_Wrist\Subjects\NORMAL\WN00106\Models\mc2_mc3\mc2_mc3_dcm_cropped.tif
Volume Name: P:\iTWA_Instrumented_Total_Wrist\Subjects\NORMAL\WN00106\Models\rad\rad_dcm_cropped.tif
Volume Name: P:\iTWA_Instrumented_Total_Wrist\Subjects\NORMAL\WN00106\Models\mc3\mc3_dcm_cropped.tif
Volume Name: P:\iTWA_Instrumented_Total_Wrist\Subjects\NORMAL\WN00106\Models\uln\uln_dcm_cropped.tif
Volume Name: P:\iTWA_Instrumented_Total_Wrist\Subjects\NORMAL\WN00106\Models\mc3_mc4\mc3_mc4_dcm_cropped.tif
C:\Users\amorton1\Documents\GitHub\Autoscoper\libautoscoper\src\View.cpp(430) : cudaSafeCall() Runtime API error : OS call failed or operation not supported on this OS.

[amymmorton]

I tried uninstalling v12.0 (revert to 11.8) and the same error on trial open persists:

C:\Users\amorton1\Documents\GitHub\Autoscoper\build_PR76_11p8\Autoscoper-build\install\bin\Release>autoscoper.exe
Autoscoper v2.8

Setting to Autoscoper_RANDOM_SEED to 100
Graphics Card Vendor: Intel
Intel(R) UHD Graphics 630
4.6.0 - Build 30.0.101.1692
Initializing OpenGL...
Graphics Card Vendor: Intel
Intel(R) UHD Graphics 630
4.6.0 - Build 30.0.101.1692
Initializing OpenGL...

Setting up a new trial...
Filename: P:/iTWA_Instrumented_Total_Wrist/Subjects/NORMAL/WN00106/Autoscoper/flx_ext.cfg
Volume Name: P:\iTWA_Instrumented_Total_Wrist\Subjects\NORMAL\WN00106\Models\mc2_mc3\mc2_mc3_dcm_cropped.tif
Volume Name: P:\iTWA_Instrumented_Total_Wrist\Subjects\NORMAL\WN00106\Models\rad\rad_dcm_cropped.tif
Volume Name: P:\iTWA_Instrumented_Total_Wrist\Subjects\NORMAL\WN00106\Models\mc3\mc3_dcm_cropped.tif
Volume Name: P:\iTWA_Instrumented_Total_Wrist\Subjects\NORMAL\WN00106\Models\uln\uln_dcm_cropped.tif
Volume Name: P:\iTWA_Instrumented_Total_Wrist\Subjects\NORMAL\WN00106\Models\mc3_mc4\mc3_mc4_dcm_cropped.tif
C:\Users\amorton1\Documents\GitHub\Autoscoper\libautoscoper\src\View.cpp(430) : cudaSafeCall() Runtime API error : OS call failed or operation not supported on this OS.

[amymmorton]

pulled from main and had the same error- I believe the instll of v12 created issues in my enviroment- attempting a reinstall of 11.8..

[amymmorton]

My reinstall of 11.8 allowed successful build of this branch. I will have our engineer Maddie and @kaitohl test v12.0 too

[NicerNewerCar]

@jcfr I got these changes to build and run successfully using the docker container on Ubuntu 22.04. Have you been able to test them yet? If so, this can be merged in.

[amymmorton]

Awesome- confirmed- mayo shoulder data loads and tracks:

[jcfr]

Suggest commit message:

ENH: Add support for 16 and 32-bit radiograph images

Prior to these changes, the supported CUDA texture types (and by extension
radiograph images) was hard-coded to 8-bit (unsigned char) and managed as
a static global variable.
  
Since texture references needed to be known at compile time and the current
type was hard-coded to 8-bit, the use of them prevented from being able to
easily support 8, 16, and 32 bit radiograph images known at runtime. 

To address this, this commit removes the use of `cudaBindTextureToArray`
(deprecated since CUDA 5.0 and removed in CUDA 12.x) and switch to using
texture object instead by introducing a convenience helper function
`createTextureObjectFromArray`.

Approach implemented here is based of recommendation the post titled
"Kepler Texture Objects Improve Performance and Flexibility" published
in 2013.
See https://developer.nvidia.com/blog/cuda-pro-tip-kepler-texture-objects-improve-performance-and-flexibility/

For reference, before this changes, the error message reported at build time
while attempting to use autoscoper built against CUDA 12.x was:

```
3 Errors under file path `...\libautoscoper\src\gpu\cuda\RadRenderer_kernels.cu:
- no instance of overloaded function "tex2D" matches the argument list 
- identifier "cudaBindTextureToArray" is undefined
- texture is not a template
```

Can you revisit to:

• properly reference the version of CUDA in which deprecation were introduced and feature removed. Reading Cuda: Convert texture references to texture objects #42 (comment), I am not sure.
• make sure the error message referenced is valid

[jcfr]

Should the outcome of the division be truncated to unsigned int ? The use of parenthesis seems off.

Also, could you confirm truncation is what we want here ?

See https://stackoverflow.com/questions/65225888/difference-among-static-cast-int-and-float2int-rn-in-cuda

[NicerNewerCar]

This is not performing truncation on the division, but converting the size_t width and height variables to prevent a compile time warning (see #42)

C:\SAM\Autoscoper\libautoscoper\src\gpu\cuda\Compositor_kernels.cu(70): warning C4267: 'argument': conversion from 'size_t' to 'unsigned int', possible loss of data [C:\SAM\Autoscoper\build-test\Autoscoper-build\libautoscoper\libautoscoper.vcxproj] [C:\SAM\Autoscoper\build-test\Autoscoper.vcxproj]