Check if the development branch which doesn't use doubles works with my GPU

[tarelli]

No description provided.

[tarelli]

@tarelli ping! Put this on the command line device=gpu. Let @skhayrulin know about outcome!

[wilzh40]

If I run with device=gpu, it breaks, on macbook air. It's uptodate with origin/development

Wz-innogen:Smoothed-Particle-Hydrodynamics wilsonzhao$ ./build/Smoothed-Particle-Hydrodynamics device=gpu
CL_PLATFORM_VERSION [0]: OpenCL 1.2 (Apr 25 2014 22:04:25)
CL_CONTEXT_PLATFORM [0]: CL_DEVICE_NAME [0]: HD Graphics 4000

CL_CONTEXT_PLATFORM [0]: CL_DEVICE_MAX_WORK_GROUP_SIZE [0]: 512
CL_CONTEXT_PLATFORM [0]: CL_DEVICE_MAX_COMPUTE_UNITS [0]: 16
CL_CONTEXT_PLATFORM [0]: CL_DEVICE_GLOBAL_MEM_SIZE [0]: 0
CL_CONTEXT_PLATFORM [0]: CL_DEVICE_GLOBAL_MEM_CACHE_SIZE [0]: 0
CL_CONTEXT_PLATFORM 0: CL_DEVICE_LOCAL_MEM_SIZE [0]: 65536
Compilation failed:
:96:6: error: "Double precision floating point not supported by OpenCL implementation."
#error "Double precision floating point not supported by OpenCL implementation."`

[skhayrulin]

Hi @wilzh40, you can try comment this lines it should turn off double precision extension. Please let me know if it works for you. On other devices on which I run Sibernetic didn't throw such error.

[wilzh40]

It works for me now, but same bad FPS. I'll try running this on my laptop later with an actual GPU.