Hello
I’m getting a great support here on forum, but just a moment ago I encountered situation which is confusing so just started another thread.
Theoritically I should be able to run my kernel with 512 work items in workgroup. But it seems that when I specify local_work_size=512 for clEnqueueNDRangeKernel then no work is done and nearly 10 000 runs of my kernel are done surprisingly fast. When I decrease value of local_work_size to 256 kernel runs normally and performs computations.
This is part of device query for OpenCL:
---------------------------------
Device Quadro NVS 140M
---------------------------------
CL_DEVICE_NAME: Quadro NVS 140M
CL_DEVICE_VENDOR: NVIDIA Corporation
CL_DRIVER_VERSION: 260.99
CL_DEVICE_VERSION: OpenCL 1.0 CUDA
CL_DEVICE_TYPE: CL_DEVICE_TYPE_GPU
CL_DEVICE_MAX_COMPUTE_UNITS: 2
CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS: 3
CL_DEVICE_MAX_WORK_ITEM_SIZES: 512 / 512 / 64
CL_DEVICE_MAX_WORK_GROUP_SIZE: 512
CL_DEVICE_MAX_CLOCK_FREQUENCY: 800 MHz
CL_DEVICE_ADDRESS_BITS: 32
CL_DEVICE_MAX_MEM_ALLOC_SIZE: 128 MByte
CL_DEVICE_GLOBAL_MEM_SIZE: 113 MByte
CL_DEVICE_ERROR_CORRECTION_SUPPORT: no
CL_DEVICE_LOCAL_MEM_TYPE: local
CL_DEVICE_LOCAL_MEM_SIZE: 16 KByte
CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE: 64 KByte
CL_DEVICE_QUEUE_PROPERTIES: CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE
CL_DEVICE_QUEUE_PROPERTIES: CL_QUEUE_PROFILING_ENABLE
CL_DEVICE_IMAGE_SUPPORT: 1
CL_DEVICE_MAX_READ_IMAGE_ARGS: 128
CL_DEVICE_MAX_WRITE_IMAGE_ARGS: 8
CL_DEVICE_SINGLE_FP_CONFIG: INF-quietNaNs round-to-nearest round-to-zero round-to-inf fma
Is it possible that driver/OpenCL is just giving me wrong information here about maximum allowed number of work items in work group?