Fixed coverity issue in acl_device.cpp

[haoxian2]

Fixed coverity issue in acl_device.cpp: Type: Argument cannot be negative (NEGATIVE_RETURNS)

acl_get_default_device_global_memory can return -1 if no device memory is found. By static_casting it to size_t, it makes the -1 into something irrationally big. Therefore, after returning from the function, I added a check to see if gmem_id is negative. If it is negative, then I return 0. I assumed that no device private memory means that by default the sizes of global memory, of max constant buffer, and of max device memory allocation are all 0.

[haoxian2]

I get this weird git warning whenever I pan over to Files changed tab:

Check warning on line 1 in src/acl_device.cpp

GitHub Actions
/ build

Coverage

82.3% (-1.1%)

I'm not sure if I'm supposed to do anything about it.

[zibaiwan]

I get this weird git warning whenever I pan over to Files changed tab:

Check warning on line 1 in src/acl_device.cpp

GitHub Actions
/ build

Coverage

82.3% (-1.1%)

I'm not sure if I'm supposed to do anything about it.

Thanks Hao. I don't see anything when I clicked on the Files changed tab. Maybe it's telling you that the "code coverage" workflow is running?

[haoxian2]

This is the log that I get from coverage status in build log:

Warning: 82.3% (-1.1%)
File Name                         	Coverage (%) 	Delta Coverage      
---------                         	------------ 	--------------      
./src/acl_auto_configure.cpp      	92.82        	0.0                 
./src/acl_bsp_io.cpp              	0.0          	0.0                 
./src/acl_command.cpp             	82.5         	0.0                 
./src/acl_command_queue.cpp       	84.82        	0.0                 
./src/acl_context.cpp             	88.21        	0.0                 
./src/acl_device.cpp              	82.33        	-1.0799999999999983 
./src/acl_device_binary.cpp       	76.8         	0.0                 
./src/acl_device_op.cpp           	88.55        	0.0                 
./src/acl_device_program_info.cpp 	44.58        	0.0                 
./src/acl_event.cpp               	91.19        	0.0                 
./src/acl_globals.cpp             	75.19        	0.0                 
./src/acl_hal.cpp                 	79.71        	0.0                 
./src/acl_hal_mmd.cpp             	12.74        	0.0                 
./src/acl_hostch.cpp              	0.88         	0.0                 
./src/acl_icd_dispatch.cpp        	0.0          	0.0                 
./src/acl_kernel.cpp              	75.62        	0.0                 
./src/acl_kernel_if.cpp           	0.0          	0.0                 
./src/acl_mem.cpp                 	74.59       	0.0                 
./src/acl_offline_hal.cpp         	26.17        	0.0                 
./src/acl_platform.cpp            	82.0         	0.0                 
./src/acl_pll.cpp                 	0.0          	0.0                 
./src/acl_printf.cpp              	45.66        	0.0                 
./src/acl_profiler.cpp            	71.88        	0.0                 
./src/acl_program.cpp             	76.89        	0.0                 
./src/acl_sampler.cpp             	1.25         	0.0                 
./src/acl_shared_aligned_ptr.cpp  	100.0        	0.0                 
./src/acl_support.cpp             	66.44        	0.0                 
./src/acl_svm.cpp                 	70.47        	0.0                 
./src/acl_thread.cpp              	70.21        	0.0                 
./src/acl_usm.cpp                 	74.71        	0.0                 
./src/check_copy_overlap.c        	81.48        	0.0                 
./src/acl_device.cpp: Coverage decreased slightly for less than 20%

I'm not sure what all of this means...

[pcolberg]

@haoxian2 This is the code coverage workflow seeing the new branch for gmem_id < 0 which is not unit tested. A dummy device would need to be prepared which makes acl_get_default_device_global_memory() return an error, which is out of scope for this change and more suited for a future initiative that improves coverage.