Return device total global memory for MaxAllocSize

The SYCL specifiction inherits an OpenCL query for the maximum single
allocaton that can be made, and requires that at least one-quarter of
the device's reported global memory size can be allocated.

CUDA doesn't really have such a limitation as far as I can tell, and
will happily allocate anything up to the total size of global memory
in one go. This means returning that size is a reasonable guess for
MaxAllocSize.

source/adapters/cuda/device.hpp

@@ -64,17 +64,11 @@ struct ur_device_handle_t_ {
  }
 
  // Max size of memory object allocation in bytes.
- // The minimum value is max(min(1024 × 1024 ×
- // 1024, 1/4th of CL_DEVICE_GLOBAL_MEM_SIZE),
- // 32 × 1024 × 1024) for devices that are not of type
- // CL_DEVICE_TYPE_CUSTOM.
- size_t Global = 0;
- UR_CHECK_ERROR(cuDeviceTotalMem(&Global, cuDevice));
-
- auto QuarterGlobal = static_cast<uint32_t>(Global / 4u);
-
- MaxAllocSize = std::max(std::min(1024u * 1024u * 1024u, QuarterGlobal),
- 32u * 1024u * 1024u);
+ // The minimum value is max (1/4th of info::device::global_mem_size,
+ // 128*1024*1024) if this SYCL device is not device_type::custom.
+ // CUDA doesn't really have this concept, and could allow almost 100% of
+ // global memory in one allocation, but is dependent on device usage.
+ UR_CHECK_ERROR(cuDeviceTotalMem(&MaxAllocSize, cuDevice));
  }
 
  ~ur_device_handle_t_() { cuDevicePrimaryCtxRelease(CuDevice); }