◆ av
hc::accelerator_view* grid_launch_parm::av |
Pointer to the accelerator_view where the kernel should execute. If NULL, the default view on the default accelerator is used.
◆ barrier_bit
enum gl_barrier_bit grid_launch_parm::barrier_bit |
Control setting of barrier bit on per-packet basis: See gl_barrier_bit description.
Placeholder, is not used to control packet dispatch yet
◆ cf
hc::completion_future* grid_launch_parm::cf |
Pointer to the completion_future used to track the status of the command. If NULL, the command does not write status. In this case, synchronization can be enforced with queue-level waits or waiting on younger commands.
◆ dynamic_group_mem_bytes
unsigned int grid_launch_parm::dynamic_group_mem_bytes |
Amount of dynamic group memory to use with the kernel launch. This memory is in addition to the amount used statically in the kernel.
◆ launch_fence
unsigned int grid_launch_parm::launch_fence |
Value of packet fences to apply to launch. The correspond to the value of bits 9:14 in the AQL packet, see HSA_PACKET_HEADER_ACQUIRE_FENCE_SCOPE and hsa_fence_scope_t.
The documentation for this struct was generated from the following file:
- /builddir/build/BUILD/clr-rocm-6.1.2/hipamd/include/hip/amd_detail/grid_launch.h