Tracing command execution in a parallel processing system
View Patent ↗Tracing command execution in a data processing system having a host processor and a co-processor. The host processor maintains a record of a plurality of commands for the co-processor, storing each of the plurality of commands is stored in a command queue. Hardware trace logic is provided to store one or more events based, at least in part, on transfer of the plurality of commands to a small memory. Software is executed to store the one or more events to a main memory, wherein the one or more events are aggregated into a single memory trace within the main memory.
1. A method for recording co-processor command execution data in a system comprising a host central processing unit (CPU) configured to execute main application code and a co-processor separate from the host CPU and configured to execute co-processor commands exclusively from the host CPU, wherein the host CPU is communicatively coupled to the co-processor via a command queue, and the co-processor is configured to receive co-processor commands via the command queue, the method comprising:
maintaining, at the host CPU, a record of the co-processor commands written into the command queue;
monitoring, via a hardware trace logic, one or more events based, at least in part, on the co-processor commands;
storing in a first memory via the hardware trace logic the one or more events; and
in response to filling the first memory, transferring, via a software element, the one or more events from the first memory to a main memory to thereby form a small memory trace data set in the main memory so that multiple small memory trace data sets in the main memory can be aggregated into a single large memory trace based, at least in part, on the record of the co-processor commands.
2. The method of claim 1 , wherein the one or more events includes a writing of one of the co-processor commands to the command queue.
3. The method of claim 1 , wherein the one or more events includes an issuing of one of the co-processor commands.
4. The method of claim 1 , wherein the one or more events includes a completion of one of the co-processor commands.
5. The method of claim 1 , wherein the first memory comprises a first-in-first-out memory register.
6. The method of claim 1 , further comprising:
executing, via the host CPU, the main application code; and
executing, via the co-processor, computation kernels within a context of the main application code.
7. The method of claim 1 , wherein the record of the co-processor commands written into the command queue is maintained independently of the one or more events stored in the first memory.
8. The method of claim 1 , wherein the software element uses a direct memory access operation to transfer the one or more events from the first memory to the main memory.
9. The method of claim 1 , wherein the host CPU and the co-processor respectively comprise master and slave processors.
10. The method of claim 1 , wherein the co-processor comprises a data-parallel co-processor including a plurality of execution lanes.
11. The method of claim 1 , wherein the co-processor comprises a Very Long Instruction Word processor.
12. The method of claim 1 , wherein the co-processor commands comprise a subset of the main application code that is predetermined to execute on the co-processor.