Deterministic replay of a multi-threaded trace on a multi-threaded processor
At least one computer-readable storage medium comprising instructions for execution by at least one graphics processing unit (GPU) that, when executed, cause the at least one GPU to: obtain program code for tracing, the program code including a plurality of instructions; identify from the plurality of instructions of the program code events to be synchronized; instrument the program code corresponding to one or more of the events identified, by inserting instructions that support monitoring code; execute the instrumented program code on at least a plurality of hardware threads of the GPU and generate trace data; replay the identified events according to an order of occurrence of the events identified; and report a GPU state indicating a utilization of the GPU based; and wherein to report the GPU state includes to indicate when the GPU executes non-graphics related tasks.
1. At least one non-transitory computer-readable storage medium comprising instructions for execution by at least one graphics processing unit (GPU) that, when executed, cause the at least one GPU to:
obtain program code for tracing, the program code including a plurality of instructions;
identify, from the plurality of instructions of the program code, events to be synchronized;
instrument the program code corresponding to one or more of the events identified, by inserting instructions that support monitoring code;
execute the instrumented program code on at least a plurality of hardware threads of the GPU and generate trace data;
replay the identified events according to an order of occurrence of the events identified; and
report a GPU state indicating a utilization of the GPU; and
wherein to report the GPU state includes to indicate when the GPU executes non-graphics related tasks.
2. The at least one non-transitory computer-readable storage medium according to claim 1 , further comprising instructions for execution by the at least one GPU that, when executed, cause the at least one GPU to:
report runtime shader errors.
3. The at least one non-transitory computer-readable storage medium of claim 1 , wherein the events include one or more of a code dispatch, a code end-of-thread event, a read or write access to global memory, a read or write access to local memory, a waiting state, or a memory fence instruction.
4. The at least one non-transitory computer-readable storage medium of claim 1 , wherein the program code is a kernel or a shader.
5. The at least one non-transitory computer-readable storage medium of claim 1 , wherein instrumenting the program code includes:
dividing the program code into a sequence of basic blocks; and
inserting a trace instruction into each basic block of the sequence of basic blocks that contains an event.
6. The at least non-transitory one computer-readable storage medium of claim 5 , wherein instrumenting the program code further includes:
inserting a dynamic instruction count relating to an original instruction in each basic block where a tracing instruction is added.
7. A method comprising:
obtaining program code for tracing, the program code including a plurality of instructions;
identifying, from the plurality of instructions of the program code, events to be synchronized;
instrumenting the program code corresponding to one or more of the events identified, by inserting instructions that support monitoring code;
executing the instrumented program code on at least a plurality of hardware threads of a graphics processing unit (GPU) and generating trace data;
replaying the identified events according to an order of occurrence of the events identified; and
reporting a GPU state indicating a utilization of the GPU; and
wherein to report the GPU state includes to indicate when the GPU executes non-graphics related tasks.
8. The method of claim 7 , further comprising:
reporting runtime shader errors.
9. The method of claim 7 , wherein the events include one or more of a code dispatch, a code end-of-thread event, a read or write access to global memory, a read or write access to local memory, a waiting state, or a memory fence instruction.
10. The method of claim 7 , wherein the program code is a kernel or a shader.
11. A system comprising:
one or more processors including at least one graphics processing unit (GPU); and
a memory to store data including instructions; and
wherein the instructions includes instructions to cause the at least one GPU to perform operations including the following:
obtaining program code for tracing, the program code including a plurality of instructions;
identifying, from the plurality of instructions of the program code, events to be synchronized;
instrumenting the program code corresponding to one or more of the events identified, by inserting instructions that support monitoring code;
executing the instrumented program code on at least a plurality of hardware threads of the GPU and generate trace data;
replaying the identified events according to an order of occurrence of the events identified; and
reporting a GPU state indicating a utilization of the GPU; and
wherein to report the GPU state includes to indicate when the GPU executes non-graphics related tasks.
12. The system of claim 11 , wherein the instructions includes instructions to cause the at least one GPU to perform operations including:
reporting runtime shader errors.
13. The system of claim 11 , wherein the events include one or more of a code dispatch, a code end-of-thread event, a read or write access to global memory, a read or write access to local memory, a waiting state, or a memory fence instruction.
14. The system of claim 11 , wherein the program code is a kernel or a shader.