GPU-executed program sequence cross-compilation
According to one embodiment of the present disclosure, a computing system is provided, including a graphical processing unit (GPU) and a processor. The processor may be configured to execute a run-time executable cross-compiler to receive a GPU-executed program of a plurality of GPU-executed programs. The processor may be further configured to receive summary data associated with the GPU-executed program. The summary data may include a sequence in which the plurality of GPU-executed programs are configured to be executed. Based at least in part on the GPU-executed program and the summary data, the processor may be further configured to generate a translated GPU-executed program.
1. A computing system, comprising:
a graphical processing unit (GPU); and
a processor configured to execute a run-time executable cross-compiler to:
receive a GPU-executed program of a plurality of GPU-executed programs;
receive summary data associated with the GPU-executed program, wherein the summary data includes a sequence in which the plurality of GPU-executed programs are configured to be executed; and
based at least in part on the GPU-executed program and the summary data, generate a translated GPU-executed program.
2. The computing system of claim 1 , wherein the cross-compiled program is configured to be executed in a GPU application binary interface (ABI) of the GPU.
3. The computing system of claim 1 , wherein the processor is further configured to determine instruction data indicating one or more instructions included in at least one other GPU-executed program of the plurality of GPU-executed programs, and wherein the cross-compiled program is generated based at least in part on the instruction data.
4. The computing system of claim 3 , wherein the processor is further configured to determine instruction data for each GPU-executed program of the plurality of GPU-executed programs.
5. The computing system of claim 3 , wherein generating the cross-compiled program includes adding to the GPU-executed program one or more instructions included in the at least one other GPU-executed program.
6. The computing system of claim 3 , wherein generating the translated GPU-executed program includes removing one or more instructions from the GPU-executed program.
7. The computing system of claim 1 , wherein each GPU-executed program of the plurality of GPU-executed programs is a shader program.
8. The computing system of claim 7 , wherein the plurality of GPU-executed programs includes at least one vertex shader program.
9. The computing system of claim 7 , wherein the plurality of GPU-executed programs includes at least one pixel shader program.
10. The computing system of claim 7 , wherein the processor is further configured to generate a plurality of translated GPU-executed programs based at least in part on the GPU-executed program and the summary data.
11. The computing system of claim 10 , wherein the processor is configured to generate a plurality of translated tessellation shader programs based at least in part on two or more GPU-executed programs of the plurality of GPU-executed programs.
12. The computing system of claim 7 , wherein generating the translated GPU-executed program includes applying a coordinate transformation to at least one output of the GPU-executed program.
13. The computing system of claim 1 , wherein the summary data further includes GPU state data that indicates one or more global properties of the GPU.
14. The computing system of claim 13 , wherein the one or more global properties of the GPU include an instruction set architecture (ISA) of the GPU.
15. A method for executing a runtime-executable cross-compiler on a processor of a computing system, the method comprising:
receiving a GPU-executed program of a plurality of GPU-executed programs;
receiving summary data associated with the GPU-executed program, wherein the summary data includes a sequence in which the plurality of GPU-executed programs are configured to be executed; and
based at least in part on the GPU-executed program and the summary data, generating a translated GPU-executed program.
16. The method of claim 15 , further comprising determining instruction data indicating one or more instructions included in at least one other GPU-executed program of the plurality of GPU-executed programs, and wherein the translated GPU-executed program is generated based at least in part on the instruction data.
17. The method of claim 16 , further comprising determining instruction data for each GPU-executed program of the plurality of GPU-executed programs.
18. The method of claim 16 , wherein generating the translated GPU-executed program includes adding to the GPU-executed program one or more instructions included in the at least one other GPU-executed program and/or removing one or more instructions from the GPU-executed program.
19. The method of claim 15 , wherein each GPU-executed program of the plurality of GPU-executed programs is a shader program.
20. A computing system, comprising:
a graphical processing unit (GPU); and
a processor configured to execute a run-time executable cross-compiler to:
receive a GPU-executed program of a plurality of GPU-executed programs, wherein each GPU-executed program of the plurality of GPU-executed programs is a shader program;
receive summary data associated with the GPU-executed program, wherein the summary data includes a sequence in which the plurality of GPU-executed programs are configured to be executed;
determine instruction data indicating instructions included in at least one other GPU-executed program of the plurality of GPU-executed programs; and
based at least in part on the GPU-executed program, the summary data, and the instruction data, generate a translated GPU-executed program.