Compiler assisted register file write reduction

A method comprising:

• identifying a group of two or more instructions that provide a partial update to an output buffer;
• including a first instruction in the group of two or more instructions to trigger a transfer that causes writing of an output from execution of the group of two or more instructions to an intermediate buffer;
• including a second instruction in the group of two or more instructions to cause copying of content of the intermediate buffer to the output buffer;
• formatting the group of two or more instructions with the first and second instructions for execution by a graphics processing unit (GPU);
• and
• storing a machine executable format of the group of two or more instructions with the first and second instructions.

A non-transitory computer-readable medium comprising instructions stored thereon, that if executed by a processor, cause the processor to:

• execute a compiler that is to:
• identify a group of two or more instructions that provide a partial but not entire update of an output buffer;
• include a first instruction in the group of two or more instructions, wherein the first instruction causes copying of content to an intermediate buffer;
• include a second instruction in the group of two or more instructions that causes copying of content of the intermediate buffer to the output buffer;
• and
• provide the group of two or more instructions with the first and second instructions for execution by a graphics processing unit (GPU).