Binary instrumentation to trace graphics processor code

A semiconductor package apparatus for use with a graphics processor, comprising:

• one or more substrates;
• and
• logic coupled to the one or more substrates, wherein the logic is at least partly implemented in one or more of configurable logic or fixed-functionality logic hardware, the logic coupled to the one or more substrates to:
• analyze, with a host processor, code that is to be instrumented to determine a size for a trace buffer based on types of memory instructions within each block of a plurality of blocks of the code,
• instrument, with the host processor, the code to generate instrumented code to be executed on a graphics processor,
• initialize, with the host processor, the trace buffer in a shared memory based on the determined size,
• provide the instrumented code to the graphics processor to be executed,
• collect data in the trace buffer from the executed instrumented code,
• bypass a storage of data from the executed instrumented code in the trace buffer in response to the trace buffer reaching a threshold amount,
• analyze the data collected in the trace buffer on the host processor, and
• generate a trace of the instrumented code on the host processor based on the analyzed data.

A method of tracing code, comprising:

• analyzing code, with a host processor, that is to be instrumented to determine a size for a trace buffer based on types of memory instructions within each block of a plurality of blocks of the code;
• instrumenting, with the host processor, the code to generate instrumented code to be executed on a graphics processor;
• initializing, with the host processor, the trace buffer in a shared memory based on the determined size;
• providing the instrumented code to the graphics processor to be executed;
• collecting data in the trace buffer from the executed instrumented code;
• bypassing a storage of data from the executed instrumented code in the trace buffer in response to the trace buffer reaching a threshold amount;
• analyzing the data collected in the trace buffer on the host processor;
• and
• generating a trace of the instrumented code on the host processor based on the analyzed data.

At least one non-transitory computer readable storage medium, comprising a set of instructions, which when executed by a computing device, cause the computing device to:

• analyze, with a host processor, code that is to be instrumented to determine a size for a trace buffer based on types of memory instructions within each block of a plurality of blocks of the code;
• instrument, with the host processor, the code to generate instrumented code to be executed on a graphics processor;
• initialize, with the host processor, the trace buffer in a shared memory based on the determined size;
• provide the instrumented code to the graphics processor to be executed;
• collect data in the trace buffer from the executed instrumented code;
• bypass a storage of data from the executed instrumented code in the trace buffer in response to the trace buffer reaching a threshold amount
• analyze the data collected in the trace buffer on the host processor;
• and
• generate a trace of the instrumented code on the host processor based on the analyzed data.