Methods and apparatus for interactive debugging on a non-preemptible graphics processing unit

A non-transitory computer-readable storage medium storing a debugger that, when executed by a processing unit, causes the processing unit to perform the steps of:

• allowing a shader program to be loaded onto a parallel processing unit by an application program, wherein the parallel processing unit is configured to render a graphical user interface for an interactive debugging environment for display on a display device coupled to the parallel processing unit;
• intercepting a command to transmit a workload to the parallel processing unit;
• issuing the workload to the parallel processing unit for processing during a first processing pass, wherein the workload is processed by a plurality of threads executing in parallel on separate input data included in the workload, and wherein the parallel processing unit is configured to halt the first processing pass when a hardware trap is triggered by any thread in the plurality of threads;
• detecting that a hardware trap has occurred during a first point in the first processing pass;
• and
• entering a re-launch loop that comprises disabling the first processing pass to allow the graphical user interface to be updated by the parallel processing unit and re-issuing the workload to the parallel processing unit for processing during one or more subsequent processing passes.

A system for performing interactive debugging on non-preemptible graphics processing units, the system comprising:

• a system memory including a debugger and an application program including one or more shader programs targeted for debugging;
• a central processing unit (CPU) coupled to the system memory;
• a parallel processing unit configured to render a graphical user interface for an interactive debugging environment for display on a display device coupled to the parallel processing unit;
• and
• a graphics memory coupled to the parallel processing unit,
• wherein the debugger, when executed by the CPU, is configured to:
• allow the shader program to be loaded onto the parallel processing unit by the application program,
• intercept a command to transmit a workload to the parallel processing unit;
• issue the workload to the parallel processing unit for processing during a first processing pass, wherein the workload is processed by a plurality of threads executing in parallel on separate input data included in the workload, and wherein the parallel processing unit is configured to halt the first processing pass when a hardware trap is triggered by any thread in the plurality of threads,
• detect that a hardware trap has occurred during a first point in the first processing pass, and
• enter a re-launch loop that comprises disabling the first processing pass to allow the graphical user interface to be updated by the parallel processing unit and re-issuing the workload to the parallel processing unit for processing during one or more subsequent processing passes.