Select a line number inside a box to plant a breakpoint in the GPU code, and select "Go" to continue the process, which will start the execution of the CUDA kernel. Once the CUDA kernel starts executing, it will hit the breakpoint planted in the GPU code, as shown in Figure 261.

The logical coordinates of the GPU focus threads are shown in the thread status title bar and the Threads pane. You can use the GPU focus thread selector to change the GPU focus thread. When you change the GPU focus thread, the logical coordinates displayed also change, and the stack trace, stack frame, and source panes are updated to reflect the state of the new GPU focus thread.

The yellow PC arrow in the source pane shows the execution location of the GPU focus thread. The GPU hardware threads, also known as "lanes", execute in parallel so multiple lanes may have the same PC value. The lanes may be part of the same warp, or in different warps.

The stack trace pane shows the stack backtrace and inlined functions. Each stack frame in the stack backtrace represents either the PC location of GPU kernel code, or the expansion of an inlined function. Inlined functions can be nested. The "return PC" of an inlined function is the address of the first instruction following the inline expansion, which is normally within the function containing the inlined-function expansion.

The stack frame pane shows the parameter, register and local variables for the function in the selected stack frame. The variables for the selected GPU kernel code or inlined function expansion are shown.