Abstract: Computer code written to execute on a multi-threaded computing environment is transformed into code designed to execute on a single-threaded computing environment and simulate concurrent executing threads. Optimization techniques during the transformation process are utilized to identify local variables for scalar expansion. A first set of local variables is defined that includes those local variables in the code identified as “Downward exposed Defined” (DD). A second set of local variables is defined that includes those local variables in the code identified as “Upward exposed Use” (UU). The intersection of the first set and the second set identifies local variables for scalar expansion.

Abstract: One embodiment of the present invention sets forth a technique for partitioning a predecessor thread program into sub-programs and dynamically spawning a thread grid of the sub-programs based on the outcome of a conditional statement in the predecessor thread program. The programming instructions for the predecessor thread program are analyzed to assess the benefit of partitioning the thread program at a conditional statement into sub-programs. If the predecessor thread program is partitioned, then each branch of the conditional statement may be used to form a separate sub-program. Predicate tables are populated at the predecessor thread program run-time to establish which possible instances of the thread sub-programs should be spawned in subsequent execution phases.

Abstract: One embodiment of the present invention sets forth a technique for partitioning a predecessor thread program into sub-programs and dynamically spawning a thread grid of the sub-programs based on the outcome of a conditional statement in the predecessor thread program. The programming instructions for the predecessor thread program are analyzed to assess the benefit of partitioning the thread program at a conditional statement into sub-programs. If the predecessor thread program is partitioned, then each branch of the conditional statement may be used to form a separate sub-program. Predicate tables are populated at the predecessor thread program run-time to establish which possible instances of the thread sub-programs should be spawned in subsequent execution phases.

Abstract: One embodiment of the present invention sets forth a technique for performing a parallel scan operation with high computational efficiency in a single-instruction multiple-data (SIMD) environment. Each participating thread initially writes an extended region of a data array to initialize the region with an identity value. For example, a value of zero is used as the identity value for addition. The initialized region of the data array includes an initialized entry for every possible out of bounds index that may be computed in the normal course of the parallel scan operation. During the parallel scan operation each thread computes data array indices according to any technically appropriate technique. When a participating thread computes an index that would conventionally be out of bounds, the thread is able to retrieve an identity value from the initialized region of the data array rather than perform a bounds check that returns the identity value.

Abstract: One embodiment of the present invention sets forth a technique for selectively spawning threads within a multiprocessing system. A computation work distributor (CWD), within the system, is responsible for performing the detailed work needed to spawn a thread grid. A request to the CWD to spawn a thread grid includes a predicate table, which includes an array of flags used to indicate which thread indices should have an associated thread block spawned and which should not. Greater efficiency is achieved by only spawning thread blocks that should perform useful computation.