Abstract: A method for executing a plurality of data-parallel threads of a processing logic on a processor core includes grouping the plurality of data-parallel threads into one or more workgroups, associating a first workgroup from the one or more workgroups with an operating system thread on the processor core, and configuring threads from the first workgroup as user-level threads within the operating system thread. In an example, a method enables the execution of GPU-kernels that has been previously configured for a GPU, to execute on a CPU such as a multi-core CPU. The mapping of the numerous data-parallel threads to the CPU is done in such a manner as to reduce the number of costly operating system threads instantiated on the CPU, and to enable efficient debugging.

Abstract: Techniques are disclosed relating to distributing workloads between processors and/or processing elements. A computer system having at least first and second processing elements may cause a request to initialize one or more memory regions to be handled by the second processing element. Initialization may be accomplished by the second processing element directly accessing a memory that includes the specified memory region to be initialized. Thus, while the second processing element is causing the memory region to be initialized, the first processing element is free to perform other computational tasks. A cache associated with the first processing element may be undisturbed as a result of the second processing element performing the initialization, which may avoid displacement of data from the cache.

Abstract: A method for executing a plurality of data-parallel threads of a processing logic on a processor core includes grouping the plurality of data-parallel threads into one or more workgroups, associating a first workgroup from the one or more workgroups with an operating system thread on the processor core, and configuring threads from the first workgroup as user-level threads within the operating system thread. In an example, a method enables the execution of GPU-kernels that has been previously configured for a GPU, to execute on a CPU such as a multi-core CPU. The mapping of the numerous data-parallel threads to the CPU is done in such a manner as to reduce the number of costly operating system threads instantiated on the CPU, and to enable efficient debugging.

Abstract: The present invention, in various embodiments, provides techniques for improving performance of programs. In one embodiment, the program is written in the Java language and runs in the Java Runtime Environment (JRE) that includes a Java Virtual Machine (JVM) having a configuration. A control panel having access to the JVM is invoked. Via the control panel, a user observes the execution of the Java program, analyzes the results of the execution, changes the configuration of the JVM, and forces re-optimization of the critical portions of the running program. By repeating the above steps of observing the program execution, analyzing the results, changing the configuration of the Java machine, and forcing re-optimization of the critical portions of the program, the user improves the program's overall performance.

Abstract: Techniques are provided for recovering from compilation errors in environments that use dynamic compilers. Application programs include Java bytecodes, and compilation includes sequential invocation of separate compilation phases on a region of bytecodes. If compilation of a region results in a fatal error, then the compiler identifies the “failed” phase. If the failed phase is a non-essential phase, then the compiler attempts to re-compile the region after skipping the failed phase. However, if the failed phase is essential, then the compiler attempts to replace that failed phase with a simpler version. Nevertheless, if the fatal error cannot be avoided or the compiler is unable to replace the failed phase with a simpler version, then the compiler prevents compilation of the code encompassing the fatal error in future attempts.

Abstract: Techniques are provided for recovering from compilation errors in environments that use dynamic compilers. Application programs include Java bytecodes, and a compiler is arranged into a plurality of compilation phases. Each phase implements a specific compiler optimization that contributes towards the total performance of the compiled code. In effect, compilation includes sequential invocation of separate compilation phases on a region of bytecodes. If compilation of a particular region results in a fatal error, then the compiler identifies the compilation phase that generated the error, which is referred to as the “failed” phase. If the failed phase is a nonessential phase, then the compiler attempts to re-compile the region after skipping the failed phase. However, if the failed phase is essential for compilation and/or contributes significantly towards the performance of the compiled code, then the compiler attempts to replace that failed phase with a simpler version.

Abstract: The present invention, in various embodiments, provides techniques for improving performance of programs. In one embodiment, the program is written in the Java language and runs in the Java Runtime Environment (JRE) that includes a Java Virtual Machine (JVM) having a configuration. A control panel having access to the JVM is invoked. Via the control panel, a user observes the execution of the Java program, analyzes the results of the execution, changes the configuration of the JVM, and forces re-optimization of the critical portions of the running program. By repeating the above steps of observing the program execution, analyzing the results, changing the configuration of the Java machine, and forcing re-optimization of the critical portions of the program, the user improves the program's overall performance.