For network computing on desktop machines, fast execution of Java bytecode programs is essential because these machines are expected to run substantial application programs written in Java. We believe higher Java performance can be achieved by exploiting instruction-level parallelism (ILP) in the context of Java JIT compilation. This paper introduces VLaTTe, a Java JIT compiler for VLIW machines that performs efficient scheduling while doing fast register allocation. It is an extended version of our previous JIT compiler for RISC machines called LaTTe whose translation overhead is low (i.e., consistently taking one or two seconds for SPECJVM98 benchmarks) due to its fast register allocation. VLaTTe adds the scheduling capability onto the same framework of register allocation, with a constraint for precise in-order exception handling which guarantees the same Java exception behavior with the original bytecode program. Our experimental results on the SPECJVM98 benchmarks show that VLaTTe achieves a geometric mean of useful IPC 1.7 (2-ALU), 2.1 (4-ALU), and 2.3 (8-ALU), while the scheduling/allocation overhead is 3.6 times longer than LaTTe's on average, which appears to be reasonable.

An efficient hybrid scheme has been developed for optimizing register allocation applicable to CISC and RISC processors, which is crucial for maximizing their execution speed. Graph-coloring at the function level is combined with a powerful local register assigner. This assigner uses accurate program flows and access patterns of variables, and optimizes a wider local range, called an extended basic-block (EBB), than other optimizing compilers. The EBB is a set of basic-blocks that constitute a tree-shaped control flow, which is suitable for the large nested branches that frequently appear in embedded system-control programs, such as those for telephone call processing. The coloring at the function level involves only the live-ranges of program variables that span EBBs. The interference graph is therefore very small even for large functions, so it can be constructed quickly. Instead of iterative live-range splitting or spilling, the unallocated live-ranges are optimized by the EBB-based register assigner, so neither load/store insertion nor code motion is needed. This facilitates generating reliable code and debug symbols. The information provided for the EBB-based assigner facilitates the priority-based heuristics, fine-grained interference checking, and deferred coloring, all of which increase the colorability. Using a thread-support package for CHILL as a sample program, performance measurement showed that local variables are successfully located in registers, and the reduction of static cycles is about 20-30%. Further improvements include using double registers and improving debuggability.