Features of Vortex

• Vortex performs static intraprocedural iterative class analysis, tracking the possible classes of expressions within a procedure. Static class information enables dynamically dispatched messages to be statically bound as regular procedure calls (if static class information proves that only one method can be called), and allows run-time class or subclass tests (such as Java's instanceof construct) to be constant-folded (if static class information proves the test is always true or false).
• Vortex performs automatic inlining, optimizing statically bound procedure calls. Inlining is interleaved with class analysis so that class analysis can track the flow of classes into and out of inlined methods.
• Vortex includes class hierarchy analysis, which augments intraprocedural analysis with information about the program's whole class hierarchy. Information from automatic class hierarchy analysis is strictly better than information about final methods or classes in Java and non-virtual member functions in C++.
• Vortex includes a wide range of interprocedural class analysis algorithms, ranging from quick linear-time algorithms to precise context-sensitive algorithms. Intraprocedural class analysis and optimization exploits information gathered from interprocedural class analysis. In addition, several other analyses rely on the call graph computed as a side-effect of interprocedural class analysis, including interprocedural constant propagation, interprocedural side-effect analysis, interprocedural escape analysis (to identify which closures may be stack-allocated), and interprocedural exception analysis (to identify which procedures never raise exceptions).
• Vortex exploits dynamic profile information about the relative frequencies of different classes at each call site to optimize dynamically dispatched messages. If one or a few receiver classes are most common at a site, then class tests are inserted in-line before the message, with each test leading to a branch where the message has been statically bound to the appropriate receiver. Even in the absence of peaked frequencies, if only a few methods can be called by a message site, Vortex can insert a few subclass tests in-line to statically bind to each of the possible methods.
• Vortex combines dynamic class profile information and static class hierarchy information to drive a selective procedure specialization algorithm.
• Vortex includes an intraprocedural optimization, splitting, which can eliminate conditional branches in procedures if earlier merges lost the information tested by the branch. This optimization is particularly useful to optimize repeated class tests.
• Vortex includes optimizations to reduce the cost of closures (first-class functions), including inlining invocations of statically known closures, delaying closure creations to those branches where the closures are needed (a kind of partial dead code elimination applied to closure creations), "shrinkwrapping" of environment creation code to the subset of the program's control flow graph where the environment is needed, and stack-allocating closures where possible.
• Vortex includes several traditional optimizations, including constant propagation and folding (including of conditional branches and switches), common subexpression elimination, and dead assignment elimination. When producing assembly code, global register allocation and simple instruction scheduling are performed.
• Vortex includes a must-alias and side-effect analysis which is used to identify and eliminate redundant loads (CSE of loads, in essense) and stores and dead stores. By interleaving the elimination of redundant loads with other analyses including class analysis and constant propagation, class and constant information can be tracked through stores and subsequent loads of memory locations. Consider an example where a callee function allocates & initializes a data structure to hold its results values, which the caller merely takes apart: if the callee function is inlined, then Vortex's optimizations can eliminate the stores into and loads from the result data structure and then eliminate the data structure allocation entirely if no more references remain to the result data structure. Other idioms using short-lived intermediate data structures can similarly be optimized.
• Vortex supports symbolic analysis of comparison outcomes, enabling it to eliminate conditional tests whose outcome is implied by some earlier test. This optimization is particularly useful to eliminate unnecessary array bounds checks.
• Vortex includes a framework for implementing iterative data flow analyses. A client writes a collection of analysis flow functions (dispatching on the different kinds of intermediate nodes), each of which specifies either a particular local flow graph transformation to perform as an optimization, or the updated data flow information to continue propagating. The framework automatically manages invoking flow functions in either a forward or backward traversal order, performing iterative approximations, applying transformations when iteration has reached fixpoint and only simulating transformations before reaching fixpoint. Clients also can specify that several analyses should be composed; the framework automatically interleaves the analyses, allowing each to benefit from the optimization and analysis effects of the others, in an optimistic manner. Vortex's interprocedural analyses are also supported by a framework that makes it easy to extend any intraprocedural analysis to an interprocedural version, with a client-specified context-sensitivity strategy.
• Vortex includes support for selective recompilation after programming changes, to make its whole-program optimization approach practical even in a normal incremental program development environment.

Some resources for Vortex

• a release of Vortex is available
  (manuals on using Vortex are also available from the release page)
• a paper and a long technical report on Vortex; other papers we have written
• Cecil/Vortex Project's home page