Hewlett-packard Development Company, L.p.

Silicon Graphics, Incorporated

Shin-Ming Liu

Sun C Chan

Raymond W Lo

Frederick Chow

Dhruva Ranjan Chakrabarti

Mark Streich

Dmitry Mikulin

Peng Tu

Robert Kennedy

Robert Hundt

Xinliang David Li

David Xinliang Li

Muralitharan Vijayasundaram

Sandya Srivilliputtur Mannarswamy

Peter Dahl

Sungdo Moon

Luis A Lozano

Mur Litharan Vijayasundaram

One embodiment relates to a computer-implemented method of compiling a software program into an executable binary file, including determining a data layout in the binary file and a data layout in the executable's dynamically allocated memory. The method includes taking into account data types of data as a factor in determining a data layout for the binary file and for the executable's dynamically allocated memory, wherein the data types include a floating-point data type and a non-floating-point data type. Other embodiments, aspects and features are also disclosed herein.

Data layout using data type information

One embodiment disclosed relates to a method of compiling a computer program. A plurality of modules of source code is received, and intermediate representations corresponding to the modules are generated. A set of data from the intermediate representations is extracted to create an inliner summary for each module. Using the inliner summaries and a globally-sorted working-list based order, an inline analysis phase determines which call sites in the modules are to be inlined by substituting code from a called module. The propagation of summaries may be done elaborately, potentially throughout the call-graph. The goodness of call sites may be computed with a view to comparing them and ordering them in a descending order of goodness. Applicants believe the compile-time effectiveness lies in the ability to work consistently with summary information in the inline analysis phase without having to touch the intermediate representation, while maintaining a high degree of run-time performance by continuously updating the summary information.

Cross-file inlining by using summaries and global worklist

A method of generating a software program executable binary file. A first file including source code therein is accessed. A second file including object code therein and further including object file summary information is accessed. The executable binary file is generated from at least the first and second files. The object file summary information is used in optimizing the executable binary file generated.

Program optimization using object file summary information

Disclosed are systems and methods for optimizing a program.

System and method for optimizing a program

One embodiment disclosed relates to a method of compiling a computer program from a plurality of files of source code. An inline analysis determines which call sites in the plurality of files to make inline. An inline transformation performs the inlining within currently opened files. The transformer dynamically determines the order of inlines independent of the analyzer by taking into account the disk input-output pressure during compilation. The resulting inline order minimizes the input and output of files from and to disk respectively by considering the inline affinity between files and maintains the best run-time performance by preserving the dependences between call sites. During the inline transformation, a determination of which files to open and close is made in dependence on an affinity weighting between the files. The affinity and the dependence relationships are continuously updated in order to generate the best transformation ordering among call sites that ensures compile-time scalability and improved run-time performance.

Scalable cross-file inlining through locality-based transformation ordering

The present disclosure relates to whole program analysis and, more particularly, short data optimization obtained through whole program analysis. In one embodiment, short data optimization is achieved by analyzing the program to estimate the size of existing short data and the size of any linkage tables, providing the size estimates to a compiler that is to compile the program, and compiling the program with the compiler in view of the size estimates such that a relatively large amount of data is allocated to a short data area.

Techniques for partitioning modules into smaller groups are disclosed. In various embodiments, cross-module optimization may be performed on these smaller groups. Further, related modules are grouped together based on a predefined algorithm. Relatedness, or closeness, or affinity, of modules are considered based on various factors including, for example, the number of calls between routines in different modules, possibility of in-lining a first routine in a first module into a second routine in a second module, characteristics of parameters being passed between routines in different modules, etc.

Partitioning modules for cross-module optimization

A method, system, and computer product uses a hashed static single assignment (SSA) form as a program representation and a medium for performing global scalar optimization. A compiler, after expressing the computer program in SSA form, can perform one or more static single assignment (SSA)-based, SSA-preserving global scalar optimization procedures on the SSA representation. Such a procedure modifies, (i.e., optimizes) the SSA representation of the program while preserving the utility of its embedded use-deprogram information for purposes of subsequent SSA-based, SSA-preserving global scalar optimizations. This saves the overhead expense of having to explicitly regenerate use-def program information for successive SSA-based, SSA-preserving global scalar optimizations.

Method, system, and computer program product for using static single assignment form as a program representation and a medium for performing global scalar optimization

A system and method for an optimizer of a compilation suite for representing aliases and indirect memory operations in static single assignment (SSA) during compilation of a program having one or more basic blocks of source code. The optimizer converts all scalar variables of said program to SSA form, wherein said SSA form includes a plurality of variable versions, zero or more occurrences of a .chi. function, zero or more occurences of a .phi. function, and zero or more occurrences of a .mu. function. The .chi. function, .phi. function, and .mu. function are inserted for the variable versions. The optimizer also determines whether a variable version can be renamed to a zero version, and upon such a determination, the optimizer renames the variable version to a zero version. The optimizer further converts all indirect variables of a program to SSA form, wherein the SSA form includes a plurality of virtual variable versions such that a virtual variable is assigned to an indirect variable, zero or more occurrences of a .chi. function, zero or more occurences of a .phi. function, and zero or more occurrences of a .mu. function. The .chi. function, .phi. function, and .mu. function are inserted for the virtual variables. The optimizer hashes a unique value number and creates a corresponding hash table entry for each variable version and each virtual variable remaining after renaming all zero versions. The optimizer also applies global value numbering to each basic block of the program.

System and method to efficiently represent aliases and indirect memory

A method, system, and computer program product for performing register promotion, that optimizes placement of load and store operations of a computer program within a compiler. Based on the observation that the circumstances for promoting a memory location's value to register coincide with situations where the program exhibits partial redundancy between accesses to the memory location, the system is an approach to register promotion that models the optimization as two separate problems: (1) the partial redundancy elimination (PRE) of loads and (2) the PRE of stores. Both of these problems are solved through a sparse approach to PRE. The static single assignment PRE (SSAPRE) method for eliminating partial redundancy using a sparse SSA representation representations the foundation in eliminating redundancy among memory accesses, enabling the achievement of both computational and live range optimality in register promotion results. A static single use (SSU) representation is defined allowing the dual of the SSAPRE algorithm, called SSUPRE, to perform the partial redundancy elimination of stores. SSUPRE is performed after the PRE of loads, taking advantage of the loads' having been converted into pseudo-register references so that there are fewer barriers to the movement of stores. Consequently, the compiler produces more efficient, register-promoted executable program code from the SSA representation.

Method, system, and computer program product for performing register

Partial redundancy elimination of a computer program is described that operates using a static single assignment (SSA) representation of a computer program. The SSA representation of the computer program is processed to eliminate partially redundant expressions in the computer program. This processing involves inserting .PHI. functions for expressions where different values of the expressions reach common points in the computer program. A result of each of the .PHI. functions is stored in a hypothetical variable h. The processing also involves a renaming step where SSA versions are assigned to hypothetical variables h in the computer program, a down safety step of determining whether each .PHI. function in the computer program is down safe, and a will be available step of determining whether each expression in the computer program will be available at each .PHI. function following eventual insertion of code into the computer program for purposes of partial redundancy elimination. The processing also includes a finalize step of transforming the SSA representation of the computer program having hypothetical variables h to a SSA graph that includes some insertion information reflecting eventual insertions of code into the computer program for purposes of partial redundancy elimination, and a code motion step of updating the SSA graph based on the insertion information to introduce real temporary variables t for the hypothetical variables h.

System, method, and computer program product for partial redundancy

A system and method for optimizing a source code representation comprising a plurality of basic blocks are described. The optimized source code representation is to be executed in a target machine. The system operates by selecting from the source code representation a basic block pair comprising a source basic block and one or more target basic blocks. An instruction in the source basic block is identified that can be moved from the source basic block to the target basic block(s) while preserving program semantics. Either the instruction or a representation of the instruction is moved from the source basic block to the target basic block(s) as a function of resource utilization of the target machine that would result from this movement.

Growing community of inventors

Saratoga, CA, United States of America

Shin-Ming Liu