Link-time optimization of address calculation on a 64-bit architecture

Abstract

Compilers for new machines with 64-bit addresses must generate code that works when the memory used by the program is large. procedures and global variables are accessed indirectly via global address tables, and calling conventions include code to establish the addressability of the appropriate tables. In the common case of a program that does not require a lot of memory, all of this can be simplified considerably, with a corresponding reduction in program size and execution time.We have used our link-time code modification system OM to perform program transformations related to globat address use on the Alpha AXP. Though simple, many of these are whole-program optimization that can be done only when we can see the entire program at once, so link-time is an ideal occasion to perform them. This paper describes the optimization performed and shows their effects on program size and performance. Relatively modest transformations, possible without moving code, improve the performance of SPEC benchmarks by an average of 1.5Y0.More ambitious transformations, requiring an understanding of program structure that is thorough but not difficult at link-time, can do even better, reducing program size by 10% or more, and improving performance by an average of 3.87i0.Even a program compiled monolithically with interprocedurat optimization can benefit nearly as much from this teehnique, if it contains statically-linked pre-compiled library code. When the benchmark sources were compiled in this way, we were still able to improve their performance by 1.35% with the modest transformations and 3.4% with the ambitious transformations.Permission to co y without fee all or part of this material is J' granted provid that the copies are not made or distributed for direct commercial advantage, the ACM copyright notice and the title of the publication and its date appear, and notice is given that copying is by permission of the Association of Computing Machinery. To copy otherwise, or to republish, requires a tee and/or specific permission. SIGPLAN 94-6/

Global Data Access Optimization Via Load/Store Instruction Extension

“…GP-addressing cannot guarantee a single instruction to access all global data in these situations. Srivastava et al [5] exploited link-time optimization to reduce the "load address". They found that many global variables nearby GAT can be accessed directly with GP.…”

Link-time binary rewriting techniques for program compaction

“…The inefficiencies of GOTs have long been recognized [Srivastava and Wall 1994]. Besides the overhead needed to ensure that the GP always points to the part of the GOT associated with the code being executed, the indirection through the GOT is itself often not necessary.…”

“…Note that this could just as well be done during constant propagation, but treating the GP-computations as a special case is more efficient. This was also done by Srivastava and Wall [1994]. Finally, during the initial compaction phase, duplicate entries in the GOT are eliminated.…”

“…The elimination of unused data from a program has been considered in Srivastava and Wall [1994] and in Sweeney and Tip [1998]. Srivastava and Wall, describing a link-time optimization technique for improving the code for subroutine calls in Alpha executables, observed that the optimization allows the elimination of most of the global address table entries in the executables.…”

Link-Time Improvement of Scheme Programs

“…We expect that the overall conclusion-that programs spend a significant amount of time in tag manipulation-holds for Scheme programs as well. Most of the prior work on link-time optimization has focused on imperative languages [7,12,15,16,19]. The differences in runtime characteristics between Scheme and C programs, as discussed above, can have a significant effect on the extent to which systems designed for executables resulting from (human-written) C programs will be effective on code generated from Scheme programs.…”