alto: a link‐time optimizer for the Compaq Alpha

Abstract

Traditional optimizing compilers are limited in the scope of their optimizations by the fact that only a single function, or possibly a single module, is available for analysis and optimization. In particular, this means that library routines cannot be optimized to specific calling contexts. Other optimization opportunities, exploiting information not available before link time, such as addresses of variables and the final code layout, are often ignored because linkers are traditionally unsophisticated. A possible solution is to carry out whole‐program optimization at link time. This paper describes alto, a link‐time optimizer for the Compaq Alpha architecture. It is able to realize significant performance improvements even for programs compiled with a good optimizing compiler with a high level of optimization. The resulting code is considerably faster than that obtained using the OM link‐time optimizer, even when the latter is used in conjunction with profile‐guided and inter‐file compile‐time optimizations. Copyright © 2001 John Wiley & Sons, Ltd.

Goal-Directed Value Profiling

“…The work described here has been implemented in the context of alto, a link-time optimizer we have constructed for the Compaq Alpha architecture [16]. Alto can be used to rewrite executable files, either to instrument them for profiling, or to carry out code optimization.…”

Reducing fetch architecture complexity using procedure inlining

“…We simulate the SPECint2000 benchmarks optimized using ALTO [10]. First, the benchmarks were compiled with the Compaq C V5.8-015 compiler on Compaq UNIX V4.0, using the -O2 optimization level without inlining.…”

“…Their results show that aggressive inlining can provide important performance improvements in some benchmarks. Likewise, the ALTO [10] optimizer is able to aggressively inline procedures, using profile-based code reordering to reduce the negative effects of inlining on the instruction cache. Aydin and Kaeli [3] take this one step further implementing cache line coloring algorithms.…”

“…The ALTO [10] optimizer is able to perform an aggressive procedure inlining. This inlining optimization is mainly controlled by the maximum resultant code size (MRCS), that is, the maximum number of instructions that an inlined portion of code should have.…”

“…We use the ALTO [10] tool to perform an aggressive procedure inlining. The main drawback of this optimization is that it increases the code size, and thus increases the number of instruction cache misses.…”