SGI Pro64/Open64 Compiler
We have this installed locally, ask Ben Breech for access, as well as access to working benchmarks
Tutorial*** on the Pro64/Open64, including introduction and framework, code generation, and case studies.
J. N. Amaral, C. Barton, A. C. Macdonell, and M. McNaughton, "Using the SGI Pro64 Open Source Compiler Infra-Structure for Teaching and Research," 13th Symposium on Computer Architecture and High Performance Computing, Pirenopolis, GO, Brazil, September, 2001. ( Abstract*) ( PostScript**)
Amaral's 680 class* (similar to this class)
Follow links to Open64***, Ski***, and Native User Environment*** to find the software, and follow installation instructions at each site. Install on a Linux box with a IA-32 (Pentium) chip.
SGI Pro64* was written for the IA-64 architecture (a.k.a. Itanium*), it is now called the Open64 Compiler***.
We will probably use the Ski IA-64 Simulator*** and the Native User Environment*** (NUE) from HP in order to run the Open64 on a IA-32 (Pentium) computer running Linux.
Problems installing? Check out these mail archives to see some troubles others have had.
Architecture (phases)
See Overview of The Pro64 Code Generator**, where the rightmost chart was found.
|
|
|
Or the Tutorial*** on the Pro64/Open64, where the leftmost chart was found.
Definitions
WHIRL = Intermediate Representation of Pro64
IGLS = Integrated Global Local Scheduling
CGIR = Code Generation Intermediate Representation
GRA/LRA = Global and Local Register Allocation
EBO = Extended Block Optimization
See Overview of The Pro64 Code Generator** and especially, the Tutorial***
The Open64 Compiler performs the following optimizations:
Inter-Procedural Analysis, alias analysis, array section, code layout analysis, inlining, cloning, dead function and variable elimination, and constant propagation.
Loop Level: loop fission, loop fusion, loop unroll and jam, loop interchange, loop peeling, loop tiling, and vector data prefetching
Global: partial redundancy elimination, support for speculative code motion