|
Accession number;99A0182298
|
| Title;Experiments of the Instruction Scheduling using Genetic Algorithm. |
| Author;
ITO ISAMU
(Shizuoka Univ., Grad. Sch.)
UMETANI YUKIO
(Shizuoka Univ., Fac. of Information)
|
Journal Title;Joho Shori Gakkai Kenkyu Hokoku
|
Journal Code:Z0031B
|
ISSN:0919-6072
|
|
VOL.98;NO.115(HPC-74);PAGE.65-70(1998)
|
| Figure&Table&Reference;FIG.6, TBL.2, REF.6 |
| Pub. Country;Japan |
| Language;Japanese |
| Abstract;An Application of Genetic Algorithm to the instruction sequence optimization is proposed to squeeze out RISC processor performance. By this way, far better results than those obtained by the scheduling of GNU-C compiler are obtained, and a stable optimization independent of the features of the machine is attained. The peformance has been evaluated using 24 Livermore kernels on 3 UNIX machines (Ultra2, Ultra1, SPARC), and higher performance up to 25.9% compared to the conventional scheduling is obtained. We also point out that conventional scheduling methods depend on a specific machine feature, and doesn't care for the date dependence by registers. (author abst.) |
|
|
|
Related Articles;
|
