Adapting code to processor architecture
From HP-SEE Wiki
(Difference between revisions)
(→Using IBM SDK for multicore acceleration) |
(→Using IBM SDK for multicore acceleration) |
||
Line 3: | Line 3: | ||
== Using IBM SDK for multicore acceleration == | == Using IBM SDK for multicore acceleration == | ||
- | + | '' Section contributed by IICT-BAS '' | |
The configure script or whatever build system is used, must be instructed to use the cross-compiler instead of the standard gcc or other, to perform compilation. Usually, it can be done by setting the environment. IBM XL compilers can produce code, specifically optimized for the PPC 450 double-hummer processor of the computing nodes. Only with IBM XL compiler it is possible to use full advantages of BlueGene/P, because it allows the execution of two times more instructions per unit time. | The configure script or whatever build system is used, must be instructed to use the cross-compiler instead of the standard gcc or other, to perform compilation. Usually, it can be done by setting the environment. IBM XL compilers can produce code, specifically optimized for the PPC 450 double-hummer processor of the computing nodes. Only with IBM XL compiler it is possible to use full advantages of BlueGene/P, because it allows the execution of two times more instructions per unit time. |
Revision as of 13:34, 10 April 2012
Contents |
Using GNU compilers
Using IBM SDK for multicore acceleration
Section contributed by IICT-BAS
The configure script or whatever build system is used, must be instructed to use the cross-compiler instead of the standard gcc or other, to perform compilation. Usually, it can be done by setting the environment. IBM XL compilers can produce code, specifically optimized for the PPC 450 double-hummer processor of the computing nodes. Only with IBM XL compiler it is possible to use full advantages of BlueGene/P, because it allows the execution of two times more instructions per unit time.