HCoop / hcoop / debian / mlton.git / blame
| Commit | Line | Data |
|---|---|---|
| 7f918cf1 CE |
1 | RunningOnSparc |
| 2 | ============== | |
| 3 | ||
| 4 | MLton runs fine on the Sparc architecture. | |
| 5 | ||
| 6 | == Notes == | |
| 7 | ||
| 8 | * When compiling for Sparc, MLton targets the 32-bit Sparc | |
| 9 | architecture (i.e., Sparc V8). | |
| 10 | ||
| 11 | * When compiling for Sparc, MLton doesn't support native code | |
| 12 | generation (`-codegen native`). Hence, performance is not as good as | |
| 13 | it might be and compile times are longer. Also, the quality of code | |
| 14 | generated by `gcc` is important. By default, MLton calls `gcc -O1`. | |
| 15 | You can change this by calling MLton with `-cc-opt -O2`. We have seen | |
| 16 | this speed up some programs by as much as 30%, especially those | |
| 17 | involving floating point; however, it can also more than double | |
| 18 | compile times. | |
| 19 | ||
| 20 | * When compiling for Sparc, MLton uses `-align 8` by default. While | |
| 21 | this speeds up reals, it also may increase object sizes. If your | |
| 22 | program does not make significant use of reals, you might see a | |
| 23 | speedup with `-align 4`. | |
| 24 | ||
| 25 | == Known issues == | |
| 26 | ||
| 27 | * Bootstrapping on the <:RunningOnSparc:Sparc architecture> is so slow | |
| 28 | as to be impractical (many hours on a 500MHz UltraSparc). For this | |
| 29 | reason, we strongly recommend building with a | |
| 30 | <:CrossCompiling:cross compiler>. | |
| 31 | ||
| 32 | == Also see == | |
| 33 | ||
| 34 | * <:RunningOnSolaris:> |