E.2 Parallel computing

Parallel computing in SINGULAR

The available ressources are given by the number of cpus (set by the command line parameter --cpus=.., or the environment variable SINGULAR_CPUS, defaults to the number of processors currently online) and the number of threads (set by the command line parameter --threads=.., defaults to 1). Threads are currently used with FLINT and NTL.
If SINGULAR is uses as a libary, the default are 1 in botrh cases.
Setting --cpus to 0 disables the creation of new processes and chooses a sequentiell algorithm if possible.
Setting --cpus to 1 chooses a sequentiell algorithm in many cases.
Using parallel algorithms can reduce the overall computing time, by can also increase the memory usage by a factor given by the number of used cpu cores.

Via modular methods, requires QQ as coefficient field:

• assPrimes: assPrimes
• chinrem: chinrem
• farey: farey
• locNormal: locNormal
• modNormal: modNormal
• modPrimdecGTZ: modPrimdecGTZ, see also: primdecGTZ
• modQuotient: modQuotient, see also: quotient
• modSat: modSat, see also: sat
• modStd: modStd, see also: std, groebner
• pdf: pfd
• syModStd: syModStd
• system("verifyGB",I): system
• zeroRadical: zeroRadical

Via modular methods, requires ZZ as coefficient domain:

• primdecZ: primdecZ

Via modular methods, requires a function field over QQ as coefficient field:

• ffmodStd: ffmodStd

Via modular methods, requires a algebraic field over QQ as coefficient field:

• nfmodStd: nfmodStd
• nfmodSyz: nfmodSyz

Via different algorithms:

• primdec: primdec
• minAss: minAss
• signatureNemethi: signatureNemethi

Frameworks for parallel algorithms:

• for modular methods: modular_lib
• for different algorithms at once: watchdog