Decimal Arithmetic Benchmark

back to index
libmpdec
decimal benchmarks
multi precision benchmarks
Python/Java benchmarks

Benchmark against other decimal arithmetic libraries

This benchmark runs a modified version of the escape time algorithm for drawing a Mandelbrot set. It was chosen simply because it contains a large number of multiplications, additions and subtractions.

Libraries and their data types

The arbitrary precision data types and the IEEE types are compatible.

	arbitrary precision	IEEE 16 digits	IEEE 34 digits
libmpdec	mpd	-	-
DecNumber	decNumber	decDouble	DecQuad
IntelRDFPMathLib	-	bid64	bid128

64-bit Ubuntu, 3.16GHz Core 2 Duo, compiled with 64 bit libs, 10000000 iterations

	mpd	decDouble	decQuad	decNumber	bid64	bid128
16 digits	4.80s	6.01s	-	7.69s	2.42s	-
34 digits	6.71s	-	11.43s	13.20s	-	8.11s
9 digits	4.45s	-	-	5.76s	-	-
19 digits	4.64s	-	-	9.49s	-	-
38 digits	6.48s	-	-	16.08s	-	-

64-bit Ubuntu, 3.16GHz Core 2 Duo, compiled with 32 bit libs, 10000000 iterations

	mpd	decDouble	decQuad	decNumber	bid64	bid128
16 digits	7.84s	6.55s	-	10.46s	4.67s	-
34 digits	11.98s	-	13.88s	17.85s	-	18.79s
9 digits	6.61s	-	-	7.52s	-	-
19 digits	8.06s	-	-	10.10s	-	-
38 digits	11.93s	-	-	15.95s	-	-

32-bit Linux, 1.2GHz Celeron, 1000000 iterations

	mpd	decDouble	decQuad	decNumber	bid64	bid128
16 digits	3.03s	2.45s	-	4.29s	1.90s	-
34 digits	5.16s	-	5.02s	7.16s	-	11.06s
9 digits	2.36s	-	-	2.93s	-	-
19 digits	2.93s	-	-	4.07s	-	-
38 digits	4.96s	-	-	6.93s	-	-

Contact:

Stefan Krah <website @ bytereef.org>