ScalaLab and GroovyLab: Comparing Scala and Groovy for Scientific Computing

Table 2

Results of some basic benchmarks.

	ScalaLab (secs)	SciLab 5.21 (secs),; SciLab 5.5	MATLAB 7.1 (secs); MATLAB 2012b	GroovyLab (secs)
Matrix multiplication with matrix sizes: (2000, 2500) × (2500, 3000)	0.9 secs using Native BLAS combined with Java multithreading	61.8, 5.05	13.05, 0.6	The same with ScalaLab
LU
1000	0.3	3.13, 2.42	0.36, 0.03	The same as ScalaLab
1500	1.2	3.82, 2.1	1.18, 0.04	As ScalaLab
2000	2.9	6.42, 1.6	2.72, 0.09	As ScalaLab
inv
1000	2.7	12.97, 1.6	1.3, 0.05	As ScalaLab
1500	7.8	13.14, 2.5	4.5, 0.15	As ScalaLab
2000	9.31	19.07, 3.2	5.9, 0.3	As ScalaLab
QR
1000	1.03	4.3, 4.2	1.2, 0.04	As ScalaLab
1500	3.7	9.96, 9.9	4.26, 0.2	As ScalaLab
2000	9.25	19.69, 19.3	9.89, 0.3	As ScalaLab
Matrix access scripting benchmark	0.03	32.16, 32.67	10.58, 0.32	0.031 static compilation,; 0.156 with primitive ops,; 0.211 with invoke dynamic
FFT; 100 ffts of 16384 sized signal	Oregon DSP: real case: 0.05,; complex case: 0.095; JTransforms: real case: 0.07; complex case: 0.11,; Apache Common Maths: complex case: 0.5; Numerical Recipes (Java Translation): real case: 0.09; complex case: 0.12	Real case: 2.32; Complex case: 4.2	Real case: 0.05; Complex case: 0.08	The Java libraries for FFT are the same as ScalaLab’s