Help / Release / J 5.04 / f/ .g Improved

>> << Pri JfC LJ Phr Dic Voc !: Rel NuVoc wd Help Release

---

f/ .g Improved

initial writing: 2004-09-01 last updated: 2005-03-18

The following inner products are now faster (and also leaner for the Boolean inner products).

+/ .*				Boolean, integer, floating point, complex
+./ .*.		+./ .*		Boolean; faster and leaner
~:/ .*.		~:/ .*		Boolean; faster and leaner

The improvement is minor with small arrays but is substantial with larger arrays, and is larger for CPUs with a large on-die cache. The following benchmarks demonstrate the improvement.

		m		p		n		Type		J5.04		J5.03		Ratio
+./ .*.		800		100		800		Boolean		0.01229828		0.06245972		5.08
		800		200		800		Boolean		0.02436215		0.12242737		5.03
		800		400		800		Boolean		0.04842964		0.48997266		10.12
		800		800		800		Boolean		0.10693234		1.19284602		11.16
+/ .*		800		100		800		Boolean		0.01522230		0.04825642		3.17
		800		200		800		Boolean		0.02714897		0.09307329		3.43
		800		400		800		Boolean		0.05281650		0.18381550		3.48
		800		800		800		Boolean		0.11380058		0.36863960		3.24
+/ .*		800		100		800		integer		0.18923574		0.16371892		0.87
		800		200		800		integer		0.38702915		0.42442516		1.10
		800		400		800		integer		0.75412938		0.86987551		1.15
		800		800		800		integer		1.49535773		4.33534356		2.90
+/ .*		800		100		800		real		0.22889839		0.74948352		3.27
		800		200		800		real		0.44844771		1.53543945		3.42
		800		400		800		real		0.89166200		4.73978968		5.32
		800		800		800		real		1.77981982		10.93269003		6.14
+/ .*		800		100		800		complex		0.65384512		0.98544341		1.51
		800		200		800		complex		1.31073778		2.74356425		2.09
		800		400		800		complex		2.61397222		5.95522645		2.28
		800		800		800		complex		5.21419328		12.33558427		2.37

---

>> << Pri JfC LJ Phr Dic Voc !: Rel NuVoc wd Help Release