gcc vs egcs -- egcs loses

Stephen Weeks sweeks@research.nj.nec.com
Tue, 9 Mar 1999 10:08:33 -0500


Here is a comparison of egcs -O9 and gcc -O1.  The first listing of
each benchmark are the egcs numbers.  The second are the gcc numbers.
In terms of run time, using egcs hurt on lexgen, mandelbrot, zern and
didn't help much on anything.

		compile	run	code
		time	time	size
		-------	-----	------
barnes-hut	2.51	10.93	47,361
barnes-hut	1.53	11.97	43,258
count-graphs	2.55	10.27	48,851
count-graphs	1.52	10.0	43,924
fft		1.8	41.04	33,995
fft		0.93	42.87	29,020
knuth-bendix	5.26	15.26	77,716
knuth-bendix	3.05	15.28	74,405
lexgen		14.41	36.36	195,861
lexgen		7.98	31.76	192,950
life		2.29	53.68	44,107
life		1.29	54.15	40,300
logic		435.33	49.45	419,675
logic		203.27	49.26	418,588
mandelbrot	0.35	19.05	19,263
mandelbrot	0.32	15.98	15,392
matrix-multiply	0.39	12.85	19,835
matrix-multiply	0.35	12.05	15,932
mlyacc		113.04	20.67	862,369
mlyacc		54.76	20.58	856,546
nucleic		2.76	15.86	57,061
nucleic		1.73	17.07	53,046
ratio-regions	4.94	24.7	58,887
ratio-regions	2.12	18.98	45,688
simple		32.71	18.33	349,030
simple		18.51	18.4	354,839
tsp		1.87	25.8	34,421
tsp		1.06	25.42	31,198
vliw		71.35	19.97	946,745
vliw		43.4	19.11	931,626
zern		0.93	50.01	27,804
zern		0.65	44.33	24,029
smlc		2185.67	102.96	11,357,937
smlc		1350.15	101.39	11,336,169