Gabriels with no-op elimination.

[Rob.MacLachlan@LISP-PMAX2.SLISP.CS.CMU.EDU]

I have times for the Gabriel benchmarks with no-op elimination
optimization (except for fprint and fread, since they are part of the
system, and have not been recompiled yet.)

The first two times are for Allegro and CMU CL on a 24meg DS3100.  The next
two columns are the CMU time normalized to the previous CMU time (i.e. the
improvement from no-op elimination) and the CMU time normalized to Allegro
(net improvment over Allegro.)

Overall, the biggest improvements were in the benchmarks where we were
weakest w.r.t. Allegro.  The geometric mean of the Allegro-normalized
numbers is 1.43.


		Allegro	CMU	X old CMU	X Allegro
________________________________________________________________________
boyer: 		10	2.3    	1.20   		4.35
browse:		11	7.7    	1.11   		1.43
ctak:		.53	0.49    1.04    	1.08
dderiv:		1.4	0.97    1.13    	1.44
deriv:		1.1	0.87    1.10    	1.26
destructive: 	.68	0.37    1.03    	1.84
iterative-div2: .31	0.36    1.06    	0.86
recursive-div2: .56	0.5    	1.08    	1.12
puzzle:		1.4	0.99    1.19    	1.41
tak:		.1	0.10    1.10    	1.00
rtak:		.1	0.10    1.10    	1.00
takl:		.62	0.56    1.21    	1.11
stak:		1.1	0.45    1.04    	2.44
init-traverse:	2.4	0.73    1.26    	3.29
run-traverse:	7.8	9.1	1.14    	0.86
triangle:	30	21.3    1.23    	1.41
fft:		0.53	0.39    1.15    	1.36