[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Highly informal UX400S benchmarks
In response to the benchmarks sent out last week, out of curiosity, I
compiled, loaded, and ran the gabriel-benchmarks on my UX400S. I did nothing
to ensure cleanliness of world, etc. Note, however, that the top-level
functions in the benchmark file call the GC before running each function and
run each function some number of times where the number appears to depend on
the detected behavior.
The results, in general, show the UX to be twice as fast as the 36xx times
posted last week. The benchmarks were run under 7.4. With Symbolic's PR
saying 8.0 will generate faster UX code, hopefully we'll see even more
competitive times. Notice, also, that I do have an FPA (although I never need
it for what I do).
The most troublesome results in terms of disproportionate time for the UX
compared to the other (non-symbolics) systems seem to be CTAK and the Div2...
results. The Div2 functions are merely list consing and traversing skipping
every other element from the initial list. This function seems to use common
enough tasks for its behavior to be worrisome, but I didn't try to see how it
differed from other functions where the UX performed well.
Lisp Type: Symbolics Common Lisp
Lisp Version: System 418.194, Ivory revision 2
Software Type: Lisp Machine
Software Version: Genera 7.4 Ivory, System 418.194, Logical Pathnames Translation Files
NEWEST, RPC 407.12, MacIvory Support 412.42, Serial 402.16, Utilities 412.28, Server Utili
ties 410.3, Hardcopy 413.8, Zmail 410.9, LMFS 411.11, Tape 412.15, Nsage 407.9, Extended H
elp 409.2, RPC Development 401.6, MacIvory Development 401.6, Documentation Database 412.9
, IP-TCP 413.6, Solstice Genera Patches 1.90, UX Support 401.13, UX Development 401.3, X R
emote Screen 406.44, CLX 406.10, Network RPC 404.15, NFS Client 404.12, Experimental Sipe
2 1.0, Ivory Revision 2 (FPA enabled), FEP 315, FEP0:>I315-Loaders.flod(4), FEP0:>I315-Inf
o.flod(4), FEP0:>I315-Debug.flod(11), FEP0:>I315-Lisp.flod(4), FEP0:>I315-KERNEL.FEP(4), B
oot PROM version 308, UX Support Server 1.3b, genera program 1.3.1.13, ivory-life program
1.3.1.7, UX kernel life-support 1.3.1.2, SunOS (SATCHMO) 4.0.3, 1526x1129 B&W Screen, 3008
x2000 B&W Screen, Machine serial number 521, Allow user-defined keyboard translations for
VT100 emulator (from SYMBOL:>don>vt100-arrowkey-patch.lisp.2), Get TCP options right. (fro
m SYMBOL:>don>tcp-patch.lisp.2)
Machine Type: Symbolics UX400S
Machine Version: Ivory Revision 2
Machine Instance: SKEPTIC
Site: TRC
Features: (:XLIB :CLX-OVERLAPPING-ARRAYS :CLX-LITTLE-ENDIAN :DEFSTORAGE :LOOP
:DEFSTRUCT :DEBUG-SCHEDULER-QUEUES :NEW-SCHEDULER :LISPM :SYMBOLICS
:GENERA :ROW-MAJOR :IMACH :CHAOS :IEEE-FLOATING-POINT :SORT :FASLOAD
:STRING :NEWIO :ROMAN :TRACE :GRINDEF :GRIND)
Boyer: CPU: 6.876 Page: 0.633 Real: 7.509 (based on 3 calls)
Browse: CPU: 11.121 Page: 1.001 Real: 12.122 (based on 3 calls)
Destruct: CPU: 0.980 Page: 0.115 Real: 1.094 (based on 7 calls)
Traverse, Initialize: CPU: 3.546 Page: 0.118 Real: 3.663 (based on 3 calls)
Traverse, Traverse: CPU: 27.806 Page: 0.178 Real: 27.984 (based on 3 calls)
Tak: CPU: 0.226 Page: 0.003 Real: 0.229 (based on 27 calls)
STak: CPU: 1.374 Page: 0.010 Real: 1.384 (based on 5 calls)
CTak: CPU: 3.352 Page: 0.021 Real: 3.374 (based on 3 calls)
TakL: CPU: 2.706 Page: 0.017 Real: 2.724 (based on 3 calls)
TakR: CPU: 0.264 Page: 0.003 Real: 0.267 (based on 23 calls)
Deriv: CPU: 1.958 Page: 0.412 Real: 2.371 (based on 4 calls)
DDeriv: CPU: 1.889 Page: 0.434 Real: 2.323 (based on 4 calls)
Div2, Iterative: CPU: 1.096 Page: 0.301 Real: 1.397 (based on 6 calls)
Div2, Recursive: CPU: 1.764 Page: 0.315 Real: 2.079 (based on 4 calls)
FFT: CPU: 2.625 Page: 0.018 Real: 2.644 (based on 3 calls)
Puzzle: CPU: 14.355 Page: 0.092 Real: 14.447 (based on 3 calls)
FRPoly, Power = 2, r = x + y + z + 1: CPU: 0.000 Page: 0.002 Real: 0.002 (based o
n 2294 calls)
FRPoly, Power = 2, r2 = 1000r: CPU: 0.002 Page: 0.002 Real: 0.004 (based on 1408
calls)
FRPoly, Power = 2, r3 = r in flonums: CPU: 0.001 Page: 0.001 Real: 0.002 (based o
n 2045 calls)
FRPoly, Power = 5, r = x + y + z + 1: CPU: 0.023 Page: 0.003 Real: 0.026 (based o
n 243 calls)
FRPoly, Power = 5, r2 = 1000r: CPU: 0.114 Page: 0.006 Real: 0.120 (based on 52 ca
lls)
FRPoly, Power = 5, r3 = r in flonums: CPU: 0.024 Page: 0.003 Real: 0.027 (based o
n 232 calls)
FRPoly, Power = 10, r = x + y + z + 1: CPU: 0.263 Page: 0.025 Real: 0.288 (based
on 23 calls)
FRPoly, Power = 10, r2 = 1000r: CPU: 1.727 Page: 0.076 Real: 1.802 (based on 4 ca
lls)
FRPoly, Power = 10, r3 = r in flonums: CPU: 0.276 Page: 0.026 Real: 0.302 (based
on 22 calls)
FRPoly, Power = 15, r = x + y + z + 1: CPU: 1.744 Page: 0.134 Real: 1.878 (based
on 4 calls)
FRPoly, Power = 15, r2 = 1000r: CPU: 11.399 Page: 0.689 Real: 12.088 (based on 3 ca
lls)
FRPoly, Power = 15, r3 = r in flonums: CPU: 1.842 Page: 0.135 Real: 1.977 (based
on 4 calls)
(DEFPARAMETER GABRIEL-CPU '(
(BOYER 6.8761516e+0)
(BROWSE 1.1120778e+1)
(DESTRU 9.7952914e-1)
(TRAVERSE-INIT 3.5455475e+0)
(TRAVERSE 2.7805527e+1)
(TAK 2.2620267e-1)
(STAK 1.3735926e+0)
(CTAK 3.3520982e+0)
(TAKL 2.7062435e+0)
(TAKR 2.640737e-1)
(DERIV 1.9580348e+0)
(DDERIV 1.8894575e+0)
(DIV2-I 1.0962129e+0)
(DIV2-R 1.763562e+0)
(FFT 2.6253839e+0)
(PUZZLE 1.4355169e+1)
(TRIANG 9.0742065e+1)
(FRPOLY2R 0.0e+0)
(FRPOLY2R2 2.0470829e-3)
(FRPOLY2R3 1.0235414e-3)
(FRPOLY5R 2.2517912e-2)
(FRPOLY5R2 1.136131e-1)
(FRPOLY5R3 2.3541454e-2)
(FRPOLY10R 2.6305014e-1)
(FRPOLY10R2 1.7267144e+0)
(FRPOLY10R3 2.763562e-1)
(FRPOLY15R 1.7441146e+0)
(FRPOLY15R2 1.1399181e+1)
(FRPOLY15R3 1.8423746e+0)
)
(DEFPARAMETER GABRIEL-REAL '(
(BOYER 7.5087e+0)
(BROWSE 1.2121801e+1)
(DESTRU 1.0941658e+0)
(TRAVERSE-INIT 3.663255e+0)
(TRAVERSE 2.7983624e+1)
(TAK 2.2927329e-1)
(STAK 1.383828e+0)
(CTAK 3.3735926e+0)
(TAKL 2.7236438e+0)
(TAKR 2.6714432e-1)
(DERIV 2.370522e+0)
(DDERIV 2.3234391e+0)
(DIV2-I 1.3971341e+0)
(DIV2-R 2.0788126e+0)
(FFT 2.6438076e+0)
(PUZZLE 1.4447288e+1)
(TRIANG 9.1335724e+1)
(FRPOLY2R 2.0470829e-3)
(FRPOLY2R2 4.0941657e-3)
(FRPOLY2R3 2.0470829e-3)
(FRPOLY5R 2.5588537e-2)
(FRPOLY5R2 1.1975435e-1)
(FRPOLY5R3 2.6612077e-2)
(FRPOLY10R 2.8761515e-1)
(FRPOLY10R2 1.8024565e+0)
(FRPOLY10R3 3.0194473e-1)
(FRPOLY15R 1.8781986e+0)
(FRPOLY15R2 1.2088024e+1)
(FRPOLY15R3 1.9774821e+0)
)
NIL