Last update: Thu Apr 12 03:37:15 MDT 2012
Top |
Symbols |
Numbers |
Math |
A |
B |
C |
D |
E |
F |
G |
H |
I |
J |
K |
L |
M |
N |
O |
P |
Q |
R |
S |
T |
U |
V |
W |
X |
Y |
Z
BibTeX entry
@Article{Lee:1991:FPP,
author = "Roland L. Lee and Alex Y. Kwok and Fay{\'e} A.
Briggs",
title = "The Floating-Point Performance of a Superscalar
{SPARC} Processor",
journal = j-SIGPLAN,
volume = "26",
number = "4",
pages = "28--37",
month = apr,
year = "1991",
CODEN = "SINODQ",
ISSN = "0362-1340 (print), 1523-2867 (print), 1558-1160 (electronic)",
ISSN-L = "0362-1340",
bibdate = "Tue Dec 12 09:20:21 MST 1995",
abstract = "The floating point performance of superscalar SPARC
processors is evaluated based on empirical data from 12
benchmarks. This evaluation is done in the context of
two software instruction scheduling optimizations: loop
unrolling and software pipelining, and for three
machine models: 1-scalar, 2-scalar and 4-scalar. The
authors also consider the effect of the memory system
on the performance improvements. Superscalar hardware
alone exhibit little performance improvement without
software optimization. Of the two scheduling methods,
software pipelining more effectively takes advantage of
increased hardware parallelism, and achieves near
optimal speedup on the 4-scalar machine model. The
performance of loop unrolling is restricted by the
limited number of floating point registers in the SPARC
architecture. The best performance level is obtained by
applying both optimization techniques. A superscalar
SPARC processor can provide improved floating point
performance but with significant software and hardware
development costs.",
acknowledgement = ack-nhfb,
affiliation = "Sun Microsyst. Inc., Mountain View, CA, USA",
classification = "C5220 (Computer architecture); C5470 (Performance
evaluation and testing)",
confdate = "8-11 April 1991",
conflocation = "Santa Clara, CA, USA",
confsponsor = "IEEE; ACM",
keywords = "Benchmarks; Development costs; Floating point
performance; Floating point registers; Hardware
parallelism; Loop unrolling; Memory system; N-scalar
machine models; Optimal speedup; Software instruction
scheduling optimizations; Software pipelining; SPARC
architecture; Superscalar SPARC processor",
thesaurus = "Optimisation; Parallel architectures; Performance
evaluation; Pipeline processing; Scheduling",
}
Related entries
- achieve,
26(6)145,
27(7)200,
27(9)248,
27(12)20,
28(6)237,
28(7)13,
28(7)64,
28(7)64-1,
28(7)83,
29(4)23,
29(6)337,
29(6)337-1,
29(11)12,
29(11)38,
29(11)208,
29(11)232,
29(11)252,
30(3)23,
30(6)67,
30(6)279,
30(8)189,
30(8)207,
30(11)1,
32(6)75
- advantage,
25(6)9,
25(6)296,
26(4)290,
27(7)116,
27(7)128,
28(6)187,
28(7)102,
29(6)107,
29(6)206,
29(6)266,
29(11)2,
29(11)38,
29(11)61,
29(11)219,
30(3)119,
30(6)79-1,
30(6)151,
30(6)205,
30(11)50,
30(11)146-1
- alone,
28(6)1,
29(6)36
- applying,
6(4)72,
27(7)212,
27(7)249,
29(6)73,
29(6)171,
30(6)56,
30(6)205
- author,
23(1)17,
23(12)728,
25(6)112,
25(10)237,
25(12)85,
26(6)145,
26(6)219,
27(1)95,
27(12)20,
27(12)61,
28(3)1,
28(3)37,
28(3)361,
28(3)365,
28(7)13,
28(7)33,
28(7)44,
28(7)54-1,
28(7)64,
28(7)83,
28(7)92,
28(7)112,
28(7)129,
28(7)149,
28(7)187,
28(7)198,
28(7)208,
28(7)229,
28(7)239,
28(8)90,
28(10)449,
31(9)290,
32(3)32,
32(10)345,
33(2)40,
33(3)37,
33(3)57,
33(3)65,
33(9)25,
33(10)421,
33(12)32,
34(2)40,
34(4)37,
34(6)36,
34(12)35
- benchmark,
6(4)159,
25(12)54,
25(12)85,
26(4)290,
27(7)44,
27(7)94,
27(7)106,
28(6)1,
28(6)258,
28(6)268,
28(7)102,
28(7)218,
28(7)249,
29(6)36,
29(6)49,
29(6)85,
29(6)171,
29(10)341,
29(11)86,
29(11)145,
29(11)196,
29(11)208,
29(11)232,
29(11)252,
29(11)286,
30(6)1,
30(6)13,
30(6)139,
30(6)218,
30(6)270,
30(8)29,
30(8)144,
31(10)83,
32(5)109,
32(5)122,
32(5)183,
32(5)235,
32(5)346-1,
33(11)35,
33(11)58,
33(11)71,
33(11)159,
33(11)170,
33(11)218,
33(11)252,
33(11)262,
33(11)283,
34(3)68,
34(7)96
- best,
25(6)16,
25(6)92,
25(6)150,
27(7)32,
27(7)283,
27(7)331,
28(7)83,
28(10)46,
29(2)44,
29(6)73,
29(11)25,
29(11)171,
29(11)252,
29(11)308,
29(11)328,
30(6)67,
30(6)79-1,
30(6)151,
30(6)279,
30(8)123,
30(8)134,
30(11)88,
33(11)252
- both,
25(1)59,
25(6)9,
25(6)85-1,
25(6)102,
25(6)112,
26(7)83,
27(5)z,
27(7)82,
27(7)94,
27(7)175,
27(7)212,
27(7)249,
27(7)283,
27(10)452,
28(3)231,
28(3)299,
28(3)353,
28(3)357,
28(6)26,
28(6)177,
28(6)248,
28(7)13,
28(7)23,
28(7)54-1,
28(7)112,
28(7)129,
29(6)1,
29(6)290,
29(6)302,
29(8)59,
29(8)119,
29(11)2,
29(11)38,
29(11)61,
29(11)98,
29(11)110,
29(11)122,
29(11)171,
29(11)183,
29(11)252,
29(11)274,
29(11)308,
30(2)25,
30(3)50,
30(6)56,
30(6)67,
30(6)93,
30(6)130,
30(6)186,
30(6)205,
30(6)279,
30(8)29,
30(8)68,
30(8)102,
30(8)156,
30(8)179,
30(8)207,
30(11)20-1,
31(5)108
- C5220,
26(4)290,
26(6)219,
29(6)186,
29(11)98,
29(11)110
- C5470,
26(4)290,
29(11)76-1,
29(11)122,
29(11)132-1,
29(11)145,
29(11)158,
29(11)196,
29(11)219,
29(11)274,
29(12)87,
30(11)20-1
- consider,
25(6)92,
27(7)273,
29(11)171,
29(11)286
- context,
24(3)34,
26(4)75,
27(4)77,
27(7)22,
27(7)311,
28(6)156,
28(7)23,
28(7)187,
29(6)24,
29(6)218,
29(6)242,
29(6)242-1,
29(8)46,
29(8)111,
29(8)119,
29(11)308,
29(11)319,
29(11)328,
30(3)50,
30(3)94,
30(6)1,
30(8)48,
31(6)239,
32(5)85,
32(12)63,
32(12)63,
33(7)83
- cost,
25(6)66,
26(12)26,
27(7)188-1,
27(7)300,
27(7)311,
27(9)262,
28(6)217,
28(6)268,
28(7)218,
29(6)61,
29(6)73,
29(9)135,
29(10)324,
29(10)341,
29(11)51,
29(11)61,
29(11)76-1,
29(11)86,
29(11)98,
29(11)110,
29(11)158,
29(11)242,
29(11)252,
29(11)263,
29(11)274,
29(11)319,
29(12)66,
30(3)35,
30(3)50,
30(6)93,
30(6)103,
30(6)301,
30(8)189,
31(6)92,
31(10)306,
32(5)320,
32(5)320-1,
32(8)292,
32(10)342,
32(10)342-1,
33(7)51,
33(7)67,
34(7)20
- done,
25(6)85-1,
27(7)152,
27(7)300,
27(12)47,
28(3)69,
28(6)207-1,
29(6)49,
29(11)12,
29(11)86,
29(11)132-1,
30(3)71,
30(3)103,
30(8)217
- effect,
26(4)75,
26(4)290,
27(3)71,
27(7)1,
27(7)32,
27(7)224,
28(3)361,
28(6)13,
28(6)56,
28(6)237,
29(6)1,
29(6)49,
29(6)107,
29(6)147,
29(6)349,
29(6)349-1,
29(11)12,
29(11)328,
30(6)1,
30(6)13,
30(6)93,
30(6)130,
30(6)233,
30(8)199,
33(7)19,
34(1)51,
34(1)63
- effectively,
27(7)106,
28(6)46,
29(11)38,
29(11)242,
30(6)301,
30(8)166
- empirical,
25(3)1,
25(6)102,
26(11)184,
27(3)54,
27(12)71,
28(6)258,
29(2)44,
29(6)242,
29(6)242-1,
30(6)1,
30(6)13,
33(7)11,
33(7)83,
33(11)35,
33(11)272
- evaluated,
27(7)200,
28(6)1,
29(6)186,
29(11)98,
30(8)48
- exhibit,
28(3)361,
29(6)266,
29(11)145,
29(11)232,
29(11)252,
29(11)274
- floating,
25(1)59,
25(6)92,
25(6)112,
25(12)85,
26(4)290,
26(6)219,
29(6)61,
29(11)98,
33(9)103
- Floating-Point,
25(6)92,
25(6)112,
31(5)108,
32(3)57,
33(5)118
- improved,
27(7)249,
28(6)26,
28(7)33,
29(11)242,
29(11)252,
30(6)151,
30(8)102,
30(8)156,
34(10)325
- improvement,
23(1)17,
24(3)34,
25(6)16,
25(6)174,
26(6)219,
27(7)68,
28(6)78-1,
28(6)278,
29(6)36,
29(6)186,
29(6)266,
29(6)313,
29(6)349,
29(6)349-1,
29(11)12,
29(11)51,
29(11)232,
29(11)242,
29(11)252,
30(3)23,
30(6)186,
30(6)196,
30(6)205,
30(8)1,
30(8)29,
30(8)80-1,
30(11)1,
30(11)7
- increased,
25(6)66,
28(6)177,
29(6)1,
29(6)97,
29(11)2,
29(11)274,
29(11)319,
30(11)20-1
- limited,
27(7)188-1,
29(6)36,
29(6)290,
29(11)51,
29(11)110,
29(11)132-1,
29(11)297,
29(11)308,
29(11)319,
29(11)328,
30(8)123,
33(7)59
- little,
25(6)78,
27(9)262,
28(6)177,
28(6)300,
28(7)64,
28(7)149,
29(6)186,
29(6)230,
29(11)297,
30(3)111,
30(6)13
- near,
30(6)246
- obtained,
27(7)128,
29(6)147,
29(11)274,
30(6)79-1,
30(6)93,
30(6)205,
30(8)134,
30(11)70,
30(11)79,
30(11)99,
30(11)125
- optimal,
25(6)85-1,
26(6)256,
26(6)256-1,
26(7)1,
27(4)68,
27(7)200,
27(7)224,
27(9)149,
28(1)68,
28(6)248,
28(7)64,
28(7)64-1,
28(12)1,
29(6)147,
29(6)313,
29(11)328,
30(3)71,
30(6)32,
30(6)186,
30(6)246,
30(8)80-1,
30(8)134,
30(8)207,
31(5)1,
32(5)194,
32(8)188,
33(11)218
- pipelining,
26(6)219,
27(7)249,
27(7)283,
28(6)100,
28(6)290,
29(6)36,
29(11)98,
30(6)139,
30(6)196,
31(5)1
- point,
25(1)59,
25(6)92,
25(6)112,
25(10)312,
25(12)85,
26(4)290,
26(6)219,
27(7)32,
27(7)224,
27(7)235,
27(9)223,
28(3)69,
28(6)68,
28(6)197,
29(6)1,
29(6)61,
29(6)121,
29(6)349,
29(6)349-1,
29(8)59,
29(10)85,
29(11)12,
29(11)98,
29(11)122,
29(11)208,
31(1)9,
31(1)9-1,
31(3)6,
31(3)6-1,
33(9)103
- Point, Floating-,
25(6)92,
25(6)112,
31(5)108,
32(3)57,
33(5)118
- restricted,
28(6)68,
28(7)112,
29(6)13,
29(11)183,
30(3)1,
30(8)48,
30(11)99
- significant,
17(9)18,
25(6)85-1,
26(4)290,
27(7)55,
27(7)68,
27(7)273,
28(3)133,
28(3)149,
28(3)177,
28(3)209,
28(3)347,
28(4)58,
28(6)36,
28(6)187,
28(6)197,
28(7)64,
28(12)169,
29(6)36,
29(6)147,
29(6)206,
29(6)218,
29(6)349,
29(6)349-1,
29(8)35,
29(11)25,
29(11)51,
29(11)183,
30(3)119,
30(6)79-1,
30(6)218,
30(6)315,
30(8)48,
30(11)125,
33(11)252
- SPARC,
26(4)290,
27(7)22,
28(6)1,
28(6)197,
28(7)218,
29(6)313,
30(3)119,
30(6)103,
30(6)139
- speedup,
17(9)18,
23(1)17,
25(6)53,
25(6)311,
28(6)36,
28(7)83,
28(7)218,
29(11)183,
30(4)29,
30(6)151,
30(6)218,
30(8)48,
30(8)112,
30(8)179,
30(11)125,
33(11)252
- superscalar,
26(3)68,
26(4)53,
26(6)241,
27(7)283,
27(7)322,
27(9)238,
27(9)248,
29(6)36,
29(6)97,
29(11)183,
34(7)35
- take,
25(6)165,
25(6)296,
28(6)187,
28(6)227,
28(6)237,
28(8)90,
29(6)107,
29(6)196,
29(11)2,
29(11)242,
30(6)79-1,
30(8)134,
30(8)189,
30(11)134,
30(11)146-1
- three,
27(7)22,
27(7)44,
27(7)200,
27(7)212,
28(6)13,
28(6)237,
28(7)33,
28(7)187,
28(7)198,
29(6)1,
29(6)186,
29(8)46,
29(11)61,
29(11)98,
29(11)286,
29(11)297,
30(3)83,
30(3)94,
30(4)29,
30(6)151,
30(8)68,
30(8)80-1,
30(11)50,
30(11)117
- unrolling,
28(6)68,
29(12)25,
30(6)151