Proceedings of the 26th Annual International Symposium on Microarchitecture 1993
DOI: 10.1109/micro.1993.282738
|View full text |Cite
|
Sign up to set email alerts
|

GPMB—software pipelining branch-intensive loops

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1

Citation Types

0
2
0

Year Published

1995
1995
2012
2012

Publication Types

Select...
4
3

Relationship

0
7

Authors

Journals

citations
Cited by 23 publications
(2 citation statements)
references
References 10 publications
0
2
0
Order By: Relevance
“…Modulo scheduling with if-conversion has been shown to perform better than GURPR* which under schedules its resources [24]. The most recent technique proposed by the GURPR* researchers, GPMB, allows multi-way branches but still uses a single I1 [22]. GPMB and modulo scheduling with predicated execution have the same 11; modulo scheduling with predicated execution has lower code expansion since it does not require code copying; and GPMB has better function unit utilization since only one path executes as opposed to all paths with predicated execution [22].…”
Section: Related Workmentioning
confidence: 99%
“…Modulo scheduling with if-conversion has been shown to perform better than GURPR* which under schedules its resources [24]. The most recent technique proposed by the GURPR* researchers, GPMB, allows multi-way branches but still uses a single I1 [22]. GPMB and modulo scheduling with predicated execution have the same 11; modulo scheduling with predicated execution has lower code expansion since it does not require code copying; and GPMB has better function unit utilization since only one path executes as opposed to all paths with predicated execution [22].…”
Section: Related Workmentioning
confidence: 99%
“…Loops that do not contain conditional branches (IF-THEN-ELSE constructs) have been both theoretically and practically analyzed in detail [2,5,9,11,20]. In contrast, there are few theoretical and empirical results for loops with conditional branches [10,12,22,27], and several effective methods for their optimization [6,7,14,16,17,23,24,25,26,28,30]. This work addresses the problem of parallelization of such loops.…”
Section: Introductionmentioning
confidence: 99%