SC16: International Conference for High Performance Computing, Networking, Storage and Analysis 2016
DOI: 10.1109/sc.2016.46
|View full text |Cite
|
Sign up to set email alerts
|

Caliper: Performance Introspection for HPC Software Stacks

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1

Citation Types

0
29
0

Year Published

2019
2019
2024
2024

Publication Types

Select...
3
3
2

Relationship

1
7

Authors

Journals

citations
Cited by 78 publications
(44 citation statements)
references
References 12 publications
0
29
0
Order By: Relevance
“…Later on, we characterize the impact of DCDB on computational resources and its scalability with various test configurations (Section 6.2.2), stressing the communication and sampling subsystems. In this part, we focus on using the shared-memory version of the High-Performance Linpack (HPL) benchmark [14], supplied with the Intel MKL library 8 . Being a compute-bound application, tests performed against HPL give us insights on the behavior of DCDB in a worst-case scenario.…”
Section: Methodsmentioning
confidence: 99%
“…Later on, we characterize the impact of DCDB on computational resources and its scalability with various test configurations (Section 6.2.2), stressing the communication and sampling subsystems. In this part, we focus on using the shared-memory version of the High-Performance Linpack (HPL) benchmark [14], supplied with the Intel MKL library 8 . Being a compute-bound application, tests performed against HPL give us insights on the behavior of DCDB in a worst-case scenario.…”
Section: Methodsmentioning
confidence: 99%
“…A variety of performance measurement and analysis tools co-exist in the HPC ecosystem. Well known examples include TAU [26], Caliper [6], HPCToolkit [1], and LIKWID [28]. Each one of these tools provide their capabilities via design abstractions around the lower-level interfaces for the hardware and generally build upon the work of more specialized libraries such as PAPI [27], CUPTI [9], and Linux perf [11].…”
Section: Need For Composite Componentsmentioning
confidence: 99%
“…i.e., Radiation shielding, particle accelerator simulations, nuclear reactor design 6. i.e., Ability to associate measurements with either the derived or abstract object.…”
mentioning
confidence: 99%
“…The second is the Caliper [31] performance system. Caliper consists of the source-code annotation API, a backend runtime component that manages blackboard buffers and the generalized context tree, add-on support services for control tasks (such as I/O and controlling snapshots), and additional data producer, measurement control, and data consumer services.…”
Section: Related Workmentioning
confidence: 99%