IPADS

User Tools

Site Tools

pub:projects:ssmalloc

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision		Previous revision
pub:projects:ssmalloc [2012/04/17 15:17] – created rootpub:projects:ssmalloc [2014/08/22 10:32] (current) – [Publications] root
Line 18: Line 18:
 Here is the figure of the overall architecture of SSMalloc. Here is the figure of the overall architecture of SSMalloc.
  
-{{ :pub:projects:ssmalloc-arch.jpg?nolink&500 |Figure 1: Overall Architecture of SSMalloc}}+{{:pub:projects:ssmalloc-arch.jpg?nolink&400|Figure 1: Overall Architecture of SSMalloc}}
  
 SSMalloc maintains a private heap for each application thread. A thread operates on its own heap for most requests, thus eliminates most of the synchronization. Each private heap holds several fixed-size memory chunks. A memory chunk contains many objects of the same size class. SSMalloc maintains a private heap for each application thread. A thread operates on its own heap for most requests, thus eliminates most of the synchronization. Each private heap holds several fixed-size memory chunks. A memory chunk contains many objects of the same size class.
Line 31: Line 31:
 Experiments using a number of commonly used allocation-benchmarks running on a 48-core machines show that for most cases SSMalloc outperforms prior systems in allocation latency and access locality, and provides more stable and scalable performance. Experiments using a number of commonly used allocation-benchmarks running on a 48-core machines show that for most cases SSMalloc outperforms prior systems in allocation latency and access locality, and provides more stable and scalable performance.
  
-Here is some of the experimental result.+Here is some experimental results. 
 + 
 +=== Sequential Performance === 
 + 
 +{{:pub:projects:seq.jpg|Figure 2: Relative Execution Time of Sequential Applications}} 
 + 
 +=== Multi-thread Scalability and Stableness === 
 + 
 +{{:pub:projects:recycle.jpg|Figure 3: Recycle Execution Time}} 
 +{{:pub:projects:shbench.jpg|Figure 4: shbench Speedup relative to glibc malloc}} 
 +{{:pub:projects:larson.jpg|Figure 5: Larson Throughput}} 
 + 
 +=== Locality Stableness === 
 + 
 +{{:pub:projects:cachemiss.jpg|Figure 6: Cache Miss Breakdown for WordCount}} 
 +{{:pub:projects:maptime.jpg|Figure 7: Execution Time of Map Phase of WordCount}}
  
 ===== Summary ===== ===== Summary =====
Line 40: Line 55:
  
 ===== Publications ===== ===== Publications =====
 +  *Ran Liu and Haibo Chen. SSMalloc: A Low-latency, Locality-conscious Memory Allocator with Stable Performance Scalability. In Proceedings of 3rd ACM SIGOPS Asia-Pacific Workshop on Systems (**APSys 2012**), Seoul, Korea. [{{:publications:ssmalloc-apsys2012.pdf|pdf}}]  
 +
  
 ===== Source Code ===== ===== Source Code =====
-(Coming soon)+[[https://github.com/Naruil/SSMalloc]]
pub/projects/ssmalloc.1334647027.txt.gz · Last modified: by root