Skip to content

Latest commit

 

History

History

distributions

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
MemcmpGoogleA.csv
MemcmpGoogleA.csv
 
 
MemcmpGoogleB.csv
MemcmpGoogleB.csv
 
 
MemcmpGoogleD.csv
MemcmpGoogleD.csv
 
 
MemcmpGoogleL.csv
MemcmpGoogleL.csv
 
 
MemcmpGoogleM.csv
MemcmpGoogleM.csv
 
 
MemcmpGoogleQ.csv
MemcmpGoogleQ.csv
 
 
MemcmpGoogleS.csv
MemcmpGoogleS.csv
 
 
MemcmpGoogleU.csv
MemcmpGoogleU.csv
 
 
MemcmpGoogleW.csv
MemcmpGoogleW.csv
 
 
MemcpyGoogleA.csv
MemcpyGoogleA.csv
 
 
MemcpyGoogleB.csv
MemcpyGoogleB.csv
 
 
MemcpyGoogleD.csv
MemcpyGoogleD.csv
 
 
MemcpyGoogleL.csv
MemcpyGoogleL.csv
 
 
MemcpyGoogleM.csv
MemcpyGoogleM.csv
 
 
MemcpyGoogleQ.csv
MemcpyGoogleQ.csv
 
 
MemcpyGoogleS.csv
MemcpyGoogleS.csv
 
 
MemcpyGoogleU.csv
MemcpyGoogleU.csv
 
 
MemcpyGoogleW.csv
MemcpyGoogleW.csv
 
 
MemmoveGoogleA.csv
MemmoveGoogleA.csv
 
 
MemmoveGoogleB.csv
MemmoveGoogleB.csv
 
 
MemmoveGoogleD.csv
MemmoveGoogleD.csv
 
 
MemmoveGoogleL.csv
MemmoveGoogleL.csv
 
 
MemmoveGoogleM.csv
MemmoveGoogleM.csv
 
 
MemmoveGoogleQ.csv
MemmoveGoogleQ.csv
 
 
MemmoveGoogleS.csv
MemmoveGoogleS.csv
 
 
MemmoveGoogleU.csv
MemmoveGoogleU.csv
 
 
MemmoveGoogleW.csv
MemmoveGoogleW.csv
 
 
MemsetGoogleA.csv
MemsetGoogleA.csv
 
 
MemsetGoogleB.csv
MemsetGoogleB.csv
 
 
MemsetGoogleD.csv
MemsetGoogleD.csv
 
 
MemsetGoogleL.csv
MemsetGoogleL.csv
 
 
MemsetGoogleM.csv
MemsetGoogleM.csv
 
 
MemsetGoogleQ.csv
MemsetGoogleQ.csv
 
 
MemsetGoogleS.csv
MemsetGoogleS.csv
 
 
MemsetGoogleU.csv
MemsetGoogleU.csv
 
 
MemsetGoogleW.csv
MemsetGoogleW.csv
 
 
README.md
README.md
 
 
Uniform384To4096.csv
Uniform384To4096.csv
 
 

README.md

Size distributions for memory functions under specific workloads

This folder contains a set of files that are included from libc/benchmarks/MemorySizeDistributions.cpp.

Offloading this data to individual files helps

  • C++ editors (large arrays are usually not well handled by editors),
  • and allows processing data by other tools to perform analysis or graph rendering.

Format

Most filenames are of the form {MemoryFunctionName}{WorkloadID}.csv. They contain a single line of comma separated real values representing the probability that a particular size occurs. e.g.

  • "0,1" indicates that only the size 1 occurs,
  • "0.5,0.5" indicates sizes 0 and 1 occur with the same frequency.

These files usually contains sizes from 0 to 4096 inclusive. To save on space trailing zeros are discarded.

Workloads

As identified in the automemcpy paper:

  • GoogleA <-> service 4
  • GoogleB <-> database 1
  • GoogleD <-> storage
  • GoogleL <-> logging
  • GoogleM <-> service 2
  • GoogleQ <-> database 2
  • GoogleS <-> database 3
  • GoogleU <-> service 3
  • GoogleW <-> service 1

Uniform384To4096 is an additional synthetic workload that simply returns a uniform repartition of the sizes from 384 to 4096 inclusive.

Note

Except for GoogleD, all distributions are gathered over one week worth of data.