Too low LLC MPKI on pointer chase benchmark

[branylagaffe]

I want to test statistics of KeenKraken So that we can assert the quality of the statistics of KeenKraken

Specification:

1. Micro benchmark are identified, and added to the base image
2. Test have been run and match micro benchmark expectation

Tasks:

• [x] Add microbenchmark

[branylagaffe]

I have added to the base images multiple micro benchmark which can be bundled with Buildroot.

• Dropwatch
• Hyperfine
• mbw
• perftest
• ramspeed
• stress
• stress-ng
• whetstone

[branylagaffe]

• [x] Use debug mode
• [x] Add VA Translation check

[branylagaffe]

@ayanchak1508 Could post maybe your graph here ? To keep a trace on how to reproduce and test for this + CLI command used to run the benchmark

[ayanchak1508]

This is the plot:

I just run: ./runq images/alp-trace where alp-trace is a personal config file that runs Alpine Linux. (attached)

The code for the microbenchmark is also present in the zip. git.zip

[branylagaffe]

@ayanchak1508 Please explain what happened

[branylagaffe]

Not using icount made QEMU use host time, therefore was privileged kernel timer interrupt instead of processing user space instruction. The benchmark was bc. of this false