Difference between revisions of "BBench"

From gem5
Jump to: navigation, search
m (Running BBench on Android with gem5)
m (Running BBench on Android with gem5)
Line 17: Line 17:
 
# The bbench.rcS file should be placed wherever you keep your rcS scripts. This is ''typically path_to_gem5_root/configs/boot/''.
 
# The bbench.rcS file should be placed wherever you keep your rcS scripts. This is ''typically path_to_gem5_root/configs/boot/''.
 
#You will also need to modify your Benchmarks.py file to include bbench, i.e., include a line like the following: '''bbench':    [SysConfig('bbench.rcS', '256', 'android-mbr.img')]''
 
#You will also need to modify your Benchmarks.py file to include bbench, i.e., include a line like the following: '''bbench':    [SysConfig('bbench.rcS', '256', 'android-mbr.img')]''
 +
# Assuming you have followed the steps above, and built a full-system version of gem5 for ARM, you should be able to type ''/path_to_gem5_root/build/ARM_FS/m5.fast -b bbench'' to run Android on ARM.
  
 
'''Note: Android only runs on the ARM ISA.'''
 
'''Note: Android only runs on the ARM ISA.'''

Revision as of 14:07, 10 November 2011

This page is under heavy construction

Source

A tarball of the BBench source can be found here:

BBench Source

Running BBench on Android with gem5

Everything required to run BBench on gem5 is provided in the next section. You will need the kernel, the disk image (containing BBench), and the rcS script.

  1. Create a directory where you will place your gem5 full-system files, call this directory system. Inside this directory create two separate directories, one called binaries and the other called disks.
  2. Point the M5_PATH environment variable to this location. E.g, type export M5_PATH=path_to_gem5_system/system/
  3. Place the kernel in the path_to_gem5_system/system/binaries/ directory
  4. Place the disk image in the path_to_gem5_system/system/disks/ directory
  5. Build the ARM bootloader and place it in the same directory as the kernel. The code for the bootloader is typically in path_to_gem5_root/system/arm/simple_bootloader/. Simply type make in this directory and the bootloader will be compiled. Copy the resulting boot.arm file to the same directory where you kernel is located. E.g., path_to_gem5_system/system/binaries/
  6. The default gem5 scripts look for a kernel called vmlinux.arm.smp.fb.2.6.38.8 and a disk image called android-mbr.img. You will have to either rename the files or create softlinks to them once you have downloaded and untarred them.
  7. The bbench.rcS file should be placed wherever you keep your rcS scripts. This is typically path_to_gem5_root/configs/boot/.
  8. You will also need to modify your Benchmarks.py file to include bbench, i.e., include a line like the following: 'bbench': [SysConfig('bbench.rcS', '256', 'android-mbr.img')]
  9. Assuming you have followed the steps above, and built a full-system version of gem5 for ARM, you should be able to type /path_to_gem5_root/build/ARM_FS/m5.fast -b bbench to run Android on ARM.

Note: Android only runs on the ARM ISA.

Android Full-System Files

These files contain everything you need to get Android, and BBench, up and running on gem5.

  • Kernel -- Pre-compiled Android kernel and config file.
  • Disk Image -- Disk image with a pre-compiled Android Gingerbread file system. This disk image contains all of the files generated during initial boot, as well as a self-terminating version of BBench and BusyBox.
  • Clean Disk Image -- Disk image containing only the pre-compiled Android Gingerbread file system. This disk image contains no benchmarks and has never been booted.
  • rcS Script -- Script for automating a BBench run on gem5.

Building Your Own Android File System and Kernel

Coming Soon!

Other Benchmarks

Coming soon!

Publications

If you use BBench or any of our benchmarks in your work please cite our IISWC 2011 paper:

A. Gutierrez, R.G. Dreslinski, T.F. Wenisch, T. Mudge, A. Saidi, C. Emmons, and N. Paver. Full-System Analysis and Characterization of Interactive Smartphone Applications. IEEE International Symposium on Workload Characterization, Austin, TX, November 2011.