Shows which features your CPU does (or does not) support. It consists of three utilities:
- dump. Prints profile: generic info about system and raw values of specific pre-defined registers. Written in C and assembly.
- convert. Checks profile syntax and translates it into the form convenient for the belowmentioned utility. Written in lex, yacc and C.
- parse. Interprets converted profile data and prints whether particular features are supported. Written in Scheme.
The following tools are required to build uProcessor:
- CMake.
- C compiler.
- Lex.
- Bison.
- Chicken Scheme with the
numbersegg.
Install prerequisites on Debian or Ubuntu:
sudo apt-get update
sudo apt-get install cmake gcc flex bison chicken-bin
chicken-install -s numbers
Install prerequisites on Fedora, RHEL or CentOS:
sudo yum install cmake gcc flex bison chicken
chicken-install -s numbers
Install prerequisites on OS X (given that you already have Xcode and Homebrew):
brew update
brew install cmake chicken
chicken-install -s numbers
Build commands for GNU/Linux, OS X and Solaris:
mkdir build
cd build
cmake ..
make
Build commands for Windows:
md build
cd build
cmake -G "NMake Makefiles" ..
nmake /nologo
All CPU-specific information is taken from MSRs (Model Specific Registers) and CPUID instruction. While CPUID is available to user space code, MSRs can be accessed only from kernel space.
The updump utility on GNU/Linux can use /dev/cpu/0/msr interface provided
by msr.ko kernel module (available out of the box in most distributions).
On OS X updump can use a similar interface provided by the companion
project. In both cases updump needs to
be run with root privileges. MSRs reading on Windows and Solaris is not
supported.
View CPU features:
sudo dump/updump | convert/upconvert | parse/upparse
Dump profile for further exploration:
sudo dump/updump > profile.txt
View CPU features from the previously saved profile:
convert/upconvert < profile.txt | parse/upparse