On 12/13/2010 02:06 AM, Jiri Denemark wrote:
So far, CPUID data were stored in two different data structures. First of them was a structure allowing direct access for CPUID data according to function number and the second was a plain array of struct cpuX86cpuid. This was a silly design which resulted in converting data from one type to the other and back again or implementing similar functionality for both data structures.
The patch leaves only the direct access structure. This makes the code both smaller and more maintainable since operations on different objects can use common low-level operations.
All 57 tests for cpu subsystem still pass after this rewrite. --- src/cpu/cpu_x86.c | 576 +++++++++++++++++++++-------------------------- src/cpu/cpu_x86_data.h | 4 +- 2 files changed, 259 insertions(+), 321 deletions(-)
Nice cleanup. ACK.
+ +static void +x86DataIteratorInit(struct data_iterator *iter, + union cpuData *data) +{ + struct data_iterator init = DATA_ITERATOR_INIT(data); + + *iter = init;
Is this any more efficient if init is marked static? Or can we even bypass the init variable and just do *iter = DATA_ITERATOR_INIT(data)? -- Eric Blake eblake@redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org