On 04/23/2015 02:13 PM, Liang Li wrote:> When compiling kernel on westmere, the performance of eager FPU> is about 0.4% faster than lazy FPU.Do you have an theory why this is? What does the regression come from?