Using the BYTE benchmark program on a 2.2 kernel compiled with gcc, and another one compiled with pgcc. The 10 % number increase was the average on FPU operations, on Integer operations the performance increase was slightly lower. Of course this was on the specific case of the AMD K6 processor, I don't know if the same percentage could be applied to other cpu's, although I think that the optimizations should even work better on pentium II and on the Pentium Pro using the -march option.
I was just trying to say that the it is possible to compile the kernel 2.2 with pgcc, cause as you might know it was'nt possible to do it with the 2.0.xx kernels, without apllying patches. Talking about stability I have a kernel 2.2pre9 compiled with amdk6 optimization, running without problems in a non-suported Super7 motherboard for one week now, this of courser is no proof of stability but it is a beggining I believe. MySQL for example runs perfectly and 12 % faster compiled with pggc. I know it is still very experimental, but as far as all the tests go pgcc makes faster code than egcs. And of course I know faster does'nt mean more stable, although I have'nt had any stability problems with pgcc compiled programs. The only problems that I have faced till now are with Mesa 3.0 and kernels of the 2.0.xx, those yes, I can't even compile them with pgcc.
Cool, I have tried compiling the kernel 2.2 with pgcc with mcpu and march optimizations and it worked, the result is an even faster kernel. I have compiled it with amdk6 optimizations and the result was a 10% faster kernel, try it cause if you thought the kernel 2.2 was fast with pgcc it gets even faster
Slashdot Mirror
Using the BYTE benchmark program on a 2.2 kernel compiled with gcc, and another one compiled with pgcc. The 10 % number increase was the average on FPU operations, on Integer operations the performance increase was slightly lower.
Of course this was on the specific case of the AMD K6 processor, I don't know if the same percentage could be applied to other cpu's, although I think that the optimizations should even work better on pentium II and on the Pentium Pro using the -march option.
I was just trying to say that the it is possible to compile the kernel 2.2 with pgcc, cause as you might know it was'nt possible to do it with the 2.0.xx kernels, without apllying patches.
Talking about stability I have a kernel 2.2pre9 compiled with amdk6 optimization, running without
problems in a non-suported Super7 motherboard for one week now, this of courser is no proof of stability but it is a beggining I believe.
MySQL for example runs perfectly and 12 % faster compiled with pggc.
I know it is still very experimental, but as far as all the tests go pgcc makes faster code than egcs.
And of course I know faster does'nt mean more stable, although I have'nt had any stability problems with pgcc compiled programs.
The only problems that I have faced till now are with Mesa 3.0 and kernels of the 2.0.xx, those yes, I can't even compile them with pgcc.
Cool, I have tried compiling the kernel 2.2 with pgcc with mcpu and march optimizations and it worked, the result is an even faster kernel.
I have compiled it with amdk6 optimizations and the result was a 10% faster kernel, try it cause if you thought the kernel 2.2 was fast with pgcc it gets even faster