Hi All,
I found in the kernel, the capability mapping and memory mapping have very similar code path. But I found their performance difference is around 1000 cycles. Is this true? If it is true, why do they have such large difference? If this gap does not make sense, it should be a bug in my test.
Thank you very much. Best, Yuxin
Hi,
On Mon Sep 29, 2014 at 13:24:56 -0400, Yuxin Ren wrote:
I found in the kernel, the capability mapping and memory mapping have very similar code path. But I found their performance difference is around 1000 cycles. Is this true? If it is true, why do they have such large difference? If this gap does not make sense, it should be a bug in my test.
Hmm, yes, there can be a difference. Of course, the number depends on the hardware. Which one was slower/faster? For example, for memory there there could be the need to TLB flushes, so it could be slower than cap mappings.
Adam
On my machine, cap mapping is faster than memory. I do not think there is tlb flush in memory mapping. It only happens when memory is unmapped, right? But I found the mapping database for them are very different. Cap mapping database is simpler than the memory's.
Thank you very much. Yuxin
On Tue, Sep 30, 2014 at 5:15 PM, Adam Lackorzynski < adam@os.inf.tu-dresden.de> wrote:
Hi,
On Mon Sep 29, 2014 at 13:24:56 -0400, Yuxin Ren wrote:
I found in the kernel, the capability mapping and memory mapping have
very
similar code path. But I found their performance difference is around 1000 cycles. Is this true? If it is true, why do they have such large difference? If this gap does not make sense, it should be a bug in my test.
Hmm, yes, there can be a difference. Of course, the number depends on the hardware. Which one was slower/faster? For example, for memory there there could be the need to TLB flushes, so it could be slower than cap mappings.
Adam
Adam adam@os.inf.tu-dresden.de Lackorzynski http://os.inf.tu-dresden.de/~adam/
l4-hackers mailing list l4-hackers@os.inf.tu-dresden.de http://os.inf.tu-dresden.de/mailman/listinfo/l4-hackers
On Wed Oct 01, 2014 at 11:02:20 -0400, Yuxin Ren wrote:
On my machine, cap mapping is faster than memory. I do not think there is tlb flush in memory mapping. It only happens when memory is unmapped, right? But I found the mapping database for them are very different. Cap mapping database is simpler than the memory's.
Ah, yes, that makes sense.
Adam
l4-hackers@os.inf.tu-dresden.de