Re: [PATCH 1/2] perf intel-pt-decoder: Report instruction bytes and length in sample

From: Andi Kleen
Date: Mon Oct 03 2016 - 20:33:12 EST


On Fri, Sep 30, 2016 at 01:07:17PM +0300, Adrian Hunter wrote:
> On 30/09/16 06:49, Andi Kleen wrote:
> > From: Andi Kleen <ak@xxxxxxxxxxxxxxx>
> >
> > Change the Intel PT decoder to pass up the length and the instruction
> > bytes of the decoded or sampled instruction in the perf sample.
> >
> > The decoder already knows this information, we just need to pass it
> > up. Since it is only a couple of movs it is not very expensive.
>
> The decoder doesn't always fill the instruction buffer because it caches the
> results from walking basic blocks. That means the bytes from the last
> instruction in the basic block would need to be added to the cache.

I fixed this and posted a new version.
>
> I wonder if we shouldn't look at something more sophisticated e.g. a python
> script that can do full disassembly.

I had an earlier version that used a disassembler library, but that
wasn't appreciated, so now switched to this simpler method.
Assembler output can be done with a sed + objdump hack now.

-Andi