== Progress ==
* GNU Cauldron (4/10) - GCC+LLVM presentation had some positive reviews - Discussed sanitizers roadmap - Very interesting meeting with QuIC
* TCWG Sprint (4/10) - Mostly about GNU tools - Team Mission streamlined, looking good - LLVM roadmap attracted some attention - We could have some bite-sized work from other team members?
* Release 3.5 testing (TCWG-476 2/10) - No test regressions - Spotted some benchmark regressions - ARMv7 is overall the same on EEMBC - AArch64 is overall 10% faster on EEMBC
* Weekend working, Friday off
== Plan ==
* Release week! - Investigate 3.5 performance regressions on v7 - Work around the lack of perf on v8? - Run SPEC on both v7 and v8 and spot regressions
Hi Renato,
On 07/26/2014 05:18 AM, Renato Golin wrote:
- Work around the lack of perf on v8?
Out of curiosity, what exactly is missing?
Christopher
On 28 July 2014 14:27, Christopher Covington cov@codeaurora.org wrote:
- Work around the lack of perf on v8?
Out of curiosity, what exactly is missing?
Hi Christopher,
As far as I know, everything. But maybe this is just for the hardware we have in the lab...
Do you guys have perf working on any piece of v8 hardware?
cheers, --renato
On 28/07/14 23:35, Renato Golin wrote:
On 28 July 2014 14:27, Christopher Covington cov@codeaurora.org wrote:
- Work around the lack of perf on v8?
Out of curiosity, what exactly is missing?
Hi Christopher,
As far as I know, everything. But maybe this is just for the hardware we have in the lab...
It could be the kernel version we use in lab. AFAIK, Charles had some luck profiling with perf on newer kernel. I am not sure about the version though.
Thanks, Kugan
On 28 July 2014 14:45, Kugan kugan.vivekanandarajah@linaro.org wrote:
It could be the kernel version we use in lab. AFAIK, Charles had some luck profiling with perf on newer kernel. I am not sure about the version though.
I have seen working performance counters on 3.15-rc8. The versions of the kernel we have on our aarch64 lab hardware have kernel oopses when hardware counters are used. On Juno, the 3.14 linaro kernel seems not to provide access to the performance counters.
Some care needs to be taken with perf reports on both arm/aarch64, as they are sometimes have bogus addresses in the profile. (sometimes large amounts of time can be attributed to anonymous hex addresses, do not ignore these. I have a workaround script)
On 07/28/2014 09:35 AM, Renato Golin wrote:
On 28 July 2014 14:27, Christopher Covington cov@codeaurora.org wrote:
- Work around the lack of perf on v8?
Out of curiosity, what exactly is missing?
Hi Christopher,
As far as I know, everything. But maybe this is just for the hardware we have in the lab...
Do you guys have perf working on any piece of v8 hardware?
So far my work has been with simulators where I've been running the likes of `perf stat -e instructions:u $APPLICATION` using kernels built from torvalds/master.
Christopher
linaro-toolchain@lists.linaro.org
-
Charles Baylis -
Christopher Covington -
Kugan -
Renato Golin