For single module build, it would call gen_libs target. then if it use
binary LIB file, it cause build failure.
Cc: Liming Gao
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Yonghong Zhu
---
Hi Feng.
Sorry, i was AFK for the past week.
We have encountered this with Platronix USB headset. This device was reporting
*USB Interface* descriptor sizes 1 byte larger than what the spec says. Headset
was simply plugged in when we booted our firmware and UsbCreateDesc failed to
parse that
On some platforms, performing cache maintenance on regions that are backed
by NOR flash result in SErrors. Since cache maintenance is unnecessary in
that case, create a PEIM specific version that only performs said cache
maintenance in its constructor if the module is shadowed in RAM. To avoid
On 13 June 2016 at 17:45, Mark Rutland wrote:
> On Mon, Jun 13, 2016 at 05:26:07PM +0200, Ard Biesheuvel wrote:
>> On some platforms, performing cache maintenance on regions that are backed
>> by NOR flash result in SErrors. Since cache maintenance is unnecessary in
>> that
On Mon, Jun 13, 2016 at 05:51:23PM +0200, Ard Biesheuvel wrote:
> On 13 June 2016 at 17:45, Mark Rutland wrote:
> > On Mon, Jun 13, 2016 at 05:26:07PM +0200, Ard Biesheuvel wrote:
> >> On some platforms, performing cache maintenance on regions that are backed
> >> by NOR
Hi Andrew,
Good news! After enhance the GenFw tool, I finally enable the clang LTO build
on edk2 and pass test on the OVMF three platforms (OvmfPkgIa32.dsc,
OvmfPkgX64.dsc, OvmfPkgIa32X64.dsc) in my side . Below is the code size
comparing data between VS2015x86 and CLANGLTO38. You can see they
On Mon, Jun 13, 2016 at 05:26:07PM +0200, Ard Biesheuvel wrote:
> On some platforms, performing cache maintenance on regions that are backed
> by NOR flash result in SErrors. Since cache maintenance is unnecessary in
> that case, create a PEIM specific version that only performs said cache
>
Liming,
UNREACHABLE() in fact may have an impact as it is a signal that the code flow
cannot reach that position. Thus a compiler or analyzer may classify everything
following as unreachable and issue warnings about it or optimize it away. For
stack cleanups, as Andrew suggested, this is a
Also for the series. Reviewed-by: Jaben Carsey
> -Original Message-
> From: edk2-devel [mailto:edk2-devel-boun...@lists.01.org] On Behalf Of
> Shia, Cinnamon
> Sent: Sunday, June 12, 2016 7:44 PM
> To: Yao, Jiewen ; Zeng, Star
Ye gods - I've slipped replying to this by a month now. Very sorry
about that. Well, I think I've managed to crawl out of the hole I fell
into.
So, I think the below sounds like it _could_ resolve most of my
concerns, with a few tweaks. Full disclosure: Personally, I would only
be actively
Star,
The PERFORMANCE_PROPERTY structure has a field called CpuFreq.
Since there are many different types of timer sources, some not
related to the CPU frequencies, should this field name just be
Frequency?
Also, one of the long standing issues with the DP command is that
all the modules
Hi,
I agree with Giri's comments here. The IntelSiliconPkg can be added using
current flat directory structure.
I will update the RFC for the proposed directory structure changes to include
target location for this new package.
Thanks,
Mike
> -Original Message-
> From: Mudusuru, Giri
Marvin,
I agree that Base.h is the right place for these new macros.
Mike
> -Original Message-
> From: edk2-devel [mailto:edk2-devel-boun...@lists.01.org] On Behalf Of Marvin
> Häuser
> Sent: Monday, June 13, 2016 11:05 AM
> To: edk2-devel@lists.01.org
> Cc: Gao, Liming
> On Jun 13, 2016, at 4:14 PM, Kinney, Michael D
> wrote:
>
> Star,
>
> The PERFORMANCE_PROPERTY structure has a field called CpuFreq.
>
> Since there are many different types of timer sources, some not
> related to the CPU frequencies, should this field name
Thanks Mike.
Another possible way I am thinking is that: Can we just record *time* instead
of *tick* in the perf record?
Then there will be no worry on which timer lib the module is using.
Thank you
Yao Jiewen
From: Kinney, Michael D
Sent: Tuesday, June 14, 2016 7:15 AM
To: Zeng, Star
Jiewen,
The reason the original design does not record time is to minimize the time
required to add a performance record to reduce the overhead of doing
performance measurements.
Reading a counter is much simpler and takes less time than converting to a time
value that usually requires
Reviewed-By: Wu Jiaxin
I will commit the patch soon. Thanks for the contribution.
Best Regards!
Jiaxin
> -Original Message-
> From: Thomas Palmer [mailto:thomas.pal...@hpe.com]
> Sent: Wednesday, June 8, 2016 2:36 AM
> To: edk2-devel@lists.01.org
> Cc:
Mike,
It is make sense to change CpuFreq field name to Frequency.
We ever thought about to catch the TimerLib mismatch case, but the frequency
returned from TimerLib may have very very small deviation, for example the
AcpiTimerLib instances in PcAtChipsetPkg that are using ACPI timer to
Cc: Ye Ting
Cc: Fu Siyuan
Cc: Zhang Lubo
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Jiaxin Wu
---
NetworkPkg/TcpDxe/SockImpl.c | 2 --
1 file changed, 2 deletions(-)
diff --git
Andrew,
I like the idea of a record type for a time source. A PerformanceLib
constructor could add that record,
and performance log entries could be tagged with a time source record indicator.
I agree that any timer source that requires an algorithm to measure the
frequency could get slightly
Reviewed-By: Wu Jiaxin
Best Regards!
Jiaxin
> -Original Message-
> From: edk2-devel [mailto:edk2-devel-boun...@lists.01.org] On Behalf Of
> Thomas Palmer
> Sent: Tuesday, June 7, 2016 10:47 AM
> To: edk2-devel@lists.01.org
> Cc: Wu, Jiaxin
>
Thanks for your info, Yakovlev.
Did you see problem with XHCI? If yes, I can provide a patch and could you help
verify it?
Thanks
Feng
-Original Message-
From: Evgeny Yakovlev [mailto:insorei...@gmail.com]
Sent: Monday, June 13, 2016 5:29 PM
To: Tian, Feng
Cc:
22 matches
Mail list logo