Hi Peter,

yes it will be part of our release.

Rgds, Yvan

On 15 October 2014 11:11, Panshilin (Peter) peter.panshilin@hisilicon.com wrote:

Dear all concerned:

ARM has reported it's 53's bug:AArch64 multiply-accumulate instruction might produce incorrect result and developed the patch descriped below. will the patch be backported to Linaro 4.9 this month's release. https://gcc.gnu.org/ml/gcc-cvs/2014-10/msg00335.html

thanks

Peter _______________________________________________ linaro-toolchain mailing list linaro-toolchain@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-toolchain

Hi Peter,

The errata fix can be turned on and off at compile time with the flags -mfix-cortex-a53-835769 and -mno-fix-cortex-a53-835769 it is also possible to turn it on by default when configuring GCC with the configure flag --enable-fix-cortex-a53-835769 (I don't know yet if the decision to turn it on by default in our binary releases was taken or not).

The possibility to workaround the hardware bug has to be and remain in the compiler to allow the users that have such a revision of the core to avoid the issue, and it will stay in the compiler even after fixed revisions of the core will be out, for legacy reason.

Thanks, Yvan

On 21 October 2014 03:46, Panshilin (Peter) peter.panshilin@hisilicon.com wrote:

Hi Yvan :

Thanks. other questions about this patch:

Will the patch stay in all arm64 compilers for ever? or someday it will be discarded for optimization reasons?

Would you provide a compiler option which enable or disable the patch during buiding and if so what the option is?

Thanks again.

Peter

---

发件人: Yvan Roux [yvan.roux@linaro.org] 发送时间: 2014年10月15日 17:17 收件人: Panshilin (Peter) 抄送: Linaro Toolchain 主题: Re: Will this ARM64 erratum patch be backported to this month's Linaro GCC 4.9 release?

Hi Peter,

yes it will be part of our release.

Rgds, Yvan

On 15 October 2014 11:11, Panshilin (Peter) peter.panshilin@hisilicon.com wrote:

...

Dear all concerned:

ARM has reported it's 53's bug:AArch64 multiply-accumulate instruction might produce incorrect result and developed the patch descriped below. will the patch be backported to Linaro 4.9 this month's release. https://gcc.gnu.org/ml/gcc-cvs/2014-10/msg00335.html

thanks

Peter _______________________________________________ linaro-toolchain mailing list linaro-toolchain@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-toolchain

Hi Peter,

From what I see in their 4.9 branch hosted in the FSF repository it's

not merged yet, and I've no information on my side when and how it will be done.

Thanks, Yvan

On 4 November 2014 07:55, Panshilin (Peter) peter.panshilin@hisilicon.com wrote:

Hi All:

Are there anyone who know whether Google has merged this patch in there toolchain release? especially in NDK's compiler( Clang 3.5 and Gcc 4.9 ) . And if so does the flag enable or disable by default?

From our perspective, anyway the flags should be enable by default , that is to say, the compiler should be with this flag by no particularly indicating .

Thanks.

Peter ________________________________________ 发件人: Yvan Roux [yvan.roux@linaro.org] 发送时间: 2014年10月21日 16:50 收件人: Panshilin (Peter) 抄送: Linaro Toolchain 主题: Re: 答复: Will this ARM64 erratum patch be backported to this month's Linaro GCC 4.9 release?

Hi Peter,

The errata fix can be turned on and off at compile time with the flags -mfix-cortex-a53-835769 and -mno-fix-cortex-a53-835769 it is also possible to turn it on by default when configuring GCC with the configure flag --enable-fix-cortex-a53-835769 (I don't know yet if the decision to turn it on by default in our binary releases was taken or not).

The possibility to workaround the hardware bug has to be and remain in the compiler to allow the users that have such a revision of the core to avoid the issue, and it will stay in the compiler even after fixed revisions of the core will be out, for legacy reason.

Thanks, Yvan

On 21 October 2014 03:46, Panshilin (Peter) peter.panshilin@hisilicon.com wrote:

...

Hi Yvan :

Thanks. other questions about this patch:

Will the patch stay in all arm64 compilers for ever? or someday it will be discarded for optimization reasons?

Would you provide a compiler option which enable or disable the patch during buiding and if so what the option is?

Thanks again.

Peter

---

发件人: Yvan Roux [yvan.roux@linaro.org] 发送时间: 2014年10月15日 17:17 收件人: Panshilin (Peter) 抄送: Linaro Toolchain 主题: Re: Will this ARM64 erratum patch be backported to this month's Linaro GCC 4.9 release?

Hi Peter,

yes it will be part of our release.

Rgds, Yvan

On 15 October 2014 11:11, Panshilin (Peter) peter.panshilin@hisilicon.com wrote:

...

Dear all concerned:

ARM has reported it's 53's bug:AArch64 multiply-accumulate instruction might produce incorrect result and developed the patch descriped below. will the patch be backported to Linaro 4.9 this month's release. https://gcc.gnu.org/ml/gcc-cvs/2014-10/msg00335.html

thanks

Peter _______________________________________________ linaro-toolchain mailing list linaro-toolchain@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-toolchain