The speed at which data is transferred from the enclave to the app

[double]

Recently I upgraded my SGX's version from 2.15.1 to 2.19. There are some small bugs when rerunning the 2.15.1 version of the code. But after solving these bugs, I found that the running speed is different from the previous version. After my analysis, I realized that the speed of transferring data from enclave to app is different from the previous version, and it is much slower than before. I do not know why. 

I transfer a matrix from the enclave to the app for execution. The transfer time of the 2.15.1 version was almost negligible, while the transfer time of the current version is higher. For example, when transmitting a double type matrix with a scale of 300*300, the time is about 120000 microseconds. This makes my experiment impossible to reproduce, please tell me the solution.

Thanks a lot!

Sincerely

Doubley

[KFPW_Intel]

Hi Doubley,

Thank you for your interest in Intel® SGX.

I am looking into the differences in time and speed for data transferred from enclave to the app. I noticed there are a lot of upgraded features and new API along the latest releases of linux-sgx, and this may cause the differences in speed here. Refer linux-sgx for more information.

I am checking with the developers for better understanding on the time/speed performance. Allow me to update you once we have any reply for the development team.

Regards,

Ken

[KFPW_Intel · Accepted solution]

Hi Doubley,

Thank you for your patience. I have checked this case with the development team.

Beginning with version 2.17.0.3, Intel has updated the Intel SGX SDK for Linux and the Edger8r tool included in the SDK to help prevent fill buffer data exposure through the code generated by the Edger8r tool. Similarly, the Intel SGX SDK now includes updates that will help prevent fill buffer data exposure through the code used by enclaves that use the switchless mode supported by the Intel SDK.

Please visit Intel® Software Guard Extensions Developer Reference for Linux OS, page 88, Mitigations for Processor MMIO Stale Data Vulnerabilities for more information.

Hope that the information that I provided is useful for you.

Regards,

Ken

[double]

Thank you for your help.

[KFPW_Intel]

Hi Doubley,

Thank you for the accepted solution. Hope the information I provided was helpful to you.

This thread has been marked as answered and Intel will no longer monitor this thread. If you want a response from Intel in a follow-up question, please open a new thread.

Regards,

Ken