sha512_block_data_order in ARM takes ~3k in stack

[Omer Anson]

Hello,

I apologise if this is the wrong list.

I have cross compiled openssl to ARM (using gcc-4.6 on Ubuntu) - both
versions 0.9.8y and versions 1.0.1e. When using objdump to view the result,
I see that sha512_block_data_order takes ~3k (2744 in 0.9.8y, and 3464 in
1.0.1e) of stack space.
Since an assembler version is available for this method, this issue is
academic. However, I am curious, and a quick google search didn't reveal
anything.
I have looked at the C code and did not see something obvious that should
cause this. Additionally, the x86 compilation product takes ~1004 (in
1.0.1e) bytes (which also seems much, but only a third of the ARM version).
Has anyone seen this before?
Does anyone know why this happens?

Thank you,
Omer Anson.

[m***@gmail.com]

Post by Omer Anson
Hello,
I apologise if this is the wrong list.
I have cross compiled openssl to ARM (using gcc-4.6 on Ubuntu) - both versions 0.9.8y and versions 1.0.1e. When using objdump to view the result, I see that sha512_block_data_order takes ~3k (2744 in 0.9.8y, and 3464 in 1.0.1e) of stack space.
Since an assembler version is available for this method, this issue is academic. However, I am curious, and a quick google search didn't reveal anything.
I have looked at the C code and did not see something obvious that should cause this. Additionally, the x86 compilation product takes ~1004 (in 1.0.1e) bytes (which also seems much, but only a third of the ARM version).
Has anyone seen this before?
Does anyone know why this happens?
Thank you,
Omer Anson.

Hi Omer,
Is there any possibility of getting crash due to this 3k memory consumption in ARM device. Basically I am getting lots of crash in this method while running app in iOS device. Do you know anything else regarding this issue ?

Regards,
Mahbub