[FEATURE REQUEST] Quit if no SSE4.2 support is detected

[MrTomRod]

Hey NCBI gang! Long time no issue :smile:

I noticed that PGAP does not crash if SSE4.2 support is not available. It just writes a few lines like this to cwltool.log:

Critical: (106.22) Application requires a CPU with SSE 4.2 support

It happened to me when I tried running it on a VM where this was not enabled. Nevertheless, PGAP went on to produce output files!

Describe the solution you'd like pgap.py could catch this early on and deliver a clear error message. Example code:

def check_sse42_support():
    with open('/proc/cpuinfo') as f:
        cpuinfo = f.read()
    assert 'sse4_2' in cpuinfo, "SSE4.2 is not supported on this system."

BTW: Are the output files produced without SSE4.2 still useable?

[azat-badretdin]

Hi, Thomas, always glad to see your Issues!

I tracked this message to our C++ Toolkit /corelib/ and I am almost sure that this message is informational only. I opened internal ticket CXX-13466 where I passed your question to our Toolkit folks

[azat-badretdin]

Here is what our expert said:

Any attempt to run any of the specific portions of code that don't support the user's (old) processor will yield an obvious "Illegal instruction" error; the up-front warning just raises awareness of the possibility.

[MrTomRod]

Not sure what that implies for my output. Does it mean that since the actual command didn't fail, it's okay? Can you then remove the following line in Quick-Start:

The CPU must have SSE 4.2 support (released in 2008).

[azat-badretdin]

Does it mean that since the actual command didn't fail, it's okay?

Probably yes, let me double check on CXX-13466

Can you then remove the following line in Quick-Start:

This lines mean that we do not support systems without SSE 4.2 support because we can't guarantee that some binaries would fail with "illegal instruction" command.

[azat-badretdin]

Does it mean that since the actual command didn't fail, it's okay?

This has been confirmed, Thomas.

I am sorry to inform you that as of original request, the answer on PGAPX-1256 was, alas, negatory. The management reaffirm our stance on SSE 4.2 support presented in Quick Start and resistance to additional /proc related processing in pgap.py

[MrTomRod]

Thanks for the quick resolution of the issue!