{toolchain} intelcuda 2017b #6709
Conversation
…b, HPL-2.2-intelcuda-2017b.eb, iccifortcuda-2017.4.196-GCC-6.4.0-2.28.eb, iimpic-2017b.eb, imkl-2017.3.196-iimpic-2017b.eb, impi-2017.3.196-iccifortcuda-2017.4.196-GCC-6.4.0-2.28.eb, intelcuda-2017b.eb
|
Test report by @vanzod |
| ('CUDA', '9.0.176', '', ('iccifort', intelver)), | ||
| ('impi', '2017.3.196', '', ('iccifortcuda', intelver)), | ||
| ('imkl', '2017.3.196', '', ('iimpic', version)), | ||
| ] |
There was a problem hiding this comment.
Any particular reason you did it this way instead of how it looks in intelcuda-2016.10.eb ?
There was a problem hiding this comment.
I simply followed the rule of encapsulating whatever gets repeated into a variable after consulting with @boegel
There was a problem hiding this comment.
Yeah, but why use
('GCCcore', gccver),
('binutils', binutilsver, ...)
instead of
('iccifort', intelver),
as in intelcuda-2016.10.eb
(That was what my question was really targeting)
There was a problem hiding this comment.
@akesandgren The reason why it is different from intelcuda-2016.10.eb is that I took intel-2017b.eb as template. I honestly do not know the reason why we prefer to call the separate components instead of the lower toolchain modules. @boegel definitely has better memory than me on that.
That said, I'm fine using the convention you used for the previous intelcuda.
There was a problem hiding this comment.
Ok, I used iccifort in the old one since it made more sense to me at the time, but the GCCcore+binutils like intel-201... uses also makes sense, and keeping it similar to intel.eb makes a lot of sense.
Keep it.
There was a problem hiding this comment.
There's actually no point in also including GCCcore and binutils in the dependencies list for intel* toolchains, other than to clarify that GCCcore + binutils is used as a base. There's no technical reason for it, everything will work just fine without it.
|
Test report by @boegel |
|
Test report by @boegel |
|
Test report by @boegel |
|
Going in, thanks @vanzod! |
(created using
eb --new-pr)