This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
building accelerators for both KNL and nvptx
- From: Mark Farnell <mark dot farnell at gmail dot com>
- To: gcc at gcc dot gnu dot org
- Date: Fri, 23 Jan 2015 13:14:20 +1300
- Subject: building accelerators for both KNL and nvptx
- Authentication-results: sourceware.org; auth=none
I have just read the updated manual:
https://gcc.gnu.org/wiki/Offloading
It says to build an accel compiler, one need to:
./configure --build=x86_64-intelmicemul-linux-gnu
--host=x86_64-intelmicemul-linux-gnu
--target=x86_64-intelmicemul-linux-gnu
--enable-as-accelerator-for=x86_64-pc-linux-gnu
make
make install DESTDIR=/install
However does such accelerator also enable acceleration on nvptx? If
not, then suppose if I have a machine that has both KNL and nvptx,
then how can I build the toolchain for nvptx compiler?
Also for the host, the website says:
./configure --build=x86_64-pc-linux-gnu --host=x86_64-pc-linux-gnu
--target=x86_64-pc-linux-gnu
--enable-offload-targets=x86_64-intelmicemul-linux-gnu=/install/prefix
make
make install DESTDIR=/install
However if the host were to support offloading for both nvptx and KNL,
then what EXTRA parameters do I need to pass in? Or does it by
default support offloading to ANY architectures?
Thanks!
Mark