Hi I tried to compile dgl for mixed precision work with
git clone --recurse-submodules https://github.com/dmlc/dgl.git
cd dgl
mkdir build
cd build
cmake -DUSE_CUDA=ON -DUSE_FP16=ON ..
make -j 8
If I compile with make -j
it return error message as follows
In file included from /home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/transport/ibv/connection_impl.h(23),
from /home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/transport/ibv/context_impl.cc(11):
/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/socket.h(252): warning #1292: unknown attribute "nodiscard"
[[nodiscard]] Error recvPayloadAndFds(T& t1, T& t2, Fds&... fds) {
^
compilation aborted for /home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/transport/shm/connection_impl.cc (code 2)
make[2]: *** [third_party/tensorpipe/tensorpipe/CMakeFiles/tensorpipe.dir/build.make:608: third_party/tensorpipe/tensorpipe/CMakeFiles/tensorpipe.dir/transport/shm/connection_impl.cc.o] Error 2
/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_role.h(144): error: unexpected designator
result[0] = {.ptr = data_ + start, .len = size};
^
detected during:
instantiation of "ssize_t={__ssize_t={long}} tensorpipe::RingBufferRole<NumRoles, RoleIdx>::readInTx<AllowPartial>(void *, size_t={unsigned long}) [with NumRoles=2, RoleIdx=0, AllowPartial=false]" at line 152 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_read_write_ops.h"
instantiation of "size_t={unsigned long} tensorpipe::RingbufferReadOperation::handleRead(tensorpipe::RingBufferRole<NumRoles, RoleIdx> &) [with NumRoles=2, RoleIdx=0]" at line 355 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/transport/ibv/connection_impl.cc"
/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_role.h(144): error: unexpected designator
result[0] = {.ptr = data_ + start, .len = size};
^
detected during:
instantiation of "ssize_t={__ssize_t={long}} tensorpipe::RingBufferRole<NumRoles, RoleIdx>::readInTx<AllowPartial>(void *, size_t={unsigned long}) [with NumRoles=2, RoleIdx=0, AllowPartial=false]" at line 152 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_read_write_ops.h"
instantiation of "size_t={unsigned long} tensorpipe::RingbufferReadOperation::handleRead(tensorpipe::RingBufferRole<NumRoles, RoleIdx> &) [with NumRoles=2, RoleIdx=0]" at line 355 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/transport/ibv/connection_impl.cc"
/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_role.h(148): error: unexpected designator
.ptr = data_ + start, .len = header_.kDataPoolByteSize - start};
^
detected during:
instantiation of "ssize_t={__ssize_t={long}} tensorpipe::RingBufferRole<NumRoles, RoleIdx>::readInTx<AllowPartial>(void *, size_t={unsigned long}) [with NumRoles=2, RoleIdx=0, AllowPartial=false]" at line 152 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_read_write_ops.h"
instantiation of "size_t={unsigned long} tensorpipe::RingbufferReadOperation::handleRead(tensorpipe::RingBufferRole<NumRoles, RoleIdx> &) [with NumRoles=2, RoleIdx=0]" at line 355 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/transport/ibv/connection_impl.cc"
/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_role.h(148): error: unexpected designator
.ptr = data_ + start, .len = header_.kDataPoolByteSize - start};
^
detected during:
instantiation of "ssize_t={__ssize_t={long}} tensorpipe::RingBufferRole<NumRoles, RoleIdx>::readInTx<AllowPartial>(void *, size_t={unsigned long}) [with NumRoles=2, RoleIdx=0, AllowPartial=false]" at line 152 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_read_write_ops.h"
instantiation of "size_t={unsigned long} tensorpipe::RingbufferReadOperation::handleRead(tensorpipe::RingBufferRole<NumRoles, RoleIdx> &) [with NumRoles=2, RoleIdx=0]" at line 355 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/transport/ibv/connection_impl.cc"
/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_role.h(149): error: unexpected designator
result[1] = {.ptr = data_, .len = end};
^
detected during:
instantiation of "ssize_t={__ssize_t={long}} tensorpipe::RingBufferRole<NumRoles, RoleIdx>::readInTx<AllowPartial>(void *, size_t={unsigned long}) [with NumRoles=2, RoleIdx=0, AllowPartial=false]" at line 152 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_read_write_ops.h"
instantiation of "size_t={unsigned long} tensorpipe::RingbufferReadOperation::handleRead(tensorpipe::RingBufferRole<NumRoles, RoleIdx> &) [with NumRoles=2, RoleIdx=0]" at line 355 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/transport/ibv/connection_impl.cc"
/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_role.h(149): error: unexpected designator
result[1] = {.ptr = data_, .len = end};
^
detected during:
instantiation of "ssize_t={__ssize_t={long}} tensorpipe::RingBufferRole<NumRoles, RoleIdx>::readInTx<AllowPartial>(void *, size_t={unsigned long}) [with NumRoles=2, RoleIdx=0, AllowPartial=false]" at line 152 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/common/ringbuffer_read_write_ops.h"
instantiation of "size_t={unsigned long} tensorpipe::RingbufferReadOperation::handleRead(tensorpipe::RingBufferRole<NumRoles, RoleIdx> &) [with NumRoles=2, RoleIdx=0]" at line 355 of "/home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/transport/ibv/connection_impl.cc"
compilation aborted for /home/dngusdnr1/lib/new_se3/dgl/third_party/tensorpipe/tensorpipe/transport/ibv/connection_impl.cc (code 2)
make[2]: *** [third_party/tensorpipe/tensorpipe/CMakeFiles/tensorpipe.dir/build.make:706: third_party/tensorpipe/tensorpipe/CMakeFiles/tensorpipe.dir/transport/ibv/connection_impl.cc.o] Error 2
[ 50%] Linking C static library libmetis.a
[ 50%] Built target metis
================================================================================
LIBXSMM master-1.16.1-1534 (Linux@galaxy3)
--------------------------------------------------------------------------------
Intel Compiler: icc 2021.5.0, icpc 2021.5.0, and ifort 2021.5.0
C / C++ target: -msse4.2
Fortran target: -msse4.2
--------------------------------------------------------------------------------
Environment: CC CXX FC
--------------------------------------------------------------------------------
BLAS dependency and fallback is removed!
--------------------------------------------------------------------------------
[ 50%] Built target libxsmm
make[1]: *** [CMakeFiles/Makefile2:361: third_party/tensorpipe/tensorpipe/CMakeFiles/tensorpipe.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
Thanks in advance.