mdouze

In [1]: import numpy as np

In [2]: import faiss
Failed to load GPU Faiss: No module named swigfaiss_gpu
Faiss falling back to CPU-only.

In [3]: import time

In [4]:

mdouze

OMP_WAIT_POLICY=PASSIVE ipython

Python 2.7.5 (default, Nov  6 2016, 00:28:07)
Type "copyright", "credits" or "license" for more information.

IPython 3.2.1 -- An enhanced Interactive Python.
?         -> Introduction and overview of IPython's features.
%quickref -> Quick reference.
help      -> Python's own help system.
object?   -> Details about 'object', use 'object??' for extra details.

mdouze

import numpy as np
import time
import faiss
import os

#################################################################
# I/O functions
#################################################################

def ivecs_read(fname):

mdouze

[matthijs@matthijs-mbp ~/Desktop/test_cmake/faiss/build]cmake -DCMAKE_CXX_COMPILER=/opt/local/bin/g++-mp-6  ..
-- The C compiler identification is AppleClang 8.0.0.8000042
-- The CXX compiler identification is GNU 6.3.0
-- Check for working C compiler: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/cc
-- Check for working C compiler: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Checking whether CXX compiler has -isysroot

mdouze

diff --git a/faiss/swigfaiss.swig b/faiss/swigfaiss.swig
--- a/faiss/swigfaiss.swig
+++ b/faiss/swigfaiss.swig
@@ -548,6 +548,10 @@
     return NULL;
 }

+float * cast_integer_to_float_ptr (long x) {
+    return (float*)x;
+}

mdouze

import numpy as np

import torch
import faiss


# make a GPU index

d = 16
nq = 5

mdouze

#! /usr/bin/env python2

import numpy as np
import faiss
import time

xd = 100
yd = 1000000

np.random.seed(1234)

mdouze

--[[ Copyright 2004-present Facebook. All Rights Reserved.
     Load the dynamic library involved in swig calls for Faiss
   + some code to convert input / output arguments

   the standard way of importing this is

   swigfaiss = require 'faiss_swig'

   if you require 'swigfaiss' directly, you get the indexing
   structures without the additions that make them easy to use from

mdouze

matthijs@devfair004:~/faiss_py3_cuda9/gpu$ make -j10
g++ -std=c++11 -fPIC -m64 -Wall -g -Wno-sign-compare -O3 -fopenmp \
         -c GpuResources.cpp -o GpuResources.o  -I/public/apps/cuda/9.0//include
g++ -std=c++11 -fPIC -m64 -Wall -g -Wno-sign-compare -O3 -fopenmp \
         -c IndexProxy.cpp -o IndexProxy.o  -I/public/apps/cuda/9.0//include
g++ -std=c++11 -fPIC -m64 -Wall -g -Wno-sign-compare -O3 -fopenmp \
         -c StandardGpuResources.cpp -o StandardGpuResources.o  -I/public/apps/cuda/9.0//include
g++ -std=c++11 -fPIC -m64 -Wall -g -Wno-sign-compare -O3 -fopenmp \
         -c GpuAutoTune.cpp -o GpuAutoTune.o  -I/public/apps/cuda/9.0//include

mdouze

matthijs@devfair0144:~/faiss_versions/alex_bug$ /public/apps/cuda/8.0//bin/nvcc -I. -I /public/apps/cuda/8.0//targets/x86_64-linux/include/ -Xcompiler -fPIC -Xcudafe --diag_suppress=unrecognized_attribute -gencode arch=compute_60,code="compute_60" -gencode arch=compute_61,code="compute_61" --std c++11 -lineinfo -ccbin g++ -DFAISS_USE_FLOAT16 -o tt tt.cpp faiss/gpu/libgpufaiss.a faiss/libfaiss.a -Xcompiler -fopenmp -lcublas -Xlinker --no-as-needed -Xlinker -rpath,/public/apps/anaconda2/5.0.1/lib/ -L/public/apps/anaconda2/5.0.1/lib/  -lmkl_intel_lp64 -lmkl_core -lmkl_gnu_thread -ldl -lpthread
matthijs@devfair0144:~/faiss_versions/alex_bug$ ./tt
Faiss assertion 'err__ == cudaSuccess' failed in faiss::gpu::DeviceTensor<T, Dim, InnerContig, IndexT, PtrTraits>::~DeviceTensor() [with T = float; int Dim = 3; bool InnerContig = true; IndexT = int; PtrTraits = faiss::gpu::traits::DefaultPtrTraits] at impl/../utils/DeviceTensor-inl.cuh:61; details: CUDA error 77
Aborted