Merge pull request #5 from matsui528/build

Updated how-to-build with Ubuntu 22.04

.github/workflows/build_from_source.yml

@@ -30,5 +30,41 @@ jobs:
         cd
         python -c "import faiss, numpy; err = faiss.Kmeans(10, 20).train(numpy.random.rand(1000, 10).astype('float32')); print(err)"
 
-        echo "##### AVX2 is activated or not #####"
-        LD_DEBUG=libs python -c "import faiss" 2>&1 | grep libfaiss_avx2.so
+        echo "##### Check AVX2 is working or not #####"      
+        LD_DEBUG=libs python -c "import faiss" 2>&1 | grep libfaiss_avx2.so  
+
+        echo "import faiss
+        import numpy as np
+        import time
+
+        np.random.seed(234)
+        D = 128
+        N = 10000
+        X = np.random.random((N, D)).astype(np.float32)
+        M = 64
+        nbits = 4
+
+        pq = faiss.IndexPQ(D, M, nbits)
+        pq.train(X)
+        pq.add(X)
+
+        pq_fast = faiss.IndexPQFastScan(D, M, nbits)
+        pq_fast.train(X)
+        pq_fast.add(X)
+
+        t0 = time.time()
+        d1, ids1 = pq.search(x=X[:3], k=5)
+        t1 = time.time()
+        print(f'pq: {(t1 - t0) * 1000} msec')
+
+        t0 = time.time()
+        d2, ids2 = pq_fast.search(x=X[:3], k=5)
+        t1 = time.time()
+        print(f'pq_fast: {(t1 - t0) * 1000} msec')
+
+        assert np.allclose(ids1, ids2)
+        " > check.py
+        python check.py
+
+
+

build.md

@@ -2,7 +2,7 @@
 This document is a step-by-step instruction for building faiss from the source. We assume:
 - x86 architecture
 - CPU
-- Ubuntu 20.04 
+- Ubuntu 22.04 
 - miniconda for python environment
 - Intel MKL (we can install it simply by `apt` for Ubuntu 20.04 or higher)
 - AVX2
@@ -49,16 +49,19 @@ If you cannot install intel-mkl, you can use open-blas by `sudo apt install -y l
 
 
 ### cmake
-Currently, cmake from apt is old (3.16). There are three options to install new cmake.
+Currently, cmake from apt is old (3.16 for Ubuntu 20.04, and 3.22 for Ubuntu 22.04). For faiss 1.7+, we need cmake 3.23+. There are three options to install new cmake.
 - Build from source
 - Install by snap. This is the easiest.
     ```bash
     sudo snap install cmake --classic
     ```
+    Note that WSL recently supported snap. See [this](https://devblogs.microsoft.com/commandline/systemd-support-is-now-available-in-wsl/#set-the-systemd-flag-set-in-your-wsl-distro-settings).
 - If you've installed conda, you can install cmake by conda.
     ```bash
     conda install -c anaconda cmake 
     ```
+- [Use APT repository](https://askubuntu.com/a/1157132). Seems easy. Not tested by myself though.
+
 
 ### miniconda
 We will use miniconda for python. See [this](https://conda.io/projects/conda/en/latest/user-guide/install/macos.html#install-macos-silent) for the instruction of the silent installation.
@@ -163,6 +166,7 @@ Then let's install the module on your python.
 cd build/faiss/python
 python setup.py install
 ```
+This will update your python environment (You can uninstall it by `pip uninstall faiss`).
 
 Finally, you need to specify the PYTHONPATH. Activate it, and write it on `~/.bashrc`.
 ```bash
@@ -202,9 +206,9 @@ import faiss
 import numpy as np
 import time
 
-np.random.seed(123)
+np.random.seed(234)
 D = 128
-N = 1000
+N = 10000
 X = np.random.random((N, D)).astype(np.float32)
 M = 64
 nbits = 4
@@ -233,8 +237,8 @@ assert np.allclose(ids1, ids2)
 Then run `python check.py`.
 If AVX2 is properly activated, pq_fast should be roughly 10x faster:
 ```bash
-pq: 0.5166530609130859 msec
-pq_fast: 0.06580352783203125 msec
+pq: 1.8916130065917969 msec
+pq_fast: 0.1723766326904297 msec
 ```