FB (Facebook) + GEMM (General Matrix-Matrix Multiplication) - https://code.fb.com/ml-applications/fbgemm/
Switch branches/tags
Clone or download
jspark1105 and facebook-github-bot instantiate more kernels for PackAmatrix (#47)
Pull Request resolved: #47

PackAMatrix (compared to PackAWithRowOffset) can be a faster alternative when B_zero_point = 0

Reviewed By: jianyuh

Differential Revision: D13413605

fbshipit-source-id: 2cac4560e8f166d19c58c65ae25400d1b0795b19
Latest commit ebbe4f4 Dec 11, 2018



Linux Build: CircleCI

FBGEMM (Facebook GEneral Matrix Multiplication) is a low-precision, high-performance matrix-matrix multiplications and convolution library for server-side inference.

The library provides efficient low-precision general matrix multiplication for small batch sizes and support for accuracy-loss minimizing techniques such as row-wise quantization and outlier-aware quantization. FBGEMM also exploits fusion opportunities in order to overcome the unique challenges of matrix multiplication at lower precision with bandwidth-bound operations.

FBGEMM is used as a backend of Caffe2 quantized operators for x86 machines (https://github.com/pytorch/pytorch/tree/master/caffe2/quantization/server). We also plan to integrate FBGEMM into PyTorch.


The tests (in test folder) and benchmarks (in bench folder) are some great examples of using FBGEMM. For instance, SpMDMTest test in test/PackedRequantizeAcc16Test.cc shows how to combine row offset calculations with packing of A (PackAWithRowOffset), how to pack B matrix (PackBMatrix) and construct output pipeline (sparse_matrix*dense_matrix --> requantization --> nop) fused with inner GEMM macro kernel.

Build Notes

FBGEMM uses the standard CMAKE-based build flow.


FBGEMM requires gcc 4.9+ and a CPU with support for avx2 instruction set or higher. It's been tested on Mac OS X and Linux.

  • asmjit

With inner kernels, FBGEMM takes a “one size doesn't fit all” approach, so the implementation dynamically generates efficient matrix-shape specific vectorized code using a third-party library called asmjit. asmjit is required to build FBGEMM.

  • cpuinfo

FBGEMM detects CPU instruction set support at runtime using cpuinfo library and dispatches optimized kernels for the detected instruction set. Therefore, cpuinfo is required to detect CPU type.

  • googletest

googletest is required to build and run FBGEMM's tests. googletest is not required if you don't want to run FBGEMM tests. By default, building of tests is on. Turn it off by setting FBGEMM_BUILD_TESTS to off.

You can download asmjit, cpuinfo, googletest and set ASMJIT_SRC_DIR, CPUINFO_SRC_DIR, GOOGLETEST_SOURCE_DIR respectively for cmake to find these libraries. If any of these variables is not set, cmake will try to download that missing library in a folder called third_party in the build directory and build it using the downloaded source code.

FBGEMM, in general, does not have any dependency on Intel MKL. However, for performance comparison, some benchmarks use MKL functions. If MKL is found or MKL path is provided with INTEL_MKL_DIR benchmarks are built with MKL and performance numbers are reported for MKL functions as well. However, if MKL is not found, the benchmarks are not built.

General build instructions are as follows:

mkdir build && cd build
cmake ..

To run the tests after building FBGEMM (if tests are built), use the following command:

make test

Installing FBGEMM

make install

How FBGEMM works

For a high-level overview, design philosophy and brief descriptions of various parts of FBGEMM please see our blog.

Full documentation

We have extensively used comments in our source files. The best and up-do-date documentation is available in the source files.

Join the FBGEMM community

See the CONTRIBUTING file for how to help out.


FBGEMM is BSD licensed, as found in the LICENSE file.