Skip to content
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
104 lines (64 sloc) 8.53 KB



By default, only non-coprocessor targets are built since both KNC/self-hosted ("native") and KNC/offload are considered legacy (OFFLOAD=0 and KNC=0). In general, the subfolders of the 'lib' directory are separating the build targets where a 'mic' folder contains the native library (KNC=1) targeting the Intel Xeon Phi coprocessor ("KNC"), and the 'intel64' folder contains either the hybrid archive made of CPU/host as well as coprocessor code (OFFLOAD=1, which also implies KNC=1), or an archive which is only containing the CPU code (default). To remove any BLAS-dependency, please follow the Link Instructions.

NOTE: The build system considers a set of given key-value pairs as a single unique build and triggers a rebuild for a distinct set of flags!

Static Specialization

By default, LIBXSMM uses the JIT backend which is automatically building optimized code (JIT=1). Matrix multiplication kernels can be also statically specialized at compile-time of the library (M, N, and K values). This mechanism also extends the interface of the library because function prototypes are included into both the C and FORTRAN interface.

make M="2 4" N="1" K="$(echo $(seq 2 5))"

The above example is generating the following set of (M,N,K) triplets:

(2,1,2), (2,1,3), (2,1,4), (2,1,5),
(4,1,2), (4,1,3), (4,1,4), (4,1,5)

The index sets are in a loop-nest relationship (M(N(K))) when generating the indexes. Moreover, an empty index set resolves to the next non-empty outer index set of the loop nest (including to wrap around from the M to K set). An empty index set does not participate in the loop-nest relationship. Here is an example of generating multiplication routines which are "squares" with respect to M and N (N inherits the current value of the "M loop"):

make M="$(echo $(seq 2 5))" K="$(echo $(seq 2 5))"

An even more flexible specialization is possible by using the MNK variable when building the library. It takes a list of indexes which are eventually grouped (using commas):

make MNK="2 3, 23"

Each group of the above indexes is combined into all possible triplets generating the following set of (M,N,K) values:

(2,2,2), (2,2,3), (2,3,2), (2,3,3),
(3,2,2), (3,2,3), (3,3,2), (3,3,3), (23,23,23)

Of course, both mechanisms (M/N/K and MNK based) can be combined using the same command line (make). Static optimization and JIT can also be combined (no need to turn off the JIT backend).

Targeted Compilation

Specifying a code path is not necessary if the JIT backend is not disabled. However, disabling JIT compilation, statically generating a collection of kernels, and targeting a specific instruction set extension for the entire library looks like:

make JIT=0 AVX=3 MNK="1 2 3 4 5"

The above example builds a library which cannot be deployed to anything else but the Intel Knights Landing processor family ("KNL") or future Intel Xeon processors supporting foundational Intel AVX‑512 instructions (AVX‑512F). The latter might be even more adjusted by supplying MIC=1 (along with AVX=3), however this does not matter since critical code is in inline assembly (and not affected). Similarly, SSE=0 (or JIT=0 without SSE or AVX build flag) employs an "arch-native" approach whereas AVX=1, AVX=2 (with FMA), and AVX=3 are specifically selecting the kind of Intel AVX code. Moreover, controlling the target flags manually or adjusting the code optimizations is also possible. The following example is GCC-specific and corresponds to OPT=3, AVX=3, and MIC=1:

make OPT=3 TARGET="-mavx512f -mavx512cd -mavx512er -mavx512pf"

An extended interface can be generated which allows to perform software prefetches. Prefetching data might be helpful when processing batches of matrix multiplications where the next operands are farther away or otherwise unpredictable in their memory location. The prefetch strategy can be specified similar as shown in the section Generator Driver i.e., by either using the number of the shown enumeration, or by exactly using the name of the prefetch strategy. The only exception is PREFETCH=1 which is automatically selecting a strategy per an internal table (navigated by CPUID flags). The following example is requesting the "AL2jpst" strategy:


The prefetch interface is extending the signature of all kernels by three arguments (pa, pb, and pc). These additional arguments are specifying the locations of the operands of the next multiplication (the next a, b, and c matrices). Providing unnecessary arguments in case of the three-argument kernels is not big a problem (beside of some additional call-overhead), however running a 3-argument kernel with more than three arguments and thereby picking up garbage data is misleading or disabling the hardware prefetcher (due to software prefetches). In this case, a misleading prefetch location is given plus an eventual page fault due to an out-of-bounds (garbage-)location.

Further, the generated configuration (template) of the library encodes the parameters for which the library was built for (static information). This helps optimizing client code related to the library's functionality. For example, the LIBXSMM_MAX_* and LIBXSMM_AVG_* information can be used with the LIBXSMM_PRAGMA_LOOP_COUNT macro to hint loop trip counts when handling matrices related to the problem domain of LIBXSMM.

To improve thread-scalability of the default memory allocation domain, the library can rely on Intel TBB, which is discovered per TBBROOT environment variable: make TBB=1.


The function libxsmm_?mmdispatch helps amortizing the cost of the dispatch when multiple calls with the same M, N, and K are needed. The automatic code dispatch is orchestrating two levels:

  1. Specialized routine (implemented in assembly code),
  2. BLAS library call (fallback).

Both levels are accessible directly, which allows to customize the code dispatch. The fallback level may be supplied by the Intel Math Kernel Library (Intel MKL) 11.2 DIRECT CALL feature.

Further, a preprocessor symbol denotes the largest problem-size (M x N x K) that belongs to the first level, and therefore determines if a matrix multiplication falls back to BLAS. The problem-size threshold can be configured by using for example:

make THRESHOLD=$((60 * 60 * 60))

The maximum of the given threshold and the largest requested specialization refines the value of the threshold. Please note that explicitly JIT'ting and executing a kernel is possible and independent of the threshold. If a problem-size is below the threshold, dispatching the code requires to figure out whether a specialized routine exists or not.

To minimize the probability of key collisions (code cache), the preferred precision of the statically generated code can be selected:


The default preference is to generate and register both single and double-precision code, and therefore no space in the dispatch table is saved (PRECISION=0). Specifying PRECISION=1|2 is only generating and registering either single-precision or double-precision code.

The automatic dispatch is highly convenient because existing GEMM calls can serve specialized kernels (even in a binary compatible fashion), however there is (and always will be) an overhead associated with looking up the code-registry and checking whether the code determined by the GEMM call is already JIT'ted or not. This lookup has been optimized using various techniques such as using specialized CPU instructions to calculate CRC32 checksums, to avoid costly synchronization (needed for thread-safety) until it is ultimately known that the requested kernel is not yet JIT'ted, and by implementing a small thread-local cache of recently dispatched kernels. The latter of which can be adjusted in size (only power-of-two sizes) but also disabled:

make CACHE=0

Please note that measuring the relative cost of automatically dispatching a requested kernel depends on the kernel size (obviously smaller matrices are multiplied faster on an absolute basis), however smaller matrix multiplications are bottlenecked by memory bandwidth rather than arithmetic intensity. The latter implies the highest relative overhead when (artificially) benchmarking the very same multiplication out of the CPU-cache.

You can’t perform that action at this time.