|
|
@@ -0,0 +1,47 @@
|
|
|
+@c -*-texinfo-*-
|
|
|
+
|
|
|
+@c This file is part of the StarPU Handbook.
|
|
|
+@c Copyright (C) 2012 University of Bordeaux
|
|
|
+@c See the file starpu.texi for copying conditions.
|
|
|
+
|
|
|
+@menu
|
|
|
+* Task size overhead:: Overhead of tasks depending on their size
|
|
|
+* Data transfer latency:: Latency of data transfers
|
|
|
+* Gemm:: Matrix-matrix multiplication
|
|
|
+* Cholesky:: Cholesky factorization
|
|
|
+* LU:: LU factorization
|
|
|
+@end menu
|
|
|
+
|
|
|
+Some interesting benchmarks are installed among examples in
|
|
|
+/usr/lib/starpu/examples . Make sure to try various schedulers, for instance
|
|
|
+STARPU_SCHED=heft
|
|
|
+
|
|
|
+@node Task size overhead
|
|
|
+@section Task size overhead
|
|
|
+
|
|
|
+This benchmark gives a glimpse into how big a size should be for StarPU overhead
|
|
|
+to be low enough. Run @code{tasks_size_overhead.sh}, it will generate a plot
|
|
|
+of the speedup of tasks of various sizes, depending on the number of CPUs being
|
|
|
+used.
|
|
|
+
|
|
|
+@node Data transfer latency
|
|
|
+@section Data transfer latency
|
|
|
+
|
|
|
+@code{local_pingpong} performs a ping-pong between the first two CUDA nodes, and
|
|
|
+prints the measured latency.
|
|
|
+
|
|
|
+@node Gemm
|
|
|
+@section Matrix-matrix multiplication
|
|
|
+
|
|
|
+@code{sgemm} and @code{dgemm} perform a blocked matrix-matrix
|
|
|
+multiplication using BLAS and cuBLAS. They output the obtained GFlops.
|
|
|
+
|
|
|
+@node Cholesky
|
|
|
+@section Cholesky factorization
|
|
|
+
|
|
|
+@code{cholesky*} perform a Cholesky factorization (single precision). They use different dependency primitives.
|
|
|
+
|
|
|
+@node LU
|
|
|
+@section LU factorization
|
|
|
+
|
|
|
+@code{lu*} perform an LU factorization. They use different dependency primitives.
|
