Updated the documentation with data related to mutliple implementations.

doc/starpu.texi

@@ -517,6 +517,7 @@ so:
 * Hello World::                 Submitting Tasks
 * Scaling a Vector::            Manipulating Data
 * Vector Scaling on an Hybrid CPU/GPU Machine::  Handling Heterogeneous Architectures
+* Using multiple implentations of a codelet::
 * Task and Worker Profiling::   
 * Partitioning Data::           Partitioning Data
 * Performance model example::   
@@ -1141,6 +1142,50 @@ see @ref{Enabling OpenCL}):
 0.000000 3.000000 6.000000 9.000000 12.000000
 @end smallexample
 
+@node Using multiple implentations of a codelet
+@section Using multiple implentations of a codelet
+One may want to write multiple implementations of a codelet for a single type of
+device and let StarPU choose which one to run. As an example, we will show how
+to use SSE to scale a vector. The codelet can be written as follows :
+
+@cartouche
+@smallexample
+#include <xmmintrin.h>
+
+void scal_sse_func(void *buffers[], void *cl_arg)
+@{
+	float *vector = (float *) STARPU_VECTOR_GET_PTR(buffers[0]);
+	unsigned int n = STARPU_VECTOR_GET_NX(buffers[0]);
+	unsigned int n_iterations = n/4;
+	if (n % 4 != 0)
+		n_iterations++;
+
+	__m128 *VECTOR = (__m128*) vector;
+	__m128 factor __attribute__((aligned(16)));
+	factor = _mm_set1_ps(*(float *) cl_arg);
+
+	unsigned int i;	
+	for (i = 0; i < n_iterations; i++)
+		VECTOR[i] = _mm_mul_ps(factor, VECTOR[i]);
+@}
+@end smallexample
+@end cartouche
+
+The @code{cpu_func} field of the @code{starpu_codelet} structure has to be set
+to the special value @code{STARPU_MULTIPLE_CPU_IMPLEMENTATIONS}. Note that
+@code{STARPU_MULTIPLE_CUDA_IMPLEMENTATIONS} and
+@code{STARPU_MULTIPLE_OPENCL_IMPLEMENTATIONS} are also available.
+
+@cartouche
+@smallexample
+starpu_codelet cl = @{
+	.where = STARPU_CPU,
+	.cpu_func = STARPU_MULTIPLE_CPU_IMPLEMENTATIONS,
+	.cpu_funcs = @{ scal_cpu_func, scal_sse_func @},
+	.nbuffers = 1
+@};
+@end smallexample
+@end cartouche
 @node Task and Worker Profiling
 @section Task and Worker Profiling
 
@@ -2616,6 +2661,7 @@ Enable flags for the @code{gcov} coverage tool.
 * --with-opencl-lib-dir::           
 * --enable-gordon::             
 * --with-gordon-dir::           
+* --enable-maximplementations::
 @end menu
 
 @node --enable-maxcpus
@@ -2739,6 +2785,14 @@ Enable the use of the Gordon runtime for Cell SPUs.
 Specify the location of the Gordon SDK.
 @end table
 
+@node --enable-maximplementations
+@subsubsection @code{--enable-maximplementations=<number>}
+@table @asis
+@item @emph{Description}:
+Defines the number of implementations that can be defined for a single kind of
+device. It is then available as the @code{STARPU_MAXIMPLEMENTATIONS} macro.
+@end table
+
 @node Advanced configuration
 @subsection Advanced configuration
 

doc/vector_scal_c.texi

@@ -11,13 +11,15 @@
 #define    NX    2048
 
 extern void scal_cpu_func(void *buffers[], void *_args);
+extern void scal_sse_func(void *buffers[], void *_args);
 extern void scal_cuda_func(void *buffers[], void *_args);
 extern void scal_opencl_func(void *buffers[], void *_args);
 
 static starpu_codelet cl = @{
     .where = STARPU_CPU | STARPU_CUDA | STARPU_OPENCL,
     /* CPU implementation of the codelet */
-    .cpu_func = scal_cpu_func,
+    .cpu_func = STARPU_MULTIPLE_CPU_IMPLEMENTATIONS,
+    .cpu_funcs = @{ scal_cpu_func, scal_sse_func @},
 #ifdef STARPU_USE_CUDA
     /* CUDA implementation of the codelet */
     .cuda_func = scal_cuda_func,

doc/vector_scal_cpu.texi

@@ -1,4 +1,5 @@
 #include <starpu.h>
+#include <xmmintrin.h>
 
 /* This kernel takes a buffer and scales it by a constant factor */
 void scal_cpu_func(void *buffers[], void *cl_arg)
@@ -30,3 +31,20 @@ void scal_cpu_func(void *buffers[], void *cl_arg)
     for (i = 0; i < n; i++)
         val[i] *= *factor;
 @}
+
+void scal_sse_func(void *buffers[], void *cl_arg)
+@{
+	float *vector = (float *) STARPU_VECTOR_GET_PTR(buffers[0]);
+	unsigned int n = STARPU_VECTOR_GET_NX(buffers[0]);
+	unsigned int n_iterations = n/4;
+	if (n % 4 != 0)
+		n_iterations++;
+
+	__m128 *VECTOR = (__m128*) vector;
+	__m128 factor __attribute__((aligned(16)));
+	factor = _mm_set1_ps(*(float *) cl_arg);
+
+	unsigned int i;	
+	for (i = 0; i < n_iterations; i++)
+		VECTOR[i] = _mm_mul_ps(factor, VECTOR[i]);
+@}