123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869 |
- @c -*-texinfo-*-
- @c This file is part of the StarPU Handbook.
- @c Copyright (C) 2009-2011 Université de Bordeaux 1
- @c Copyright (C) 2010, 2011, 2012 Centre National de la Recherche Scientifique
- @c See the file starpu.texi for copying conditions.
- @smallexample
- #include <starpu.h>
- #include <xmmintrin.h>
- void scal_cpu_func(void *buffers[], void *cl_arg)
- @{
- unsigned i;
- float *factor = cl_arg;
-
- struct starpu_vector_interface *vector = buffers[0];
-
- unsigned n = STARPU_VECTOR_GET_NX(vector);
-
- float *val = (float *)STARPU_VECTOR_GET_PTR(vector);
-
- 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;
- __m128 *VECTOR = (__m128*) vector;
- __m128 FACTOR __attribute__((aligned(16)));
- float factor = *(float *) cl_arg;
- FACTOR = _mm_set1_ps(factor);
- unsigned int i;
- for (i = 0; i < n_iterations; i++)
- VECTOR[i] = _mm_mul_ps(FACTOR, VECTOR[i]);
- unsigned int remainder = n%4;
- if (remainder != 0)
- @{
- unsigned int start = 4 * n_iterations;
- for (i = start; i < start+remainder; ++i)
- @{
- vector[i] = factor * vector[i];
- @}
- @}
- @}
- @end smallexample
|