diff options
author | CNugteren <web@cedricnugteren.nl> | 2015-08-22 17:11:20 +0200 |
---|---|---|
committer | CNugteren <web@cedricnugteren.nl> | 2015-08-22 17:11:20 +0200 |
commit | ff0c54c3865b45eff807315262e73d3f01cb19c3 (patch) | |
tree | 839e9def73fb068f988b07e1e879ecce48d884c8 /src/kernels/level1/xaxpy.opencl | |
parent | 75517353d505de1d3979866060261a666aebfd36 (diff) |
Added the XSWAP, XSCAL and XCOPY level-1 routines
Diffstat (limited to 'src/kernels/level1/xaxpy.opencl')
-rw-r--r-- | src/kernels/level1/xaxpy.opencl | 6 |
1 files changed, 4 insertions, 2 deletions
diff --git a/src/kernels/level1/xaxpy.opencl b/src/kernels/level1/xaxpy.opencl index 3d926d9e..1f1e8ce0 100644 --- a/src/kernels/level1/xaxpy.opencl +++ b/src/kernels/level1/xaxpy.opencl @@ -11,6 +11,8 @@ // strides (incx=incy=1) and no offsets (offx=offy=0). Another version is more general, but doesn't // support vector data-types. // +// This kernel uses the level-1 BLAS common tuning parameters. +// // ================================================================================================= // Enables loading of this file using the C++ pre-processor's #include (C++11 standard raw string @@ -38,8 +40,8 @@ __kernel void Xaxpy(const int n, const real alpha, // dividable by 'VW', 'WGS' and 'WPT'. __attribute__((reqd_work_group_size(WGS, 1, 1))) __kernel void XaxpyFast(const int n, const real alpha, - const __global realV* restrict xgm, - __global realV* ygm) { + const __global realV* restrict xgm, + __global realV* ygm) { #pragma unroll for (int w=0; w<WPT; ++w) { const int id = w*get_global_size(0) + get_global_id(0); |