From 93ffb876c60838bee75d3bb25ebbcbfce02e2cc7 Mon Sep 17 00:00:00 2001 From: Cedric Nugteren Date: Wed, 29 Nov 2017 20:21:08 +0100 Subject: Reformatted unrollable kernel loops and added the new promote_to_registers pragma for several kernels --- src/kernels/level1/xaxpy.opencl | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src/kernels/level1/xaxpy.opencl') diff --git a/src/kernels/level1/xaxpy.opencl b/src/kernels/level1/xaxpy.opencl index 3a574ec2..74e49930 100644 --- a/src/kernels/level1/xaxpy.opencl +++ b/src/kernels/level1/xaxpy.opencl @@ -45,8 +45,8 @@ void XaxpyFaster(const int n, const real_arg arg_alpha, if (get_global_id(0) < n / (VW)) { #pragma unroll - for (int w = 0; w < WPT; w += 1) { - const int id = w*get_global_size(0) + get_global_id(0); + for (int _w = 0; _w < WPT; _w += 1) { + const int id = _w*get_global_size(0) + get_global_id(0); realV xvalue = xgm[id]; realV yvalue = ygm[id]; ygm[id] = MultiplyAddVector(yvalue, alpha, xvalue); @@ -63,8 +63,8 @@ void XaxpyFastest(const int n, const real_arg arg_alpha, const real alpha = GetRealArg(arg_alpha); #pragma unroll - for (int w = 0; w < WPT; w += 1) { - const int id = w*get_global_size(0) + get_global_id(0); + for (int _w = 0; _w < WPT; _w += 1) { + const int id = _w*get_global_size(0) + get_global_id(0); realV xvalue = xgm[id]; realV yvalue = ygm[id]; ygm[id] = MultiplyAddVector(yvalue, alpha, xvalue); -- cgit v1.2.3