diff options
Diffstat (limited to 'src/kernels/level1/xaxpy.opencl')
| -rw-r--r-- | src/kernels/level1/xaxpy.opencl | 24 |
1 files changed, 20 insertions, 4 deletions
diff --git a/src/kernels/level1/xaxpy.opencl b/src/kernels/level1/xaxpy.opencl index 772b57f3..b20ad200 100644 --- a/src/kernels/level1/xaxpy.opencl +++ b/src/kernels/level1/xaxpy.opencl @@ -22,7 +22,11 @@ R"( // ================================================================================================= // Full version of the kernel with offsets and strided accesses -__kernel __attribute__((reqd_work_group_size(WGS, 1, 1))) +#if RELAX_WORKGROUP_SIZE == 1 + __kernel +#elif + __kernel __attribute__((reqd_work_group_size(WGS, 1, 1))) +#endif void Xaxpy(const int n, const real_arg arg_alpha, const __global real* restrict xgm, const int x_offset, const int x_inc, __global real* ygm, const int y_offset, const int y_inc) { @@ -37,7 +41,11 @@ void Xaxpy(const int n, const real_arg arg_alpha, // Faster version of the kernel without offsets and strided accesses but with if-statement. Also // assumes that 'n' is dividable by 'VW' and 'WPT'. -__kernel __attribute__((reqd_work_group_size(WGS, 1, 1))) +#if RELAX_WORKGROUP_SIZE == 1 + __kernel +#elif + __kernel __attribute__((reqd_work_group_size(WGS, 1, 1))) +#endif void XaxpyFaster(const int n, const real_arg arg_alpha, const __global realV* restrict xgm, __global realV* ygm) { @@ -57,7 +65,11 @@ void XaxpyFaster(const int n, const real_arg arg_alpha, // Faster version of the kernel without offsets and strided accesses. Also assumes that 'n' is // dividable by 'VW', 'WGS' and 'WPT'. -__kernel __attribute__((reqd_work_group_size(WGS, 1, 1))) +#if RELAX_WORKGROUP_SIZE == 1 + __kernel +#elif + __kernel __attribute__((reqd_work_group_size(WGS, 1, 1))) +#endif void XaxpyFastest(const int n, const real_arg arg_alpha, const __global realV* restrict xgm, __global realV* ygm) { @@ -75,7 +87,11 @@ void XaxpyFastest(const int n, const real_arg arg_alpha, // ================================================================================================= // Full version of the kernel with offsets and strided accesses: batched version -__kernel __attribute__((reqd_work_group_size(WGS, 1, 1))) +#if RELAX_WORKGROUP_SIZE == 1 + __kernel +#elif + __kernel __attribute__((reqd_work_group_size(WGS, 1, 1))) +#endif void XaxpyBatched(const int n, const __constant real_arg* arg_alphas, const __global real* restrict xgm, const __constant int* x_offsets, const int x_inc, __global real* ygm, const __constant int* y_offsets, const int y_inc) { |