pme_gpu_stage_atom_data(sm_coordinates, gm_coordinates, DIM);
barrier(CLK_LOCAL_MEM_FENCE);
- calculate_splines(kernelParams, atomIndexOffset, sm_coordinates, sm_coefficients, sm_theta,
- sm_gridlineIndices, sm_fractCoords, gm_theta, gm_dtheta,
- gm_gridlineIndices, gm_fractShiftsTable, gm_gridlineIndicesTable);
+ calculate_splines(kernelParams,
+ atomIndexOffset,
+ sm_coordinates,
+ sm_coefficients,
+ sm_theta,
+ sm_gridlineIndices,
+ sm_fractCoords,
+ gm_theta,
+ gm_dtheta,
+ gm_gridlineIndices,
+ gm_fractShiftsTable,
+ gm_gridlineIndicesTable);
#if !defined(_AMD_SOURCE_) && !defined(_NVIDIA_SOURCE_)
/* This is only here for execution of e.g. 32-sized warps on 16-wide hardware; this was
* __syncwarp() in CUDA. #2519
/* Spline data - only thetas (dthetas will only be needed in gather) */
pme_gpu_stage_atom_data(sm_theta, gm_theta, DIM * order);
/* Gridline indices - they're actually int and not float, but C99 is angry about overloads */
- pme_gpu_stage_atom_data((__local float*)sm_gridlineIndices,
- (__global const float*)gm_gridlineIndices, DIM);
+ pme_gpu_stage_atom_data(
+ (__local float*)sm_gridlineIndices, (__global const float*)gm_gridlineIndices, DIM);
barrier(CLK_LOCAL_MEM_FENCE);
}