I am implementing a compute shader that uses 16-bit SSBO, defined as
layout(std430, binding=10) writeonly buffer _chunkDesc{
uint16_t chunkDesc[];};
using
#extension GL_AMD_gpu_shader_int16 : enable
The problem is that when I read the data on the CPU, it looks like they are indexed as they were 32-bit and not 16, like:
value, 0, value, 0, value, 0, value, 0,...
This works perfectly OK on nVidia cards (values not interleaved by zeros), what am I missing here? The compute shader header is here:
#version 450 core
#extension GL_ARB_gpu_shader_int64 : enable
#extension GL_AMD_gpu_shader_int64 : enable
#extension GL_ARB_shader_ballot : enable
#extension GL_NV_gpu_shader5 : enable //16 bit nV
#extension GL_AMD_gpu_shader_int16 : enable //16bit AMD
Testing on 8GB RX 480 and Vega64, same result, latest driver.
Thanks for help.
