Hi guys. The following compute shader takes over 40 seconds to link. Halving the tempStorage size reduces it to around 8 seconds (still unacceptable), hinting that the linking time somehow relates to the specified SSBO size.
#version 430
layout (local_size_x = 1, local_size_y = 1) in;
layout (binding = 0, std430) buffer values
{
uint tempStorage[480 * 1080];
};
void main()
{
}
Using an unsized buffer causes the linking to be almost instant. Is this behavior correct? I mean, it’s not a huge error, I can work around it. However, the SSBO minimum size requirement was not exceeded at all.
Might be a driver bug. Here’s my specs:
Toshiba laptop
Core i7
GPUs: Intel HD4600 (on chip) + NVIDIA GT 740M
Fully updated Windows 10 64bit
Nvidia driver version 358.50 (most recent)
#pragma optimize(off) has no effect, not even changing work group size. Any ideas?