Hi!
Cuda profiler shows 2 divergent branches in my kernel. But I don’t use a if or switch instruction. What is the reason?
best regards!
Hi!
Cuda profiler shows 2 divergent branches in my kernel. But I don’t use a if or switch instruction. What is the reason?
best regards!
May be you used the ternary operator ( i == j ? 4 : 2)
OR
You used internal math functions that translate to conditionals and so on…