with tow op like
a,b,c size is n1
d,e,f size is n2
op1:c = a + b
op2:f = d + e
block (0,n1) cal op1;
block (n1, n1 + n2) cal op2
It is very import for small network when the num of block is small than the num of sm
with tow op like
a,b,c size is n1
d,e,f size is n2
op1:c = a + b
op2:f = d + e
block (0,n1) cal op1;
block (n1, n1 + n2) cal op2
It is very import for small network when the num of block is small than the num of sm
A HorizontalFusionPass can resolve these problem. Thanks!
hello, I wonder which HorizontalFusionPass you mean
We have implemented a Horizontal Fusion pass in SparseTIR, and we will create a PR to TVM mainline soon.