Hi, I've been working on generating CUDA code for sparse matrix multiplication using TVM in order to save me from the laborious work of designing and tuning the kernel. Since I'm new to TVM , I can't figure out why topi cannot express the computation. In this [code](https://github.com/amazon-research/FeatGraph/blob/master/python/featgraph/op/vanilla_spmm.py), it seems that the compute and schedule method can also be used in the SpMM. I suppose that maybe I can also use cache_read or other techniques described in this [tutorial](https://tvm.apache.org/docs/tutorials/optimize/opt_conv_cuda.html#sphx-glr-tutorials-optimize-opt-conv-cuda-py) to achieve higher performance? I've also read the [code](https://github.com/apache/tvm/blob/main/python/tvm/topi/cuda/sparse.py#L134) you mentioned above, and I'm wondering if there is a shorted way to express such operation. Thanks!
--- [Visit Topic](https://discuss.tvm.apache.org/t/sparse-opencl-error-scheduling-sparse-computations-that-use-tir-ir-builder/8585/8) to respond. You are receiving this because you enabled mailing list mode. To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/48c2d306f825ad308695b3cd7c36a681756201ce077f8f898ffd8d447a9b842a).
