Most notably, this PR: - changes the traits (and assembly format) of addptr so it can handle offsets that have arbitrary integer width. - adds support for `cat`
Example: ``` auto encoding = triton::gpu::BlockedEncodingAttr::get( &getContext(), {8, 32}, {2, 2}, {1, 0}, 2); //shape = [32 x 8], order = [1, 0], sizePerThread=[2, 2], numWarps=2 ``` Expected output: ``` //#triton_gpu.blocked_layout<{ // sizePerThread = {2, 2} // threadsPerWarp = {8, 4} // warpsPerCTA = {2, 1} //}> ``` Incorrect output by the current branch ``` //#triton_gpu.blocked_layout<{ // sizePerThread = {2, 2} // threadsPerWarp = {16, 2} // warpsPerCTA = {2, 1} //}> ```
This PR adds basic support for scalar-type inputs to some ops (cast and pointer arithmetics) for Triton-MLIR. Also renames getelementptr -> addptr
num-warps
Improved design to avoid inconsistent `num-warps` value between the pass and the parent module of the operation it processes.
[Triton] make other optional and remove isOtherUnspecified
ExpandDimsOp
