[Mlir-commits] [mlir] [MLIR][NVGPU] Move max threads/blocks size to dialect (NFC) (PR #124454)

[Matthias Springer]

================
@@ -22,8 +22,20 @@
 
 #include "mlir/Dialect/NVGPU/IR/NVGPUEnums.h.inc"
 
+// Maximum warp size
 constexpr int kWarpSize = 32;
 
+// Maximum number of threads in a block and block in a grid
+// https://docs.nvidia.com/cuda/cuda-c-programming-guide/#features-and-technical-specifications-technical-specifications-per-compute-capability
+constexpr int kMaxTotalBlockdim = 1024;
+constexpr int kMaxBlockdimx = 1024;
----------------
matthias-springer wrote:

nit: in camel case they would be `kMaxBlockDimX` etc.

https://github.com/llvm/llvm-project/pull/124454