[Dev][Bugfix] Add RMS Normalization Kernels and Fix Reduce Bug (#188)

* [Dev][Bugfix] Add RMS Normalization Kernels and Fix Reduce Bug

- Implement two RMS normalization implementations in TileLang:
  * `rms_norm_splitk`: Split-K reduction approach for large matrices
  * `rms_norm`: Full reduction kernel with simplified implementation
- Add reference implementation using PyTorch for validation
- Include performance benchmarking for both kernel variants
- Demonstrate flexible block size and matrix size configurations

* [Examples] Simplify RMS Normalization Kernel Compilation

- Remove commented-out code for split-K RMS normalization
- Simplify kernel compilation by removing explicit TMA lowering configuration
- Update copyright header to Tile-AI Corporation
- Streamline main script for RMS normalization example

examples/norm/rms_norm.py

+import torch
+import tilelang
+import tilelang.language as T
+
+
+def rms_norm_splitk(M, N, blk_m, blk_k):
+    dtype = "float"
+
+    @T.prim_func
+    def main(A: T.Buffer((M, N), dtype), B: T.Buffer((M, N), dtype)):
+        with T.Kernel(T.ceildiv(M, blk_m), threads=128) as bx:
+            A_shared = T.alloc_shared((blk_m, blk_k), dtype)
+            A_local = T.alloc_fragment((blk_m, blk_k), dtype)
+            A_powsum = T.alloc_fragment((blk_m,), dtype)
+
+            num_k_step = T.ceildiv(N, blk_k)
+            T.clear(A_local)
+            for k in range(num_k_step):
+                T.copy(A[bx * blk_m, k * blk_k], A_shared)
+                for i, j in T.Parallel(blk_m, blk_k):
+                    A_local[i, j] += A_shared[i, j] * A_shared[i, j]
+            T.reduce_sum(A_local, A_powsum, dim=1)
+            for i in T.Parallel(blk_m):
+                A_powsum[i] = T.rsqrt(A_powsum[i] / N) + 1e-12
+
+            for k in range(num_k_step):
+                # reverse, better cache hit rate
+                T.copy(A[bx * blk_m, (num_k_step - 1 - k) * blk_k], A_shared)
+                for i, j in T.Parallel(blk_m, blk_k):
+                    A_shared[i, j] *= A_powsum[i]
+                T.copy(A_shared, B[bx * blk_m, (num_k_step - 1 - k) * blk_k])
+
+    return main
+
+
+def rms_norm(M, N, blk_m):
+    dtype = "float"
+
+    @T.prim_func
+    def main(A: T.Buffer((M, N), dtype), B: T.Buffer((M, N), dtype)):
+        with T.Kernel(T.ceildiv(M, blk_m), threads=128) as bx:
+            A_shared = T.alloc_shared((blk_m, N), dtype)
+            A_local = T.alloc_fragment((blk_m, N), dtype)
+            A_powsum = T.alloc_fragment((blk_m,), dtype)
+
+            T.copy(A[bx * blk_m:(bx + 1) * blk_m, :], A_shared)
+            for i, j in T.Parallel(blk_m, N):
+                A_local[i, j] = A_shared[i, j] * A_shared[i, j]
+            T.reduce_sum(A_local, A_powsum, dim=1)
+            for i in T.Parallel(blk_m):
+                A_powsum[i] = T.rsqrt(A_powsum[i] / N) + 1e-12
+            for i, j in T.Parallel(blk_m, N):
+                A_shared[i, j] *= A_powsum[i]
+            T.copy(A_shared, B[bx * blk_m:(bx + 1) * blk_m, :])
+
+    return main
+
+
+def ref_program(x):
+    return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + 1e-12)
+
+
+if __name__ == "__main__":
+    M, N, blk_m, blk_k = 8192, 8192, 1, 512
+    program = rms_norm(M, N, blk_m)
+    kernel = tilelang.compile(program, out_idx=-1, target="cuda", execution_backend="cython")
+    profiler = kernel.get_profiler()
+    profiler.assert_allclose(ref_program, rtol=0.01, atol=0.01)
+    print("All checks pass.")
+
+    latency = profiler.do_bench(ref_program, warmup=500)
+    print("Ref: {:.2f} ms".format(latency))
+    latency = profiler.do_bench(profiler.mod, warmup=500)
+    print("Tile-lang: {:.2f} ms".format(latency))

src/tl_templates/cuda/reduce.h

@@ -24,7 +24,8 @@ struct MinOp {
   }
 };
 
-template <class Reducer, int threads, int scale> struct AllReduce {
+template <class Reducer, int threads, int scale, int all_threads = threads>
+struct AllReduce {
   static_assert(threads == 1024 or threads == 512 or threads == 256 or
                 threads == 128 or threads == 64 or threads == 32 or
                 threads == 16 or threads == 8 or threads == 4 or threads == 2);
@@ -50,9 +51,9 @@ template <class Reducer, int threads, int scale> struct AllReduce {
   static TL_DEVICE T run_hopper(T x, T *red_buf = nullptr) {
     constexpr int offset = threads / 2;
     if constexpr (offset >= 32) {
-      asm volatile("bar.sync %0, %1;" : : "r"(1), "r"(threads));
+      asm volatile("bar.sync %0, %1;" : : "r"(1), "r"(all_threads));
       red_buf[threadIdx.x] = x;
-      asm volatile("bar.sync %0, %1;" : : "r"(2), "r"(threads));
+      asm volatile("bar.sync %0, %1;" : : "r"(2), "r"(all_threads));
       x = Reducer()(x, red_buf[threadIdx.x ^ offset]);
     } else {
       x = Reducer()(x, T(__shfl_xor_sync(uint32_t(-1), x, offset)));
@@ -60,7 +61,8 @@ template <class Reducer, int threads, int scale> struct AllReduce {
     if constexpr (offset == scale) {
       return x;
     } else {
-      return AllReduce<Reducer, offset, scale>::run_hopper(x, red_buf);
+      return AllReduce<Reducer, offset, scale, all_threads>::run_hopper(
+          x, red_buf);
     }
   }
 };