[Common] Added an optimized gated rowwise MXFP8 SwiGLU kernel

transformer_engine/common/cast/core/common.cuh

@@ -30,11 +30,17 @@ inline bool full_tile_1D_tensor(const Tensor *const t, const size_t elems_per_bl
 
 inline bool dimensions_supported_by_TMA(const Tensor *const t) {
   const size_t cols = t->flat_last_dim();
-  constexpr size_t TMA_bytes = 16;
-  const size_t alignment_requirement = (TMA_bytes * 8) / typeToNumBits(t->dtype());
+  const size_t alignment_requirement = (TMA_GMEM_ALIGNMENT * 8) / typeToNumBits(t->dtype());
   return cols % alignment_requirement == 0;
 }
 
+__device__ __forceinline__ unsigned char*
+align_smem_ptr_per_TMA_requirements(unsigned char* p) {
+  size_t addr = reinterpret_cast<size_t>(p);
+  addr = (addr + TMA_SHMEM_ALIGNMENT - 1) & ~(TMA_SHMEM_ALIGNMENT - 1);
+  return reinterpret_cast<unsigned char*>(addr);
+}
+
 namespace kernel {
 
 constexpr size_t THREADS_PER_BLOCK = 256;

transformer_engine/common/cast/mxfp8/gated_mxfp8.cuh

@@ -20,6 +20,7 @@
 #include "../../util/math.h"
 #include "../../util/ptx.cuh"
 #include "../../utils.cuh"
+#include "./specialized/gated_mxfp8_rowwise_swiglu.cuh"
 
 namespace transformer_engine {
 namespace dispatch {
@@ -696,6 +697,18 @@ void quantize_gated(const Tensor &gated_input, const Tensor &grad, Tensor *outpu
     scaling_type = ScalingType::BIDIMENSIONAL;
   }
 
+  // Optimized BWD/FWD SwiGLU MXFP8 Rowwise kernels for BF16/FP16 inputs
+  if constexpr (!std::is_same<ParamOP, ClampedSwiGLUParam>::value) {
+    if constexpr ((!IS_BWD && (ActOP == &silu<fp32, fp32>))
+                  || (IS_BWD && (ActOP == &silu<fp32, fp32>) && (DActOP == &dsilu<fp32, fp32>))) {
+      if (((gated_input.dtype() == DType::kFloat16) || (gated_input.dtype() == DType::kBFloat16)) 
+          && (scaling_type == ScalingType::ROWWISE)) {
+        quantize_gated_rowwise<IS_BWD, ParamOP, ActOP, DActOP>(grad, gated_input, output, p, stream); 
+        return;
+      }
+    }
+  }
+
   const size_t rows = gated_input.flat_first_dim();
   const size_t cols = gated_input.flat_last_dim() / 2;
   const size_t output_cols = (IS_BWD ? 2 : 1) * cols;