data: complete SGEMM upstream from 3 repos (siboehm+wangzyon+edtallison) + xllm fused_qknorm_rope + xattention kernels

SGEMM repos (upstream_ref/sgemm_cuda/, 41 files):
  siboehm/SGEMM_CUDA: kernel 1-12, runner, CMake, cuBLAS benchmark
  wangzyon/NVIDIA_SGEMM_PRACTICE: kernel 1-7 (Chinese comments), utils
  edtallison/sgemm-cuda: kernel 01-09 (learning notes), Makefile

xllm kernels (ex_engine/xllm_kernels/cuda/):
  fused_qknorm_rope.cu + bind — saves 128 kernel launches/fwd
  xattention/ — 6 files from upstream xllm
  headers: corex_compat_utils.h, topk_last_dim.cuh
  ilu/CMakeLists.txt

SO_BUILD_MANIFEST.md — complete .so inventory and call chain analysis

upstream_ref/sgemm_cuda/kernel_2.cuh

@@ -0,0 +1,45 @@
+#pragma once
+
+#include <cuda_runtime.h>
+#include <cublas_v2.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+template<const int BLOCK_SIZE>
+__global__ void mysgemm_v2(int M, int N, int K, float alpha, float *A, float *B, float beta, float *C) {
+    int bx = blockIdx.x;
+    int by = blockIdx.y;
+
+    const int BM = BLOCK_SIZE;
+    const int BN = BLOCK_SIZE;
+    const int BK = BLOCK_SIZE;
+
+    int tx = threadIdx.x % BN;
+    int ty = threadIdx.x / BN;
+
+    // 申请共享内存空间
+    __shared__ float As[BM * BK];
+    __shared__ float Bs[BK * BN];
+
+    // 移动到当前block
+    A = &A[by * BM * K];
+    B = &B[bx * BN];
+    C = &C[by * BM * N + bx * BN];
+
+    float tmp = 0.;
+    for (int k = 0; k < K; k += BK) {
+        // 缓存A_tile和B_tile
+        As[ty * BK + tx] = A[ty * K + tx];
+        Bs[ty * BN + tx] = B[ty * N + tx];
+        // 同步所有线程缓存完成
+        __syncthreads();
+        A += BK;
+        B += BK * N;
+        for (int i = 0; i < BK; i++) {
+            tmp += As[ty * BK + i] * Bs[i * BN + tx];
+        }
+        // FMA计算需要读取缓存数据，在新一轮写入缓存前进行同步，确保所有线程计算完成
+        __syncthreads();
+    }
+    C[ty * N + tx] = alpha * tmp + beta * C[ty * N + tx];
+}