Add fp8 shared_expert kernel for CPU in sgl-kernel and add UT (#6339)

Co-authored-by: Jiang, Yanbing <yanbing.jiang@intel.com>
Co-authored-by: mingfeima <mingfei.ma@intel.com>

sgl-kernel/csrc/cpu/gemm.h

@@ -104,6 +104,24 @@ void shared_expert_int8_kernel_impl(
     int64_t N,
     int64_t K);
 
+template <typename scalar_t>
+void shared_expert_fp8_kernel_impl(
+    scalar_t* __restrict__ output,
+    scalar_t* __restrict__ ic0,
+    scalar_t* __restrict__ ic1,
+    const scalar_t* __restrict__ input,
+    const at::Float8_e4m3fn* __restrict__ packed_w1,
+    const at::Float8_e4m3fn* __restrict__ packed_w2,
+    const float* __restrict__ w1s,
+    const float* __restrict__ w2s,
+    int64_t block_size_N,
+    int64_t block_size_K,
+    const scalar_t* __restrict__ fused_experts_out,
+    float routed_scaling_factor,
+    int64_t M,
+    int64_t N,
+    int64_t K);
+
 // tinygemm interface
 template <typename scalar_t>
 void tinygemm_kernel(
@@ -134,3 +152,20 @@ void tinygemm_kernel(
     int64_t ldb,
     int64_t ldc,
     bool brg);
+
+template <typename scalar_t>
+void tinygemm_kernel(
+    const scalar_t* __restrict__ A,
+    const at::Float8_e4m3fn* __restrict__ B,
+    scalar_t* __restrict__ C,
+    scalar_t* __restrict__ Btmp,
+    float* __restrict__ Ctmp,
+    const float* __restrict__ scale,
+    int64_t M,
+    int64_t N,
+    int64_t K,
+    int64_t lda,
+    int64_t ldb,
+    int64_t ldc,
+    bool brg,
+    int64_t block_size_K);