优化nmz tp8

csrc/extension/flash_fwd_mla_kernel_fp8.h

@@ -863,7 +863,12 @@ __forceinline__ __device__ void compute_attn_1rowblock_splitkv_mla_fp8_gfx938(co
     
     constexpr static int STAGE = 8;
 #if 1
-    uint8_t* kv_lds_write_ptr_base = reinterpret_cast<uint8_t*>(&tSsK(0, 0 ,0));
+    extern __shared__ char shared_memory[];
+    int row_ = lane_idx / 8;
+    int col_ = lane_idx % 8;
+    int swizzle_col_ = row_ ^ col_;
+    uint8_t* kv_lds_write_ptr_base = reinterpret_cast<uint8_t*>(shared_memory) + 
+        row_ * 128 + swizzle_col_ * 16 + warp_idx * 16 * 64;
     
     v4f c0_0, c0_1, c1_0, c1_1, c2_0, c2_1, c3_0, c3_1;
     c0_0.x = 0.0f; c0_0.y = 0.0f; c0_0.z = 0.0f; c0_0.w = 0.0f;  
@@ -877,7 +882,6 @@ __forceinline__ __device__ void compute_attn_1rowblock_splitkv_mla_fp8_gfx938(co
 
     c3_0.x = 0.0f; c3_0.y = 0.0f; c3_0.z = 0.0f; c3_0.w = 0.0f;  
     c3_1.x = 0.0f; c3_1.y = 0.0f; c3_1.z = 0.0f; c3_1.w = 0.0f;  
-    extern __shared__ char shared_memory[];
     struct IsMaskBlock {};
     struct IsFirstMaskBlock {};
     struct IsNoMaskBlock {};

csrc/extension/utils.h

@@ -642,8 +642,8 @@ buffer_load_copy_qkvfp8(
         global_addr[3] = 0x00020000;
 
         int mma_k = 32*64;
-        int row = lane % 16;
-        int col = lane / 16;
+        int row = lane / 4;
+        int col = lane % 4;
         int row_offset = row +  (warp_id * 16);
         int col_offset = col * elements_per_thread + k_idx * 64;
         int offset_v = (row_offset * row_stride + col_offset) * element_size; // bytes