Conv3d new (#94)

* conv3d compiles but has memory error

* conv3d works

* fix performance issue by using __builtin_amdgc_readfirstlane

* change MakeBlock2CTileMap to MakeDefaultBlock2CTileMap; change c_blockid_to* to cblockid_to*

* clang-format

* remove CK_EXPERIMENTAL_PASS_TENSOR_DECRIPTOR_BY_*; moved wrapper into DeviceConv3d

* format

* remove useless marc

* add comment
Co-authored-by: Chao Liu <chao.liu2@amd.com>

test/magic_number_division/main.cpp

@@ -41,6 +41,19 @@ gpu_naive_division(int32_t divisor, const int32_t* p_dividend, int32_t* p_result
     }
 }
 
+__host__ void cpu_magic_number_division(uint32_t magic_multiplier,
+                                        uint32_t magic_shift,
+                                        const int32_t* p_dividend,
+                                        int32_t* p_result,
+                                        uint64_t num)
+{
+    for(uint64_t data_id = 0; data_id < num; ++data_id)
+    {
+        p_result[data_id] =
+            ck::MagicDivision::DoMagicDivision(p_dividend[data_id], magic_multiplier, magic_shift);
+    }
+}
+
 template <typename T>
 T check_error(const std::vector<T>& ref, const std::vector<T>& result)
 {
@@ -90,6 +103,7 @@ int main(int, char*[])
 
     std::vector<int32_t> naive_result_host(num_dividend);
     std::vector<int32_t> magic_result_host(num_dividend);
+    std::vector<int32_t> magic_result_host2(num_dividend);
 
     dividends_dev_buf.ToDevice(dividends_host.data());
 
@@ -128,6 +142,20 @@ int main(int, char*[])
             pass = false;
             continue;
         }
+
+        cpu_magic_number_division(magic_multiplier,
+                                  magic_shift,
+                                  dividends_host.data(),
+                                  magic_result_host2.data(),
+                                  num_dividend);
+
+        max_diff = check_error(naive_result_host, magic_result_host2);
+
+        if(max_diff != 0)
+        {
+            pass = false;
+            continue;
+        }
     }
 
     if(pass)