Polish windows compile for Ninja, fix UT random compile (#34237)

* polish windows compile for Ninja, fix random compile fail

* polish windows compile for Ninja, fix random compile fail

CMakeLists.txt

@@ -119,8 +119,6 @@ if(WIN32)
         endforeach(flag_var)
     endif()
 
-    math(EXPR PROCESS_MAX "${CPU_CORES} * 2 / 3")
-
     # windows build turn off warnings, use parallel compiling.
     foreach(flag_var
         CMAKE_CXX_FLAGS CMAKE_CXX_FLAGS_DEBUG CMAKE_CXX_FLAGS_RELEASE
@@ -128,8 +126,12 @@ if(WIN32)
         CMAKE_C_FLAGS CMAKE_C_FLAGS_DEBUG CMAKE_C_FLAGS_RELEASE
         CMAKE_C_FLAGS_MINSIZEREL CMAKE_C_FLAGS_RELWITHDEBINFO)
         string(REGEX REPLACE "/W[1-4]" " /W0 " ${flag_var} "${${flag_var}}")
-        # NOTE(zhouwei25): GPU compile have too high memory utilization when parallel compiling
-        if(NOT WITH_GPU)
+        
+        # NOTE(zhouwei25): GPU compile have too high memory utilization when parallel compiling,
+        # For Visual Studio generators, /MP should be added. 
+        # For other generators like Ninja, it is not need to add /MP.
+        if("${CMAKE_GENERATOR}" STREQUAL "Visual Studio" AND NOT WITH_GPU)
+            math(EXPR PROCESS_MAX "${CPU_CORES} * 2 / 3")
             set(${flag_var} "${${flag_var}} /MP${PROCESS_MAX}")
         endif()
     endforeach(flag_var)

cmake/ccache.cmake

@@ -18,7 +18,7 @@ elseif("${CMAKE_GENERATOR}" STREQUAL "Ninja")
 
     if(SCCACHE_PATH)
         execute_process(COMMAND sccache -V OUTPUT_VARIABLE sccache_version)
-        message(STATUS "${sccache_version} is founded, use [${SCCACHE_PATH}] to speed up compile on Windows.")
+        message(STATUS "sccache is founded, use [${SCCACHE_PATH}] to speed up compile on Windows.")
 
         set(CMAKE_C_COMPILER_LAUNCHER ${SCCACHE_PATH})
         set(CMAKE_CXX_COMPILER_LAUNCHER ${SCCACHE_PATH})

paddle/scripts/paddle_build.bat

@@ -324,14 +324,17 @@ if %day_now% NEQ %day_before% (
     echo %day_now% > %cache_dir%\day.txt
     type %cache_dir%\day.txt
     if %day_now% EQU 21 (
+        del D:\sccache\sccache_log.txt
         rmdir %cache_dir%\third_party_GPU /s/q
         rmdir %cache_dir%\third_party /s/q
     )
     if %day_now% EQU 11 (
+        del D:\sccache\sccache_log.txt
         rmdir %cache_dir%\third_party_GPU /s/q
         rmdir %cache_dir%\third_party /s/q
     )
     if %day_now% EQU 01 (
+        del D:\sccache\sccache_log.txt
         rmdir %cache_dir%\third_party_GPU /s/q
         rmdir %cache_dir%\third_party /s/q
     )

tools/parallel_UT_rule.py

@@ -685,7 +685,6 @@ TWO_PARALLEL_JOB = [
     'test_nn_functional_hot_op',
     'test_op_name_conflict',
     'test_imperative_gan',
-    'test_simnet',
     'test_amp_check_finite_and_scale_op',
     'test_random_seed',
     'test_histogram_op',
@@ -819,7 +818,6 @@ TWO_PARALLEL_JOB = [
     'test_prelu_op',
     'test_fill_zeros_like_op',
     'test_pool2d_op',
-    'test_for_enumerate',
     'test_gather_op',
     'test_partial_concat_op',
     'test_gaussian_random_op',
@@ -883,7 +881,6 @@ TWO_PARALLEL_JOB = [
     'test_empty_like_op',
     'test_rank_loss_op',
     'test_elementwise_mod_op',
-    'test_reinforcement_learning',
     'test_elementwise_max_op',
     'test_retain_graph',
     'test_edit_distance_op',
@@ -1001,7 +998,6 @@ TWO_PARALLEL_JOB = [
     'test_static_save_load',
     'test_coalesce_tensor_op',
     'test_fuse_bn_act_pass',
-    'test_simnet_v2',
     'test_shard_index_op',
     'test_cuda_random_seed',
     'test_dequantize_log_op',
@@ -1023,7 +1019,6 @@ TWO_PARALLEL_JOB = [
     'test_py_reader_pin_memory',
     'test_train_recognize_digits',
     'test_parallel_executor_feed_persistable_var',
-    'test_mnist',
     'test_update_loss_scaling_op',
     'test_rnn_cell_api',
     'test_imperative_load_static_param',