[aot] Add test for shared array

c_api/tests/c_api_aot_test.cpp

@@ -122,6 +122,41 @@ void texture_aot_kernel_test(TiArch arch) {
   }
 }
 
+static void shared_array_aot_test(TiArch arch) {
+  uint32_t kArrLen = 8192;
+
+  const auto folder_dir = getenv("TAICHI_AOT_FOLDER_PATH");
+
+  std::stringstream aot_mod_ss;
+  aot_mod_ss << folder_dir;
+
+  ti::Runtime runtime(arch);
+
+  ti::NdArray<float> v_array =
+      runtime.allocate_ndarray<float>({kArrLen}, {}, true);
+  ti::NdArray<float> d_array =
+      runtime.allocate_ndarray<float>({kArrLen}, {}, true);
+  ti::NdArray<float> a_array =
+      runtime.allocate_ndarray<float>({kArrLen}, {}, true);
+  ti::AotModule aot_mod = runtime.load_aot_module(aot_mod_ss.str().c_str());
+  ti::Kernel k_run = aot_mod.get_kernel("run");
+
+  k_run.push_arg(v_array);
+  k_run.push_arg(d_array);
+  k_run.push_arg(a_array);
+  k_run.launch();
+  runtime.wait();
+
+  // Check Results
+  float *data = reinterpret_cast<float *>(a_array.map());
+
+  for (int i = 0; i < kArrLen; ++i) {
+    EXPECT_EQ(data[i], kArrLen);
+  }
+
+  a_array.unmap();
+}
+
 TEST_F(CapiTest, AotTestCpuField) {
   TiArch arch = TiArch::TI_ARCH_X64;
   field_aot_test(arch);
@@ -166,3 +201,17 @@ TEST_F(CapiTest, GraphTestVulkanTextureKernel) {
     texture_aot_kernel_test(arch);
   }
 }
+
+TEST_F(CapiTest, AotTestCudaSharedArray) {
+  if (ti::is_arch_available(TI_ARCH_CUDA)) {
+    TiArch arch = TiArch::TI_ARCH_CUDA;
+    shared_array_aot_test(arch);
+  }
+}
+
+TEST_F(CapiTest, AotTestVulkanSharedArray) {
+  if (ti::is_arch_available(TI_ARCH_VULKAN)) {
+    TiArch arch = TiArch::TI_ARCH_VULKAN;
+    shared_array_aot_test(arch);
+  }
+}

tests/cpp/aot/python_scripts/shared_array_aot_test.py

@@ -0,0 +1,60 @@
+import argparse
+import os
+
+import numpy as np
+
+import taichi as ti
+
+
+def shared_array_aot_test(arch):
+    ti.init(arch=arch)
+
+    if ti.lang.impl.current_cfg().arch != arch:
+        return
+    block_dim = 128
+    nBlocks = 64
+    N = nBlocks * block_dim
+    v_arr = np.zeros(N).astype(np.float32)
+    d_arr = np.zeros(N).astype(np.float32)
+    a_arr = np.zeros(N).astype(np.float32)
+
+    @ti.kernel
+    def run(v: ti.types.ndarray(ndim=1), d: ti.types.ndarray(ndim=1),
+            a: ti.types.ndarray(ndim=1)):
+        for i in range(nBlocks * block_dim):
+            v[i] = 1.0
+            d[i] = 1.0
+
+        ti.loop_config(block_dim=block_dim)
+        for i in range(nBlocks * block_dim):
+            tid = i % block_dim
+            pad = ti.simt.block.SharedArray((block_dim, ), ti.f32)
+            acc = 0.0
+            v_val = v[i]
+            for k in range(nBlocks):
+                pad[tid] = d[k * block_dim + tid]
+                ti.simt.block.sync()
+                for j in range(block_dim):
+                    acc += v_val * pad[j]
+                ti.simt.block.sync()
+            a[i] = acc
+
+    assert "TAICHI_AOT_FOLDER_PATH" in os.environ.keys()
+    dir_name = str(os.environ["TAICHI_AOT_FOLDER_PATH"])
+
+    m = ti.aot.Module()
+    m.add_kernel(run, template_args={'v': v_arr, 'd': d_arr, 'a': a_arr})
+    m.save(dir_name)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--arch", type=str)
+    args = parser.parse_args()
+
+    if args.arch == "cuda":
+        shared_array_aot_test(arch=ti.cuda)
+    elif args.arch == "vulkan":
+        shared_array_aot_test(arch=ti.vulkan)
+    else:
+        assert False

tests/test_config.json

@@ -183,10 +183,18 @@
         ["cpp", "aot", "python_scripts", "kernel_aot_test1.py"],
         "--arch=cuda"
     ],
+    "CapiTest.AotTestCudaSharedArray": [
+        ["cpp", "aot", "python_scripts", "shared_array_aot_test.py"],
+        "--arch=cuda"
+    ],
     "CapiTest.AotTestVulkanKernel": [
         ["cpp", "aot", "python_scripts", "kernel_aot_test1.py"],
         "--arch=vulkan"
     ],
+    "CapiTest.AotTestVulkanSharedArray": [
+        ["cpp", "aot", "python_scripts", "shared_array_aot_test.py"],
+        "--arch=vulkan"
+    ],
     "CapiTest.AotTestOpenglKernel": [
         ["cpp", "aot", "python_scripts", "kernel_aot_test1.py"],
         "--arch=opengl"