[Beignet] [PATCH 2/2] add utest for intel_sub_group_shuffle

[Guo, Yejun]

Please ignore this patch set, I'll send a new patch set, thanks.

-----Original Message-----
From: Guo, Yejun 
Sent: Tuesday, May 12, 2015 1:34 PM
To: beignet at lists.freedesktop.org
Cc: Guo, Yejun
Subject: [PATCH 2/2] add utest for intel_sub_group_shuffle

Signed-off-by: Guo Yejun <yejun.guo at intel.com>
---
 kernels/compiler_sub_group_shuffle.cl | 15 ++++++++++++
 utests/CMakeLists.txt                 |  3 ++-
 utests/compiler_sub_group_shuffle.cpp | 44 +++++++++++++++++++++++++++++++++++
 3 files changed, 61 insertions(+), 1 deletion(-)  create mode 100644 kernels/compiler_sub_group_shuffle.cl
 create mode 100644 utests/compiler_sub_group_shuffle.cpp

diff --git a/kernels/compiler_sub_group_shuffle.cl b/kernels/compiler_sub_group_shuffle.cl
new file mode 100644
index 0000000..f139920
--- /dev/null
+++ b/kernels/compiler_sub_group_shuffle.cl
@@ -0,0 +1,15 @@
+__kernel void compiler_sub_group_shuffle(global int *dst, int c) {
+  int i = get_global_id(0);
+  if (i == 0)
+    dst[0] = __gen_ocl_get_simd_size();
+  dst++;
+
+  int from = i;
+  int o0 = __gen_ocl_get_simd_id();
+  int o1 = intel_sub_group_shuffle(from, c);
+  int o2 = intel_sub_group_shuffle(from, 5);
+  dst[i*3] = o0;
+  dst[i*3+1] = o1;
+  dst[i*3+2] = o2;
+}
diff --git a/utests/CMakeLists.txt b/utests/CMakeLists.txt index dcb3385..b68eb7b 100644
--- a/utests/CMakeLists.txt
+++ b/utests/CMakeLists.txt
@@ -210,7 +210,8 @@ set (utests_sources
   runtime_use_host_ptr_buffer.cpp
   runtime_alloc_host_ptr_buffer.cpp
   compiler_get_simd_size.cpp
-  compiler_get_simd_id.cpp)
+  compiler_get_simd_id.cpp
+  compiler_sub_group_shuffle.cpp)
 
 if (LLVM_VERSION_NODOT VERSION_GREATER 34)
   SET(utests_sources
diff --git a/utests/compiler_sub_group_shuffle.cpp b/utests/compiler_sub_group_shuffle.cpp
new file mode 100644
index 0000000..a746d75
--- /dev/null
+++ b/utests/compiler_sub_group_shuffle.cpp
@@ -0,0 +1,44 @@
+#include "utest_helper.hpp"
+
+void compiler_sub_group_shuffle(void)
+{
+  const size_t n = 32;
+  const int32_t buf_size = 3 * n + 1;
+
+  // Setup kernel and buffers
+  OCL_CREATE_KERNEL("compiler_sub_group_shuffle");
+  OCL_CREATE_BUFFER(buf[0], 0, buf_size * sizeof(int), NULL);  
+ OCL_SET_ARG(0, sizeof(cl_mem), &buf[0]);
+
+  int c = 3;
+  OCL_SET_ARG(1, sizeof(int), &c);
+
+  globals[0] = n;
+  locals[0] = 16;
+
+  OCL_MAP_BUFFER(0);
+  for (int32_t i = 0; i < buf_size; ++i)
+    ((int*)buf_data[0])[i] = -1;
+  OCL_UNMAP_BUFFER(0);
+
+  // Run the kernel on GPU
+  OCL_NDRANGE(1);
+
+  // Compare
+  OCL_MAP_BUFFER(0);
+  int* dst = (int *)buf_data[0];
+  int simdsize = dst[0];
+  OCL_ASSERT(simdsize == 8 || simdsize == 16);
+
+  dst++;
+  for (int32_t i = 0; i < (int32_t) n; ++i){
+    int round = i / simdsize;
+    int index = i % simdsize;
+    OCL_ASSERT(index == dst[3*i]);
+    OCL_ASSERT((round * simdsize + c) == dst[3*i+1]);
+    OCL_ASSERT((round * simdsize + 5) == dst[3*i+2]);
+  }
+  OCL_UNMAP_BUFFER(0);
+}
+
+MAKE_UTEST_FROM_FUNCTION(compiler_sub_group_shuffle);
--
1.9.1