cleaned up comments, removed logging code. sb gtg

Signed-off-by: Chris Perkins <[email protected]>

Author: cperkinsintel

sycl/include/CL/sycl/detail/cl.h

@@ -14,11 +14,6 @@
 #define CL_TARGET_OPENCL_VERSION 220
 #endif
 
-//#define CPOUT  std::clog
-#define CPOUT                                                                  \
-  std::clog.rdbuf(NULL);                                                       \
-  std::clog
-//#define CP_CHANGE_IMAGE 1
 
 #include <CL/cl.h>
 #include <CL/cl_ext.h>

sycl/include/CL/sycl/detail/image_impl.hpp

@@ -79,7 +79,6 @@ class __SYCL_EXPORT image_impl final : public SYCLMemObjT {
                 "Dimensions of cl::sycl::image can be 1, 2 or 3");
 
   void setPitches() {
-    CPOUT << "setPitches()!!" << std::endl;
     size_t WHD[3] = {1, 1, 1}; // Width, Height, Depth.
 #ifndef CP_CHANGE_IMAGE
     for (int I = 0; I < Dimensions; I++)
@@ -106,8 +105,6 @@ class __SYCL_EXPORT image_impl final : public SYCLMemObjT {
 
   template <bool B = (Dimensions > 1)>
   void setPitches(const EnableIfPitchT<B> Pitch) {
-    CPOUT << "setPitched() DIM BIG! " << Pitch[0] << " / "
-          << ((Dimensions == 3) ? Pitch[1] : 0) << std::endl;
     MRowPitch = Pitch[0];
     MSlicePitch =
         (Dimensions == 3) ? Pitch[1] : MRowPitch; // Dimensions will be 2/3.
@@ -268,45 +265,19 @@ class __SYCL_EXPORT image_impl final : public SYCLMemObjT {
   RT::PiMemImageDesc getImageDesc(bool InitFromHostPtr) {
     RT::PiMemImageDesc Desc;
     Desc.image_type = getImageType();
-    // CP
-    // MRange<> is [width], [height,width], or [depth,height,width]
-    if (Dimensions == 3) {
-      CPOUT << "getImageDesc: MRange: {" << MRange[0] << "," << MRange[1] << ","
-            << MRange[2] << "}" << std::endl;
-    } else if (Dimensions == 2) {
-      CPOUT << "getImageDesc: MRange: {" << MRange[0] << "," << MRange[1] << "}"
-            << std::endl;
-    } else if (Dimensions == 1) {
-      CPOUT << "getImageDesc: MRange: {" << MRange[0] << "}" << std::endl;
-    }
 
-#ifdef CP_CHANGE_IMAGE
-    int x_term_pos = 2, y_term_pos = 1, z_term_pos = 0;
-    if (Dimensions == 2) {
-      y_term_pos = 0;
-      x_term_pos = 1;
-    }
-    if (Dimensions == 1) {
-      x_term_pos = 0;
-    }
-#else
+    // MRange<> is [width], [width,height,width], or [width,height,depth] (which
+    // is different than MAccessRange, etc in bufffers)
     int x_term_pos = 0, y_term_pos = 1, z_term_pos = 2;
-#endif
-
     Desc.image_width = MRange[x_term_pos];
     Desc.image_height = Dimensions > 1 ? MRange[y_term_pos] : 1;
     Desc.image_depth = Dimensions > 2 ? MRange[z_term_pos] : 1;
-    CPOUT << "    Desc.image_ w/h/d: " << Desc.image_width << " / "
-          << Desc.image_height << " / " << Desc.image_depth << std::endl;
 
     // TODO handle cases with IMAGE1D_ARRAY and IMAGE2D_ARRAY
     Desc.image_array_size = 0;
     // Pitches must be 0 if host ptr is not provided.
     Desc.image_row_pitch = InitFromHostPtr ? MRowPitch : 0;
     Desc.image_slice_pitch = InitFromHostPtr ? MSlicePitch : 0;
-    CPOUT << "    Desc.image_ row_pitch/slice: " << Desc.image_row_pitch
-          << " / " << Desc.image_slice_pitch << std::endl;
-
     Desc.num_mip_levels = 0;
     Desc.num_samples = 0;
     Desc.buffer = nullptr;

sycl/plugins/opencl/pi_opencl.cpp

@@ -529,20 +529,12 @@ pi_result piMemImageCreate(pi_context context, pi_mem_flags flags,
                            const pi_image_format *image_format,
                            const pi_image_desc *image_desc, void *host_ptr,
                            pi_mem *ret_mem) {
-  CPOUT << "image_desc w/h/d: " << image_desc->image_width << " / "
-        << image_desc->image_height << " / " << image_desc->image_depth
-        << std::endl;
-  CPOUT << "           arrSz/row/slice: " << image_desc->image_array_size
-        << " / " << image_desc->image_row_pitch << " / "
-        << image_desc->image_slice_pitch << std::endl;
-
   pi_result ret_err = PI_INVALID_OPERATION;
   *ret_mem = cast<pi_mem>(
       clCreateImage(cast<cl_context>(context), cast<cl_mem_flags>(flags),
                     cast<const cl_image_format *>(image_format),
                     cast<const cl_image_desc *>(image_desc), host_ptr,
                     cast<cl_int *>(&ret_err)));
-  CPOUT << "ret_mem :: err: " << *ret_mem << " :: " << ret_err << std::endl;
   return ret_err;
 }
 

sycl/source/detail/memory_manager.cpp

@@ -256,9 +256,7 @@ void prepTermPositions(term_positions &pos, int Dimensions,
   //  1 ==>  {width, 1, 1}
   //  2 ==>  {height, width, 1}
   //  3 ==>  {depth, height, width}
-  //  If the Dimension is 1, and being called by ~SYCLMemObjT, then
-  //  DstAccessRange[0] and DstSize[0] will already sized to bytes with
-  //  DstElemSize of 1. Some callers enqueue 0 as DimDst/DimSrc.
+  // Some callers enqueue 0 as DimDst/DimSrc.
 #ifdef CP_CHANGE_IMAGE
   if (true) {
 #else
@@ -296,38 +294,14 @@ void copyH2D(SYCLMemObjI *SYCLMemObj, char *SrcMem, QueueImplPtr,
   const RT::PiQueue Queue = TgtQueue->getHandleRef();
   const detail::plugin &Plugin = TgtQueue->getPlugin();
 
-  CPOUT << "copyH2D " << std::endl;
-  CPOUT << "DimSrc: " << DimSrc << "  SrcAccessRange: " << SrcAccessRange[0]
-        << " / " << SrcAccessRange[1] << " / " << SrcAccessRange[2]
-        << std::endl;
-  // CPOUT << "        " << DimSrc << "  SrcOffset: " << SrcOffset[0] << " / "
-  // << SrcOffset[1] << " / " << SrcOffset[2] << std::endl;
-  CPOUT << "DimDst: " << DimDst << "  DstAccessRange: " << DstAccessRange[0]
-        << " / " << DstAccessRange[1] << " / " << DstAccessRange[2]
-        << std::endl;
-  // CPOUT << "        " << DimSrc << "  DstOffset: " << DstOffset[0] << " / "
-  // << DstOffset[1] << " / " << DstOffset[2] << std::endl;
-  CPOUT << "SrcElemSize / DstElemSize: " << SrcElemSize << " / " << DstElemSize
-        << std::endl;
-  CPOUT << std::endl;
-
   detail::SYCLMemObjI::MemObjType MemType = SYCLMemObj->getType();
   term_positions SrcPos, DstPos;
   prepTermPositions(SrcPos, DimSrc, MemType);
   prepTermPositions(DstPos, DimDst, MemType);
 
-  // int src_x_term_pos = 2, src_y_term_pos = 1, src_z_term_pos = 0;
-  // if(DimSrc == 2){
-  //   src_x_term_pos = 1; src_y_term_pos = 0; src_z_term_pos = 2;
-  // } else if (DimSrc == 1){
-  //   src_x_term_pos = 0; src_y_term_pos = 1; src_z_term_pos = 2;
-  // }
-  // int dst_x_term_pos = 2, dst_y_term_pos = 1, dst_z_term_pos = 0;
-  // if(DimDst == 2){
-  //   dst_x_term_pos = 1; dst_y_term_pos = 0; dst_z_term_pos = 2;
-  // } else if (DimDst == 1){
-  //   dst_x_term_pos = 0; dst_y_term_pos = 1; dst_z_term_pos = 2;
-  // }
+  //  If the Dimension is 1, and being called by ~SYCLMemObjT,
+  //  thenDstAccessRange[0] and DstSize[0] will already sized to bytes with
+  //  DstElemSize of 1.
   size_t DstXOffBytes = DstOffset[DstPos.x_term] * DstElemSize;
   size_t SrcXOffBytes = SrcOffset[SrcPos.x_term] * SrcElemSize;
   size_t DstARWidthBytes = DstAccessRange[DstPos.x_term] * DstElemSize;
@@ -395,44 +369,14 @@ void copyD2H(SYCLMemObjI *SYCLMemObj, RT::PiMem SrcMem, QueueImplPtr SrcQueue,
   const RT::PiQueue Queue = SrcQueue->getHandleRef();
   const detail::plugin &Plugin = SrcQueue->getPlugin();
 
-  // the offsets/ranges coming from accessor are always id<3>/range<3>
-  // But their organization varies by dimension:
-  //  1 ==>  {width, 1, 1}
-  //  2 ==>  {height, width, 1}
-  //  3 ==>  {depth, height, width}
-
-  CPOUT << "copyD2H " << std::endl;
-  CPOUT << "DimSrc: " << DimSrc << "  SrcAccessRange: " << SrcAccessRange[0]
-        << " / " << SrcAccessRange[1] << " / " << SrcAccessRange[2]
-        << std::endl;
-  // CPOUT << "        " << DimSrc << "  SrcOffset: " << SrcOffset[0] << " / "
-  // << SrcOffset[1] << " / " << SrcOffset[2] << std::endl;
-  CPOUT << "DimDst: " << DimDst << "  DstAccessRange: " << DstAccessRange[0]
-        << " / " << DstAccessRange[1] << " / " << DstAccessRange[2]
-        << std::endl;
-  // CPOUT << "        " << DimSrc << "  DstOffset: " << DstOffset[0] << " / "
-  // << DstOffset[1] << " / " << DstOffset[2] << std::endl;
-  CPOUT << "SrcElemSize / DstElemSize: " << SrcElemSize << " / " << DstElemSize
-        << std::endl;
-  CPOUT << std::endl;
-
   detail::SYCLMemObjI::MemObjType MemType = SYCLMemObj->getType();
   term_positions SrcPos, DstPos;
   prepTermPositions(SrcPos, DimSrc, MemType);
   prepTermPositions(DstPos, DimDst, MemType);
 
-  // int src_x_term_pos = 2, src_y_term_pos = 1, src_z_term_pos = 0;
-  // if(DimSrc == 2){
-  //   src_x_term_pos = 1; src_y_term_pos = 0; src_z_term_pos = 2;
-  // } else if (DimSrc == 1){
-  //   src_x_term_pos = 0; src_y_term_pos = 1; src_z_term_pos = 2;
-  // }
-  // int dst_x_term_pos = 2, dst_y_term_pos = 1, dst_z_term_pos = 0;
-  // if(DimDst == 2){
-  //   dst_x_term_pos = 1; dst_y_term_pos = 0; dst_z_term_pos = 2;
-  // } else if (DimDst == 1){
-  //   dst_x_term_pos = 0; dst_y_term_pos = 1; dst_z_term_pos = 2;
-  // }
+  //  If the Dimension is 1, and being called by ~SYCLMemObjT,
+  //  thenDstAccessRange[0] and DstSize[0] will already sized to bytes with
+  //  DstElemSize of 1.
   size_t DstXOffBytes = DstOffset[DstPos.x_term] * DstElemSize;
   size_t SrcXOffBytes = SrcOffset[SrcPos.x_term] * SrcElemSize;
   // size_t DstARWidthBytes = DstAccessRange[DstPos.x_term] * DstElemSize;
@@ -498,44 +442,14 @@ void copyD2D(SYCLMemObjI *SYCLMemObj, RT::PiMem SrcMem, QueueImplPtr SrcQueue,
   const RT::PiQueue Queue = SrcQueue->getHandleRef();
   const detail::plugin &Plugin = SrcQueue->getPlugin();
 
-  // the offsets/ranges coming from accessor are always id<3>/range<3>
-  // But their organization varies by dimension:
-  //  1 ==>  {width, 1, 1}
-  //  2 ==>  {height, width, 1}
-  //  3 ==>  {depth, height, width}
-
-  CPOUT << "copyD2D " << std::endl;
-  CPOUT << "DimSrc: " << DimSrc << "  SrcAccessRange: " << SrcAccessRange[0]
-        << " / " << SrcAccessRange[1] << " / " << SrcAccessRange[2]
-        << std::endl;
-  // CPOUT << "        " << DimSrc << "  SrcOffset: " << SrcOffset[0] << " / "
-  // << SrcOffset[1] << " / " << SrcOffset[2] << std::endl;
-  CPOUT << "DimDst: " /* << DimDst << "  DstAccessRange: " << DstAccessRange[0]
-                         << " / " << DstAccessRange[1] << " / " <<
-                         DstAccessRange[2] */
-        << std::endl;
-  // CPOUT << "        " << DimSrc << "  DstOffset: " << DstOffset[0] << " / "
-  // << DstOffset[1] << " / " << DstOffset[2] << std::endl;
-  CPOUT << "SrcElemSize / DstElemSize: " << SrcElemSize << " / " << DstElemSize
-        << std::endl;
-  CPOUT << std::endl;
-
   detail::SYCLMemObjI::MemObjType MemType = SYCLMemObj->getType();
   term_positions SrcPos, DstPos;
   prepTermPositions(SrcPos, DimSrc, MemType);
   prepTermPositions(DstPos, DimDst, MemType);
-  // int src_x_term_pos = 2, src_y_term_pos = 1, src_z_term_pos = 0;
-  // if(DimSrc == 2){
-  //   src_x_term_pos = 1; src_y_term_pos = 0; src_z_term_pos = 2;
-  // } else if (DimSrc == 1){
-  //   src_x_term_pos = 0; src_y_term_pos = 1; src_z_term_pos = 2;
-  // }
-  // int dst_x_term_pos = 2, dst_y_term_pos = 1, dst_z_term_pos = 0;
-  // if(DimDst == 2){
-  //   dst_x_term_pos = 1; dst_y_term_pos = 0; dst_z_term_pos = 2;
-  // } else if (DimDst == 1){
-  //   dst_x_term_pos = 0; dst_y_term_pos = 1; dst_z_term_pos = 2;
-  // }
+
+  //  If the Dimension is 1, and being called by ~SYCLMemObjT,
+  //  thenDstAccessRange[0] and DstSize[0] will already sized to bytes with
+  //  DstElemSize of 1.
   size_t DstXOffBytes = DstOffset[DstPos.x_term] * DstElemSize;
   size_t SrcXOffBytes = SrcOffset[SrcPos.x_term] * SrcElemSize;
   // size_t DstARWidthBytes = DstAccessRange[DstPos.x_term] * DstElemSize;

sycl/test/plugins/enqueue-arg-order-buffer.cpp

@@ -2,20 +2,6 @@
 // RUN: env SYCL_PI_TRACE=2 %GPU_RUN_PLACEHOLDER %t.out %GPU_CHECK_PLACEHOLDER
 // RUN: env SYCL_PI_TRACE=2 %CPU_RUN_PLACEHOLDER %t.out %CPU_CHECK_PLACEHOLDER
 
-// TFAIL: *
-
-/*
-  Manual
-    clang++ -fsycl -o eaob.bin enqueue-arg-order-buffer.cpp
-    clang++ -fsycl -g -o eaob.d enqueue-arg-order-buffer.cpp
-    SYCL_PI_TRACE=2 ./eaob.bin
-
-    clang++ --driver-mode=g++ -fsycl -fsycl-targets=nvptx64-nvidia-cuda-sycldevice -o eaob.bin enqueue-arg-order-buffer.cpp
-    SYCL_PI_TRACE=2 SYCL_BE=PI_CUDA ./eaob.bin
-
-    llvm-lit --param SYCL_BE=PI_CUDA -v enqueue-arg-order-buffer.cpp
-*/
-
 #include <CL/sycl.hpp>
 #include <CL/sycl/accessor.hpp>
 #include <iostream>
@@ -60,10 +46,6 @@ void remind() {
   std::cout << "SlicePitch SHOULD be 0 or : " << width * sizeof(float) * height
             << std::endl
             << std::endl; // 0 or 320
-
-  // NOTE: presently we see 20/16/1 for Region and 20 for row pitch.  both
-  // incorrect.
-
 }
 // ----------- FUNCTIONAL
 template <template <int> class T>
@@ -72,31 +54,38 @@ static void printRangeId(T<3> arr) {
 }
 
 void testDetailConvertToArrayOfN(){
-  //ranges
+  // ranges, as used with buffers (args reverse order for images)
   range<1> range_1D(width);
   range<2> range_2D(height, width);
   range<3> range_3D(depth, height, width);
 
-  range<3> arr1 = sycl::detail::convertToArrayOfN<3,1>(range_1D); 
-  //should be: {1,1,16}
+  range<3> arr1 = sycl::detail::convertToArrayOfN<3,1>(range_1D);
+  // {16,1,1}
   printRangeId(arr1);
-  //assert(arr1[0] == 1 && arr1[1] == 1 && arr1[2] == width && "arr1 should be {1,1,16} ");
+  assert(arr1[0] == width && arr1[1] == 1 && arr1[2] == 1 &&
+         "arr1 expected as {16,1,1}");
 
   range<3> arr2 = sycl::detail::convertToArrayOfN<3,1>(range_2D);
-  //should be: {1, 5, 16}
+  //{5, 16, 1}
   printRangeId(arr2);
-  //assert(arr2[0] == 1 && arr2[1] == height && arr2[2] == width && "arr2 should be {1,5,16} ");
+  assert(arr2[0] == height && arr2[1] == width && arr2[2] == 1 &&
+         "arr2 expected as {5, 16, 1}");
 
   range<3> arr3 = sycl::detail::convertToArrayOfN<3,1>(range_3D);
-  //should be: {3, 5, 16}
+  //{3, 5, 16}
   printRangeId(arr3);
-  //assert(arr3[0] == depth && arr3[1] == height && arr3[2] == width && "arr3 should be {3,5,16} ");
+  assert(arr3[0] == depth && arr3[1] == height && arr3[2] == width &&
+         "arr3 expected as {3,5,16}");
 
   range<2> smaller2 = sycl::detail::convertToArrayOfN<2,1>(range_3D);
-  //assert(smaller2[0] == height && smaller2[1] == width  && "smaller2 should be {5,16} ");
+  //{3,5}
+  std::cout << "{" << smaller2[0] << "," << smaller2[1] << "}" << std::endl;
+  assert(smaller2[0] == depth && smaller2[1] == height &&
+         "smaller2 expected {3,5} ");
 
   range<1> smaller1 = sycl::detail::convertToArrayOfN<1,1>(range_3D);
-  //assert(smaller1[0] == width && "smaller1 should be {16} ");
+  //{3}
+  assert(smaller1[0] == depth && "smaller1 expected {3} ");
 }
 
 // class to give access to protected function getLinearIndex

sycl/test/plugins/enqueue-arg-order-image.cpp

@@ -2,18 +2,6 @@
 // RUN: env SYCL_PI_TRACE=2 %GPU_RUN_PLACEHOLDER %t.out %GPU_CHECK_PLACEHOLDER
 // RUN: env SYCL_PI_TRACE=2 %CPU_RUN_PLACEHOLDER %t.out %CPU_CHECK_PLACEHOLDER
 
-/*
-  Manual
-    clang++ -fsycl -o eaoi.bin enqueue-arg-order-image.cpp
-    clang++ -fsycl -g -o eaoi.d enqueue-arg-order-image.cpp
-    SYCL_PI_TRACE=2 ./eaoi.bin
-
-    clang++ --driver-mode=g++ -fsycl -fsycl-targets=nvptx64-nvidia-cuda-sycldevice -o eaoi.bin enqueue-arg-order-image.cpp
-    SYCL_PI_TRACE=2 SYCL_BE=PI_CUDA ./eaoi.bin
-
-    llvm-lit --param SYCL_BE=PI_CUDA -v enqueue-arg-order-image.cpp
-*/
-
 #include <CL/sycl.hpp>
 #include <CL/sycl/accessor.hpp>
 #include <iostream>
@@ -52,8 +40,6 @@ void remind() {
             << width * sizeof(sycl::float4) * height << std::endl
             << std::endl; // 0 or 1280
 
-  // NOTE: presently we see 5/16/1 for image Region and 80 for row pitch.  both
-  // incorrect
 }