From c10d16fa10f5931cdc1c9edff033f45ef0196847 Mon Sep 17 00:00:00 2001
From: foreman <dl.swbuild@amd.com>
Date: Fri, 20 Dec 2019 11:50:41 -0500
Subject: [PATCH] P4 to Git Change 2049201 by
 skudchad@skudchad_test2_win_opencl on 2019/12/20 11:44:13

	SWDEV-216708 - [hipclang-vdi-rocm][FBA-77]hipGetDeviceCount() query should not trigger any queue creation.
	- KFD queues get created when we call hsa_executable_load_agent_code_object when creating a blitProgram at init. Delay blit creation until when needed
	- Queue also gets created when we init a xferQueue. Delay that too until when needed.

	If we want to have a lite init, there are multiple changes needed in both OpenCL and HIP. Thats rather a feature to work on later.

	ReviewBoardURL = http://ocltc.amd.com/reviews/r/18416/diff/

Affected files ...

... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocblit.cpp#34 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.cpp#148 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocdevice.hpp#47 edit


[ROCm/clr commit: 1384e82e6017723cfa4164dbc16fa8711485df88]
---
 .../rocclr/runtime/device/rocm/rocblit.cpp    |  4 +-
 .../rocclr/runtime/device/rocm/rocdevice.cpp  | 49 ++++++++++---------
 .../rocclr/runtime/device/rocm/rocdevice.hpp  |  3 ++
 3 files changed, 32 insertions(+), 24 deletions(-)

diff --git a/projects/clr/rocclr/runtime/device/rocm/rocblit.cpp b/projects/clr/rocclr/runtime/device/rocm/rocblit.cpp
index 5704e05b2f..c06a90fbf2 100644
--- a/projects/clr/rocclr/runtime/device/rocm/rocblit.cpp
+++ b/projects/clr/rocclr/runtime/device/rocm/rocblit.cpp
@@ -733,7 +733,9 @@ bool KernelBlitManager::create(amd::Device& device) {
 
 bool KernelBlitManager::createProgram(Device& device) {
   if (device.blitProgram() == nullptr) {
-    return false;
+    if (!device.createBlitProgram()) {
+      return false;
+    }
   }
 
   std::vector<amd::Device*> devices;
diff --git a/projects/clr/rocclr/runtime/device/rocm/rocdevice.cpp b/projects/clr/rocclr/runtime/device/rocm/rocdevice.cpp
index 2cc53cd4b8..49eadce97b 100644
--- a/projects/clr/rocclr/runtime/device/rocm/rocdevice.cpp
+++ b/projects/clr/rocclr/runtime/device/rocm/rocdevice.cpp
@@ -656,18 +656,6 @@ bool Device::create(bool sramEccEnabled) {
     return false;
   }
 
-  const char* scheduler = nullptr;
-
-#if defined(USE_COMGR_LIBRARY)
-  std::string sch = SchedulerSourceCode;
-  if (settings().useLightning_) {
-    if (info().cooperativeGroups_) {
-      sch.append(GwsInitSourceCode);
-    }
-    scheduler = sch.c_str();
-  }
-#endif  // USE_COMGR_LIBRARY
-
   amd::Context::Info info = {0};
   std::vector<amd::Device*> devices;
   devices.push_back(this);
@@ -678,15 +666,6 @@ bool Device::create(bool sramEccEnabled) {
     return false;
   }
 
-  blitProgram_ = new BlitProgram(context_);
-  // Create blit programs
-  if (blitProgram_ == nullptr || !blitProgram_->create(this, scheduler)) {
-    delete blitProgram_;
-    blitProgram_ = nullptr;
-    LogError("Couldn't create blit kernels!");
-    return false;
-  }
-
   mapCacheOps_ = new amd::Monitor("Map Cache Lock", true);
   if (nullptr == mapCacheOps_) {
     return false;
@@ -757,8 +736,6 @@ bool Device::create(bool sramEccEnabled) {
     }
   }
 
-  xferQueue();
-
   return true;
 }
 
@@ -809,6 +786,32 @@ void Device::ReleaseExclusiveGpuAccess(VirtualGPU& vgpu) const {
   vgpusAccess().unlock();
 }
 
+bool Device::createBlitProgram() {
+  bool result = true;
+  const char* scheduler = nullptr;
+
+#if defined(USE_COMGR_LIBRARY)
+  std::string sch = SchedulerSourceCode;
+  if (settings().useLightning_) {
+    if (info().cooperativeGroups_) {
+      sch.append(GwsInitSourceCode);
+    }
+    scheduler = sch.c_str();
+  }
+#endif  // USE_COMGR_LIBRARY
+
+  blitProgram_ = new BlitProgram(context_);
+  // Create blit programs
+  if (blitProgram_ == nullptr || !blitProgram_->create(this, scheduler)) {
+    delete blitProgram_;
+    blitProgram_ = nullptr;
+    LogError("Couldn't create blit kernels!");
+    return false;
+  }
+
+  return result;
+}
+
 device::Program* Device::createProgram(amd::Program& owner, amd::option::Options* options) {
   device::Program* program;
   if (settings().useLightning_) {
diff --git a/projects/clr/rocclr/runtime/device/rocm/rocdevice.hpp b/projects/clr/rocclr/runtime/device/rocm/rocdevice.hpp
index b257e1cf6d..0e499845d4 100644
--- a/projects/clr/rocclr/runtime/device/rocm/rocdevice.hpp
+++ b/projects/clr/rocclr/runtime/device/rocm/rocdevice.hpp
@@ -380,6 +380,9 @@ class Device : public NullDevice {
 
   amd::Context& context() const { return *context_; }
 
+  //! Create internal blit program
+  bool createBlitProgram();
+
   // Returns AMD GPU Pro interfaces
   const IProDevice& iPro() const { return *pro_device_; }
   bool ProEna() const  { return pro_ena_; }