RCP 5.5 release

Build/Linux/Common.mk

@@ -58,7 +58,7 @@ COMMON_SRC = $(COMMON_DIR)/Src
 COMMON_LIB = $(COMMON_DIR)/Lib
 COMMON_LIB_AMD = $(COMMON_DIR)/Lib/AMD
 COMMON_LIB_EXT = $(COMMON_DIR)/Lib/Ext
-GPU_PERF_API_DIR = $(COMMON_LIB_AMD)/GPUPerfAPI/3_1/Include
+GPU_PERF_API_DIR = $(COMMON_LIB_AMD)/GPUPerfAPI/3_2/Include
 APPSDK_DIR = $(COMMON_LIB_AMD)/APPSDK/3-0
 CAL_DIR = $(COMMON_LIB_AMD)/CAL/8.95
 ACL_DIR = $(COMMON_LIB_AMD)/ACL/TOT
@@ -77,7 +77,6 @@ UTF8CPP_DIR = $(COMMON_LIB_EXT)/utf8cpp/source
 ADL_DIR = $(COMMON_LIB_AMD)/ADL
 ADLUTIL_DIR = $(COMMON_SRC)/ADLUtil
 COMMON_PROJ_DIR = $(COMMON_DIR)/Src
-CELF_DIR = $(COMMON_SRC)/CElf
 HSAUTILS_DIR = $(COMMON_SRC)/HSAUtils
 
 GPU_PROFILER_LIB_PREFIX=RCP

Build/Linux/CommonTargets.mk

@@ -94,9 +94,6 @@ $(OBJ_DIR)/%.o: $(SRC_SPROFILE_DIR)/%.cpp
 $(OBJ_DIR)/%.o: $(SRC_SANALYZE_DIR)/%.cpp
 	$(BUILD_SRC)
 
-$(OBJ_DIR)/%.o: $(CELF_DIR)/Src/%.cpp
-	$(BUILD_SRC)
-
 $(OBJ_DIR)/%.o: $(HSATESTCOMMON_DIR)/%.cpp
 	$(BUILD_SRC)
 

Build/Linux/build_rcp.sh

@@ -161,7 +161,7 @@ HSAFDNTRACE="$SRCDIR/HSAFdnTrace"
 PRELOADXINITTHREADS="$SRCDIR/PreloadXInitThreads"
 ACTIVITYLOGGER="CXLActivityLogger"
 ACTIVITYLOGGERDIR="$COMMONSRC/AMDTActivityLogger/"
-GPA="$COMMON/Lib/AMD/GPUPerfAPI/3_1"
+GPA="$COMMON/Lib/AMD/GPUPerfAPI/3_2"
 VKSTABLECLOCKS="$COMMON/Lib/AMD/VKStableClocks/VKStableClocks/VkStableClocks"
 JQPLOT_PATH="$SRCCOMMON/jqPlot"
 PROFILEDATAPARSERSRC="$SRCDIR/ProfileDataParser"

Build/VS2015/RCP-GPUPerfAPI.props

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="utf-8"?>
 <Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
   <PropertyGroup Label="UserMacros">
-    <GPUPerfAPIDir>$(CommonDir)\Lib\AMD\GPUPerfAPI\3_1</GPUPerfAPIDir>
+    <GPUPerfAPIDir>$(CommonDir)\Lib\AMD\GPUPerfAPI\3_2</GPUPerfAPIDir>
   </PropertyGroup>
   <PropertyGroup>
     <_ProjectFileVersion>10.0.30319.1</_ProjectFileVersion>

Build/VS2015/RCPCLProfileAgent.vcxproj

@@ -40,10 +40,8 @@
     <Import Project="RCP-VKStableClocks.props" />
     <Import Project="..\..\..\RCP-Internal\Build\VS2015\RCP-Internal.props" Condition="exists('..\..\..\RCP-Internal\Build\VS2015\RCP-Internal.props')" />
     <Import Project="$(CommonDir)\Lib\AMD\ACL\Global-ACL.props" />
-    <Import Project="$(CommonDir)\Src\CElf\Global-CElf.props" />
     <Import Project="$(CommonDir)\Src\ADLUtil\Global-ADLUtil.props" />
     <Import Project="$(CommonDir)\Src\ACLModuleManager\Global-ACLModuleManager.props" />
-    <Import Project="$(CommonDir)\Lib\AMD\CAL\Global-CAL-NoLib.props" />
   </ImportGroup>
   <PropertyGroup Label="UserMacros" />
   <PropertyGroup>
@@ -63,10 +61,6 @@
   </ItemDefinitionGroup>
   <ItemGroup>
     <ClCompile Include="..\..\..\Common\Src\ADLUtil\ADLUtil.cpp" />
-    <ClCompile Include="..\..\..\Common\Src\CElf\Src\CElf.cpp" />
-    <ClCompile Include="..\..\..\Common\Src\CElf\Src\CElfSection.cpp" />
-    <ClCompile Include="..\..\..\Common\Src\CElf\Src\CElfStringTable.cpp" />
-    <ClCompile Include="..\..\..\Common\Src\CElf\Src\CElfSymbolTable.cpp" />
     <ClCompile Include="..\..\..\Common\Src\DynamicLibraryModule\ACLModule.cpp" />
     <ClCompile Include="..\..\..\Common\Src\ACLModuleManager\ACLModuleManager.cpp" />
     <ClCompile Include="..\..\Src\CLProfileAgent\CLBuffer.cpp" />

Build/VS2015/RCPCLProfileAgent.vcxproj.filters

@@ -10,18 +10,6 @@
     <ClCompile Include="..\..\..\Common\Src\ADLUtil\ADLUtil.cpp">
       <Filter>ADLUtil</Filter>
     </ClCompile>
-    <ClCompile Include="..\..\..\Common\Src\CElf\Src\CElf.cpp">
-      <Filter>CElf</Filter>
-    </ClCompile>
-    <ClCompile Include="..\..\..\Common\Src\CElf\Src\CElfSection.cpp">
-      <Filter>CElf</Filter>
-    </ClCompile>
-    <ClCompile Include="..\..\..\Common\Src\CElf\Src\CElfStringTable.cpp">
-      <Filter>CElf</Filter>
-    </ClCompile>
-    <ClCompile Include="..\..\..\Common\Src\CElf\Src\CElfSymbolTable.cpp">
-      <Filter>CElf</Filter>
-    </ClCompile>
     <ClCompile Include="..\..\Src\CLProfileAgent\CLBuffer.cpp">
       <Filter>Source Files</Filter>
     </ClCompile>
@@ -98,9 +86,6 @@
     <Filter Include="Resource Files">
       <UniqueIdentifier>{7124187b-7b76-4cc9-bdd2-0ff0af4cbbdc}</UniqueIdentifier>
     </Filter>
-    <Filter Include="CElf">
-      <UniqueIdentifier>{59c4443f-a055-4755-8763-238265e9bc85}</UniqueIdentifier>
-    </Filter>
     <Filter Include="Header Files">
       <UniqueIdentifier>{22d938b8-1adb-4011-ac9a-850e37321634}</UniqueIdentifier>
     </Filter>

README.md

@@ -32,20 +32,13 @@ RCP was formerly delivered as part of CodeXL with the executable name
 * When used with CodeXL, all profiler data can be visualized in a user-friendly graphical user interface.
 
 ## What's New
-* Version 5.4 (6/22/18)
-  * Adds support for additional GPUs and APUs.
-  * Support for profiling OpenCL applications running on ROCm
-  * OpenCL: Support for tracing OpenCL 2.1 and 2.2 APIs
-  * ROCm/HSA: Support for ROCm 1.8.
-  * ROCm/HSA: Support for tracing AMD vendor extensions.
-  * Fixes an issue parsing occupancy data collected on systems with certain locale settings.
-  * ROCm/HSA: Fixes an issue with garbage characters in the .atp file for some HSA API string parameters.
-  * OpenCL: Fixes profiling on recent amdgpu-pro drivers using the legacy OpenCL stack.
-  * OpenCL: Works around a driver issue where GPU clock frequencies remain fixed after profiling on GFX9-based GPUs.
+* Version 5.5 (8/22/18)
+ * Adds support for additional GPUs and APUs.
+ * ROCm/HSA: Fixes several issues with incorrect or missing data transfer timestamps.
 
 ## System Requirements
 * An AMD Radeon GCN-based GPU or APU
-* Radeon Software Crimson ReLive Edition 18.5.1 or later (Driver Packaging Version 18.10 or later).
+* Radeon Software Adrenaline Edition 18.8.1 or later (Driver Packaging Version 18.30 or later).
 * ROCm 1.8. See system requirements for ROCm: https://rocm.github.io/install.html and https://rocm.github.io/hardware.html
 * Windows 7, 8.1, and 10
   * For Windows, the `Visual C++ Redistributable for Visual Studio 2015` is required. It can be downloaded from https://www.microsoft.com/en-us/download/details.aspx?id=48145

ReleaseNotes.md

@@ -1,5 +1,9 @@
 # Radeon Compute Profiler Release Notes
 ---
+## Version 5.5 (8/22/18)
+ * Adds support for additional GPUs and APUs.
+ * ROCm/HSA: Fixes several issues with incorrect or missing data transfer timestamps.
+
 ## Version 5.4 (6/22/18)
  * Adds support for additional GPUs and APUs.
  * Support for profiling OpenCL applications running on ROCm

Scripts/UpdateCommonMap.py

@@ -9,7 +9,6 @@
     "common-lib-amd-ADL"                     : "../Common/Lib/AMD/ADL",
     "common-lib-AMD-ACL"                     : "../Common/Lib/AMD/ACL",
     "common-lib-amd-APPSDK-3.0"              : "../Common/Lib/AMD/APPSDK",
-    "common-lib-AMD-CAL-8.95"                : "../Common/Lib/AMD/CAL",
     "common-lib-ext-Boost-1.59"              : "../Common/Lib/Ext/Boost",
     "common-lib-ext-mhook"                   : "../Common/Lib/Ext/mhook",
     "common-lib-ext-tinyxml-2.6.2"           : "../Common/Lib/Ext/tinyxml",
@@ -21,7 +20,6 @@
     "common-src-AMDTBaseTools"               : "../Common/Src/AMDTBaseTools",
     "common-src-AMDTInterceptor"             : "../Common/Src/AMDTInterceptor",
     "common-src-AMDTOSWrappers"              : "../Common/Src/AMDTOSWrappers",
-    "common-src-CElf"                        : "../Common/Src/CElf",
     "common-src-DeviceInfo"                  : "../Common/Src/DeviceInfo",
     "common-src-DynamicLibraryModule"        : "../Common/Src/DynamicLibraryModule",
     "common-src-HSAUtils"                    : "../Common/Src/HSAUtils",
@@ -33,12 +31,12 @@
 }
 
 downloadMappingWin = {
-    "https://github.com/GPUOpen-Tools/GPA/releases/download/v3.1a/GPUPerfAPI-3.1.571.0.zip" : "../Common/Lib/AMD/GPUPerfAPI",
+    "https://github.com/GPUOpen-Tools/GPA/releases/download/v3.2/GPUPerfAPI-3.2.623.0.zip" : "../Common/Lib/AMD/GPUPerfAPI",
     "https://github.com/GPUOpen-Tools/RCP/releases/download/v5.4/RadeonComputeProfiler-v5.4.6937.zip" : "../Common/Lib/AMD/VKStableClocks"
 }
 
 downloadMappingLin = {
-    "https://github.com/GPUOpen-Tools/GPA/releases/download/v3.1a/GPUPerfAPI.3.1.793-lnx.tgz" : "../Common/Lib/AMD/GPUPerfAPI",
+    "https://github.com/GPUOpen-Tools/GPA/releases/download/v3.2/GPUPerfAPI.3.2.858-lnx.tgz" : "../Common/Lib/AMD/GPUPerfAPI",
     "https://github.com/GPUOpen-Tools/RCP/releases/download/v5.4/RadeonComputeProfiler-v5.4.6906.tgz" : "../Common/Lib/AMD/VKStableClocks"
 }
 

Src/CLProfileAgent/CLKernelAssembly.cpp

@@ -8,8 +8,8 @@
 
 #include <sstream>
 #include <algorithm>
+#include <fstream>
 #include <boost/algorithm/string.hpp>
-#include <CElf.h>
 
 // ADL headers
 #include <ADLUtil.h>
@@ -43,8 +43,7 @@ KernelAssembly::KernelAssembly() : m_strFilePrefix(KERNEL_ASSEMBLY_FILE_PREFIX),
     m_bOutputIL(false),
     m_bOutputISA(false),
     m_bOutputCL(false),
-    m_bOutputHSAIL(false),
-    m_bInitCAL(false)
+    m_bOutputHSAIL(false)
 {
 };
 
@@ -196,25 +195,6 @@ bool SaveBifToFile(const char* pszFileName, ACLModule* mod, aclBinary* pBin)
     return true;
 }
 
-bool GetCPUISA(CElf& elf, string& strISAOut)
-{
-    const CElfSection* pAstextSection = elf.GetSection(".astext");
-
-    if (pAstextSection == NULL)
-    {
-        // ERROR
-        Log(logERROR, "Failed to retrieve astext section.\n");
-        return false;
-    }
-    else
-    {
-        const vector<char> data(pAstextSection->GetData());
-        strISAOut = string(data.begin(), data.end());
-    }
-
-    return true;
-}
-
 bool KernelAssembly::GenerateKernelFilesFromACLModule(ACLModule*         pAclModule,
                                                       aclCompiler*       pAclCompiler,
                                                       std::vector<char>& vBinary,
@@ -297,16 +277,6 @@ bool KernelAssembly::GenerateKernelFilesFromACLModule(ACLModule*         pAclMod
                 strISA = pAsText;
                 bISA = true;
             }
-            else
-            {
-                // Fallback to CELF path
-                CElf elf(vBinary);
-
-                if (elf.good())
-                {
-                    bISA = GetCPUISA(elf, strISA);
-                }
-            }
 
             if (bISA)
             {

Src/CLProfileAgent/CLKernelAssembly.h

@@ -181,7 +181,6 @@ class KernelAssembly
     bool                                       m_bOutputISA;                 ///< flag indicating whether or not to write out the ISA file
     bool                                       m_bOutputCL;                  ///< flag indicating whether or not to write out the CL file
     bool                                       m_bOutputHSAIL;               ///< flag indicating whether or not to write out the HSAIL file
-    bool                                       m_bInitCAL;                   ///< flag indicating whether or not CALRT is initialized
     static std::string                         m_sTmpDisassemblyLoggerISA;   ///< string to hold the ISA text contents from the disassembly logger
     static std::string                         m_sTmpDisassemblyLoggerHSAIL; ///< string to hold the HSAIL text contents from the disassembly logger
     static unsigned int                        m_sDisassembleCount;          ///< count of the number of unique text items from the disassembly logger

Src/CLProfileAgent/makefile

@@ -17,7 +17,6 @@ INCLUDES = \
 	-I$(DEVICEINFO_DIR) \
 	-I$(TSINGLETON_DIR) \
 	-isystem$(BOOST_DIR) \
-	-isystem$(CELF_DIR)/Include \
 	-I$(UTF8CPP_DIR) \
 	-I$(ACL_DIR)/include \
 	-I$(ADL_DIR)/include \
@@ -46,10 +45,6 @@ SO_OBJS =  \
 	./$(OBJ_DIR)/CLDeferredKernel.o \
 	./$(OBJ_DIR)/CLProfilerMineCLEntry.o \
 	./$(OBJ_DIR)/CLProfilerMineCLMemory.o \
-	./$(OBJ_DIR)/CElf.o \
-	./$(OBJ_DIR)/CElfSection.o \
-	./$(OBJ_DIR)/CElfStringTable.o \
-	./$(OBJ_DIR)/CElfSymbolTable.o \
 	./$(OBJ_DIR)/GPUPerfAPICounterLoader.o \
 
 include $(DEPTH)/Build/Linux/CommonTargets.mk

Src/Common/Version.h

@@ -23,7 +23,7 @@
 // @{
 
 #define RCP_MAJOR_VERSION 5
-#define RCP_MINOR_VERSION 4
+#define RCP_MINOR_VERSION 5
 #define RCP_BUILD_NUMBER 0
 #define RCP_UPDATE_VERSION 0
 

Src/HSAFdnCommon/makefile

@@ -23,7 +23,6 @@ LIB_OBJS =  \
 	./$(OBJ_DIR)/HSAToolsRTModule.o \
 	./$(OBJ_DIR)/HSAAgentUtils.o \
 	./$(OBJ_DIR)/HSAAgentIterateReplacer.o \
-	./$(OBJ_DIR)/HSAAPITableVersions.o \
 	./$(OBJ_DIR)/HSAKernelDemangler.o \
 
 include $(DEPTH)/Build/Linux/CommonTargets.mk

Src/HSAFdnPMC/HSAPMCAgent.cpp

@@ -14,9 +14,6 @@
 #include "FileUtils.h"
 #include "GlobalSettings.h"
 
-#include "HSAPMCInterceptionTable1_0.h"
-#include "HSAPMCInterceptionTable1_2.h"
-#include "HSAAPITableVersions.h"
 #include "HSAAgentUtils.h"
 
 #include "HSAGPAProfiler.h"
@@ -52,26 +49,9 @@ extern "C" bool DLL_PUBLIC OnLoad(void* pTable, uint64_t runtimeVersion, uint64_
 
     std::cout << RCP_PRODUCT_NAME " " << RCP_VERSION_STRING << " is enabled\n";
 
-    if (ROCM_1_1_X_AND_EARLIER_ROOT_RUNTIME_VERSION == runtimeVersion)
-    {
-        // ROCm versions 1.1.1 and earlier
-        InitHSAAPIInterceptPMC1_0(reinterpret_cast<ApiTable1_0*>(pTable));
-    }
-    else
-    {
-        HsaApiTable* pHsaTable = reinterpret_cast<HsaApiTable*>(pTable);
+    HsaApiTable* pHsaTable = reinterpret_cast<HsaApiTable*>(pTable);
 
-        if (IsROCm12(pHsaTable))
-        {
-            // ROCm 1.2 backwards compatibility
-            HsaApiTable1_2* pHsaTable1_2 = reinterpret_cast<HsaApiTable1_2*>(pTable);
-            InitHSAAPIInterceptPMC1_2(pHsaTable1_2);
-        }
-        else
-        {
-            InitHSAAPIInterceptPMC(pHsaTable);
-        }
-    }
+    InitHSAAPIInterceptPMC(pHsaTable);
 
     Parameters params;
     FileUtils::GetParametersFromFile(params);

Src/HSAFdnPMC/makefile

@@ -33,8 +33,6 @@ LIBPATH = $(COMMON_LIB_PATHS)
 SO_OBJS = \
 	./$(OBJ_DIR)/HSAPMCAgent.o \
 	./$(OBJ_DIR)/HSAPMCInterceptionHelpers.o \
-	./$(OBJ_DIR)/HSAPMCInterceptionTable1_0.o \
-	./$(OBJ_DIR)/HSAPMCInterceptionTable1_2.o \
 	./$(OBJ_DIR)/HSAGPAProfiler.o \
 	./$(OBJ_DIR)/HSAPMCInterception.o \
 	./$(OBJ_DIR)/CLCUInfoBase.o \

Src/HSAFdnTrace/AutoGenerated/HSAAMDExtensionAPITraceClasses.cpp

@@ -833,7 +833,8 @@ void HSA_APITrace_hsa_amd_memory_async_copy::Create(
     uint32_t num_dep_signals,
     const hsa_signal_t* dep_signals,
     hsa_signal_t completion_signal,
-    hsa_status_t retVal)
+    hsa_status_t retVal,
+    ULONGLONG asyncCopyIdentifier)
 {
     m_ullStart = ullStartTime;
     m_ullEnd = ullEndTime;
@@ -853,6 +854,19 @@ void HSA_APITrace_hsa_amd_memory_async_copy::Create(
 
     m_completion_signal = completion_signal;
     m_retVal = retVal;
+
+    m_asyncCopyIdentifier = asyncCopyIdentifier;
+}
+
+bool HSA_APITrace_hsa_amd_memory_async_copy::WriteTimestampEntry(std::ostream& sout, bool bTimeout)
+{
+    if (HSAAPIBase::WriteTimestampEntry(sout, bTimeout))
+    {
+        // async copy identifier
+        sout << std::left << std::setw(21) << m_asyncCopyIdentifier;
+    }
+
+    return true;
 }
 
 ///////////////////////////////////////////////////

Src/HSAFdnTrace/AutoGenerated/HSAAMDExtensionAPITraceClasses.h

@@ -840,6 +840,7 @@ class HSA_APITrace_hsa_amd_memory_async_copy : public HSAAPIBase
     /// \param dep_signals Parameter passed to hsa_amd_memory_async_copy
     /// \param completion_signal Parameter passed to hsa_amd_memory_async_copy
     /// \param retVal the return value for hsa_amd_memory_async_copy
+    /// \param asyncCopyIdentifier Parameter passed to hsa_amd_memory_async_copy
     void Create(ULONGLONG ullStartTime,
                 ULONGLONG ullEndTime,
                 void* dst,
@@ -850,7 +851,14 @@ class HSA_APITrace_hsa_amd_memory_async_copy : public HSAAPIBase
                 uint32_t num_dep_signals,
                 const hsa_signal_t* dep_signals,
                 hsa_signal_t completion_signal,
-                hsa_status_t retVal);
+                hsa_status_t retVal,
+                ULONGLONG asyncCopyIdentifier);
+
+    /// Write timestamp entry
+    /// \param sout output stream
+    /// \param bTimeout a flag indicating output mode
+    /// \return True if timestamps are ready
+    bool WriteTimestampEntry(std::ostream& sout, bool bTimeout);
 
 private:
     /// Disabled copy constructor
@@ -871,6 +879,7 @@ class HSA_APITrace_hsa_amd_memory_async_copy : public HSAAPIBase
     hsa_signal_t m_dep_signalsVal; ///< Member to hold value passed to hsa_amd_memory_async_copy in dep_signals parameter
     hsa_signal_t m_completion_signal; ///< Parameter passed to hsa_amd_memory_async_copy
     hsa_status_t m_retVal; ///< Parameter passed to hsa_amd_memory_async_copy
+    ULONGLONG m_asyncCopyIdentifier; ///< async copy identifier
 };
 
 ///////////////////////////////////////////////////

Src/HSAFdnTrace/AutoGenerated/HSATraceInterception.cpp

@@ -3758,7 +3758,9 @@ hsa_status_t HSA_API_Trace_hsa_amd_memory_pool_free(void* ptr)
 hsa_status_t HSA_API_Trace_hsa_amd_memory_async_copy(void* dst, hsa_agent_t dst_agent, const void* src, hsa_agent_t src_agent, size_t size, uint32_t num_dep_signals, const hsa_signal_t* dep_signals, hsa_signal_t completion_signal)
 {
     hsa_signal_t origSignal = completion_signal;
-    HSA_APITrace_hsa_amd_memory_async_copy_PreCallHelper(dst, dst_agent, src, src_agent, size, num_dep_signals, dep_signals, completion_signal);
+    ULONGLONG asyncCopyIdentifier = OSUtils::Instance()->GetTimeNanos();
+    HSA_APITrace_hsa_amd_memory_async_copy_PreCallHelper(dst, dst_agent, src, src_agent, size, num_dep_signals, dep_signals, completion_signal, asyncCopyIdentifier);
+
     ULONGLONG ullStart = OSUtils::Instance()->GetTimeNanos();
     hsa_status_t retVal = g_pRealAmdExtFunctions->hsa_amd_memory_async_copy_fn(dst, dst_agent, src, src_agent, size, num_dep_signals, dep_signals, completion_signal);
     ULONGLONG ullEnd = OSUtils::Instance()->GetTimeNanos();
@@ -3777,7 +3779,8 @@ hsa_status_t HSA_API_Trace_hsa_amd_memory_async_copy(void* dst, hsa_agent_t dst_
         num_dep_signals,
         dep_signals,
         origSignal,
-        retVal);
+        retVal,
+        asyncCopyIdentifier);
 
     RECORD_STACK_TRACE_FOR_API(pAPIInfo);
     HSAAPIInfoManager::Instance()->AddAPIInfoEntry(pAPIInfo);