Add support of variant

.github/workflows/main.yml

@@ -26,6 +26,7 @@ jobs:
         git branch -a
         sudo apt-get update --yes
         sudo apt-get install --yes git clang-format-14
+        sudo apt-get install python3-mako --yes
 
     - name: Run lint script
       run: ./scripts/clangformat.sh
@@ -51,6 +52,10 @@ jobs:
           version: 1.10.0
       - uses: ilammy/msvc-dev-cmd@v1
       - uses: ilammy/setup-nasm@v1
+      - name: Install dependency
+        run: |
+          C:/hostedtoolcache/windows/Python/3.11.6/x64/python3.exe -m pip install mako
+          C:/hostedtoolcache/windows/Python/3.12.0/x64/python3.exe -m pip install mako
       - name: Build the UI
         run: |
           mkdir build && cd build && cmake -DCMAKE_BUILD_TYPE=Debug -G "Visual Studio 16 2019" -A x64 ..
@@ -77,6 +82,10 @@ jobs:
           version: 1.10.0
       - uses: ilammy/msvc-dev-cmd@v1
       - uses: ilammy/setup-nasm@v1
+      - name: Install dependency
+        run: |
+          C:/hostedtoolcache/windows/Python/3.11.6/x64/python3.exe -m pip install mako
+          C:/hostedtoolcache/windows/Python/3.12.0/x64/python3.exe -m pip install mako
       - name: Build the UI
         run: |
           mkdir build && cd build && cmake -DCMAKE_BUILD_TYPE=Release -G "Visual Studio 16 2019" -A x64 ..
@@ -96,6 +105,8 @@ jobs:
         run: |
           sudo apt-get update --yes
           sudo apt-get install --yes cmake gcc-10  clang-14 libsystemd-dev libbsd-dev ninja-build
+          /opt/hostedtoolcache/Python/3.11.6/x64/bin/pip install mako
+          /opt/hostedtoolcache/Python/3.12.0/x64/bin/pip install mako
           which gcc-10
           which clang-14
       - name: Prepare Vulkan SDK
@@ -120,6 +131,8 @@ jobs:
         run: |
           sudo apt-get update --yes
           sudo apt-get install --yes cmake gcc-10 clang-14 libsystemd-dev libbsd-dev ninja-build
+          /opt/hostedtoolcache/Python/3.11.6/x64/bin/pip install mako
+          /opt/hostedtoolcache/Python/3.12.0/x64/bin/pip install mako
           which gcc-10
           which clang-14
       - name: Prepare Vulkan SDK
@@ -142,6 +155,7 @@ jobs:
         run: |
           sudo apt-get update --yes
           sudo apt-get install --yes cmake ninja-build
+          sudo apt-get install python3-mako --yes
       - uses: nttld/setup-ndk@v1
         id: setup-ndk
         with:

dive_core/CMakeLists.txt

@@ -29,6 +29,9 @@ else()
   set(NATVIS_FILES "")
 endif()
 
+# Disable following warnings for compiling .c files in mesa 
+set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wno-override-init -Wno-sign-compare")
+
 set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/lib)
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
 
@@ -37,6 +40,8 @@ set(PM4_INFO_FILE "${DIVE_CORE_GENERATED_FILE_DIR}/pm4_info")
 set(PM4_GENERATED_SRC_FILE "${DIVE_CORE_GENERATED_FILE_DIR}/pm4_info.cpp")
 set(PM4_GENERATED_HDR_FILE "${DIVE_CORE_GENERATED_FILE_DIR}/pm4_info.h")
 set(ADRENO_HDR_FILE "${DIVE_CORE_GENERATED_FILE_DIR}/adreno.h")
+set(A6XX_PY_FILE "${DIVE_CORE_GENERATED_FILE_DIR}/a6xx.py")
+set(FREEDRENO_DEVICES_HDR_FILE "${DIVE_CORE_GENERATED_FILE_DIR}/freedreno_devices.h")
 
 if(${CMAKE_VERSION} VERSION_LESS "3.12.0")
   find_program (Python3_EXECUTABLE python3)
@@ -72,6 +77,33 @@ add_custom_command(
 )
 add_custom_target(adreno_header ALL DEPENDS ${ADRENO_HDR_FILE})
 
+add_custom_command(
+    OUTPUT ${A6XX_PY_FILE}
+    COMMAND ${Python3_EXECUTABLE} ${THIRDPARTY_DIRECTORY}/mesa/src/freedreno/registers/gen_header.py
+                   "--rnn"
+                   "${THIRDPARTY_DIRECTORY}/mesa/src/freedreno/registers"
+                   "--xml"
+                   "${THIRDPARTY_DIRECTORY}/mesa/src/freedreno/registers/adreno/a6xx.xml"
+                   "py-defines"
+                   > ${A6XX_PY_FILE}
+    DEPENDS "${THIRDPARTY_DIRECTORY}/mesa/src/freedreno/registers/gen_header.py"
+    WORKING_DIRECTORY ${THIRDPARTY_DIRECTORY}/mesa/src/freedreno/registers
+    VERBATIM
+)
+add_custom_target(a6xx_py ALL DEPENDS ${A6XX_PY_FILE})
+
+add_custom_command(
+    OUTPUT ${FREEDRENO_DEVICES_HDR_FILE}
+    COMMAND ${Python3_EXECUTABLE} ${THIRDPARTY_DIRECTORY}/mesa/src/freedreno/common/freedreno_devices.py
+                   "-p"
+                   "${DIVE_CORE_GENERATED_FILE_DIR}"
+                   > ${FREEDRENO_DEVICES_HDR_FILE}
+    DEPENDS "${THIRDPARTY_DIRECTORY}/mesa/src/freedreno/common/freedreno_devices.py"
+            "${A6XX_PY_FILE}"
+    VERBATIM
+)
+add_custom_target(freedreno_devices_header ALL DEPENDS ${FREEDRENO_DEVICES_HDR_FILE})
+
 file(GLOB_RECURSE HDR_FILES "${CMAKE_CURRENT_SOURCE_DIR}/*.h")
 file(GLOB_RECURSE SRC_FILES "${CMAKE_CURRENT_SOURCE_DIR}/*.cpp")
 list(FILTER SRC_FILES EXCLUDE REGEX "_tests.cpp$")
@@ -85,20 +117,27 @@ endif()
 file(GLOB_RECURSE COMMON_HDR_FILES "${CMAKE_SOURCE_DIR}/dive_core/common/*.h")
 file(GLOB_RECURSE COMMON_SRC_FILES "${CMAKE_SOURCE_DIR}/dive_core/common/*.cpp")
 
+file(GLOB_RECURSE FREEDRENO_SRC_FILES "${THIRDPARTY_DIRECTORY}/mesa/src/freedreno/common/freedreno_dev_info.c")
+
 include_directories(${THIRDPARTY_DIRECTORY}/Vulkan-Headers/include ${CMAKE_SOURCE_DIR} ${DIVE_CORE_GENERATED_FILE_DIR})
 
 include_directories(${CMAKE_SOURCE_DIR}/capture_layer)
 include_directories(${CMAKE_SOURCE_DIR}/capture_layer/generated)
 
+include_directories(${THIRDPARTY_DIRECTORY}/mesa/src/freedreno/common)
+include_directories(${THIRDPARTY_DIRECTORY}/mesa/src/)
+
 
 add_definitions(-DLITTLEENDIAN_CPU)     # Required for some PAL files
 
 add_library(${PROJECT_NAME} STATIC ${HDR_FILES} ${SRC_FILES}
-  ${COMMON_HDR_FILES} ${COMMON_SRC_FILES} ${PM4_GENERATED_SRC_FILE} ${NATVIS_FILES}
+  ${COMMON_HDR_FILES} ${COMMON_SRC_FILES} ${PM4_GENERATED_SRC_FILE} ${NATVIS_FILES} ${FREEDRENO_SRC_FILES}
   "${CMAKE_SOURCE_DIR}/capture_layer/generated/command_printer.cpp" "${CMAKE_SOURCE_DIR}/capture_layer/command_printer_brief.cpp"
   "${CMAKE_SOURCE_DIR}/capture_layer/generated/command_decoder.cpp" )
 add_dependencies(${PROJECT_NAME} pm4_info)
 add_dependencies(${PROJECT_NAME} adreno_header)
+add_dependencies(${PROJECT_NAME} a6xx_py)
+add_dependencies(${PROJECT_NAME} freedreno_devices_header)
 
 
 if("${CMAKE_SYSTEM_NAME}" STREQUAL "Linux")

dive_core/capture_data.cpp

@@ -22,6 +22,8 @@
 #include "capture_layer/generated/command_decoder.h"
 #include "capture_layer/generated/command_printer.h"
 #include "dive_core/command_hierarchy.h"
+#include "freedreno_dev_info.h"
+#include "pm4_info.h"
 
 namespace Dive
 {
@@ -946,18 +948,30 @@ CaptureData::LoadResult CaptureData::LoadAdrenoRdFile(std::istream &capture_file
         case RD_NONE:
         case RD_TEST:
         case RD_CMD:
-        // case RD_GPUADDR:
         case RD_CONTEXT:
         case RD_CMDSTREAM:
-        // case RD_CMDSTREAM_ADDR:
         case RD_PARAM:
         case RD_FLUSH:
         case RD_PROGRAM:
         case RD_VERT_SHADER:
-        case RD_FRAG_SHADER:
-        // case RD_BUFFER_CONTENTS:
+        case RD_FRAG_SHADER: capture_file.seekg(block_info.m_data_size, std::ios::cur); break;
         case RD_GPU_ID:
-        case RD_CHIP_ID: capture_file.seekg(block_info.m_data_size, std::ios::cur);
+        {
+            DIVE_ASSERT(block_info.m_data_size == 4);
+            uint32_t gpu_id = 0;
+            capture_file.read(reinterpret_cast<char *>(&gpu_id), block_info.m_data_size);
+            SetGPUID(gpu_id / 100);
+        }
+        break;
+        case RD_CHIP_ID:
+        {
+            DIVE_ASSERT(block_info.m_data_size == 8);
+            fd_dev_id dev_id;
+            capture_file.read(reinterpret_cast<char *>(&dev_id.chip_id), block_info.m_data_size);
+            auto info = fd_dev_info(&dev_id);
+            SetGPUID(info->chip);
+        }
+        break;
         }
     }
     m_memory.Finalize(true, true);

dive_core/common.py

@@ -1,4 +1,72 @@
 # Common functions
+import ctypes
+import re
+c_uint8 = ctypes.c_uint8
+
+class BitfieldsBits(ctypes.LittleEndianStructure):
+    _fields_ = [
+            ("A2XX", c_uint8, 1),
+            ("A3XX", c_uint8, 1),
+            ("A4XX", c_uint8, 1),
+            ("A5XX", c_uint8, 1),
+            ("A6XX", c_uint8, 1),
+            ("A7XX", c_uint8, 1),
+        ]
+
+class Bitfields(ctypes.Union):
+    _fields_ = [("bits", BitfieldsBits),
+                ("asbyte", c_uint8)]
+
+# convert "variants" string to bitfield 
+def GetGPUVariantsBitField(str):
+    pattern = r"A\d{1}XX"
+    gpu_list = re.findall(pattern, str)
+    bitfields = Bitfields()
+    bitfields.asbyte = 0
+    a_count = len(gpu_list)
+    if a_count == 0:
+        return 0
+    dash_count = len(re.findall(r"-", str))
+
+    # if there is a "-""
+    # 1. if there are 2 A?xx, we need to add the ones between the 2 
+    # 2. if there is only 1 A?XX, we need to add all following ones
+    if dash_count != 0:
+        begin = end = gpu_list[0]
+        skip = False
+        if a_count == 2:
+            end = gpu_list[1]
+        else:
+            end = bitfields._fields_[0][1]._fields_[-1][0]
+            if end != begin:
+                gpu_list.append(end)
+            else:
+                skip = True
+
+        # iterate all elements in BitfieldsBits to add the missing ones into "gpu_list"
+        # note that the begin and end have already added to "gpu_list"  
+        if not skip:  
+            need_add = False
+            for e in bitfields._fields_[0][1]._fields_:
+                if e[0] == end:
+                    need_add = False
+                    break
+
+                if need_add:
+                    gpu_list.append(e[0])
+
+                if e[0] == begin:
+                    need_add = True     
+
+    for var in gpu_list:
+        match var:
+            case "A2XX": bitfields.bits.A2XX = 1
+            case "A3XX": bitfields.bits.A3XX = 1
+            case "A4XX": bitfields.bits.A4XX = 1
+            case "A5XX": bitfields.bits.A5XX = 1
+            case "A6XX": bitfields.bits.A6XX = 1
+            case "A7XX": bitfields.bits.A7XX = 1
+    return bitfields.asbyte
 
 # ---------------------------------------------------------------------------------------
 def isBuiltInType(type):

dive_core/data_core.cpp

@@ -713,13 +713,17 @@ void CaptureMetadataCreator::FillHardwareSpecificStates(EventStateInfo::Iterator
         const uint32_t         binw = bitfields.BINW << reg_field_w->m_shr;
         const uint32_t         binh = bitfields.BINH << reg_field_h->m_shr;
         const a6xx_render_mode render_mode = bitfields.RENDER_MODE;
-        // TODO(wangra): this is only available on a6xx, should disable this on a7xx captures
-        const a6xx_buffers_location buffers_location = bitfields.BUFFERS_LOCATION;
 
         event_state_it->SetBinW(binw);
         event_state_it->SetBinH(binh);
         event_state_it->SetRenderMode(render_mode);
-        event_state_it->SetBuffersLocation(buffers_location);
+
+        // this is only available on a6xx
+        if (IsFieldEnabled(GetRegFieldByName("BUFFERS_LOCATION", reg_info)))
+        {
+            const a6xx_buffers_location buffers_location = bitfields.BUFFERS_LOCATION;
+            event_state_it->SetBuffersLocation(buffers_location);
+        }
     }
 
     // helper lane related