rocm-systems/rocclr/runtime/device/pal/palprogram.cpp

//
// Copyright (c) 2015 Advanced Micro Devices, Inc. All rights reserved.
//
#include "os/os.hpp"
#include "utils/flags.hpp"
#include "include/aclTypes.h"
#include "utils/amdilUtils.hpp"
#include "utils/bif_section_labels.hpp"
#include "device/pal/palprogram.hpp"
#include "device/pal/palblit.hpp"
#include "macrodata.h"
#include "MDParser/AMDILMDInterface.h"
#include <fstream>
#include <sstream>
#include <cstdio>
#include <algorithm>
#include <iterator>
#include "utils/options.hpp"
#include "hsa.h"
#include "hsa_ext_image.h"
#include "amd_hsa_loader.hpp"
#if defined(WITH_LIGHTNING_COMPILER) || defined(USE_COMGR_LIBRARY)
#ifndef USE_COMGR_LIBRARY
#include "driver/AmdCompiler.h"
#endif
#include "llvm/Support/AMDGPUMetadata.h"
#include "libraries.amdgcn.inc"
#include "gelf.h"
#endif  // defined(WITH_LIGHTNING_COMPILER) || defined(USE_COMGR_LIBRARY)

namespace pal {

Segment::Segment() : gpuAccess_(nullptr), cpuAccess_(nullptr), cpuMem_(nullptr) {}

Segment::~Segment() {
  delete gpuAccess_;
  DestroyCpuAccess();
}

void Segment::DestroyCpuAccess() {
  if (cpuAccess_ != nullptr) {
    cpuAccess_->unmap(nullptr);
    delete cpuAccess_;
    cpuAccess_ = nullptr;
  }
  if (cpuMem_ != nullptr) {
    delete[] cpuMem_;
    cpuMem_ = nullptr;
  }
}

bool Segment::alloc(HSAILProgram& prog, amdgpu_hsa_elf_segment_t segment, size_t size, size_t align,
                    bool zero) {
  align = amd::alignUp(align, sizeof(uint32_t));
  gpuAccess_ = new pal::Memory(prog.dev(), amd::alignUp(size, align));
  if ((gpuAccess_ == nullptr) || !gpuAccess_->create(pal::Resource::Shader)) {
    delete gpuAccess_;
    gpuAccess_ = nullptr;
    return false;
  }
  if (segment == AMDGPU_HSA_SEGMENT_CODE_AGENT) {
    void* ptr = nullptr;
    cpuAccess_ = new pal::Memory(prog.dev(), amd::alignUp(size, align));
    if ((cpuAccess_ == nullptr) || !cpuAccess_->create(pal::Resource::Remote)) {
      delete cpuAccess_;
      cpuAccess_ = nullptr;
      ptr = cpuMem_ = reinterpret_cast<address>(new char[amd::alignUp(size, align)]);
      if (cpuMem_ == nullptr) {
        return false;
      }
    } else {
      ptr = cpuAccess_->map(nullptr, 0);
    }
    if (zero) {
      memset(ptr, 0, size);
    }
  }

  if (zero && !prog.isInternal()) {
    uint64_t pattern = 0;
    size_t   patternSize = ((size % sizeof(pattern)) == 0) ? sizeof(pattern) : 1;
    prog.dev().xferMgr().fillBuffer(*gpuAccess_, &pattern, patternSize,
        amd::Coord3D(0), amd::Coord3D(size));
  }

  switch (segment) {
    case AMDGPU_HSA_SEGMENT_GLOBAL_PROGRAM:
    case AMDGPU_HSA_SEGMENT_GLOBAL_AGENT:
    case AMDGPU_HSA_SEGMENT_READONLY_AGENT:
      prog.addGlobalStore(gpuAccess_);
      prog.setGlobalVariableTotalSize(prog.globalVariableTotalSize() + size);
      break;
    case AMDGPU_HSA_SEGMENT_CODE_AGENT:
      prog.setCodeObjects(this, gpuAccess_, reinterpret_cast<address>(cpuAddress(0)));
      break;
    default:
      break;
  }
  return true;
}

void Segment::copy(size_t offset, const void* src, size_t size) {
  if (cpuAccess_ != nullptr) {
    amd::Os::fastMemcpy(cpuAddress(offset), src, size);
  } else {
    if (cpuMem_ != nullptr) {
      amd::Os::fastMemcpy(cpuAddress(offset), src, size);
    }
    amd::ScopedLock k(gpuAccess_->dev().xferMgr().lockXfer());
    VirtualGPU& gpu = *gpuAccess_->dev().xferQueue();
    Memory& xferBuf = gpu.xferWrite().Acquire(size);
    size_t tmpSize = std::min(static_cast<size_t>(xferBuf.size()), size);
    size_t srcOffs = 0;
    while (size != 0) {
      xferBuf.hostWrite(&gpu, reinterpret_cast<const_address>(src) + srcOffs, 0, tmpSize);
      xferBuf.partialMemCopyTo(gpu, 0, (offset + srcOffs), tmpSize, *gpuAccess_, false, true);
      size -= tmpSize;
      srcOffs += tmpSize;
      tmpSize = std::min(static_cast<size_t>(xferBuf.size()), size);
    }
    gpu.xferWrite().Release(xferBuf);
    gpu.waitAllEngines();
  }
}

bool Segment::freeze(bool destroySysmem) {
  VirtualGPU& gpu = *gpuAccess_->dev().xferQueue();
  bool result = true;
  if (cpuAccess_ != nullptr) {
    assert(gpuAccess_->size() == cpuAccess_->size() && "Backing store size mismatch!");
    amd::ScopedLock k(gpuAccess_->dev().xferMgr().lockXfer());
    result = cpuAccess_->partialMemCopyTo(gpu, 0, 0, gpuAccess_->size(), *gpuAccess_, false, true);
    gpu.waitAllEngines();
  }
  assert(!destroySysmem || (cpuAccess_ == nullptr));
  return result;
}

const static char* Carrizo = "Carrizo";
HSAILProgram::HSAILProgram(Device& device)
    : Program(device),
      rawBinary_(nullptr),
      kernels_(nullptr),
      codeSegGpu_(nullptr),
      codeSegment_(nullptr),
      maxScratchRegs_(0),
      executable_(nullptr),
      loaderContext_(this) {
  xnackEnabled_ = dev().hwInfo()->xnackEnabled_;
  sramEccEnabled_ = dev().info().sramEccEnabled_;
  if (dev().asicRevision() == Pal::AsicRevision::Bristol) {
    machineTarget_ = Carrizo;
  } else {
    machineTarget_ = dev().hwInfo()->targetName_;
  }
  loader_ = amd::hsa::loader::Loader::Create(&loaderContext_);
}

HSAILProgram::HSAILProgram(NullDevice& device)
    : Program(device),
      rawBinary_(nullptr),
      kernels_(nullptr),
      codeSegGpu_(nullptr),
      codeSegment_(nullptr),
      maxScratchRegs_(0),
      executable_(nullptr),
      loaderContext_(this) {
  isNull_ = true;
  xnackEnabled_ = dev().hwInfo()->xnackEnabled_;
  sramEccEnabled_ = dev().info().sramEccEnabled_;
  if (dev().asicRevision() == Pal::AsicRevision::Bristol) {
    machineTarget_ = Carrizo;
  } else {
    machineTarget_ = dev().hwInfo()->targetName_;
  }
  loader_ = amd::hsa::loader::Loader::Create(&loaderContext_);
}

HSAILProgram::~HSAILProgram() {
  // Destroy internal static samplers
  for (auto& it : staticSamplers_) {
    delete it;
  }
#if defined(WITH_COMPILER_LIB)
  if (rawBinary_ != nullptr) {
    aclFreeMem(binaryElf_, rawBinary_);
  }
  acl_error error;
  // Free the elf binary
  if (binaryElf_ != nullptr) {
    error = aclBinaryFini(binaryElf_);
    if (error != ACL_SUCCESS) {
      LogWarning("Error while destroying the acl binary \n");
    }
  }
#endif  // defined(WITH_COMPILER_LIB)
  releaseClBinary();
  if (executable_ != nullptr) {
    loader_->DestroyExecutable(executable_);
  }
  delete kernels_;
  amd::hsa::loader::Loader::Destroy(loader_);
}


inline static std::vector<std::string> splitSpaceSeparatedString(char* str) {
  std::string s(str);
  std::stringstream ss(s);
  std::istream_iterator<std::string> beg(ss), end;
  std::vector<std::string> vec(beg, end);
  return vec;
}

bool HSAILProgram::setKernels(amd::option::Options* options, void* binary, size_t binSize) {
#if  defined(WITH_COMPILER_LIB)
  // ACL_TYPE_CG stage is not performed for offline compilation
  hsa_agent_t agent;
  agent.handle = 1;
  executable_ = loader_->CreateExecutable(HSA_PROFILE_FULL, nullptr);
  if (executable_ == nullptr) {
    buildLog_ += "Error: Executable for AMD HSA Code Object isn't created.\n";
    return false;
  }
  size_t size = binSize;
  hsa_code_object_t code_object;
  code_object.handle = reinterpret_cast<uint64_t>(binary);

  hsa_status_t status = executable_->LoadCodeObject(agent, code_object, nullptr);
  if (status != HSA_STATUS_SUCCESS) {
    buildLog_ += "Error: AMD HSA Code Object loading failed.\n";
    return false;
  }
  status = executable_->Freeze(nullptr);
  if (status != HSA_STATUS_SUCCESS) {
    buildLog_ += "Error: AMD HSA Code Object freeze failed.\n";
    return false;
  }

  size_t kernelNamesSize = 0;
  acl_error errorCode = aclQueryInfo(dev().compiler(), binaryElf_, RT_KERNEL_NAMES,
    nullptr, nullptr, &kernelNamesSize);
  if (errorCode != ACL_SUCCESS) {
    buildLog_ += "Error: Querying of kernel names size from the binary failed.\n";
    return false;
  }
  if (kernelNamesSize > 0) {
    char* kernelNames = new char[kernelNamesSize];
    errorCode = aclQueryInfo(dev().compiler(), binaryElf_, RT_KERNEL_NAMES, nullptr, kernelNames,
                             &kernelNamesSize);
    if (errorCode != ACL_SUCCESS) {
      buildLog_ += "Error: Querying of kernel names from the binary failed.\n";
      delete [] kernelNames;
      return false;
    }
    std::vector<std::string> vKernels = splitSpaceSeparatedString(kernelNames);
    delete [] kernelNames;
    bool dynamicParallelism = false;
    for (const auto& it : vKernels) {
      std::string kernelName(it);
      std::string openclKernelName = device::Kernel::openclMangledName(kernelName);

      HSAILKernel* aKernel =
          new HSAILKernel(kernelName, this, options->origOptionStr + ProcessOptions(options));
      kernels()[kernelName] = aKernel;

      amd::hsa::loader::Symbol* sym = executable_->GetSymbol(openclKernelName.c_str(), &agent);
      if (!sym) {
        buildLog_ += "Error: Getting kernel ISA code symbol '" + openclKernelName +
            "' from AMD HSA Code Object failed. Kernel initialization failed.\n";
        return false;
      }
      if (!aKernel->init(sym, false)) {
        buildLog_ += "Error: Kernel '" + openclKernelName + "' initialization failed.\n";
        return false;
      }
      buildLog_ += aKernel->buildLog();
      aKernel->setUniformWorkGroupSize(options->oVariables->UniformWorkGroupSize);
      dynamicParallelism |= aKernel->dynamicParallelism();
      // Find max scratch regs used in the program. It's used for scratch buffer preallocation
      // with dynamic parallelism, since runtime doesn't know which child kernel will be called
      maxScratchRegs_ =
          std::max(static_cast<uint>(aKernel->workGroupInfo()->scratchRegs_), maxScratchRegs_);
    }
    // Allocate kernel table for device enqueuing
    if (!isNull() && dynamicParallelism && !allocKernelTable()) {
      return false;
    }
  }

  DestroySegmentCpuAccess();
#endif  // defined(WITH_COMPILER_LIB)
  return true;
}

bool HSAILProgram::createBinary(amd::option::Options* options) { return true; }

bool HSAILProgram::allocKernelTable() {
  uint size = kernels().size() * sizeof(size_t);

  kernels_ = new pal::Memory(dev(), size);
  // Initialize kernel table
  if ((kernels_ == nullptr) || !kernels_->create(Resource::RemoteUSWC)) {
    delete kernels_;
    return false;
  } else {
    size_t* table = reinterpret_cast<size_t*>(kernels_->map(nullptr, pal::Resource::WriteOnly));
    for (auto& it : kernels()) {
      HSAILKernel* kernel = static_cast<HSAILKernel*>(it.second);
      table[kernel->index()] = static_cast<size_t>(kernel->gpuAqlCode());
    }
    kernels_->unmap(nullptr);
  }
  return true;
}

void HSAILProgram::fillResListWithKernels(VirtualGPU& gpu) const {
  gpu.addVmMemory(&codeSegGpu());
}

const aclTargetInfo& HSAILProgram::info(const char* str) {
#if  defined(WITH_COMPILER_LIB)
  acl_error err;
  std::string arch = "hsail";
  if (dev().settings().use64BitPtr_) {
    arch = "hsail64";
  }
  info_ = aclGetTargetInfo(arch.c_str(),
                           (str && str[0] == '\0' ? dev().hwInfo()->targetName_ : str), &err);
  if (err != ACL_SUCCESS) {
    LogWarning("aclGetTargetInfo failed");
  }
#endif  // defined(WITH_COMPILER_LIB)
  return info_;
}

bool HSAILProgram::saveBinaryAndSetType(type_t type) {
#if  defined(WITH_COMPILER_LIB)
  // Write binary to memory
  if (rawBinary_ != nullptr) {
    // Free memory containing rawBinary
    aclFreeMem(binaryElf_, rawBinary_);
    rawBinary_ = nullptr;
  }
  size_t size = 0;
  if (aclWriteToMem(binaryElf_, &rawBinary_, &size) != ACL_SUCCESS) {
    buildLog_ += "Failed to write binary to memory \n";
    return false;
  }
  setBinary(static_cast<char*>(rawBinary_), size);
  // Set the type of binary
  setType(type);
#endif  // defined(WITH_COMPILER_LIB)
  return true;
}

hsa_isa_t PALHSALoaderContext::IsaFromName(const char* name) {
  hsa_isa_t isa = {0};
  uint32_t gfxip  = 0;
  std::string gfx_target(name);
  if (gfx_target.find("amdgcn-") == 0) {
    std::string gfxip_version_str = gfx_target.substr(gfx_target.find("gfx") + 3);
    gfxip = std::atoi(gfxip_version_str.c_str());
  }
  else {
    // FIXME: Old way. To be remove.
    uint32_t shift = 1;
    size_t last = gfx_target.length();
    std::string ver;
    do {
      size_t first = gfx_target.find_last_of(':', last);
      ver = gfx_target.substr(first + 1, last - first);
      last = first - 1;
      gfxip += static_cast<uint32_t>(atoi(ver.c_str())) * shift;
      shift *= 10;
    } while (shift <= 100);
  }
  isa.handle = gfxip;
  return isa;
}

bool PALHSALoaderContext::IsaSupportedByAgent(hsa_agent_t agent, hsa_isa_t isa) {
  uint32_t gfxipVersion = program_->dev().settings().useLightning_ ?
    program_->dev().hwInfo()->gfxipVersionLC_ :
    program_->dev().hwInfo()->gfxipVersion_;
  uint32_t majorSrc = gfxipVersion / 10;
  uint32_t minorSrc = gfxipVersion % 10;

  uint32_t majorTrg = isa.handle / 10;
  uint32_t minorTrg = isa.handle % 10;

  if (majorSrc != majorTrg) {
    return false;
  }
  else if (minorTrg == minorSrc) {
    return true;
  }
  else if (minorTrg < minorSrc) {
    LogWarning("ISA downgrade for execution!");
    return true;
  }

  return false;
}

void* PALHSALoaderContext::SegmentAlloc(amdgpu_hsa_elf_segment_t segment, hsa_agent_t agent,
                                        size_t size, size_t align, bool zero) {
  assert(size);
  assert(align);
  if (program_->isNull()) {
    // Note: In Linux ::posix_memalign() requires at least 16 bytes for the alignment.
    align = amd::alignUp(align, 16);
    void* ptr = amd::Os::alignedMalloc(size, align);
    if ((ptr != nullptr) && zero) {
      memset(ptr, 0, size);
    }
    return ptr;
  }
  Segment* seg = new Segment();
  if (seg != nullptr && !seg->alloc(*program_, segment, size, align, zero)) {
    return nullptr;
  }
  return seg;
}

bool PALHSALoaderContext::SegmentCopy(amdgpu_hsa_elf_segment_t segment, hsa_agent_t agent,
                                      void* dst, size_t offset, const void* src, size_t size) {
  if (program_->isNull()) {
    amd::Os::fastMemcpy(reinterpret_cast<address>(dst) + offset, src, size);
    return true;
  }
  Segment* s = reinterpret_cast<Segment*>(dst);
  s->copy(offset, src, size);
  return true;
}

void PALHSALoaderContext::SegmentFree(amdgpu_hsa_elf_segment_t segment, hsa_agent_t agent,
                                      void* seg, size_t size) {
  if (program_->isNull()) {
    amd::Os::alignedFree(seg);
  } else {
    Segment* s = reinterpret_cast<Segment*>(seg);
    delete s;
  }
}

void* PALHSALoaderContext::SegmentAddress(amdgpu_hsa_elf_segment_t segment, hsa_agent_t agent,
                                          void* seg, size_t offset) {
  assert(seg);
  if (program_->isNull()) {
    return (reinterpret_cast<address>(seg) + offset);
  }
  Segment* s = reinterpret_cast<Segment*>(seg);
  return reinterpret_cast<void*>(s->gpuAddress(offset));
}

void* PALHSALoaderContext::SegmentHostAddress(amdgpu_hsa_elf_segment_t segment, hsa_agent_t agent,
                                              void* seg, size_t offset) {
  assert(seg);
  if (program_->isNull()) {
    return (reinterpret_cast<address>(seg) + offset);
  }
  Segment* s = reinterpret_cast<Segment*>(seg);
  return s->cpuAddress(offset);
}

bool PALHSALoaderContext::SegmentFreeze(amdgpu_hsa_elf_segment_t segment, hsa_agent_t agent,
                                        void* seg, size_t size) {
  if (program_->isNull()) {
    return true;
  }
  Segment* s = reinterpret_cast<Segment*>(seg);
  return s->freeze((segment == AMDGPU_HSA_SEGMENT_CODE_AGENT) ? false : true);
}

hsa_status_t PALHSALoaderContext::SamplerCreate(
    hsa_agent_t agent, const hsa_ext_sampler_descriptor_t* sampler_descriptor,
    hsa_ext_sampler_t* sampler_handle) {
  if (!agent.handle) {
    return HSA_STATUS_ERROR_INVALID_AGENT;
  }
  if (!sampler_descriptor || !sampler_handle) {
    return HSA_STATUS_ERROR_INVALID_ARGUMENT;
  }
  if (program_->isNull()) {
    // Offline compilation. Provide a fake handle to avoid an assert
    sampler_handle->handle = 1;
    return HSA_STATUS_SUCCESS;
  }
  uint32_t state = 0;
  switch (sampler_descriptor->coordinate_mode) {
    case HSA_EXT_SAMPLER_COORDINATE_MODE_UNNORMALIZED:
      state = amd::Sampler::StateNormalizedCoordsFalse;
      break;
    case HSA_EXT_SAMPLER_COORDINATE_MODE_NORMALIZED:
      state = amd::Sampler::StateNormalizedCoordsTrue;
      break;
    default:
      assert(false);
      return HSA_STATUS_ERROR_INVALID_ARGUMENT;
  }
  switch (sampler_descriptor->filter_mode) {
    case HSA_EXT_SAMPLER_FILTER_MODE_NEAREST:
      state |= amd::Sampler::StateFilterNearest;
      break;
    case HSA_EXT_SAMPLER_FILTER_MODE_LINEAR:
      state |= amd::Sampler::StateFilterLinear;
      break;
    default:
      assert(false);
      return HSA_STATUS_ERROR_INVALID_ARGUMENT;
  }
  switch (sampler_descriptor->address_mode) {
    case HSA_EXT_SAMPLER_ADDRESSING_MODE_CLAMP_TO_EDGE:
      state |= amd::Sampler::StateAddressClampToEdge;
      break;
    case HSA_EXT_SAMPLER_ADDRESSING_MODE_CLAMP_TO_BORDER:
      state |= amd::Sampler::StateAddressClamp;
      break;
    case HSA_EXT_SAMPLER_ADDRESSING_MODE_REPEAT:
      state |= amd::Sampler::StateAddressRepeat;
      break;
    case HSA_EXT_SAMPLER_ADDRESSING_MODE_MIRRORED_REPEAT:
      state |= amd::Sampler::StateAddressMirroredRepeat;
      break;
    case HSA_EXT_SAMPLER_ADDRESSING_MODE_UNDEFINED:
      state |= amd::Sampler::StateAddressNone;
      break;
    default:
      assert(false);
      return HSA_STATUS_ERROR_INVALID_ARGUMENT;
  }
  pal::Sampler* sampler = new pal::Sampler(program_->dev());
  if (!sampler || !sampler->create(state)) {
    delete sampler;
    return HSA_STATUS_ERROR;
  }
  program_->addSampler(sampler);
  sampler_handle->handle = sampler->hwSrd();
  return HSA_STATUS_SUCCESS;
}

hsa_status_t PALHSALoaderContext::SamplerDestroy(hsa_agent_t agent,
                                                 hsa_ext_sampler_t sampler_handle) {
  if (!agent.handle) {
    return HSA_STATUS_ERROR_INVALID_AGENT;
  }
  if (!sampler_handle.handle) {
    return HSA_STATUS_ERROR_INVALID_ARGUMENT;
  }
  return HSA_STATUS_SUCCESS;
}

#if defined(WITH_LIGHTNING_COMPILER) || defined(USE_COMGR_LIBRARY)

static hsa_status_t GetKernelNamesCallback(hsa_executable_t hExec, hsa_executable_symbol_t hSymbol,
                                           void* data) {
  auto symbol = Symbol::Object(hSymbol);
  auto symbolNameList = reinterpret_cast<std::vector<std::string>*>(data);

  hsa_symbol_kind_t type;
  if (!symbol->GetInfo(HSA_EXECUTABLE_SYMBOL_INFO_TYPE, &type)) {
    return HSA_STATUS_ERROR;
  }

  if (type == HSA_SYMBOL_KIND_KERNEL) {
    uint32_t length;
    if (!symbol->GetInfo(HSA_EXECUTABLE_SYMBOL_INFO_NAME_LENGTH, &length)) {
      return HSA_STATUS_ERROR;
    }

    char* name = reinterpret_cast<char*>(alloca(length + 1));
    if (!symbol->GetInfo(HSA_EXECUTABLE_SYMBOL_INFO_NAME, name)) {
      return HSA_STATUS_ERROR;
    }
    name[length] = '\0';

    symbolNameList->push_back(std::string(name));
  }
  return HSA_STATUS_SUCCESS;
}

#endif // defined(WITH_LIGHTNING_COMPILER) || defined(USE_COMGR_LIBRARY)

bool LightningProgram::createBinary(amd::option::Options* options) {
#if defined(WITH_LIGHTNING_COMPILER) || defined(USE_COMGR_LIBRARY)
  if (!clBinary()->createElfBinary(options->oVariables->BinEncrypt, type())) {
    LogError("Failed to create ELF binary image!");
    return false;
  }
#endif // defined(WITH_LIGHTNING_COMPILER) || defined(USE_COMGR_LIBRARY)
  return true;
}

bool LightningProgram::setKernels(amd::option::Options* options, void* binary, size_t binSize) {
#if defined(WITH_LIGHTNING_COMPILER) || defined(USE_COMGR_LIBRARY)
  hsa_agent_t agent;
  agent.handle = 1;

  executable_ = loader_->CreateExecutable(HSA_PROFILE_FULL, nullptr);
  if (executable_ == nullptr) {
    buildLog_ += "Error: Executable for AMD HSA Code Object isn't created.\n";
    return false;
  }

  hsa_code_object_t code_object;
  code_object.handle = reinterpret_cast<uint64_t>(binary);

  hsa_status_t status = executable_->LoadCodeObject(agent, code_object, nullptr);
  if (status != HSA_STATUS_SUCCESS) {
    buildLog_ += "Error: AMD HSA Code Object loading failed.\n";
    return false;
  }

  status = executable_->Freeze(nullptr);
  if (status != HSA_STATUS_SUCCESS) {
    buildLog_ += "Error: Freezing the executable failed: ";
    return false;
  }

  // Find the size of global variables from the binary
  if (!FindGlobalVarSize(binary, binSize)) {
    return false;
  }

  // Get the list of kernels
  std::vector<std::string> kernelNameList;
  status = executable_->IterateSymbols(GetKernelNamesCallback, &kernelNameList);
  if (status != HSA_STATUS_SUCCESS) {
    buildLog_ += "Error: Failed to get kernel names\n";
    return false;
  }

  for (const auto& kernelName : kernelNameList) {
    auto kernel =
        new LightningKernel(kernelName, this, options->origOptionStr + ProcessOptions(options));

    kernels()[kernelName] = kernel;

    auto symbol = executable_->GetSymbol(kernelName.c_str(), &agent);
    if (!symbol) {
      buildLog_ += "Error: Getting kernel symbol '" + kernelName +
          "' from AMD HSA Code Object failed. "
          "Kernel initialization failed.\n";
      return false;
    }
    if (!kernel->init(symbol)) {
      buildLog_ += "Error: Kernel '" + kernelName + "' initialization failed.\n";
      return false;
    }
    buildLog_ += kernel->buildLog();

    kernel->setUniformWorkGroupSize(options->oVariables->UniformWorkGroupSize);

    // Find max scratch regs used in the program. It's used for scratch buffer preallocation
    // with dynamic parallelism, since runtime doesn't know which child kernel will be called
    maxScratchRegs_ =
        std::max(static_cast<uint>(kernel->workGroupInfo()->scratchRegs_), maxScratchRegs_);
  }

  DestroySegmentCpuAccess();
#endif // defined(WITH_LIGHTNING_COMPILER) || defined(USE_COMGR_LIBRARY)
  return true;
}

}  // namespace pal