SWDEV-515426 - Remove HIP_USE_RUNTIME_UNBUNDLER (#205)
* remove HIP_USE_RUNTIME_UNBUNDLER * clang-format * Generic to use comgr * Remove HIP_ALWAYS_USE_NEW_COMGR_UNBUNDLING_ACTION flag * Removes runtime unbundling unused and debug Code * Removes stale functions
This commit is contained in:
Belton-Schure, Aidan
committed by
GitHub
GitHub
parent
22b1ca4d8c
commit
81238db679
@@ -128,314 +128,6 @@ bool CodeObject::containGenericTarget(const void *data) {
|
||||
|
||||
uint64_t CodeObject::ElfSize(const void* emi) { return amd::Elf::getElfSize(emi); }
|
||||
|
||||
static bool getProcName(uint32_t EFlags, std::string& proc_name, bool& xnackSupported,
|
||||
bool& sramEccSupported) {
|
||||
switch (EFlags & EF_AMDGPU_MACH) {
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX700:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx700";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX701:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx701";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX702:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx702";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX703:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx703";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX704:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx704";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX705:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx705";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX801:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx801";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX802:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx802";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX803:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx803";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX805:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx805";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX810:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx810";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX900:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx900";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX902:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx902";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX904:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx904";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX906:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = true;
|
||||
proc_name = "gfx906";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX908:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = true;
|
||||
proc_name = "gfx908";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX909:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx909";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX90A:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = true;
|
||||
proc_name = "gfx90a";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX90C:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx90c";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX942:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = true;
|
||||
proc_name = "gfx942";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX950:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = true;
|
||||
proc_name = "gfx950";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1010:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1010";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1011:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1011";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1012:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1012";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1013:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1013";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1030:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1030";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1031:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1031";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1032:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1032";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1033:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1033";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1034:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1034";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1035:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1035";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1036:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1036";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1100:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1100";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1101:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1101";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1102:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1102";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1103:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1103";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1150:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1150";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1151:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1151";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1200:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1200";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX1201:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx1201";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX9_GENERIC:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx9-generic";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX9_4_GENERIC:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = true;
|
||||
proc_name = "gfx9-4-generic";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX10_1_GENERIC:
|
||||
xnackSupported = true;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx10-1-generic";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX10_3_GENERIC:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx10-3-generic";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX11_GENERIC:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx11-generic";
|
||||
break;
|
||||
case EF_AMDGPU_MACH_AMDGCN_GFX12_GENERIC:
|
||||
xnackSupported = false;
|
||||
sramEccSupported = false;
|
||||
proc_name = "gfx12-generic";
|
||||
break;
|
||||
default:
|
||||
return false;
|
||||
}
|
||||
return true;
|
||||
}
|
||||
|
||||
static bool getTripleTargetIDFromCodeObject(const void* code_object, std::string& target_id) {
|
||||
if (!code_object) return false;
|
||||
const Elf64_Ehdr* ehdr = reinterpret_cast<const Elf64_Ehdr*>(code_object);
|
||||
if (ehdr->e_machine != EM_AMDGPU) return false;
|
||||
if (ehdr->e_ident[EI_OSABI] != ELFOSABI_AMDGPU_HSA) return false;
|
||||
|
||||
bool isXnackSupported{false}, isSramEccSupported{false};
|
||||
const char* vstr = nullptr;
|
||||
std::string proc_name;
|
||||
if (!getProcName(ehdr->e_flags, proc_name, isXnackSupported, isSramEccSupported)) return false;
|
||||
target_id = std::string(kAmdgcnTargetTriple) + '-' + proc_name;
|
||||
|
||||
switch (ehdr->e_ident[EI_ABIVERSION]) {
|
||||
case ELFABIVERSION_AMDGPU_HSA_V2: {
|
||||
ClPrint(amd::LOG_INFO, amd::LOG_COMGR, "[Code Object V2, target id:%s]", target_id.c_str());
|
||||
return false;
|
||||
}
|
||||
|
||||
case ELFABIVERSION_AMDGPU_HSA_V3: {
|
||||
ClPrint(amd::LOG_INFO, amd::LOG_COMGR, "[Code Object V3, target id:%s]", target_id.c_str());
|
||||
if (isSramEccSupported) {
|
||||
if (ehdr->e_flags & EF_AMDGPU_FEATURE_SRAMECC_V3)
|
||||
target_id += ":sramecc+";
|
||||
else
|
||||
target_id += ":sramecc-";
|
||||
}
|
||||
if (isXnackSupported) {
|
||||
if (ehdr->e_flags & EF_AMDGPU_FEATURE_XNACK_V3)
|
||||
target_id += ":xnack+";
|
||||
else
|
||||
target_id += ":xnack-";
|
||||
}
|
||||
break;
|
||||
}
|
||||
|
||||
case ELFABIVERSION_AMDGPU_HSA_V4:
|
||||
case ELFABIVERSION_AMDGPU_HSA_V5:
|
||||
case ELFABIVERSION_AMDGPU_HSA_V6: {
|
||||
if (ehdr->e_ident[EI_ABIVERSION] & ELFABIVERSION_AMDGPU_HSA_V4) {
|
||||
vstr = "V4";
|
||||
} else if (ehdr->e_ident[EI_ABIVERSION] & ELFABIVERSION_AMDGPU_HSA_V5) {
|
||||
vstr = "V5";
|
||||
} else if (ehdr->e_ident[EI_ABIVERSION] & ELFABIVERSION_AMDGPU_HSA_V6) {
|
||||
vstr = "V6";
|
||||
}
|
||||
unsigned co_sram_value = (ehdr->e_flags) & EF_AMDGPU_FEATURE_SRAMECC_V4;
|
||||
if (co_sram_value == EF_AMDGPU_FEATURE_SRAMECC_OFF_V4)
|
||||
target_id += ":sramecc-";
|
||||
else if (co_sram_value == EF_AMDGPU_FEATURE_SRAMECC_ON_V4)
|
||||
target_id += ":sramecc+";
|
||||
|
||||
unsigned co_xnack_value = (ehdr->e_flags) & EF_AMDGPU_FEATURE_XNACK_V4;
|
||||
if (co_xnack_value == EF_AMDGPU_FEATURE_XNACK_OFF_V4)
|
||||
target_id += ":xnack-";
|
||||
|
||||
else if (co_xnack_value == EF_AMDGPU_FEATURE_XNACK_ON_V4)
|
||||
target_id += ":xnack+";
|
||||
ClPrint(amd::LOG_INFO, amd::LOG_COMGR,
|
||||
"[Code Object %s, target id: %s]", vstr, target_id.c_str());
|
||||
break;
|
||||
}
|
||||
|
||||
default: {
|
||||
return false;
|
||||
}
|
||||
}
|
||||
return true;
|
||||
}
|
||||
|
||||
// Consumes the string 'consume_' from the starting of the given input
|
||||
// eg: input = amdgcn-amd-amdhsa--gfx908 and consume_ is amdgcn-amd-amdhsa--
|
||||
// input will become gfx908.
|
||||
@@ -493,21 +185,6 @@ static bool getTargetIDValue(std::string& input, std::string& processor, char& s
|
||||
return true;
|
||||
}
|
||||
|
||||
static bool getTripleTargetID(std::string bundled_co_entry_id, const void* code_object,
|
||||
std::string& co_triple_target_id) {
|
||||
std::string offload_kind = trimName(bundled_co_entry_id, '-');
|
||||
if (offload_kind != kOffloadKindHipv4 && offload_kind != kOffloadKindHip &&
|
||||
offload_kind != kOffloadKindHcc)
|
||||
return false;
|
||||
if (offload_kind != kOffloadKindHipv4)
|
||||
return getTripleTargetIDFromCodeObject(code_object, co_triple_target_id);
|
||||
|
||||
// For code object V4 onwards the bundled code object entry ID correctly
|
||||
// specifies the target triple.
|
||||
co_triple_target_id = bundled_co_entry_id.substr(1);
|
||||
return true;
|
||||
}
|
||||
|
||||
static bool isCodeObjectCompatibleWithDevice(std::string co_triple_target_id,
|
||||
std::string agent_triple_target_id, unsigned int genericVersion) {
|
||||
// Primitive Check
|
||||
@@ -574,125 +251,6 @@ bool CodeObject::QueryGenericTarget(std::string agentTarget, std::string& proces
|
||||
return true;
|
||||
}
|
||||
|
||||
// This will be moved to COMGR eventually
|
||||
hipError_t CodeObject::ExtractCodeObjectFromFile(
|
||||
amd::Os::FileDesc fdesc, size_t fsize, const void** image,
|
||||
const std::vector<std::string>& device_names,
|
||||
std::vector<std::pair<const void*, size_t>>& code_objs, size_t foffset) {
|
||||
if (!amd::Os::isValidFileDesc(fdesc)) {
|
||||
return hipErrorFileNotFound;
|
||||
}
|
||||
|
||||
// Map the file to memory, with offset 0.
|
||||
// file will be unmapped in ModuleUnload
|
||||
// const void* image = nullptr;
|
||||
if (!amd::Os::MemoryMapFileDesc(fdesc, fsize, foffset, image)) {
|
||||
return hipErrorInvalidValue;
|
||||
}
|
||||
|
||||
// retrieve code_objs{binary_image, binary_size} for devices
|
||||
return extractCodeObjectFromFatBinary(*image, device_names, code_objs);
|
||||
}
|
||||
|
||||
// This will be moved to COMGR eventually
|
||||
hipError_t CodeObject::ExtractCodeObjectFromMemory(
|
||||
const void* data, const std::vector<std::string>& device_names,
|
||||
std::vector<std::pair<const void*, size_t>>& code_objs, std::string& uri) {
|
||||
// Get the URI from memory
|
||||
if (!amd::Os::GetURIFromMemory(data, 0, uri)) {
|
||||
return hipErrorInvalidValue;
|
||||
}
|
||||
|
||||
return extractCodeObjectFromFatBinary(data, device_names, code_objs);
|
||||
}
|
||||
|
||||
// This will be moved to COMGR eventually
|
||||
hipError_t CodeObject::extractCodeObjectFromFatBinary(
|
||||
const void* data, const std::vector<std::string>& agent_triple_target_ids,
|
||||
std::vector<std::pair<const void*, size_t>>& code_objs) {
|
||||
bool isCompressed = false;
|
||||
if (!IsClangOffloadMagicBundle(data, isCompressed) || isCompressed) {
|
||||
LogPrintfError("IsClangOffloadMagicBundle(%p) return false or isCompressed is true", data);
|
||||
return hipErrorInvalidKernelFile;
|
||||
}
|
||||
|
||||
// Initialize Code objects
|
||||
code_objs.reserve(agent_triple_target_ids.size());
|
||||
for (size_t i = 0; i < agent_triple_target_ids.size(); i++) {
|
||||
code_objs.push_back(std::make_pair(nullptr, 0));
|
||||
}
|
||||
|
||||
const auto obheader = reinterpret_cast<const __ClangOffloadBundleUncompressedHeader*>(data);
|
||||
const auto* desc = &obheader->desc[0];
|
||||
size_t num_code_objs = code_objs.size();
|
||||
for (uint64_t i = 0; i < obheader->numOfCodeObjects; ++i,
|
||||
desc = reinterpret_cast<const __ClangOffloadBundleInfo*>(
|
||||
reinterpret_cast<uintptr_t>(&desc->bundleEntryId[0]) +
|
||||
desc->bundleEntryIdSize)) {
|
||||
const void* image =
|
||||
reinterpret_cast<const void*>(reinterpret_cast<uintptr_t>(obheader) + desc->offset);
|
||||
const size_t image_size = desc->size;
|
||||
|
||||
if (num_code_objs == 0) break;
|
||||
std::string bundleEntryId{desc->bundleEntryId, desc->bundleEntryIdSize};
|
||||
|
||||
std::string co_triple_target_id;
|
||||
uint32_t genericVersion = getGenericVersion(image);
|
||||
if (!getTripleTargetID(bundleEntryId, image, co_triple_target_id)) continue;
|
||||
ClPrint(amd::LOG_INFO, amd::LOG_COMGR,
|
||||
"bundleEntryId=%s, co_triple_target_id=%s, genericVersion=%u\n",
|
||||
bundleEntryId.c_str(), co_triple_target_id.c_str(), genericVersion);
|
||||
|
||||
for (size_t dev = 0; dev < agent_triple_target_ids.size(); ++dev) {
|
||||
if (code_objs[dev].first != nullptr) {
|
||||
if (!isGenericTarget(code_objs[dev].first)) {
|
||||
continue; // Specific target already found
|
||||
} else if(genericVersion >= EF_AMDGPU_GENERIC_VERSION_MIN) {
|
||||
continue; // Generic target already found, no need to check another generic
|
||||
}
|
||||
}
|
||||
if (isCodeObjectCompatibleWithDevice(co_triple_target_id, agent_triple_target_ids[dev],
|
||||
genericVersion)) {
|
||||
if (code_objs[dev].first == nullptr) --num_code_objs;
|
||||
code_objs[dev] = std::make_pair(image, image_size);
|
||||
}
|
||||
}
|
||||
}
|
||||
if (num_code_objs == 0) {
|
||||
return hipSuccess;
|
||||
} else {
|
||||
LogPrintfError("%s",
|
||||
"hipErrorNoBinaryForGpu: Unable to find code object for all current devices!");
|
||||
LogPrintfError("%s", " Devices:");
|
||||
for (size_t i = 0; i < agent_triple_target_ids.size(); i++) {
|
||||
LogPrintfError(" %s - [%s]", agent_triple_target_ids[i].c_str(),
|
||||
((code_objs[i].first) ? "Found" : "Not Found"));
|
||||
}
|
||||
const auto obheader = reinterpret_cast<const __ClangOffloadBundleUncompressedHeader*>(data);
|
||||
const auto* desc = &obheader->desc[0];
|
||||
LogPrintfError("%s", " Bundled Code Objects:");
|
||||
for (uint64_t i = 0; i < obheader->numOfCodeObjects; ++i,
|
||||
desc = reinterpret_cast<const __ClangOffloadBundleInfo*>(
|
||||
reinterpret_cast<uintptr_t>(&desc->bundleEntryId[0]) +
|
||||
desc->bundleEntryIdSize)) {
|
||||
std::string bundleEntryId{desc->bundleEntryId, desc->bundleEntryIdSize};
|
||||
const void* image =
|
||||
reinterpret_cast<const void*>(reinterpret_cast<uintptr_t>(obheader) + desc->offset);
|
||||
|
||||
std::string co_triple_target_id;
|
||||
bool valid_co = getTripleTargetID(bundleEntryId, image, co_triple_target_id);
|
||||
|
||||
if (valid_co) {
|
||||
LogPrintfError(" %s - [Code object targetID is %s]", bundleEntryId.c_str(),
|
||||
co_triple_target_id.c_str());
|
||||
} else {
|
||||
LogPrintfError(" %s - [Unsupported]", bundleEntryId.c_str());
|
||||
}
|
||||
}
|
||||
return hipErrorNoBinaryForGpu;
|
||||
}
|
||||
}
|
||||
|
||||
// ================================================================================================
|
||||
size_t CodeObject::getFatbinSize(const void* data, const bool isCompressed) {
|
||||
if (isCompressed) {
|
||||
@@ -1012,41 +570,6 @@ hipError_t CodeObject::extractCodeObjectFromFatBinaryUsingComgr(
|
||||
}
|
||||
} while (0);
|
||||
|
||||
if (hipStatus == hipSuccess && num_code_objs != 0) {
|
||||
hipStatus = hipErrorNoBinaryForGpu;
|
||||
|
||||
// Leave it for debug purpose in uncompressed mode.
|
||||
if (!isCompressed) {
|
||||
LogPrintfError("%s",
|
||||
"hipErrorNoBinaryForGpu: Unable to find code object for all current devices!");
|
||||
LogPrintfError("%s", " Devices:");
|
||||
for (size_t i = 0; i < agent_triple_target_ids.size(); i++) {
|
||||
LogPrintfError(" %s - [%s]", agent_triple_target_ids[i].c_str(),
|
||||
((code_objs[i].first) ? "Found" : "Not Found"));
|
||||
}
|
||||
const auto obheader = reinterpret_cast<const __ClangOffloadBundleUncompressedHeader*>(data);
|
||||
const auto* desc = &obheader->desc[0];
|
||||
LogPrintfError("%s", " Bundled Code Objects:");
|
||||
for (uint64_t i = 0; i < obheader->numOfCodeObjects; ++i,
|
||||
desc = reinterpret_cast<const __ClangOffloadBundleInfo*>(
|
||||
reinterpret_cast<uintptr_t>(&desc->bundleEntryId[0]) +
|
||||
desc->bundleEntryIdSize)) {
|
||||
std::string bundleEntryId{desc->bundleEntryId, desc->bundleEntryIdSize};
|
||||
const void* image =
|
||||
reinterpret_cast<const void*>(reinterpret_cast<uintptr_t>(obheader) + desc->offset);
|
||||
|
||||
std::string co_triple_target_id;
|
||||
bool valid_co = getTripleTargetID(bundleEntryId, image, co_triple_target_id);
|
||||
if (valid_co) {
|
||||
LogPrintfError(" %s - [Code object targetID is %s]", bundleEntryId.c_str(),
|
||||
co_triple_target_id.c_str());
|
||||
} else {
|
||||
LogPrintfError(" %s - [Unsupported]", bundleEntryId.c_str());
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Cleanup
|
||||
if (actionInfoUnbundle.handle) {
|
||||
comgrStatus = amd::Comgr::destroy_action_info(actionInfoUnbundle);
|
||||
|
||||
Reference in New Issue
Block a user