Minor optimizations of debug::WriteCFISection.
The method might be passed method infos without CFI in some cases.
Use the sorting phase as a chance to filter them out.
This makes sure we do not allocate memory and sort methods if there
is in fact no work to do.
Also change the sort to stable - for the sake of determinism.
Change-Id: I97d57d77e8b709d0d49d6971f66b955efcbb57b0
diff --git a/compiler/debug/elf_debug_frame_writer.h b/compiler/debug/elf_debug_frame_writer.h
index badbd93..f9d33c1 100644
--- a/compiler/debug/elf_debug_frame_writer.h
+++ b/compiler/debug/elf_debug_frame_writer.h
@@ -175,18 +175,6 @@
CHECK(format == dwarf::DW_DEBUG_FRAME_FORMAT || format == dwarf::DW_EH_FRAME_FORMAT);
typedef typename ElfTypes::Addr Elf_Addr;
- if (method_infos.empty()) {
- return;
- }
-
- std::vector<uint32_t> binary_search_table;
- std::vector<uintptr_t> patch_locations;
- if (format == dwarf::DW_EH_FRAME_FORMAT) {
- binary_search_table.reserve(2 * method_infos.size());
- } else {
- patch_locations.reserve(method_infos.size());
- }
-
// The methods can be written in any order.
// Let's therefore sort them in the lexicographical order of the opcodes.
// This has no effect on its own. However, if the final .debug_frame section is
@@ -194,9 +182,14 @@
std::vector<const MethodDebugInfo*> sorted_method_infos;
sorted_method_infos.reserve(method_infos.size());
for (size_t i = 0; i < method_infos.size(); i++) {
- sorted_method_infos.push_back(&method_infos[i]);
+ if (!method_infos[i].cfi.empty() && !method_infos[i].deduped) {
+ sorted_method_infos.push_back(&method_infos[i]);
+ }
}
- std::sort(
+ if (sorted_method_infos.empty()) {
+ return;
+ }
+ std::stable_sort(
sorted_method_infos.begin(),
sorted_method_infos.end(),
[](const MethodDebugInfo* lhs, const MethodDebugInfo* rhs) {
@@ -205,6 +198,14 @@
return std::lexicographical_compare(l.begin(), l.end(), r.begin(), r.end());
});
+ std::vector<uint32_t> binary_search_table;
+ std::vector<uintptr_t> patch_locations;
+ if (format == dwarf::DW_EH_FRAME_FORMAT) {
+ binary_search_table.reserve(2 * sorted_method_infos.size());
+ } else {
+ patch_locations.reserve(sorted_method_infos.size());
+ }
+
// Write .eh_frame/.debug_frame section.
auto* cfi_section = (format == dwarf::DW_DEBUG_FRAME_FORMAT
? builder->GetDebugFrame()
@@ -221,26 +222,21 @@
buffer_address += buffer.size();
buffer.clear();
for (const MethodDebugInfo* mi : sorted_method_infos) {
- if (!mi->deduped) { // Only one FDE per unique address.
- ArrayRef<const uint8_t> opcodes = mi->cfi;
- if (!opcodes.empty()) {
- const Elf_Addr code_address = mi->code_address +
- (mi->is_code_address_text_relative ? builder->GetText()->GetAddress() : 0);
- if (format == dwarf::DW_EH_FRAME_FORMAT) {
- binary_search_table.push_back(
- dchecked_integral_cast<uint32_t>(code_address));
- binary_search_table.push_back(
- dchecked_integral_cast<uint32_t>(buffer_address));
- }
- WriteFDE(is64bit, cfi_address, cie_address,
- code_address, mi->code_size,
- opcodes, format, buffer_address, &buffer,
- &patch_locations);
- cfi_section->WriteFully(buffer.data(), buffer.size());
- buffer_address += buffer.size();
- buffer.clear();
- }
+ DCHECK(!mi->deduped);
+ DCHECK(!mi->cfi.empty());
+ const Elf_Addr code_address = mi->code_address +
+ (mi->is_code_address_text_relative ? builder->GetText()->GetAddress() : 0);
+ if (format == dwarf::DW_EH_FRAME_FORMAT) {
+ binary_search_table.push_back(dchecked_integral_cast<uint32_t>(code_address));
+ binary_search_table.push_back(dchecked_integral_cast<uint32_t>(buffer_address));
}
+ WriteFDE(is64bit, cfi_address, cie_address,
+ code_address, mi->code_size,
+ mi->cfi, format, buffer_address, &buffer,
+ &patch_locations);
+ cfi_section->WriteFully(buffer.data(), buffer.size());
+ buffer_address += buffer.size();
+ buffer.clear();
}
cfi_section->End();
}