[llvm] 69c38be - Reapply "Reland "[llvm-profgen] Add support for ETM trace decoding"" … (#194730)
via llvm-commits
llvm-commits at lists.llvm.org
Tue Apr 28 20:46:57 PDT 2026
Author: gulfemsavrun
Date: 2026-04-28T20:46:51-07:00
New Revision: 69c38be839913749fb34b681fe05a053f222e311
URL: https://github.com/llvm/llvm-project/commit/69c38be839913749fb34b681fe05a053f222e311
DIFF: https://github.com/llvm/llvm-project/commit/69c38be839913749fb34b681fe05a053f222e311.diff
LOG: Reapply "Reland "[llvm-profgen] Add support for ETM trace decoding"" … (#194730)
…(#194695)
This relands the original commit
e3bd61890e68303a33fdd33fbdd9abeda1037450 (#191584).
The original change was reverted in the following commits:
1) ec9d7d18bdfe21c30c94c02f14f3613f7b69a17b (#194087)
2) c26ae41c8765fdd073c71061f21a83a161598901 (#194695)
This reland incorporates the following fixes:
1) Change LLVM_ENABLE_OPENCSD to default to OFF (opt-in).
2) Parse the OpenCSD version from ocsd_if_version.h and gate OpenCSD
support on a minimum version of 1.5.4 to avoid compilation errors.
3) Refactor the test feature detection to use a
configure-time LLVM_HAVE_OPENCSD variable instead of runtime checks in
lit.local.cfg.
Added:
llvm/include/llvm/ProfileData/ETMTraceDecoder.h
llvm/lib/ProfileData/ETMTraceDecoder.cpp
llvm/test/tools/llvm-profgen/Inputs/etm-opencsd.yaml
llvm/test/tools/llvm-profgen/etm-arch.test
llvm/test/tools/llvm-profgen/etm-opencsd.test
Modified:
llvm/docs/CommandGuide/llvm-profgen.rst
llvm/docs/ReleaseNotes.md
llvm/lib/ProfileData/CMakeLists.txt
llvm/test/CMakeLists.txt
llvm/test/lit.cfg.py
llvm/test/lit.site.cfg.py.in
llvm/tools/llvm-profgen/PerfReader.cpp
llvm/tools/llvm-profgen/PerfReader.h
llvm/tools/llvm-profgen/ProfiledBinary.cpp
llvm/tools/llvm-profgen/ProfiledBinary.h
llvm/tools/llvm-profgen/llvm-profgen.cpp
Removed:
################################################################################
diff --git a/llvm/docs/CommandGuide/llvm-profgen.rst b/llvm/docs/CommandGuide/llvm-profgen.rst
index 90ad729ec10b2..006cdfecf4f88 100644
--- a/llvm/docs/CommandGuide/llvm-profgen.rst
+++ b/llvm/docs/CommandGuide/llvm-profgen.rst
@@ -24,6 +24,11 @@ At least one of the following commands are required:
Path of perf-script trace created by Linux perf tool with `script`
command(the raw perf.data should be profiled with -b).
+.. option:: --etm=<string>
+
+ Path of the ETM trace file created by ARM CoreSight trace tools.
+ Requires the OpenCSD library version 1.5.4 or higher to be enabled during the build.
+
.. option:: --perfdata=<perfdata>, --pd
Path of raw perf data created by Linux perf tool (it should be profiled
@@ -41,7 +46,7 @@ At least one of the following commands are required:
.. note::
Only one of ``--perfscript``, ``--perfdata``, ``--unsymbolized-profile``,
- or ``--llvm-sample-profile`` may be specified at a time.
+ ``--llvm-sample-profile``, or ``--etm`` may be specified at a time.
.. option:: --binary=<string[,string,...]>
@@ -79,6 +84,11 @@ OPTIONS
Path of debug info binary. ``llvm-profgen`` will load the DWARF info from
it instead of the executable binary.
+
+.. option:: --target-triple=<triple>
+
+ Override the target triple for the binary. This is useful for ETM trace
+ decoding to specify the correct Arm M-profile target.
.. option:: --pid
diff --git a/llvm/docs/ReleaseNotes.md b/llvm/docs/ReleaseNotes.md
index d540cbf31211e..fc0114342e38e 100644
--- a/llvm/docs/ReleaseNotes.md
+++ b/llvm/docs/ReleaseNotes.md
@@ -245,6 +245,8 @@ Changes to the Debug Info
Changes to the LLVM tools
-------------------------
+* `llvm-profgen` now supports ETM trace decoding using the OpenCSD library for Cortex-M targets. OpenCSD version 1.5.4 or higher is required.
+
* `llvm-objcopy` no longer corrupts the symbol table when `--update-section` is called for ELF files.
* `FileCheck` option `-check-prefix` now accepts a comma-separated list of
prefixes, making it an alias of the existing `-check-prefixes` option.
diff --git a/llvm/include/llvm/ProfileData/ETMTraceDecoder.h b/llvm/include/llvm/ProfileData/ETMTraceDecoder.h
new file mode 100644
index 0000000000000..86599fa2fb6d2
--- /dev/null
+++ b/llvm/include/llvm/ProfileData/ETMTraceDecoder.h
@@ -0,0 +1,46 @@
+//===-- ETMTraceDecoder.h - ETM Trace Decoder -------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_PROFILEDATA_ETMTRACEDECODER_H
+#define LLVM_PROFILEDATA_ETMTRACEDECODER_H
+
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Support/Error.h"
+
+#include <cstdint>
+#include <memory>
+
+namespace llvm {
+
+class Triple;
+namespace object {
+class Binary;
+}
+
+class ETMDecoder {
+public:
+ virtual ~ETMDecoder() = default;
+
+ class Callback {
+ public:
+ virtual ~Callback() = default;
+ virtual void processInstructionRange(uint64_t Start, uint64_t End) = 0;
+ };
+
+ static Expected<std::unique_ptr<ETMDecoder>>
+ create(const object::Binary &Binary, const Triple &TargetTriple,
+ uint8_t TraceID = 0x10);
+
+ virtual Error processTrace(ArrayRef<uint8_t> TraceData,
+ Callback &TraceCallback) = 0;
+};
+
+} // namespace llvm
+
+#endif // LLVM_PROFILEDATA_ETMTRACEDECODER_H
diff --git a/llvm/lib/ProfileData/CMakeLists.txt b/llvm/lib/ProfileData/CMakeLists.txt
index c2366dc0ae03e..08e5e760252e1 100644
--- a/llvm/lib/ProfileData/CMakeLists.txt
+++ b/llvm/lib/ProfileData/CMakeLists.txt
@@ -1,5 +1,6 @@
add_llvm_component_library(LLVMProfileData
DataAccessProf.cpp
+ ETMTraceDecoder.cpp
GCOV.cpp
IndexedMemProfData.cpp
InstrProf.cpp
@@ -39,4 +40,60 @@ add_llvm_component_library(LLVMProfileData
TargetParser
)
+set(LLVM_ENABLE_OPENCSD "OFF" CACHE STRING "Enable OpenCSD support in LLVMProfileData")
+
+set(LLVM_HAVE_OPENCSD OFF CACHE BOOL "Is OpenCSD available" FORCE)
+
+# Minimum OpenCSD release that exposes every symbol referenced by
+# ETMTraceDecoder.cpp. OCSD_OPFLG_CHK_RANGE_CONTINUE was introduced in
+# OpenCSD 1.5.4; older releases (e.g. the 1.2.0 shipped by Debian/Ubuntu)
+# build the OpenCSD library but cannot satisfy our use of the C API.
+set(LLVM_OPENCSD_MIN_VERSION "1.5.4")
+
+if(NOT LLVM_ENABLE_OPENCSD STREQUAL "OFF")
+ find_library(OPENCSD_LIB NAMES opencsd_c_api opencsd)
+ find_path(OPENCSD_INCLUDE NAMES opencsd/ocsd_if_types.h)
+
+ set(OPENCSD_VERSION "")
+ set(OPENCSD_VERSION_OK OFF)
+ if(OPENCSD_LIB AND OPENCSD_INCLUDE)
+ # The header exposes an OCSD_VER_STRING macro of the form "M.m.p"
+ # which we extract for a robust version comparison.
+ file(READ "${OPENCSD_INCLUDE}/opencsd/ocsd_if_version.h" _opencsd_ver_h)
+ string(REGEX MATCH "OCSD_VER_STRING[ \t]+\"([0-9]+\\.[0-9]+\\.[0-9]+)\""
+ _ver_match "${_opencsd_ver_h}")
+ set(OPENCSD_VERSION "${CMAKE_MATCH_1}")
+ if(OPENCSD_VERSION AND
+ NOT OPENCSD_VERSION VERSION_LESS LLVM_OPENCSD_MIN_VERSION)
+ set(OPENCSD_VERSION_OK ON)
+ endif()
+ endif()
+
+ if(OPENCSD_LIB AND OPENCSD_INCLUDE AND OPENCSD_VERSION_OK)
+ target_compile_definitions(LLVMProfileData PRIVATE HAVE_OPENCSD)
+ target_include_directories(LLVMProfileData PRIVATE ${OPENCSD_INCLUDE})
+ target_link_libraries(LLVMProfileData PRIVATE ${OPENCSD_LIB})
+ message(STATUS "LLVMProfileData: OpenCSD support enabled (found ${OPENCSD_VERSION}).")
+ set(LLVM_HAVE_OPENCSD ON CACHE BOOL "Is OpenCSD available" FORCE)
+ elseif(OPENCSD_LIB AND OPENCSD_INCLUDE)
+ if(LLVM_ENABLE_OPENCSD STREQUAL "ON")
+ message(FATAL_ERROR
+ "OpenCSD found at ${OPENCSD_LIB} but its version "
+ "(${OPENCSD_VERSION}) is older than the required "
+ "${LLVM_OPENCSD_MIN_VERSION}. Install a newer OpenCSD or "
+ "configure with -DLLVM_ENABLE_OPENCSD=OFF.")
+ else()
+ message(STATUS
+ "LLVMProfileData: OpenCSD ${OPENCSD_VERSION} is older than "
+ "${LLVM_OPENCSD_MIN_VERSION}; ETM decoding support disabled.")
+ endif()
+ else()
+ if(LLVM_ENABLE_OPENCSD STREQUAL "ON")
+ message(FATAL_ERROR "OpenCSD enabled but library or headers not found.")
+ else()
+ message(STATUS "LLVMProfileData: OpenCSD not found; ETM decoding support disabled.")
+ endif()
+ endif()
+endif()
+
add_subdirectory(Coverage)
diff --git a/llvm/lib/ProfileData/ETMTraceDecoder.cpp b/llvm/lib/ProfileData/ETMTraceDecoder.cpp
new file mode 100644
index 0000000000000..3d6a9846ac25c
--- /dev/null
+++ b/llvm/lib/ProfileData/ETMTraceDecoder.cpp
@@ -0,0 +1,251 @@
+//===-- ETMTraceDecoder.cpp - ETM Trace Decoder -----------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/ProfileData/ETMTraceDecoder.h"
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/Object/ELFObjectFile.h"
+#include "llvm/Object/ObjectFile.h"
+#include "llvm/Support/Error.h"
+#include "llvm/TargetParser/ARMTargetParser.h"
+
+#ifdef HAVE_OPENCSD
+#include "opencsd/c_api/opencsd_c_api.h"
+
+namespace llvm {
+
+namespace {
+
+class HardwareTraceConfig {
+public:
+ virtual ~HardwareTraceConfig() = default;
+};
+
+class ETMTraceConfig : public HardwareTraceConfig {
+public:
+ ocsd_etmv4_cfg Cfg{};
+ uint8_t TraceID;
+
+ ETMTraceConfig(const Triple &TargetTriple, uint8_t TraceID)
+ : TraceID(TraceID) {
+ ocsd_arch_version_t ArchVer = ARCH_UNKNOWN;
+ if (TargetTriple.isArmMClass()) {
+ unsigned ArchVersion = ARM::parseArchVersion(TargetTriple.getArchName());
+ if (ArchVersion >= 8)
+ ArchVer = ARCH_V8;
+ else if (ArchVersion == 7)
+ ArchVer = ARCH_V7;
+ else
+ // For version 6 (Cortex-M0) and others.
+ ArchVer = ARCH_UNKNOWN;
+ }
+ // Initialize the decoder for Arm M-profile targets.
+ Cfg.arch_ver = ArchVer;
+ Cfg.core_prof = profile_CortexM;
+
+ // The CoreSight Trace ID (CSID) is a hardware-assigned 7-bit identifier
+ // used to route trace data.
+ Cfg.reg_traceidr = TraceID;
+ }
+
+ Error validate() const {
+ if (Cfg.arch_ver == ARCH_UNKNOWN)
+ return createStringError(
+ inconvertibleErrorCode(),
+ "OpenCSD: Unsupported processor architecture. Only Arm M-profile "
+ "(Cortex-M) with ETM support is currently supported.");
+ return Error::success();
+ }
+};
+
+class ETMDecoderImpl : public ETMDecoder {
+ dcd_tree_handle_t DcdTree = 0;
+ const object::Binary &Binary;
+ const Triple &TargetTriple;
+
+ // Trace processing and Callback handling.
+ static ocsd_datapath_resp_t
+ processTrace(const void *PContext, const ocsd_trc_index_t /*IndexSOP*/,
+ const uint8_t /*TrcChanID*/,
+ const ocsd_generic_trace_elem *Element) {
+ auto *Decoder = static_cast<ETMDecoderImpl *>(const_cast<void *>(PContext));
+ if (!Decoder || !Element)
+ return OCSD_RESP_FATAL_SYS_ERR;
+
+ // Process instruction ranges reconstructed by the decoder.
+ if (Element->elem_type == OCSD_GEN_TRC_ELEM_INSTR_RANGE) {
+ uint64_t Start = Element->st_addr;
+ uint64_t End = Element->en_addr;
+ if (End > Start) {
+ // OpenCSD ranges are exclusive at the end [Start, End).
+ // llvm-profgen range counters expect inclusive bounds [Start, End].
+ // Adjust the exclusive end address provided by OpenCSD to include
+ // the last executed instruction within the reported range.
+ Decoder->CurrentCallback->processInstructionRange(Start, End - 1);
+ }
+ }
+ return OCSD_RESP_CONT;
+ }
+
+ Callback *CurrentCallback = nullptr;
+
+ // Iterate through the ELF program headers to collect all executable LOAD
+ // segments. These are registered as a single transaction to the OpenCSD
+ // memory manager to prevent overlap/collision errors between
diff erent
+ // memory regions.
+ Error mapELFSegments(dcd_tree_handle_t DcdTree,
+ const object::Binary &SourceBin) {
+ SmallVector<ocsd_file_mem_region_t, 4> Regions;
+ auto ProcessHeaders = [&](const auto &ElfFile) {
+ auto ProgramHeaders = ElfFile.program_headers();
+ if (!ProgramHeaders)
+ return;
+
+ for (const auto &Phdr : *ProgramHeaders) {
+ if (Phdr.p_type == llvm::ELF::PT_LOAD &&
+ (Phdr.p_flags & llvm::ELF::PF_X)) {
+ ocsd_file_mem_region_t Region{};
+ Region.start_address = (uint64_t)Phdr.p_vaddr;
+ Region.file_offset = (uint64_t)Phdr.p_offset;
+ Region.region_size = (uint64_t)Phdr.p_filesz;
+ Regions.push_back(Region);
+ }
+ }
+ };
+
+ if (auto *O = dyn_cast<object::ELF32LEObjectFile>(&SourceBin))
+ ProcessHeaders(O->getELFFile());
+ else if (auto *O = dyn_cast<object::ELF64LEObjectFile>(&SourceBin))
+ ProcessHeaders(O->getELFFile());
+ else if (auto *O = dyn_cast<object::ELF32BEObjectFile>(&SourceBin))
+ ProcessHeaders(O->getELFFile());
+ else if (auto *O = dyn_cast<object::ELF64BEObjectFile>(&SourceBin))
+ ProcessHeaders(O->getELFFile());
+
+ if (!Regions.empty()) {
+ std::string Path = SourceBin.getFileName().str();
+ if (ocsd_dt_add_binfile_region_mem_acc(
+ DcdTree, Regions.data(), (uint32_t)Regions.size(),
+ OCSD_MEM_SPACE_ANY, Path.c_str()) != 0) {
+ return createStringError(
+ inconvertibleErrorCode(),
+ "OpenCSD: Failed to map ELF executable segments.");
+ }
+ }
+ return Error::success();
+ }
+
+public:
+ uint8_t TraceID;
+
+ ETMDecoderImpl(const object::Binary &Binary, const Triple &Triple,
+ uint8_t TraceID)
+ : Binary(Binary), TargetTriple(Triple), TraceID(TraceID) {}
+
+ ~ETMDecoderImpl() override {
+ if (DcdTree)
+ // Deallocate the decoder tree resources.
+ ocsd_destroy_dcd_tree(DcdTree);
+ }
+
+ // Initialize the decoder by auto-detecting the target architecture and
+ // configuring the OpenCSD decoder.
+ Error initialize() {
+ DcdTree = ocsd_create_dcd_tree(OCSD_TRC_SRC_SINGLE, 0);
+ if (!DcdTree)
+ return createStringError(inconvertibleErrorCode(),
+ "Failed to create OpenCSD decoder tree.");
+
+ // Configure and initialize the instruction-level decoder.
+ ETMTraceConfig Config(TargetTriple, TraceID);
+ if (Error E = Config.validate())
+ return E;
+
+ uint32_t Flags =
+ OCSD_CREATE_FLG_FULL_DECODER | OCSD_OPFLG_CHK_RANGE_CONTINUE;
+ if (ocsd_dt_create_decoder(DcdTree, OCSD_BUILTIN_DCD_ETMV4I, Flags,
+ (void *)&Config.Cfg, &Config.TraceID) != 0)
+ return createStringError(
+ inconvertibleErrorCode(),
+ "OpenCSD: Failed to initialize the instruction decoder.");
+
+ // Extract and map executable segments from the ELF binary.
+ if (Error E = mapELFSegments(DcdTree, Binary))
+ return E;
+
+ // Register the high-level packet callback. The 'processTrace' function
+ // will be invoked for every decoded instruction range.
+ ocsd_dt_set_gen_elem_outfn(DcdTree, processTrace, this);
+ return Error::success();
+ }
+
+ Error processTrace(ArrayRef<uint8_t> TraceData,
+ Callback &TraceCallback) override {
+ CurrentCallback = &TraceCallback;
+ // Initial reset to prime the decoder.
+ ocsd_dt_process_data(DcdTree, OCSD_OP_RESET, 0, 0, nullptr, nullptr);
+
+ const uint8_t *DataPtr = TraceData.data();
+ uint32_t TotalSize = TraceData.size();
+ uint32_t Processed = 0;
+
+ // Core Decoding Loop.
+ while (Processed < TotalSize) {
+ uint32_t Consumed = 0;
+ uint32_t Remaining = TotalSize - Processed;
+ ocsd_datapath_resp_t Response =
+ ocsd_dt_process_data(DcdTree, OCSD_OP_DATA, Processed, Remaining,
+ DataPtr + Processed, &Consumed);
+
+ if (Response == OCSD_RESP_WAIT) {
+ // Decoder buffers are full; flush to drain internal states.
+ ocsd_dt_process_data(DcdTree, OCSD_OP_FLUSH, 0, 0, nullptr, nullptr);
+ } else if (Consumed == 0 && Processed < TotalSize) {
+ // Decoder stalled; skip byte and reset to find next sync point.
+ Processed++;
+ ocsd_dt_process_data(DcdTree, OCSD_OP_RESET, 0, 0, nullptr, nullptr);
+ } else {
+ // Successfully consumed bytes of the bitstream.
+ Processed += Consumed;
+ }
+
+ if (Response >= OCSD_RESP_FATAL_INVALID_DATA)
+ return createStringError(inconvertibleErrorCode(),
+ "OpenCSD: Fatal decoding error.");
+ }
+
+ // Finalize the decoding session by flushing the EOT (End of Trace) marker.
+ ocsd_dt_process_data(DcdTree, OCSD_OP_EOT, 0, 0, nullptr, nullptr);
+ return Error::success();
+ }
+};
+} // namespace
+
+Expected<std::unique_ptr<ETMDecoder>>
+ETMDecoder::create(const object::Binary &Binary, const Triple &Triple,
+ uint8_t TraceID) {
+ auto Decoder = std::make_unique<ETMDecoderImpl>(Binary, Triple, TraceID);
+ if (Error E = Decoder->initialize())
+ return std::move(E);
+ return std::unique_ptr<ETMDecoder>(std::move(Decoder));
+}
+
+} // namespace llvm
+
+#else // !HAVE_OPENCSD
+
+namespace llvm {
+
+Expected<std::unique_ptr<ETMDecoder>>
+ETMDecoder::create(const object::Binary & /*Binary*/, const Triple & /*Triple*/,
+ uint8_t /*TraceID*/) {
+ return createStringError(inconvertibleErrorCode(), "OpenCSD not enabled.");
+}
+
+} // namespace llvm
+
+#endif // HAVE_OPENCSD
diff --git a/llvm/test/CMakeLists.txt b/llvm/test/CMakeLists.txt
index a027cd754135c..54094b02604f9 100644
--- a/llvm/test/CMakeLists.txt
+++ b/llvm/test/CMakeLists.txt
@@ -31,6 +31,7 @@ llvm_canonicalize_cmake_booleans(
LLVM_APPEND_VC_REV
LLVM_HAS_LOGF128
LLVM_ENABLE_ONDISK_CAS
+ LLVM_HAVE_OPENCSD
)
configure_lit_site_cfg(
diff --git a/llvm/test/lit.cfg.py b/llvm/test/lit.cfg.py
index 704151e0228e4..e44079aa5cd26 100644
--- a/llvm/test/lit.cfg.py
+++ b/llvm/test/lit.cfg.py
@@ -589,6 +589,9 @@ def enable_ptxas(ptxas_executable):
if config.have_tf_aot:
config.available_features.add("have_tf_aot")
+if getattr(config, "have_opencsd", False):
+ config.available_features.add("opencsd")
+
if config.have_tflite:
config.available_features.add("have_tflite")
diff --git a/llvm/test/lit.site.cfg.py.in b/llvm/test/lit.site.cfg.py.in
index 6fe48f21cc685..944779c0d7f2f 100644
--- a/llvm/test/lit.site.cfg.py.in
+++ b/llvm/test/lit.site.cfg.py.in
@@ -68,6 +68,7 @@ config.have_vc_rev = @LLVM_APPEND_VC_REV@
config.force_vc_rev = "@LLVM_FORCE_VC_REVISION@"
config.has_logf128 = @LLVM_HAS_LOGF128@
config.have_ondisk_cas = @LLVM_ENABLE_ONDISK_CAS@
+config.have_opencsd = @LLVM_HAVE_OPENCSD@
import lit.llvm
lit.llvm.initialize(lit_config, config)
diff --git a/llvm/test/tools/llvm-profgen/Inputs/etm-opencsd.yaml b/llvm/test/tools/llvm-profgen/Inputs/etm-opencsd.yaml
new file mode 100644
index 0000000000000..6824b4a9c108b
--- /dev/null
+++ b/llvm/test/tools/llvm-profgen/Inputs/etm-opencsd.yaml
@@ -0,0 +1,48 @@
+# This file contains binary data extracted from the OpenCSD repository sample tests
+# (specifically from decoder/tests/snapshots/armv8_1m_branches/).
+# It is used to verify ETM trace decoding in llvm-profgen.
+--- !ELF
+FileHeader:
+ Class: ELFCLASS32
+ Data: ELFDATA2LSB
+ Type: ET_DYN
+ Machine: EM_ARM
+ Flags: [ EF_ARM_EABI_VER5 ]
+ProgramHeaders:
+ - Type: PT_LOAD
+ Flags: [ PF_X, PF_R ]
+ VAddr: 0x1E008
+ FirstSec: .text1
+ LastSec: .text1
+ - Type: PT_LOAD
+ Flags: [ PF_X, PF_R ]
+ VAddr: 0x4F9A02
+ FirstSec: .text2
+ LastSec: .text2
+Sections:
+ - Name: .text1
+ Type: SHT_PROGBITS
+ Flags: [ SHF_ALLOC, SHF_EXECINSTR ]
+ Address: 0x1E008
+ AddressAlign: 0x4
+ Content: 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
+ - Name: .text2
+ Type: SHT_PROGBITS
+ Flags: [ SHF_ALLOC, SHF_EXECINSTR ]
+ Address: 0x4F9A02
+ AddressAlign: 0x4
+ Content: 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
+ - Name: .ARM.attributes
+ Type: SHT_ARM_ATTRIBUTES
+ Content: 4140000000616561626900013600000043322E30390005636F727465782D6D3300060A074D080009020E0011011204140115001703180119011A021E0622012601
+Symbols:
+ - Name: foo
+ Type: STT_FUNC
+ Section: .text1
+ Value: 0x1E008
+ Size: 436
+ - Name: bar
+ Type: STT_FUNC
+ Section: .text2
+ Value: 0x4F9A02
+ Size: 2776
diff --git a/llvm/test/tools/llvm-profgen/etm-arch.test b/llvm/test/tools/llvm-profgen/etm-arch.test
new file mode 100644
index 0000000000000..418a91fec265c
--- /dev/null
+++ b/llvm/test/tools/llvm-profgen/etm-arch.test
@@ -0,0 +1,81 @@
+; REQUIRES: opencsd, arm-registered-target
+; RUN: split-file %s %t
+; RUN: printf '\x80' > %t.etm
+
+; Test 1: ARM binary with Cortex-M attributes should pass.
+; RUN: yaml2obj %t/arm-valid.yaml -o %t.arm.valid.bin
+; RUN: llvm-profgen --etm=%t.etm --binary=%t.arm.valid.bin --output=%t.out 2>&1 | FileCheck %s --check-prefix=ARM-VALID
+
+; Test 2: ARM binary without Cortex-M attributes should fail.
+; RUN: yaml2obj %t/arm-no-attr.yaml -o %t.arm.noattr.bin
+; RUN: not llvm-profgen --etm=%t.etm --binary=%t.arm.noattr.bin --output=%t.out 2>&1 | FileCheck %s --check-prefix=ARM-NOT-VALID
+
+; Test 3: Non-ARM binary (X86) should fail.
+; RUN: yaml2obj %t/x86.yaml -o %t.x86.bin
+; RUN: not llvm-profgen --etm=%t.etm --binary=%t.x86.bin --output=%t.out 2>&1 | FileCheck %s --check-prefix=X86
+
+; ARM-VALID-NOT: OpenCSD: Unsupported processor architecture
+
+; ARM-NOT-VALID: OpenCSD: Unsupported processor architecture. Only Arm M-profile (Cortex-M) with ETM support is currently supported.
+
+; X86: OpenCSD: Unsupported processor architecture. Only Arm M-profile (Cortex-M) with ETM support is currently supported.
+
+;--- arm-valid.yaml
+--- !ELF
+FileHeader:
+ Class: ELFCLASS32
+ Data: ELFDATA2LSB
+ Type: ET_DYN
+ Machine: EM_ARM
+ProgramHeaders:
+ - Type: PT_LOAD
+ Flags: [ PF_X, PF_R ]
+ VAddr: 0x1000
+Sections:
+ - Name: .text
+ Type: SHT_PROGBITS
+ Flags: [ SHF_ALLOC, SHF_EXECINSTR ]
+ Address: 0x1000
+ AddressAlign: 0x4
+ Content: C3
+ - Name: .ARM.attributes
+ Type: SHT_ARM_ATTRIBUTES
+ Content: 4140000000616561626900013600000043322E30390005636F727465782D6D3300060A074D080009020E0011011204140115001703180119011A021E0622012601
+
+;--- arm-no-attr.yaml
+--- !ELF
+FileHeader:
+ Class: ELFCLASS32
+ Data: ELFDATA2LSB
+ Type: ET_DYN
+ Machine: EM_ARM
+ProgramHeaders:
+ - Type: PT_LOAD
+ Flags: [ PF_X, PF_R ]
+ VAddr: 0x1000
+Sections:
+ - Name: .text
+ Type: SHT_PROGBITS
+ Flags: [ SHF_ALLOC, SHF_EXECINSTR ]
+ Address: 0x1000
+ AddressAlign: 0x4
+ Content: C3
+
+;--- x86.yaml
+--- !ELF
+FileHeader:
+ Class: ELFCLASS64
+ Data: ELFDATA2LSB
+ Type: ET_DYN
+ Machine: EM_X86_64
+ProgramHeaders:
+ - Type: PT_LOAD
+ Flags: [ PF_X, PF_R ]
+ VAddr: 0x1000
+Sections:
+ - Name: .text
+ Type: SHT_PROGBITS
+ Flags: [ SHF_ALLOC, SHF_EXECINSTR ]
+ Address: 0x1000
+ AddressAlign: 0x4
+ Content: C3
diff --git a/llvm/test/tools/llvm-profgen/etm-opencsd.test b/llvm/test/tools/llvm-profgen/etm-opencsd.test
new file mode 100644
index 0000000000000..5148455684a88
--- /dev/null
+++ b/llvm/test/tools/llvm-profgen/etm-opencsd.test
@@ -0,0 +1,19 @@
+; REQUIRES: opencsd, arm-registered-target
+; RUN: yaml2obj %S/Inputs/etm-opencsd.yaml -o %t.bin
+
+; Verify ETM trace decoding using OpenCSD with inlined trace data.
+; RUN: printf "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x80\x01\x00\x83\x01\x9a\x4f\x00\x00\x96\x81\x9a\xfc\xfe\xfa\xf8\xfd\xfd\xd6\xfd\xfd\xc1" > %t.trace
+
+; Run the generator to produce the output file.
+; RUN: llvm-profgen --etm=%t.trace --binary=%t.bin --output=%t.prof --format=text
+; RUN: ls %t.prof
+
+; Verify target-triple override works.
+; RUN: llvm-profgen --etm=%t.trace --binary=%t.bin --output=%t2.prof --format=text --target-triple=thumbv7m-none-eabi
+; RUN: ls %t2.prof
+
+; Verify error handling when no synchronization header (0x80) is found.
+; RUN: printf "\x00\x00\x00" > %t.nosync
+; RUN: not llvm-profgen --etm=%t.nosync --binary=%t.bin --output=%t3.prof 2>&1 | FileCheck %s --check-prefix=NOSYNC
+
+; NOSYNC: error: No synchronization header (0x80) found in the bitstream.
diff --git a/llvm/tools/llvm-profgen/PerfReader.cpp b/llvm/tools/llvm-profgen/PerfReader.cpp
index 7c51c08afdbe0..101530b24cdba 100644
--- a/llvm/tools/llvm-profgen/PerfReader.cpp
+++ b/llvm/tools/llvm-profgen/PerfReader.cpp
@@ -11,12 +11,14 @@
#include "ProfileGenerator.h"
#include "llvm/ADT/SmallString.h"
#include "llvm/DebugInfo/Symbolize/SymbolizableModule.h"
+#include "llvm/ProfileData/ETMTraceDecoder.h"
#include "llvm/Support/FileSystem.h"
#include "llvm/Support/LineIterator.h"
#include "llvm/Support/MemoryBuffer.h"
#include "llvm/Support/Process.h"
#include "llvm/Support/Timer.h"
#include "llvm/Support/ToolOutputFile.h"
+#include "llvm/TargetParser/Triple.h"
#define DEBUG_TYPE "perf-reader"
@@ -351,32 +353,31 @@ bool VirtualUnwinder::unwind(const PerfSample *Sample, uint64_t Repeat) {
}
std::unique_ptr<PerfReaderBase>
-PerfReaderBase::create(ProfiledBinary *Binary, PerfInputFile &PerfInput,
+PerfReaderBase::create(ProfiledBinary *Binary, InputFile &Input,
std::optional<int32_t> PIDFilter) {
std::unique_ptr<PerfReaderBase> PerfReader;
- if (PerfInput.Format == PerfFormat::UnsymbolizedProfile) {
+ if (Input.Format == InputFormat::UnsymbolizedProfile) {
PerfReader.reset(
- new UnsymbolizedProfileReader(Binary, PerfInput.InputFile));
+ new UnsymbolizedProfileReader(Binary, Input.InputFilePath));
return PerfReader;
}
// For perf data input, we need to convert them into perf script first.
// If this is a kernel perf file, there is no need for retrieving PIDs.
- if (PerfInput.Format == PerfFormat::PerfData)
- PerfInput = PerfScriptReader::convertPerfDataToTrace(
- Binary, Binary->isKernel(), PerfInput, PIDFilter);
+ if (Input.Format == InputFormat::PerfData)
+ Input = PerfScriptReader::convertPerfDataToTrace(Binary, Binary->isKernel(),
+ Input, PIDFilter);
- assert((PerfInput.Format == PerfFormat::PerfScript) &&
+ assert((Input.Format == InputFormat::PerfScript) &&
"Should be a perfscript!");
- PerfInput.Content =
- PerfScriptReader::checkPerfScriptType(PerfInput.InputFile);
- if (PerfInput.Content == PerfContent::LBRStack) {
+ Input.Content = PerfScriptReader::checkPerfScriptType(Input.InputFilePath);
+ if (Input.Content == PerfContent::LBRStack) {
PerfReader.reset(
- new HybridPerfReader(Binary, PerfInput.InputFile, PIDFilter));
- } else if (PerfInput.Content == PerfContent::LBR) {
- PerfReader.reset(new LBRPerfReader(Binary, PerfInput.InputFile, PIDFilter));
+ new HybridPerfReader(Binary, Input.InputFilePath, PIDFilter));
+ } else if (Input.Content == PerfContent::LBR) {
+ PerfReader.reset(new LBRPerfReader(Binary, Input.InputFilePath, PIDFilter));
} else {
exitWithError("Unsupported perfscript!");
}
@@ -455,11 +456,11 @@ Error PerfReaderBase::parseDataAccessPerfTraces(
return Error::success();
}
-PerfInputFile
+InputFile
PerfScriptReader::convertPerfDataToTrace(ProfiledBinary *Binary, bool SkipPID,
- PerfInputFile &File,
+ InputFile &File,
std::optional<int32_t> PIDFilter) {
- StringRef PerfData = File.InputFile;
+ StringRef PerfData = File.InputFilePath;
// Run perf script to retrieve PIDs matching binary we're interested in.
auto PerfExecutable = sys::Process::FindInEnvPath("PATH", "perf");
if (!PerfExecutable) {
@@ -521,7 +522,7 @@ PerfScriptReader::convertPerfDataToTrace(ProfiledBinary *Binary, bool SkipPID,
}
sys::ExecuteAndWait(PerfPath, ScriptSampleArgs, std::nullopt, Redirects);
- return {std::string(PerfTraceFile), PerfFormat::PerfScript,
+ return {std::string(PerfTraceFile), InputFormat::PerfScript,
PerfContent::UnknownContent};
}
@@ -1423,5 +1424,59 @@ void PerfScriptReader::parsePerfTraces() {
SmallVector<CleanupInstaller, 2> PerfScriptReader::TempFileCleanups;
+void ETMReader::recordProcessedRange(uint64_t Start, uint64_t End,
+ uint64_t Count) {
+ assert(!Counters.empty() && "Counters should not be empty!");
+ auto &Counter = Counters.begin()->second;
+ Counter.recordRangeCount(Start, End, Count);
+}
+
+class ETMCallback : public ETMDecoder::Callback {
+ ETMReader *Reader;
+
+public:
+ ETMCallback(ETMReader *R) : Reader(R) {}
+ void processInstructionRange(uint64_t Start, uint64_t End) override {
+ Reader->recordProcessedRange(Start, End, 1);
+ }
+};
+
+void ETMReader::parseETMTraces() {
+ auto BufferOrErr = MemoryBuffer::getFile(TraceFile);
+ if (std::error_code EC = BufferOrErr.getError())
+ exitWithError("Could not open ETM trace file: " + EC.message());
+
+ ArrayRef<uint8_t> Data(
+ reinterpret_cast<const uint8_t *>((*BufferOrErr)->getBufferStart()),
+ (*BufferOrErr)->getBufferSize());
+
+ // There is no context for ETM instruction traces.
+ // Initialize the SampleCounters map with a single empty context key
+ // to aggregate all instruction hits into a global bucket.
+ auto Key = std::make_shared<StringBasedCtxKey>();
+ Counters.emplace(Hashable<ContextKey>(Key), SampleCounter());
+
+ // The protocol utilizes a 0x80 byte as an initial synchronization header.
+ // Perform a manual search for this sync point to discard any leading
+ // padding or truncated packets before decoding begins.
+ size_t StartIdx = 0;
+ while (StartIdx < Data.size() && Data[StartIdx] != 0x80)
+ StartIdx++;
+ if (StartIdx >= Data.size())
+ exitWithError("No synchronization header (0x80) found in the bitstream.");
+ ArrayRef<uint8_t> TraceSlice = Data.slice(StartIdx);
+
+ auto DecoderOrErr = ETMDecoder::create(
+ Binary->getBinary(), Binary->getTriple(), static_cast<uint8_t>(TraceID));
+
+ if (!DecoderOrErr)
+ exitWithError(toString(DecoderOrErr.takeError()));
+ auto Decoder = std::move(*DecoderOrErr);
+
+ ETMCallback CB(this);
+ if (Error E = Decoder->processTrace(TraceSlice, CB))
+ exitWithError(toString(std::move(E)));
+}
+
} // end namespace sampleprof
} // end namespace llvm
diff --git a/llvm/tools/llvm-profgen/PerfReader.h b/llvm/tools/llvm-profgen/PerfReader.h
index 9011e0b644261..d7b527e1fe549 100644
--- a/llvm/tools/llvm-profgen/PerfReader.h
+++ b/llvm/tools/llvm-profgen/PerfReader.h
@@ -58,12 +58,12 @@ class TraceStream {
};
// The type of input format.
-enum PerfFormat {
+enum InputFormat {
UnknownFormat = 0,
PerfData = 1, // Raw linux perf.data.
PerfScript = 2, // Perf script create by `perf script` command.
UnsymbolizedProfile = 3, // Unsymbolized profile generated by llvm-profgen.
-
+ ETMFormat = 4, // Raw ETM format.
};
// The type of perfscript content.
@@ -73,9 +73,9 @@ enum PerfContent {
LBRStack = 2, // Hybrid sample including call stack and LBR stack.
};
-struct PerfInputFile {
- std::string InputFile;
- PerfFormat Format = PerfFormat::UnknownFormat;
+struct InputFile {
+ std::string InputFilePath;
+ InputFormat Format = InputFormat::UnknownFormat;
PerfContent Content = PerfContent::UnknownContent;
};
@@ -573,7 +573,7 @@ class PerfReaderBase {
};
virtual ~PerfReaderBase() = default;
static std::unique_ptr<PerfReaderBase>
- create(ProfiledBinary *Binary, PerfInputFile &PerfInput,
+ create(ProfiledBinary *Binary, InputFile &Input,
std::optional<int32_t> PIDFilter);
// Entry of the reader to parse multiple perf traces
@@ -620,9 +620,9 @@ class PerfScriptReader : public PerfReaderBase {
MMapEvent &MMap);
// Generate perf script from perf data
- static PerfInputFile convertPerfDataToTrace(ProfiledBinary *Binary,
- bool SkipPID, PerfInputFile &File,
- std::optional<int32_t> PIDFilter);
+ static InputFile convertPerfDataToTrace(ProfiledBinary *Binary, bool SkipPID,
+ InputFile &File,
+ std::optional<int32_t> PIDFilter);
// Extract perf script type by peaking at the input
static PerfContent checkPerfScriptType(StringRef FileName);
@@ -750,6 +750,21 @@ class UnsymbolizedProfileReader : public PerfReaderBase {
std::unordered_set<std::string> ContextStrSet;
};
+class ETMReader {
+public:
+ ETMReader(ProfiledBinary *Binary, StringRef TraceFile, uint8_t TraceID)
+ : Binary(Binary), TraceFile(TraceFile), TraceID(TraceID) {}
+ void parseETMTraces();
+ void recordProcessedRange(uint64_t Start, uint64_t End, uint64_t Count);
+ const ContextSampleCounterMap &getSampleCounters() const { return Counters; }
+
+private:
+ ProfiledBinary *Binary = nullptr;
+ StringRef TraceFile;
+ uint8_t TraceID;
+ ContextSampleCounterMap Counters;
+};
+
} // end namespace sampleprof
} // end namespace llvm
diff --git a/llvm/tools/llvm-profgen/ProfiledBinary.cpp b/llvm/tools/llvm-profgen/ProfiledBinary.cpp
index 6d355b6141e2c..fb7dd1d01ba93 100644
--- a/llvm/tools/llvm-profgen/ProfiledBinary.cpp
+++ b/llvm/tools/llvm-profgen/ProfiledBinary.cpp
@@ -195,7 +195,6 @@ ProfiledBinary::ProfiledBinary(const StringRef ExeBinPath,
SymbolizerPath = DebugBinPath.empty() ? ExeBinPath : DebugBinPath;
if (InferMissingFrames)
MissingContextInferrer = std::make_unique<MissingFrameInferrer>(this);
- load();
}
ProfiledBinary::~ProfiledBinary() = default;
@@ -228,9 +227,9 @@ void ProfiledBinary::warnNoFuncEntry() {
"inconsistent name from symbol table and dwarf info.");
}
-void ProfiledBinary::load() {
+void ProfiledBinary::load(StringRef TripleStr) {
// Attempt to open the binary.
- OwningBinary<Binary> OBinary = unwrapOrError(createBinary(Path), Path);
+ OBinary = unwrapOrError(createBinary(Path), Path);
Binary &ExeBinary = *OBinary.getBinary();
IsCOFF = isa<COFFObjectFile>(&ExeBinary);
@@ -238,7 +237,10 @@ void ProfiledBinary::load() {
exitWithError("not a valid ELF/COFF image", Path);
auto *Obj = cast<ObjectFile>(&ExeBinary);
- TheTriple = Obj->makeTriple();
+ if (!TripleStr.empty())
+ TheTriple = Triple(TripleStr);
+ else
+ TheTriple = Obj->makeTriple();
LLVM_DEBUG(dbgs() << "Loading " << Path << "\n");
diff --git a/llvm/tools/llvm-profgen/ProfiledBinary.h b/llvm/tools/llvm-profgen/ProfiledBinary.h
index af8d2dbb8eb06..4a2c4cec8b49f 100644
--- a/llvm/tools/llvm-profgen/ProfiledBinary.h
+++ b/llvm/tools/llvm-profgen/ProfiledBinary.h
@@ -200,6 +200,8 @@ struct MMapEvent {
};
class ProfiledBinary {
+ // The executable binary file.
+ object::OwningBinary<object::Binary> OBinary;
// Absolute path of the executable binary.
std::string Path;
// Path of the debug info binary.
@@ -414,6 +416,11 @@ class ProfiledBinary {
SampleContextFrameVector symbolize(const InstructionPointer &IP,
bool UseCanonicalFnName = false,
bool UseProbeDiscriminator = false);
+
+public:
+ ProfiledBinary(const StringRef ExeBinPath, const StringRef DebugBinPath);
+ ~ProfiledBinary();
+
/// Decode the interesting parts of the binary and build internal data
/// structures. On high level, the parts of interest are:
/// 1. Text sections, including the main code section and the PLT
@@ -421,11 +428,7 @@ class ProfiledBinary {
/// 2. The .debug_line section, used by Dwarf-based profile generation.
/// 3. Pseudo probe related sections, used by probe-based profile
/// generation.
- void load();
-
-public:
- ProfiledBinary(const StringRef ExeBinPath, const StringRef DebugBinPath);
- ~ProfiledBinary();
+ void load(StringRef TripleStr = "");
/// Symbolize an address and return the symbol name. The returned StringRef is
/// owned by this ProfiledBinary object.
@@ -435,6 +438,8 @@ class ProfiledBinary {
StringRef getPath() const { return Path; }
StringRef getName() const { return llvm::sys::path::filename(Path); }
+ const Triple &getTriple() const { return TheTriple; }
+ const object::Binary &getBinary() const { return *OBinary.getBinary(); }
uint64_t getBaseAddress() const { return BaseAddress; }
void setBaseAddress(uint64_t Address) { BaseAddress = Address; }
diff --git a/llvm/tools/llvm-profgen/llvm-profgen.cpp b/llvm/tools/llvm-profgen/llvm-profgen.cpp
index 821e3e4613c7c..cba5f1cd3d878 100644
--- a/llvm/tools/llvm-profgen/llvm-profgen.cpp
+++ b/llvm/tools/llvm-profgen/llvm-profgen.cpp
@@ -86,6 +86,20 @@ static cl::opt<std::string> DataAccessProfileFilename(
"-D`) consisting of memory access events."),
cl::cat(ProfGenCategory));
+static cl::opt<std::string> ETMPath("etm", cl::value_desc("etm"),
+ cl::desc("Path of raw ETM trace file"),
+ cl::cat(ProfGenCategory));
+
+static cl::opt<unsigned> ETMTraceID(
+ "etm-trace-id", cl::init(0x10),
+ cl::desc("CoreSight Trace ID (CSID) used to route ETM trace data."),
+ cl::cat(ProfGenCategory));
+
+static cl::opt<std::string>
+ TargetTriple("target-triple", cl::value_desc("triple"),
+ cl::desc("Override the target triple for the binary"),
+ cl::cat(ProfGenCategory));
+
// Validate the command line input.
static void validateCommandLine() {
// Allow the missing perfscript if we only use to show binary disassembly.
@@ -96,15 +110,18 @@ static void validateCommandLine() {
bool HasUnsymbolizedProfile =
UnsymbolizedProfFilename.getNumOccurrences() > 0;
bool HasSampleProfile = SampleProfFilename.getNumOccurrences() > 0;
- uint16_t S =
- HasPerfData + HasPerfScript + HasUnsymbolizedProfile + HasSampleProfile;
+ bool HasEtm = ETMPath.getNumOccurrences() > 0;
+ uint16_t S = HasPerfData + HasPerfScript + HasUnsymbolizedProfile +
+ HasSampleProfile + HasEtm;
if (S != 1) {
std::string Msg =
- S > 1
- ? "`--perfscript`, `--perfdata` and `--unsymbolized-profile` "
- "cannot be used together."
- : "Perf input file is missing, please use one of `--perfscript`, "
- "`--perfdata` and `--unsymbolized-profile` for the input.";
+ S > 1 ? "Only one of `--perfscript`, `--perfdata`, "
+ "`--unsymbolized-profile`, "
+ "`--sample-profile` or `--etm` can be used."
+ : "Perf input file is missing. Please provide one of "
+ "`--perfscript`, "
+ "`--perfdata`, `--unsymbolized-profile`, `--sample-profile`, "
+ "`--etm`.";
exitWithError(Msg);
}
@@ -119,6 +136,7 @@ static void validateCommandLine() {
CheckFileExists(HasPerfScript, PerfScriptFilename);
CheckFileExists(HasUnsymbolizedProfile, UnsymbolizedProfFilename);
CheckFileExists(HasSampleProfile, SampleProfFilename);
+ CheckFileExists(HasEtm, ETMPath);
}
if (!llvm::sys::fs::exists(BinaryPath)) {
@@ -135,17 +153,20 @@ static void validateCommandLine() {
}
}
-static PerfInputFile getPerfInputFile() {
- PerfInputFile File;
+static InputFile getInputFile() {
+ InputFile File;
if (PerfDataFilename.getNumOccurrences()) {
- File.InputFile = PerfDataFilename;
- File.Format = PerfFormat::PerfData;
+ File.InputFilePath = PerfDataFilename;
+ File.Format = InputFormat::PerfData;
} else if (PerfScriptFilename.getNumOccurrences()) {
- File.InputFile = PerfScriptFilename;
- File.Format = PerfFormat::PerfScript;
+ File.InputFilePath = PerfScriptFilename;
+ File.Format = InputFormat::PerfScript;
} else if (UnsymbolizedProfFilename.getNumOccurrences()) {
- File.InputFile = UnsymbolizedProfFilename;
- File.Format = PerfFormat::UnsymbolizedProfile;
+ File.InputFilePath = UnsymbolizedProfFilename;
+ File.Format = InputFormat::UnsymbolizedProfile;
+ } else if (ETMPath.getNumOccurrences()) {
+ File.InputFilePath = ETMPath;
+ File.Format = InputFormat::ETMFormat;
}
return File;
}
@@ -167,6 +188,8 @@ int main(int argc, const char *argv[]) {
// Load symbols and disassemble the code of a given binary.
std::unique_ptr<ProfiledBinary> Binary =
std::make_unique<ProfiledBinary>(BinaryPath, DebugBinPath);
+ Binary->load(TargetTriple);
+
if (ShowDisassemblyOnly)
return EXIT_SUCCESS;
@@ -189,35 +212,47 @@ int main(int argc, const char *argv[]) {
std::optional<uint32_t> PIDFilter;
if (ProcessId.getNumOccurrences())
PIDFilter = ProcessId;
- PerfInputFile PerfFile = getPerfInputFile();
- std::unique_ptr<PerfReaderBase> Reader =
- PerfReaderBase::create(Binary.get(), PerfFile, PIDFilter);
- // Parse perf events and samples
- Reader->parsePerfTraces();
-
- if (!DataAccessProfileFilename.empty()) {
- if (Reader->profileIsCS() || Binary->usePseudoProbes()) {
- exitWithError("Symbolizing vtables from data access profiles is not "
- "yet supported for context-sensitive perf traces or "
- "when pseudo-probe based mapping is enabled. ");
- }
- // Parse the data access perf traces into <ip, data-addr> pairs, symbolize
- // the data-addr to data-symbol. If the data-addr is a vtable, increment
- // counters for the <ip, data-symbol> pair.
- if (Error E = Reader->parseDataAccessPerfTraces(DataAccessProfileFilename,
- PIDFilter)) {
- handleAllErrors(std::move(E), [&](const StringError &SE) {
- exitWithError(SE.getMessage());
- });
+ InputFile File = getInputFile();
+ const ContextSampleCounterMap *Counters = nullptr;
+ bool ProfileIsCS = false;
+ std::unique_ptr<ETMReader> EtmReader;
+ std::unique_ptr<PerfReaderBase> PerfReader;
+
+ if (File.Format == InputFormat::ETMFormat) {
+ EtmReader = std::make_unique<ETMReader>(Binary.get(), File.InputFilePath,
+ static_cast<uint8_t>(ETMTraceID));
+ EtmReader->parseETMTraces();
+ Counters = &EtmReader->getSampleCounters();
+ } else {
+ PerfReader = PerfReaderBase::create(Binary.get(), File, PIDFilter);
+ // Parse perf events and samples
+ PerfReader->parsePerfTraces();
+
+ if (!DataAccessProfileFilename.empty()) {
+ if (PerfReader->profileIsCS() || Binary->usePseudoProbes()) {
+ exitWithError("Symbolizing vtables from data access profiles is not "
+ "yet supported for context-sensitive perf traces or "
+ "when pseudo-probe based mapping is enabled. ");
+ }
+ // Parse the data access perf traces into <ip, data-addr> pairs,
+ // symbolize the data-addr to data-symbol. If the data-addr is a vtable,
+ // increment counters for the <ip, data-symbol> pair.
+ if (Error E = PerfReader->parseDataAccessPerfTraces(
+ DataAccessProfileFilename, PIDFilter)) {
+ handleAllErrors(std::move(E), [&](const StringError &SE) {
+ exitWithError(SE.getMessage());
+ });
+ }
}
+ Counters = &PerfReader->getSampleCounters();
+ ProfileIsCS = PerfReader->profileIsCS();
}
if (SkipSymbolization)
return EXIT_SUCCESS;
std::unique_ptr<ProfileGeneratorBase> Generator =
- ProfileGeneratorBase::create(Binary.get(), &Reader->getSampleCounters(),
- Reader->profileIsCS());
+ ProfileGeneratorBase::create(Binary.get(), Counters, ProfileIsCS);
Generator->generateProfile();
Generator->write();
}
More information about the llvm-commits
mailing list