[clang] [clang][perf-training] Support excluding LLVM build from PGO training (PR #126876)

[Petr Hosek via cfe-commits]

https://github.com/petrhosek created https://github.com/llvm/llvm-project/pull/126876

Using LLVM build itself for PGO training is convenient and a great starting point but it also has several issues:

* LLVM build implicitly depends on tools other than CMake and C/C++ compiler and if those tools aren't available in PATH, the build will fail.
* LLVM build also requires standard headers and libraries which may not always be available in the default location requiring an explicit sysroot.
* Building a single configuration (-DCMAKE_BUILD_TYPE=Release) only exercises the -O3 pipeline and can pesimize other configurations.
* Building for the host target doesn't exercise all other targets.
* Since LLVMSupport is a static library, this doesn't exercise the linker (beyond what the CMake itself does).

Rather than using LLVM build, ideally we would provide a more minimal, purpose built corpus. While we're working on building such a corpus, provide a CMake option that lets vendors disable the use LLVM build for PGO training.

>From 7297d3bf62624f57579d7a1facf7cac36fc75dd4 Mon Sep 17 00:00:00 2001
From: Petr Hosek <phosek at google.com>
Date: Tue, 11 Feb 2025 08:28:05 +0000
Subject: [PATCH] [clang][perf-training] Support excluding LLVM build from PGO
 training

Using LLVM build itself for PGO training is convenient and a great
starting point but it also has several issues:

* LLVM build implicitly depends on tools other than CMake and C/C++
  compiler and if those tools aren't available in PATH, the build
  will fail.
* LLVM build also requires standard headers and libraries which
  may not always be available in the default location requiring an
  explicit sysroot.
* Building a single configuration (-DCMAKE_BUILD_TYPE=Release) only
  exercises the -O3 pipeline and can pesimize other configurations.
* Building for the host target doesn't exercise all other targets.
* Since LLVMSupport is a static library, this doesn't exercise the
  linker (beyond what the CMake itself does).

Rather than using LLVM build, ideally we would provide a more minimal,
purpose built corpus. While we're working on building such a corpus,
provide a CMake option that lets vendors disable the use LLVM build for
PGO training.
---
 clang/utils/perf-training/CMakeLists.txt  | 6 ++++++
 clang/utils/perf-training/lit.cfg         | 3 +++
 clang/utils/perf-training/lit.site.cfg.in | 1 +
 3 files changed, 10 insertions(+)

diff --git a/clang/utils/perf-training/CMakeLists.txt b/clang/utils/perf-training/CMakeLists.txt
index 4aed086563ee9..0c1cdd9a1fb60 100644
--- a/clang/utils/perf-training/CMakeLists.txt
+++ b/clang/utils/perf-training/CMakeLists.txt
@@ -6,6 +6,12 @@ set(CLANG_PGO_TRAINING_DATA "${CMAKE_CURRENT_SOURCE_DIR}" CACHE PATH
 set(CLANG_PGO_TRAINING_DATA_SOURCE_DIR OFF CACHE STRING "Path to source directory containing cmake project with source files to use for generating pgo data")
 set(CLANG_PGO_TRAINING_DEPS "" CACHE STRING "Extra dependencies needed to build the PGO training data.")
 
+option(CLANG_PGO_TRAINING_USE_LLVM_BUILD "Use LLVM build for generating PGO data" ON)
+
+llvm_canonicalize_cmake_booleans(
+  CLANG_PGO_TRAINING_USE_LLVM
+)
+
 if(LLVM_BUILD_INSTRUMENTED)
   configure_lit_site_cfg(
     ${CMAKE_CURRENT_SOURCE_DIR}/lit.site.cfg.in
diff --git a/clang/utils/perf-training/lit.cfg b/clang/utils/perf-training/lit.cfg
index adefc7893ac44..3f6089b7139a7 100644
--- a/clang/utils/perf-training/lit.cfg
+++ b/clang/utils/perf-training/lit.cfg
@@ -27,6 +27,9 @@ config.clang = lit.util.which('clang', config.clang_tools_dir).replace('\\', '/'
 config.name = 'Clang Perf Training'
 config.suffixes = ['.c', '.cc', '.cpp', '.m', '.mm', '.cu', '.ll', '.cl', '.s', '.S', '.modulemap', '.test']
 
+if not config.use_llvm_build:
+    config.excludes = ['llvm-support']
+
 cc1_wrapper = '%s %s/perf-helper.py cc1' % (config.python_exe, config.perf_helper_dir)
 
 use_lit_shell = os.environ.get("LIT_USE_INTERNAL_SHELL")
diff --git a/clang/utils/perf-training/lit.site.cfg.in b/clang/utils/perf-training/lit.site.cfg.in
index 9d279d552919a..da81ec21a28a6 100644
--- a/clang/utils/perf-training/lit.site.cfg.in
+++ b/clang/utils/perf-training/lit.site.cfg.in
@@ -11,6 +11,7 @@ config.python_exe = "@Python3_EXECUTABLE@"
 config.cmake_exe = "@CMAKE_COMMAND@"
 config.llvm_src_dir ="@CMAKE_SOURCE_DIR@"
 config.cmake_generator ="@CMAKE_GENERATOR@"
+config.use_llvm_build = @CLANG_PGO_TRAINING_USE_LLVM_BUILD@
 
 # Let the main config do the real work.
 lit_config.load_config(config, "@CLANG_SOURCE_DIR@/utils/perf-training/lit.cfg")