From aa91871539ca518e81af58485df895f8db30496a Mon Sep 17 00:00:00 2001 From: Augie Fackler Date: Fri, 19 Jan 2024 14:42:43 -0500 Subject: [PATCH] rustc_codegen_llvm: add support for writing summary bitcode Typical uses of ThinLTO don't have any use for this as a standalone file, but distributed ThinLTO uses this to make the linker phase more efficient. With clang you'd do something like `clang -flto=thin -fthin-link-bitcode=foo.indexing.o -c foo.c` and then get both foo.o (full of bitcode) and foo.indexing.o (just the summary or index part of the bitcode). That's then usable by a two-stage linking process that's more friendly to distributed build systems like bazel, which is why I'm working on this area. I talked some to @teresajohnson about naming in this area, as things seem to be a little confused between various blog posts and build systems. "bitcode index" and "bitcode summary" tend to be a little too ambiguous, and she tends to use "thin link bitcode" and "minimized bitcode" (which matches the descriptions in LLVM). Since the clang option is thin-link-bitcode, I went with that to try and not add a new spelling in the world. Per @dtolnay, you can work around the lack of this by using `lld --thinlto-index-only` to do the indexing on regular .o files of bitcode, but that is a bit wasteful on actions when we already have all the information in rustc and could just write out the matching minimized bitcode. I didn't test that at all in our infrastructure, because by the time I learned that I already had this patch largely written. --- .../rustc_codegen_cranelift/src/driver/aot.rs | 2 +- compiler/rustc_codegen_gcc/src/lib.rs | 4 +++ compiler/rustc_codegen_llvm/src/back/lto.rs | 8 +++++ compiler/rustc_codegen_llvm/src/back/write.rs | 32 ++++++++++++++++--- compiler/rustc_codegen_llvm/src/llvm/ffi.rs | 2 ++ compiler/rustc_codegen_ssa/src/back/write.rs | 9 ++++++ .../rustc_codegen_ssa/src/traits/write.rs | 1 + .../rustc_llvm/llvm-wrapper/PassWrapper.cpp | 14 +++++++- compiler/rustc_session/src/config.rs | 24 +++++++++++--- 9 files changed, 85 insertions(+), 11 deletions(-) diff --git a/compiler/rustc_codegen_cranelift/src/driver/aot.rs b/compiler/rustc_codegen_cranelift/src/driver/aot.rs index fce4690f97d..394c810176a 100644 --- a/compiler/rustc_codegen_cranelift/src/driver/aot.rs +++ b/compiler/rustc_codegen_cranelift/src/driver/aot.rs @@ -200,7 +200,7 @@ fn produce_final_output_artifacts( // to get rid of it. for output_type in crate_output.outputs.keys() { match *output_type { - OutputType::Bitcode => { + OutputType::Bitcode | OutputType::ThinLinkBitcode => { // Cranelift doesn't have bitcode // user_wants_bitcode = true; // // Copy to .bc, but always keep the .0.bc. There is a later diff --git a/compiler/rustc_codegen_gcc/src/lib.rs b/compiler/rustc_codegen_gcc/src/lib.rs index 3f2fadce9e4..d786d5116ce 100644 --- a/compiler/rustc_codegen_gcc/src/lib.rs +++ b/compiler/rustc_codegen_gcc/src/lib.rs @@ -335,6 +335,10 @@ impl ThinBufferMethods for ThinBuffer { fn data(&self) -> &[u8] { unimplemented!(); } + + fn thin_link_data(&self) -> &[u8] { + unimplemented!(); + } } pub struct GccContext { diff --git a/compiler/rustc_codegen_llvm/src/back/lto.rs b/compiler/rustc_codegen_llvm/src/back/lto.rs index e61af863dc0..740939eb6bb 100644 --- a/compiler/rustc_codegen_llvm/src/back/lto.rs +++ b/compiler/rustc_codegen_llvm/src/back/lto.rs @@ -687,6 +687,14 @@ impl ThinBufferMethods for ThinBuffer { slice::from_raw_parts(ptr, len) } } + + fn thin_link_data(&self) -> &[u8] { + unsafe { + let ptr = llvm::LLVMRustThinLTOBufferThinLinkDataPtr(self.0) as *const _; + let len = llvm::LLVMRustThinLTOBufferThinLinkDataLen(self.0); + slice::from_raw_parts(ptr, len) + } + } } impl Drop for ThinBuffer { diff --git a/compiler/rustc_codegen_llvm/src/back/write.rs b/compiler/rustc_codegen_llvm/src/back/write.rs index 49f9d7ddab6..de12ae4ecc8 100644 --- a/compiler/rustc_codegen_llvm/src/back/write.rs +++ b/compiler/rustc_codegen_llvm/src/back/write.rs @@ -708,6 +708,8 @@ pub(crate) unsafe fn codegen( // asm from LLVM and use `gcc` to create the object file. let bc_out = cgcx.output_filenames.temp_path(OutputType::Bitcode, module_name); + let bc_index_out = + cgcx.output_filenames.temp_path(OutputType::ThinLinkBitcode, module_name); let obj_out = cgcx.output_filenames.temp_path(OutputType::Object, module_name); if config.bitcode_needed() { @@ -716,6 +718,7 @@ pub(crate) unsafe fn codegen( .generic_activity_with_arg("LLVM_module_codegen_make_bitcode", &*module.name); let thin = ThinBuffer::new(llmod, config.emit_thin_lto); let data = thin.data(); + let index_data = thin.thin_link_data(); if let Some(bitcode_filename) = bc_out.file_name() { cgcx.prof.artifact_size( @@ -725,12 +728,31 @@ pub(crate) unsafe fn codegen( ); } + if let Some(thin_link_bitcode_filename) = bc_index_out.file_name() { + cgcx.prof.artifact_size( + "llvm_bitcode_summary", + thin_link_bitcode_filename.to_string_lossy(), + index_data.len() as u64, + ); + + let _timer = cgcx.prof.generic_activity_with_arg( + "LLVM_module_codegen_emit_bitcode_index", + &*module.name, + ); + if let Err(err) = fs::write(&bc_index_out, index_data) { + dcx.emit_err(WriteBytecode { path: &bc_index_out, err }); + } + } + if config.emit_bc || config.emit_obj == EmitObj::Bitcode { - let _timer = cgcx - .prof - .generic_activity_with_arg("LLVM_module_codegen_emit_bitcode", &*module.name); - if let Err(err) = fs::write(&bc_out, data) { - dcx.emit_err(WriteBytecode { path: &bc_out, err }); + { + let _timer = cgcx.prof.generic_activity_with_arg( + "LLVM_module_codegen_emit_bitcode", + &*module.name, + ); + if let Err(err) = fs::write(&bc_out, data) { + dcx.emit_err(WriteBytecode { path: &bc_out, err }); + } } } diff --git a/compiler/rustc_codegen_llvm/src/llvm/ffi.rs b/compiler/rustc_codegen_llvm/src/llvm/ffi.rs index 7a34e21628d..648e7f12890 100644 --- a/compiler/rustc_codegen_llvm/src/llvm/ffi.rs +++ b/compiler/rustc_codegen_llvm/src/llvm/ffi.rs @@ -2354,6 +2354,8 @@ extern "C" { pub fn LLVMRustThinLTOBufferFree(M: &'static mut ThinLTOBuffer); pub fn LLVMRustThinLTOBufferPtr(M: &ThinLTOBuffer) -> *const c_char; pub fn LLVMRustThinLTOBufferLen(M: &ThinLTOBuffer) -> size_t; + pub fn LLVMRustThinLTOBufferThinLinkDataPtr(M: &ThinLTOBuffer) -> *const c_char; + pub fn LLVMRustThinLTOBufferThinLinkDataLen(M: &ThinLTOBuffer) -> size_t; pub fn LLVMRustCreateThinLTOData( Modules: *const ThinLTOModule, NumModules: c_uint, diff --git a/compiler/rustc_codegen_ssa/src/back/write.rs b/compiler/rustc_codegen_ssa/src/back/write.rs index 53ba0da7d04..cee177d5eec 100644 --- a/compiler/rustc_codegen_ssa/src/back/write.rs +++ b/compiler/rustc_codegen_ssa/src/back/write.rs @@ -107,6 +107,7 @@ pub struct ModuleConfig { pub emit_asm: bool, pub emit_obj: EmitObj, pub emit_thin_lto: bool, + pub emit_thin_lto_index: bool, pub bc_cmdline: String, // Miscellaneous flags. These are mostly copied from command-line @@ -231,6 +232,10 @@ impl ModuleConfig { ), emit_obj, emit_thin_lto: sess.opts.unstable_opts.emit_thin_lto, + emit_thin_lto_index: if_regular!( + sess.opts.output_types.contains_key(&OutputType::ThinLinkBitcode), + false + ), bc_cmdline: sess.target.bitcode_llvm_cmdline.to_string(), verify_llvm_ir: sess.verify_llvm_ir(), @@ -282,6 +287,7 @@ impl ModuleConfig { pub fn bitcode_needed(&self) -> bool { self.emit_bc + || self.emit_thin_lto_index || self.emit_obj == EmitObj::Bitcode || self.emit_obj == EmitObj::ObjectCode(BitcodeSection::Full) } @@ -629,6 +635,9 @@ fn produce_final_output_artifacts( // them for making an rlib. copy_if_one_unit(OutputType::Bitcode, true); } + OutputType::ThinLinkBitcode => { + copy_if_one_unit(OutputType::ThinLinkBitcode, false); + } OutputType::LlvmAssembly => { copy_if_one_unit(OutputType::LlvmAssembly, false); } diff --git a/compiler/rustc_codegen_ssa/src/traits/write.rs b/compiler/rustc_codegen_ssa/src/traits/write.rs index 048540894ac..b4b23812d4d 100644 --- a/compiler/rustc_codegen_ssa/src/traits/write.rs +++ b/compiler/rustc_codegen_ssa/src/traits/write.rs @@ -62,6 +62,7 @@ pub trait WriteBackendMethods: 'static + Sized + Clone { pub trait ThinBufferMethods: Send + Sync { fn data(&self) -> &[u8]; + fn thin_link_data(&self) -> &[u8]; } pub trait ModuleBufferMethods: Send + Sync { diff --git a/compiler/rustc_llvm/llvm-wrapper/PassWrapper.cpp b/compiler/rustc_llvm/llvm-wrapper/PassWrapper.cpp index 067374c0261..c210d2531e7 100644 --- a/compiler/rustc_llvm/llvm-wrapper/PassWrapper.cpp +++ b/compiler/rustc_llvm/llvm-wrapper/PassWrapper.cpp @@ -1488,6 +1488,7 @@ LLVMRustPrepareThinLTOImport(const LLVMRustThinLTOData *Data, LLVMModuleRef M, // a ThinLTO summary attached. struct LLVMRustThinLTOBuffer { std::string data; + std::string thin_link_data; }; extern "C" LLVMRustThinLTOBuffer* @@ -1495,6 +1496,7 @@ LLVMRustThinLTOBufferCreate(LLVMModuleRef M, bool is_thin) { auto Ret = std::make_unique(); { auto OS = raw_string_ostream(Ret->data); + auto ThinLinkOS = raw_string_ostream(Ret->thin_link_data); { if (is_thin) { PassBuilder PB; @@ -1508,7 +1510,7 @@ LLVMRustThinLTOBufferCreate(LLVMModuleRef M, bool is_thin) { PB.registerLoopAnalyses(LAM); PB.crossRegisterProxies(LAM, FAM, CGAM, MAM); ModulePassManager MPM; - MPM.addPass(ThinLTOBitcodeWriterPass(OS, nullptr)); + MPM.addPass(ThinLTOBitcodeWriterPass(OS, &ThinLinkOS)); MPM.run(*unwrap(M), MAM); } else { WriteBitcodeToFile(*unwrap(M), OS); @@ -1533,6 +1535,16 @@ LLVMRustThinLTOBufferLen(const LLVMRustThinLTOBuffer *Buffer) { return Buffer->data.length(); } +extern "C" const void* +LLVMRustThinLTOBufferThinLinkDataPtr(const LLVMRustThinLTOBuffer *Buffer) { + return Buffer->thin_link_data.data(); +} + +extern "C" size_t +LLVMRustThinLTOBufferThinLinkDataLen(const LLVMRustThinLTOBuffer *Buffer) { + return Buffer->thin_link_data.length(); +} + // This is what we used to parse upstream bitcode for actual ThinLTO // processing. We'll call this once per module optimized through ThinLTO, and // it'll be called concurrently on many threads. diff --git a/compiler/rustc_session/src/config.rs b/compiler/rustc_session/src/config.rs index 7dd9fdf60f9..5ac4d194e87 100644 --- a/compiler/rustc_session/src/config.rs +++ b/compiler/rustc_session/src/config.rs @@ -465,6 +465,7 @@ impl FromStr for SplitDwarfKind { #[derive(Encodable, Decodable)] pub enum OutputType { Bitcode, + ThinLinkBitcode, Assembly, LlvmAssembly, Mir, @@ -492,6 +493,7 @@ impl OutputType { match *self { OutputType::Exe | OutputType::DepInfo | OutputType::Metadata => true, OutputType::Bitcode + | OutputType::ThinLinkBitcode | OutputType::Assembly | OutputType::LlvmAssembly | OutputType::Mir @@ -502,6 +504,7 @@ impl OutputType { pub fn shorthand(&self) -> &'static str { match *self { OutputType::Bitcode => "llvm-bc", + OutputType::ThinLinkBitcode => "thin-link-bitcode", OutputType::Assembly => "asm", OutputType::LlvmAssembly => "llvm-ir", OutputType::Mir => "mir", @@ -518,6 +521,7 @@ impl OutputType { "llvm-ir" => OutputType::LlvmAssembly, "mir" => OutputType::Mir, "llvm-bc" => OutputType::Bitcode, + "thin-link-bitcode" => OutputType::ThinLinkBitcode, "obj" => OutputType::Object, "metadata" => OutputType::Metadata, "link" => OutputType::Exe, @@ -528,8 +532,9 @@ impl OutputType { fn shorthands_display() -> String { format!( - "`{}`, `{}`, `{}`, `{}`, `{}`, `{}`, `{}`, `{}`", + "`{}`, `{}`, `{}`, `{}`, `{}`, `{}`, `{}`, `{}`, `{}`", OutputType::Bitcode.shorthand(), + OutputType::ThinLinkBitcode.shorthand(), OutputType::Assembly.shorthand(), OutputType::LlvmAssembly.shorthand(), OutputType::Mir.shorthand(), @@ -543,6 +548,7 @@ impl OutputType { pub fn extension(&self) -> &'static str { match *self { OutputType::Bitcode => "bc", + OutputType::ThinLinkBitcode => "indexing.o", OutputType::Assembly => "s", OutputType::LlvmAssembly => "ll", OutputType::Mir => "mir", @@ -559,9 +565,11 @@ impl OutputType { | OutputType::LlvmAssembly | OutputType::Mir | OutputType::DepInfo => true, - OutputType::Bitcode | OutputType::Object | OutputType::Metadata | OutputType::Exe => { - false - } + OutputType::Bitcode + | OutputType::ThinLinkBitcode + | OutputType::Object + | OutputType::Metadata + | OutputType::Exe => false, } } } @@ -644,6 +652,7 @@ impl OutputTypes { pub fn should_codegen(&self) -> bool { self.0.keys().any(|k| match *k { OutputType::Bitcode + | OutputType::ThinLinkBitcode | OutputType::Assembly | OutputType::LlvmAssembly | OutputType::Mir @@ -657,6 +666,7 @@ impl OutputTypes { pub fn should_link(&self) -> bool { self.0.keys().any(|k| match *k { OutputType::Bitcode + | OutputType::ThinLinkBitcode | OutputType::Assembly | OutputType::LlvmAssembly | OutputType::Mir @@ -1769,6 +1779,12 @@ fn parse_output_types( display = OutputType::shorthands_display(), )) }); + if output_type == OutputType::ThinLinkBitcode && !unstable_opts.unstable_options { + early_dcx.early_fatal(format!( + "{} requested but -Zunstable-options not specified", + OutputType::ThinLinkBitcode.shorthand() + )); + } output_types.insert(output_type, path); } }