diff options
Diffstat (limited to 'compiler/rustc_codegen_llvm/src')
| -rw-r--r-- | compiler/rustc_codegen_llvm/src/back/lto.rs | 49 | ||||
| -rw-r--r-- | compiler/rustc_codegen_llvm/src/debuginfo/gdb.rs | 3 | ||||
| -rw-r--r-- | compiler/rustc_codegen_llvm/src/debuginfo/metadata.rs | 3 | ||||
| -rw-r--r-- | compiler/rustc_codegen_llvm/src/lib.rs | 1 | ||||
| -rw-r--r-- | compiler/rustc_codegen_llvm/src/llvm/ffi.rs | 6 | ||||
| -rw-r--r-- | compiler/rustc_codegen_llvm/src/llvm_util.rs | 166 |
6 files changed, 131 insertions, 97 deletions
diff --git a/compiler/rustc_codegen_llvm/src/back/lto.rs b/compiler/rustc_codegen_llvm/src/back/lto.rs index d2e01708a37..604f68eb6a4 100644 --- a/compiler/rustc_codegen_llvm/src/back/lto.rs +++ b/compiler/rustc_codegen_llvm/src/back/lto.rs @@ -25,7 +25,6 @@ use std::fs::File; use std::io; use std::iter; use std::path::Path; -use std::ptr; use std::slice; use std::sync::Arc; @@ -709,17 +708,6 @@ pub unsafe fn optimize_thin_module( let llmod = module.module_llvm.llmod(); save_temp_bitcode(cgcx, &module, "thin-lto-input"); - // Before we do much else find the "main" `DICompileUnit` that we'll be - // using below. If we find more than one though then rustc has changed - // in a way we're not ready for, so generate an ICE by returning - // an error. - let mut cu1 = ptr::null_mut(); - let mut cu2 = ptr::null_mut(); - llvm::LLVMRustThinLTOGetDICompileUnit(llmod, &mut cu1, &mut cu2); - if !cu2.is_null() { - return Err(write::llvm_err(&diag_handler, LlvmError::MultipleSourceDiCompileUnit)); - } - // Up next comes the per-module local analyses that we do for Thin LTO. // Each of these functions is basically copied from the LLVM // implementation and then tailored to suit this implementation. Ideally @@ -766,43 +754,6 @@ pub unsafe fn optimize_thin_module( save_temp_bitcode(cgcx, &module, "thin-lto-after-import"); } - // Ok now this is a bit unfortunate. This is also something you won't - // find upstream in LLVM's ThinLTO passes! This is a hack for now to - // work around bugs in LLVM. - // - // First discovered in #45511 it was found that as part of ThinLTO - // importing passes LLVM will import `DICompileUnit` metadata - // information across modules. This means that we'll be working with one - // LLVM module that has multiple `DICompileUnit` instances in it (a - // bunch of `llvm.dbg.cu` members). Unfortunately there's a number of - // bugs in LLVM's backend which generates invalid DWARF in a situation - // like this: - // - // https://bugs.llvm.org/show_bug.cgi?id=35212 - // https://bugs.llvm.org/show_bug.cgi?id=35562 - // - // While the first bug there is fixed the second ended up causing #46346 - // which was basically a resurgence of #45511 after LLVM's bug 35212 was - // fixed. - // - // This function below is a huge hack around this problem. The function - // below is defined in `PassWrapper.cpp` and will basically "merge" - // all `DICompileUnit` instances in a module. Basically it'll take all - // the objects, rewrite all pointers of `DISubprogram` to point to the - // first `DICompileUnit`, and then delete all the other units. - // - // This is probably mangling to the debug info slightly (but hopefully - // not too much) but for now at least gets LLVM to emit valid DWARF (or - // so it appears). Hopefully we can remove this once upstream bugs are - // fixed in LLVM. - { - let _timer = cgcx - .prof - .generic_activity_with_arg("LLVM_thin_lto_patch_debuginfo", thin_module.name()); - llvm::LLVMRustThinLTOPatchDICompileUnit(llmod, cu1); - save_temp_bitcode(cgcx, &module, "thin-lto-after-patch"); - } - // Alright now that we've done everything related to the ThinLTO // analysis it's time to run some optimizations! Here we use the same // `run_pass_manager` as the "fat" LTO above except that we tell it to diff --git a/compiler/rustc_codegen_llvm/src/debuginfo/gdb.rs b/compiler/rustc_codegen_llvm/src/debuginfo/gdb.rs index aaf5dbd9930..37f30917609 100644 --- a/compiler/rustc_codegen_llvm/src/debuginfo/gdb.rs +++ b/compiler/rustc_codegen_llvm/src/debuginfo/gdb.rs @@ -9,10 +9,9 @@ use rustc_ast::attr; use rustc_codegen_ssa::base::collect_debugger_visualizers_transitive; use rustc_codegen_ssa::traits::*; use rustc_hir::def_id::LOCAL_CRATE; -use rustc_middle::bug; +use rustc_middle::{bug, middle::debugger_visualizer::DebuggerVisualizerType}; use rustc_session::config::{CrateType, DebugInfo}; use rustc_span::symbol::sym; -use rustc_span::DebuggerVisualizerType; /// Inserts a side-effect free instruction sequence that makes sure that the /// .debug_gdb_scripts global is referenced, so it isn't removed by the linker. diff --git a/compiler/rustc_codegen_llvm/src/debuginfo/metadata.rs b/compiler/rustc_codegen_llvm/src/debuginfo/metadata.rs index 25fe3cb265d..bd2fba12602 100644 --- a/compiler/rustc_codegen_llvm/src/debuginfo/metadata.rs +++ b/compiler/rustc_codegen_llvm/src/debuginfo/metadata.rs @@ -806,8 +806,7 @@ pub fn build_compile_unit_di_node<'ll, 'tcx>( name_in_debuginfo.push(codegen_unit_name); debug!("build_compile_unit_di_node: {:?}", name_in_debuginfo); - let rustc_producer = - format!("rustc version {}", option_env!("CFG_VERSION").expect("CFG_VERSION"),); + let rustc_producer = format!("rustc version {}", tcx.sess.cfg_version); // FIXME(#41252) Remove "clang LLVM" if we can get GDB and LLVM to play nice. let producer = format!("clang LLVM ({})", rustc_producer); diff --git a/compiler/rustc_codegen_llvm/src/lib.rs b/compiler/rustc_codegen_llvm/src/lib.rs index 6a86237d79e..805843e5863 100644 --- a/compiler/rustc_codegen_llvm/src/lib.rs +++ b/compiler/rustc_codegen_llvm/src/lib.rs @@ -10,6 +10,7 @@ #![feature(iter_intersperse)] #![feature(let_chains)] #![feature(never_type)] +#![feature(impl_trait_in_assoc_type)] #![recursion_limit = "256"] #![allow(rustc::potential_query_instability)] #![deny(rustc::untranslatable_diagnostic)] diff --git a/compiler/rustc_codegen_llvm/src/llvm/ffi.rs b/compiler/rustc_codegen_llvm/src/llvm/ffi.rs index aefd5b2a13c..de93a64c0d6 100644 --- a/compiler/rustc_codegen_llvm/src/llvm/ffi.rs +++ b/compiler/rustc_codegen_llvm/src/llvm/ffi.rs @@ -2484,12 +2484,6 @@ extern "C" { len: usize, out_len: &mut usize, ) -> *const u8; - pub fn LLVMRustThinLTOGetDICompileUnit( - M: &Module, - CU1: &mut *mut c_void, - CU2: &mut *mut c_void, - ); - pub fn LLVMRustThinLTOPatchDICompileUnit(M: &Module, CU: *mut c_void); pub fn LLVMRustLinkerNew(M: &Module) -> &mut Linker<'_>; pub fn LLVMRustLinkerAdd( diff --git a/compiler/rustc_codegen_llvm/src/llvm_util.rs b/compiler/rustc_codegen_llvm/src/llvm_util.rs index 994addf12eb..03be0654b50 100644 --- a/compiler/rustc_codegen_llvm/src/llvm_util.rs +++ b/compiler/rustc_codegen_llvm/src/llvm_util.rs @@ -16,7 +16,6 @@ use rustc_session::config::PrintRequest; use rustc_session::Session; use rustc_span::symbol::Symbol; use rustc_target::spec::{MergeFunctions, PanicStrategy}; -use smallvec::{smallvec, SmallVec}; use std::ffi::{CStr, CString}; use std::path::Path; @@ -132,6 +131,60 @@ pub fn time_trace_profiler_finish(file_name: &Path) { } } +pub enum TargetFeatureFoldStrength<'a> { + // The feature is only tied when enabling the feature, disabling + // this feature shouldn't disable the tied feature. + EnableOnly(&'a str), + // The feature is tied for both enabling and disabling this feature. + Both(&'a str), +} + +impl<'a> TargetFeatureFoldStrength<'a> { + fn as_str(&self) -> &'a str { + match self { + TargetFeatureFoldStrength::EnableOnly(feat) => feat, + TargetFeatureFoldStrength::Both(feat) => feat, + } + } +} + +pub struct LLVMFeature<'a> { + pub llvm_feature_name: &'a str, + pub dependency: Option<TargetFeatureFoldStrength<'a>>, +} + +impl<'a> LLVMFeature<'a> { + pub fn new(llvm_feature_name: &'a str) -> Self { + Self { llvm_feature_name, dependency: None } + } + + pub fn with_dependency( + llvm_feature_name: &'a str, + dependency: TargetFeatureFoldStrength<'a>, + ) -> Self { + Self { llvm_feature_name, dependency: Some(dependency) } + } + + pub fn contains(&self, feat: &str) -> bool { + self.iter().any(|dep| dep == feat) + } + + pub fn iter(&'a self) -> impl Iterator<Item = &'a str> { + let dependencies = self.dependency.iter().map(|feat| feat.as_str()); + std::iter::once(self.llvm_feature_name).chain(dependencies) + } +} + +impl<'a> IntoIterator for LLVMFeature<'a> { + type Item = &'a str; + type IntoIter = impl Iterator<Item = &'a str>; + + fn into_iter(self) -> Self::IntoIter { + let dependencies = self.dependency.into_iter().map(|feat| feat.as_str()); + std::iter::once(self.llvm_feature_name).chain(dependencies) + } +} + // WARNING: the features after applying `to_llvm_features` must be known // to LLVM or the feature detection code will walk past the end of the feature // array, leading to crashes. @@ -147,36 +200,65 @@ pub fn time_trace_profiler_finish(file_name: &Path) { // Though note that Rust can also be build with an external precompiled version of LLVM // which might lead to failures if the oldest tested / supported LLVM version // doesn't yet support the relevant intrinsics -pub fn to_llvm_features<'a>(sess: &Session, s: &'a str) -> SmallVec<[&'a str; 2]> { +pub fn to_llvm_features<'a>(sess: &Session, s: &'a str) -> LLVMFeature<'a> { let arch = if sess.target.arch == "x86_64" { "x86" } else { &*sess.target.arch }; match (arch, s) { - ("x86", "sse4.2") => smallvec!["sse4.2", "crc32"], - ("x86", "pclmulqdq") => smallvec!["pclmul"], - ("x86", "rdrand") => smallvec!["rdrnd"], - ("x86", "bmi1") => smallvec!["bmi"], - ("x86", "cmpxchg16b") => smallvec!["cx16"], - ("aarch64", "rcpc2") => smallvec!["rcpc-immo"], - ("aarch64", "dpb") => smallvec!["ccpp"], - ("aarch64", "dpb2") => smallvec!["ccdp"], - ("aarch64", "frintts") => smallvec!["fptoint"], - ("aarch64", "fcma") => smallvec!["complxnum"], - ("aarch64", "pmuv3") => smallvec!["perfmon"], - ("aarch64", "paca") => smallvec!["pauth"], - ("aarch64", "pacg") => smallvec!["pauth"], - // Rust ties fp and neon together. In LLVM neon implicitly enables fp, - // but we manually enable neon when a feature only implicitly enables fp - ("aarch64", "f32mm") => smallvec!["f32mm", "neon"], - ("aarch64", "f64mm") => smallvec!["f64mm", "neon"], - ("aarch64", "fhm") => smallvec!["fp16fml", "neon"], - ("aarch64", "fp16") => smallvec!["fullfp16", "neon"], - ("aarch64", "jsconv") => smallvec!["jsconv", "neon"], - ("aarch64", "sve") => smallvec!["sve", "neon"], - ("aarch64", "sve2") => smallvec!["sve2", "neon"], - ("aarch64", "sve2-aes") => smallvec!["sve2-aes", "neon"], - ("aarch64", "sve2-sm4") => smallvec!["sve2-sm4", "neon"], - ("aarch64", "sve2-sha3") => smallvec!["sve2-sha3", "neon"], - ("aarch64", "sve2-bitperm") => smallvec!["sve2-bitperm", "neon"], - (_, s) => smallvec![s], + ("x86", "sse4.2") => { + LLVMFeature::with_dependency("sse4.2", TargetFeatureFoldStrength::EnableOnly("crc32")) + } + ("x86", "pclmulqdq") => LLVMFeature::new("pclmul"), + ("x86", "rdrand") => LLVMFeature::new("rdrnd"), + ("x86", "bmi1") => LLVMFeature::new("bmi"), + ("x86", "cmpxchg16b") => LLVMFeature::new("cx16"), + ("aarch64", "rcpc2") => LLVMFeature::new("rcpc-immo"), + ("aarch64", "dpb") => LLVMFeature::new("ccpp"), + ("aarch64", "dpb2") => LLVMFeature::new("ccdp"), + ("aarch64", "frintts") => LLVMFeature::new("fptoint"), + ("aarch64", "fcma") => LLVMFeature::new("complxnum"), + ("aarch64", "pmuv3") => LLVMFeature::new("perfmon"), + ("aarch64", "paca") => LLVMFeature::new("pauth"), + ("aarch64", "pacg") => LLVMFeature::new("pauth"), + // Rust ties fp and neon together. + ("aarch64", "neon") => { + LLVMFeature::with_dependency("neon", TargetFeatureFoldStrength::Both("fp-armv8")) + } + // In LLVM neon implicitly enables fp, but we manually enable + // neon when a feature only implicitly enables fp + ("aarch64", "f32mm") => { + LLVMFeature::with_dependency("f32mm", TargetFeatureFoldStrength::EnableOnly("neon")) + } + ("aarch64", "f64mm") => { + LLVMFeature::with_dependency("f64mm", TargetFeatureFoldStrength::EnableOnly("neon")) + } + ("aarch64", "fhm") => { + LLVMFeature::with_dependency("fp16fml", TargetFeatureFoldStrength::EnableOnly("neon")) + } + ("aarch64", "fp16") => { + LLVMFeature::with_dependency("fullfp16", TargetFeatureFoldStrength::EnableOnly("neon")) + } + ("aarch64", "jsconv") => { + LLVMFeature::with_dependency("jsconv", TargetFeatureFoldStrength::EnableOnly("neon")) + } + ("aarch64", "sve") => { + LLVMFeature::with_dependency("sve", TargetFeatureFoldStrength::EnableOnly("neon")) + } + ("aarch64", "sve2") => { + LLVMFeature::with_dependency("sve2", TargetFeatureFoldStrength::EnableOnly("neon")) + } + ("aarch64", "sve2-aes") => { + LLVMFeature::with_dependency("sve2-aes", TargetFeatureFoldStrength::EnableOnly("neon")) + } + ("aarch64", "sve2-sm4") => { + LLVMFeature::with_dependency("sve2-sm4", TargetFeatureFoldStrength::EnableOnly("neon")) + } + ("aarch64", "sve2-sha3") => { + LLVMFeature::with_dependency("sve2-sha3", TargetFeatureFoldStrength::EnableOnly("neon")) + } + ("aarch64", "sve2-bitperm") => LLVMFeature::with_dependency( + "sve2-bitperm", + TargetFeatureFoldStrength::EnableOnly("neon"), + ), + (_, s) => LLVMFeature::new(s), } } @@ -274,18 +356,17 @@ fn print_target_features(sess: &Session, tm: &llvm::TargetMachine) { let mut rustc_target_features = supported_target_features(sess) .iter() .map(|(feature, _gate)| { - let desc = if let Some(llvm_feature) = to_llvm_features(sess, *feature).first() { - // LLVM asserts that these are sorted. LLVM and Rust both use byte comparison for these strings. + // LLVM asserts that these are sorted. LLVM and Rust both use byte comparison for these strings. + let llvm_feature = to_llvm_features(sess, *feature).llvm_feature_name; + let desc = match llvm_target_features.binary_search_by_key(&llvm_feature, |(f, _d)| f).ok() { Some(index) => { known_llvm_target_features.insert(llvm_feature); llvm_target_features[index].1 } None => "", - } - } else { - "" - }; + }; + (*feature, desc) }) .collect::<Vec<_>>(); @@ -469,10 +550,19 @@ pub(crate) fn global_llvm_features(sess: &Session, diagnostics: bool) -> Vec<Str // passing requests down to LLVM. This means that all in-language // features also work on the command line instead of having two // different names when the LLVM name and the Rust name differ. + let llvm_feature = to_llvm_features(sess, feature); + Some( - to_llvm_features(sess, feature) - .into_iter() - .map(move |f| format!("{}{}", enable_disable, f)), + std::iter::once(format!("{}{}", enable_disable, llvm_feature.llvm_feature_name)) + .chain(llvm_feature.dependency.into_iter().filter_map(move |feat| { + match (enable_disable, feat) { + ('-' | '+', TargetFeatureFoldStrength::Both(f)) + | ('+', TargetFeatureFoldStrength::EnableOnly(f)) => { + Some(format!("{}{}", enable_disable, f)) + } + _ => None, + } + })), ) }) .flatten(); |