[rust.git] / src / librustc_trans / back / lto.rs

// Copyright 2013 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.

use back::bytecode::{DecodedBytecode, RLIB_BYTECODE_EXTENSION};
use back::symbol_export;
use back::write::{ModuleConfig, with_llvm_pmb, CodegenContext};
use back::write;
use errors::{FatalError, Handler};
use llvm::archive_ro::ArchiveRO;
use llvm::{ModuleRef, TargetMachineRef, True, False};
use llvm;
use rustc::hir::def_id::LOCAL_CRATE;
use rustc::middle::exported_symbols::SymbolExportLevel;
use rustc::session::config;
use rustc::util::common::time;
use time_graph::Timeline;
use {ModuleTranslation, ModuleLlvm, ModuleKind, ModuleSource};

use libc;

use std::ffi::CString;
use std::ptr;
use std::slice;
use std::sync::Arc;

pub fn crate_type_allows_lto(crate_type: config::CrateType) -> bool {
    match crate_type {
        config::CrateTypeExecutable |
        config::CrateTypeStaticlib  |
        config::CrateTypeCdylib     => true,

        config::CrateTypeDylib     |
        config::CrateTypeRlib      |
        config::CrateTypeProcMacro => false,
    }
}

pub(crate) enum LtoModuleTranslation {
    Fat {
        module: Option<ModuleTranslation>,
        _serialized_bitcode: Vec<SerializedModule>,
    },

    Thin(ThinModule),
}

impl LtoModuleTranslation {
    pub fn name(&self) -> &str {
        match *self {
            LtoModuleTranslation::Fat { .. } => "everything",
            LtoModuleTranslation::Thin(ref m) => m.name(),
        }
    }

    /// Optimize this module within the given codegen context.
    ///
    /// This function is unsafe as it'll return a `ModuleTranslation` still
    /// points to LLVM data structures owned by this `LtoModuleTranslation`.
    /// It's intended that the module returned is immediately code generated and
    /// dropped, and then this LTO module is dropped.
    pub(crate) unsafe fn optimize(&mut self,
                                  cgcx: &CodegenContext,
                                  timeline: &mut Timeline)
        -> Result<ModuleTranslation, FatalError>
    {
        match *self {
            LtoModuleTranslation::Fat { ref mut module, .. } => {
                let trans = module.take().unwrap();
                let config = cgcx.config(trans.kind);
                let llmod = trans.llvm().unwrap().llmod;
                let tm = trans.llvm().unwrap().tm;
                run_pass_manager(cgcx, tm, llmod, config, false);
                timeline.record("fat-done");
                Ok(trans)
            }
            LtoModuleTranslation::Thin(ref mut thin) => thin.optimize(cgcx, timeline),
        }
    }

    /// A "guage" of how costly it is to optimize this module, used to sort
    /// biggest modules first.
    pub fn cost(&self) -> u64 {
        match *self {
            // Only one module with fat LTO, so the cost doesn't matter.
            LtoModuleTranslation::Fat { .. } => 0,
            LtoModuleTranslation::Thin(ref m) => m.cost(),
        }
    }
}

pub enum LTOMode {
    WholeCrateGraph,
    JustThisCrate,
}

pub(crate) fn run(cgcx: &CodegenContext,
           modules: Vec<ModuleTranslation>,
           mode: LTOMode,
           timeline: &mut Timeline)
    -> Result<Vec<LtoModuleTranslation>, FatalError>
{
    let diag_handler = cgcx.create_diag_handler();
    let export_threshold = match mode {
        LTOMode::WholeCrateGraph => {
            symbol_export::crates_export_threshold(&cgcx.crate_types)
        }
        LTOMode::JustThisCrate => {
            SymbolExportLevel::Rust
        }
    };

    let symbol_filter = &|&(ref name, _, level): &(String, _, SymbolExportLevel)| {
        if level.is_below_threshold(export_threshold) {
            let mut bytes = Vec::with_capacity(name.len() + 1);
            bytes.extend(name.bytes());
            Some(CString::new(bytes).unwrap())
        } else {
            None
        }
    };

    let mut symbol_white_list = cgcx.exported_symbols[&LOCAL_CRATE]
        .iter()
        .filter_map(symbol_filter)
        .collect::<Vec<CString>>();
    timeline.record("whitelist");
    info!("{} symbols to preserve in this crate", symbol_white_list.len());

    // If we're performing LTO for the entire crate graph, then for each of our
    // upstream dependencies, find the corresponding rlib and load the bitcode
    // from the archive.
    //
    // We save off all the bytecode and LLVM module ids for later processing
    // with either fat or thin LTO
    let mut upstream_modules = Vec::new();
    if let LTOMode::WholeCrateGraph = mode {
        if cgcx.opts.cg.prefer_dynamic {
            diag_handler.struct_err("cannot prefer dynamic linking when performing LTO")
                        .note("only 'staticlib', 'bin', and 'cdylib' outputs are \
                               supported with LTO")
                        .emit();
            return Err(FatalError)
        }

        // Make sure we actually can run LTO
        for crate_type in cgcx.crate_types.iter() {
            if !crate_type_allows_lto(*crate_type) {
                let e = diag_handler.fatal("lto can only be run for executables, cdylibs and \
                                            static library outputs");
                return Err(e)
            }
        }

        for &(cnum, ref path) in cgcx.each_linked_rlib_for_lto.iter() {
            symbol_white_list.extend(
                cgcx.exported_symbols[&cnum]
                    .iter()
                    .filter_map(symbol_filter));

            let archive = ArchiveRO::open(&path).expect("wanted an rlib");
            let bytecodes = archive.iter().filter_map(|child| {
                child.ok().and_then(|c| c.name().map(|name| (name, c)))
            }).filter(|&(name, _)| name.ends_with(RLIB_BYTECODE_EXTENSION));
            for (name, data) in bytecodes {
                info!("adding bytecode {}", name);
                let bc_encoded = data.data();

                let (bc, id) = time(cgcx.time_passes, &format!("decode {}", name), || {
                    match DecodedBytecode::new(bc_encoded) {
                        Ok(b) => Ok((b.bytecode(), b.identifier().to_string())),
                        Err(e) => Err(diag_handler.fatal(&e)),
                    }
                })?;
                let bc = SerializedModule::FromRlib(bc);
                upstream_modules.push((bc, CString::new(id).unwrap()));
            }
            timeline.record(&format!("load: {}", path.display()));
        }
    }

    let arr = symbol_white_list.iter().map(|c| c.as_ptr()).collect::<Vec<_>>();
    match mode {
        LTOMode::WholeCrateGraph if !cgcx.thinlto => {
            fat_lto(cgcx, &diag_handler, modules, upstream_modules, &arr, timeline)
        }
        _ => {
            thin_lto(&diag_handler, modules, upstream_modules, &arr, timeline)
        }
    }
}

fn fat_lto(cgcx: &CodegenContext,
           diag_handler: &Handler,
           mut modules: Vec<ModuleTranslation>,
           mut serialized_modules: Vec<(SerializedModule, CString)>,
           symbol_white_list: &[*const libc::c_char],
           timeline: &mut Timeline)
    -> Result<Vec<LtoModuleTranslation>, FatalError>
{
    info!("going for a fat lto");

    // Find the "costliest" module and merge everything into that codegen unit.
    // All the other modules will be serialized and reparsed into the new
    // context, so this hopefully avoids serializing and parsing the largest
    // codegen unit.
    //
    // Additionally use a regular module as the base here to ensure that various
    // file copy operations in the backend work correctly. The only other kind
    // of module here should be an allocator one, and if your crate is smaller
    // than the allocator module then the size doesn't really matter anyway.
    let (_, costliest_module) = modules.iter()
        .enumerate()
        .filter(|&(_, module)| module.kind == ModuleKind::Regular)
        .map(|(i, module)| {
            let cost = unsafe {
                llvm::LLVMRustModuleCost(module.llvm().unwrap().llmod)
            };
            (cost, i)
        })
        .max()
        .expect("must be trans'ing at least one module");
    let module = modules.remove(costliest_module);
    let llmod = module.llvm().expect("can't lto pre-translated modules").llmod;
    info!("using {:?} as a base module", module.llmod_id);

    // For all other modules we translated we'll need to link them into our own
    // bitcode. All modules were translated in their own LLVM context, however,
    // and we want to move everything to the same LLVM context. Currently the
    // way we know of to do that is to serialize them to a string and them parse
    // them later. Not great but hey, that's why it's "fat" LTO, right?
    for module in modules {
        let llvm = module.llvm().expect("can't lto pre-translated modules");
        let buffer = ModuleBuffer::new(llvm.llmod);
        let llmod_id = CString::new(&module.llmod_id[..]).unwrap();
        serialized_modules.push((SerializedModule::Local(buffer), llmod_id));
    }

    // For all serialized bitcode files we parse them and link them in as we did
    // above, this is all mostly handled in C++. Like above, though, we don't
    // know much about the memory management here so we err on the side of being
    // save and persist everything with the original module.
    let mut serialized_bitcode = Vec::new();
    for (bc_decoded, name) in serialized_modules {
        info!("linking {:?}", name);
        time(cgcx.time_passes, &format!("ll link {:?}", name), || unsafe {
            let data = bc_decoded.data();
            if llvm::LLVMRustLinkInExternalBitcode(llmod,
                                                   data.as_ptr() as *const libc::c_char,
                                                   data.len() as libc::size_t) {
                Ok(())
            } else {
                let msg = format!("failed to load bc of {:?}", name);
                Err(write::llvm_err(&diag_handler, msg))
            }
        })?;
        timeline.record(&format!("link {:?}", name));
        serialized_bitcode.push(bc_decoded);
    }
    cgcx.save_temp_bitcode(&module, "lto.input");

    // Internalize everything that *isn't* in our whitelist to help strip out
    // more modules and such
    unsafe {
        let ptr = symbol_white_list.as_ptr();
        llvm::LLVMRustRunRestrictionPass(llmod,
                                         ptr as *const *const libc::c_char,
                                         symbol_white_list.len() as libc::size_t);
        cgcx.save_temp_bitcode(&module, "lto.after-restriction");
    }

    if cgcx.no_landing_pads {
        unsafe {
            llvm::LLVMRustMarkAllFunctionsNounwind(llmod);
        }
        cgcx.save_temp_bitcode(&module, "lto.after-nounwind");
    }
    timeline.record("passes");

    Ok(vec![LtoModuleTranslation::Fat {
        module: Some(module),
        _serialized_bitcode: serialized_bitcode,
    }])
}

/// Prepare "thin" LTO to get run on these modules.
///
/// The general structure of ThinLTO is quite different from the structure of
/// "fat" LTO above. With "fat" LTO all LLVM modules in question are merged into
/// one giant LLVM module, and then we run more optimization passes over this
/// big module after internalizing most symbols. Thin LTO, on the other hand,
/// avoid this large bottleneck through more targeted optimization.
///
/// At a high level Thin LTO looks like:
///
///     1. Prepare a "summary" of each LLVM module in question which describes
///        the values inside, cost of the values, etc.
///     2. Merge the summaries of all modules in question into one "index"
///     3. Perform some global analysis on this index
///     4. For each module, use the index and analysis calculated previously to
///        perform local transformations on the module, for example inlining
///        small functions from other modules.
///     5. Run thin-specific optimization passes over each module, and then code
///        generate everything at the end.
///
/// The summary for each module is intended to be quite cheap, and the global
/// index is relatively quite cheap to create as well. As a result, the goal of
/// ThinLTO is to reduce the bottleneck on LTO and enable LTO to be used in more
/// situations. For example one cheap optimization is that we can parallelize
/// all codegen modules, easily making use of all the cores on a machine.
///
/// With all that in mind, the function here is designed at specifically just
/// calculating the *index* for ThinLTO. This index will then be shared amongst
/// all of the `LtoModuleTranslation` units returned below and destroyed once
/// they all go out of scope.
fn thin_lto(diag_handler: &Handler,
            modules: Vec<ModuleTranslation>,
            serialized_modules: Vec<(SerializedModule, CString)>,
            symbol_white_list: &[*const libc::c_char],
            timeline: &mut Timeline)
    -> Result<Vec<LtoModuleTranslation>, FatalError>
{
    unsafe {
        info!("going for that thin, thin LTO");

        let mut thin_buffers = Vec::new();
        let mut module_names = Vec::new();
        let mut thin_modules = Vec::new();

        // FIXME: right now, like with fat LTO, we serialize all in-memory
        //        modules before working with them and ThinLTO. We really
        //        shouldn't do this, however, and instead figure out how to
        //        extract a summary from an in-memory module and then merge that
        //        into the global index. It turns out that this loop is by far
        //        the most expensive portion of this small bit of global
        //        analysis!
        for (i, module) in modules.iter().enumerate() {
            info!("local module: {} - {}", i, module.llmod_id);
            let llvm = module.llvm().expect("can't lto pretranslated module");
            let name = CString::new(module.llmod_id.clone()).unwrap();
            let buffer = ThinBuffer::new(llvm.llmod);
            thin_modules.push(llvm::ThinLTOModule {
                identifier: name.as_ptr(),
                data: buffer.data().as_ptr(),
                len: buffer.data().len(),
            });
            thin_buffers.push(buffer);
            module_names.push(name);
            timeline.record(&module.llmod_id);
        }

        // FIXME: All upstream crates are deserialized internally in the
        //        function below to extract their summary and modules. Note that
        //        unlike the loop above we *must* decode and/or read something
        //        here as these are all just serialized files on disk. An
        //        improvement, however, to make here would be to store the
        //        module summary separately from the actual module itself. Right
        //        now this is store in one large bitcode file, and the entire
        //        file is deflate-compressed. We could try to bypass some of the
        //        decompression by storing the index uncompressed and only
        //        lazily decompressing the bytecode if necessary.
        //
        //        Note that truly taking advantage of this optimization will
        //        likely be further down the road. We'd have to implement
        //        incremental ThinLTO first where we could actually avoid
        //        looking at upstream modules entirely sometimes (the contents,
        //        we must always unconditionally look at the index).
        let mut serialized = Vec::new();
        for (module, name) in serialized_modules {
            info!("foreign module {:?}", name);
            thin_modules.push(llvm::ThinLTOModule {
                identifier: name.as_ptr(),
                data: module.data().as_ptr(),
                len: module.data().len(),
            });
            serialized.push(module);
            module_names.push(name);
        }

        // Delegate to the C++ bindings to create some data here. Once this is a
        // tried-and-true interface we may wish to try to upstream some of this
        // to LLVM itself, right now we reimplement a lot of what they do
        // upstream...
        let data = llvm::LLVMRustCreateThinLTOData(
            thin_modules.as_ptr(),
            thin_modules.len() as u32,
            symbol_white_list.as_ptr(),
            symbol_white_list.len() as u32,
        );
        if data.is_null() {
            let msg = format!("failed to prepare thin LTO context");
            return Err(write::llvm_err(&diag_handler, msg))
        }
        let data = ThinData(data);
        info!("thin LTO data created");
        timeline.record("data");

        // Throw our data in an `Arc` as we'll be sharing it across threads. We
        // also put all memory referenced by the C++ data (buffers, ids, etc)
        // into the arc as well. After this we'll create a thin module
        // translation per module in this data.
        let shared = Arc::new(ThinShared {
            data,
            thin_buffers,
            serialized_modules: serialized,
            module_names,
        });
        Ok((0..shared.module_names.len()).map(|i| {
            LtoModuleTranslation::Thin(ThinModule {
                shared: shared.clone(),
                idx: i,
            })
        }).collect())
    }
}

fn run_pass_manager(cgcx: &CodegenContext,
                    tm: TargetMachineRef,
                    llmod: ModuleRef,
                    config: &ModuleConfig,
                    thin: bool) {
    // Now we have one massive module inside of llmod. Time to run the
    // LTO-specific optimization passes that LLVM provides.
    //
    // This code is based off the code found in llvm's LTO code generator:
    //      tools/lto/LTOCodeGenerator.cpp
    debug!("running the pass manager");
    unsafe {
        let pm = llvm::LLVMCreatePassManager();
        llvm::LLVMRustAddAnalysisPasses(tm, pm, llmod);
        let pass = llvm::LLVMRustFindAndCreatePass("verify\0".as_ptr() as *const _);
        assert!(!pass.is_null());
        llvm::LLVMRustAddPass(pm, pass);

        // When optimizing for LTO we don't actually pass in `-O0`, but we force
        // it to always happen at least with `-O1`.
        //
        // With ThinLTO we mess around a lot with symbol visibility in a way
        // that will actually cause linking failures if we optimize at O0 which
        // notable is lacking in dead code elimination. To ensure we at least
        // get some optimizations and correctly link we forcibly switch to `-O1`
        // to get dead code elimination.
        //
        // Note that in general this shouldn't matter too much as you typically
        // only turn on ThinLTO when you're compiling with optimizations
        // otherwise.
        let opt_level = config.opt_level.unwrap_or(llvm::CodeGenOptLevel::None);
        let opt_level = match opt_level {
            llvm::CodeGenOptLevel::None => llvm::CodeGenOptLevel::Less,
            level => level,
        };
        with_llvm_pmb(llmod, config, opt_level, &mut |b| {
            if thin {
                if !llvm::LLVMRustPassManagerBuilderPopulateThinLTOPassManager(b, pm) {
                    panic!("this version of LLVM does not support ThinLTO");
                }
            } else {
                llvm::LLVMPassManagerBuilderPopulateLTOPassManager(b, pm,
                    /* Internalize = */ False,
                    /* RunInliner = */ True);
            }
        });

        let pass = llvm::LLVMRustFindAndCreatePass("verify\0".as_ptr() as *const _);
        assert!(!pass.is_null());
        llvm::LLVMRustAddPass(pm, pass);

        time(cgcx.time_passes, "LTO passes", ||
             llvm::LLVMRunPassManager(pm, llmod));

        llvm::LLVMDisposePassManager(pm);
    }
    debug!("lto done");
}

pub enum SerializedModule {
    Local(ModuleBuffer),
    FromRlib(Vec<u8>),
}

impl SerializedModule {
    fn data(&self) -> &[u8] {
        match *self {
            SerializedModule::Local(ref m) => m.data(),
            SerializedModule::FromRlib(ref m) => m,
        }
    }
}

pub struct ModuleBuffer(*mut llvm::ModuleBuffer);

unsafe impl Send for ModuleBuffer {}
unsafe impl Sync for ModuleBuffer {}

impl ModuleBuffer {
    pub fn new(m: ModuleRef) -> ModuleBuffer {
        ModuleBuffer(unsafe {
            llvm::LLVMRustModuleBufferCreate(m)
        })
    }

    pub fn data(&self) -> &[u8] {
        unsafe {
            let ptr = llvm::LLVMRustModuleBufferPtr(self.0);
            let len = llvm::LLVMRustModuleBufferLen(self.0);
            slice::from_raw_parts(ptr, len)
        }
    }
}

impl Drop for ModuleBuffer {
    fn drop(&mut self) {
        unsafe { llvm::LLVMRustModuleBufferFree(self.0); }
    }
}

pub struct ThinModule {
    shared: Arc<ThinShared>,
    idx: usize,
}

struct ThinShared {
    data: ThinData,
    thin_buffers: Vec<ThinBuffer>,
    serialized_modules: Vec<SerializedModule>,
    module_names: Vec<CString>,
}

struct ThinData(*mut llvm::ThinLTOData);

unsafe impl Send for ThinData {}
unsafe impl Sync for ThinData {}

impl Drop for ThinData {
    fn drop(&mut self) {
        unsafe {
            llvm::LLVMRustFreeThinLTOData(self.0);
        }
    }
}

pub struct ThinBuffer(*mut llvm::ThinLTOBuffer);

unsafe impl Send for ThinBuffer {}
unsafe impl Sync for ThinBuffer {}

impl ThinBuffer {
    pub fn new(m: ModuleRef) -> ThinBuffer {
        unsafe {
            let buffer = llvm::LLVMRustThinLTOBufferCreate(m);
            ThinBuffer(buffer)
        }
    }

    pub fn data(&self) -> &[u8] {
        unsafe {
            let ptr = llvm::LLVMRustThinLTOBufferPtr(self.0) as *const _;
            let len = llvm::LLVMRustThinLTOBufferLen(self.0);
            slice::from_raw_parts(ptr, len)
        }
    }
}

impl Drop for ThinBuffer {
    fn drop(&mut self) {
        unsafe {
            llvm::LLVMRustThinLTOBufferFree(self.0);
        }
    }
}

impl ThinModule {
    fn name(&self) -> &str {
        self.shared.module_names[self.idx].to_str().unwrap()
    }

    fn cost(&self) -> u64 {
        // Yes, that's correct, we're using the size of the bytecode as an
        // indicator for how costly this codegen unit is.
        self.data().len() as u64
    }

    fn data(&self) -> &[u8] {
        let a = self.shared.thin_buffers.get(self.idx).map(|b| b.data());
        a.unwrap_or_else(|| {
            let len = self.shared.thin_buffers.len();
            self.shared.serialized_modules[self.idx - len].data()
        })
    }

    unsafe fn optimize(&mut self, cgcx: &CodegenContext, timeline: &mut Timeline)
        -> Result<ModuleTranslation, FatalError>
    {
        let diag_handler = cgcx.create_diag_handler();
        let tm = (cgcx.tm_factory)().map_err(|e| {
            write::llvm_err(&diag_handler, e)
        })?;

        // Right now the implementation we've got only works over serialized
        // modules, so we create a fresh new LLVM context and parse the module
        // into that context. One day, however, we may do this for upstream
        // crates but for locally translated modules we may be able to reuse
        // that LLVM Context and Module.
        let llcx = llvm::LLVMRustContextCreate(cgcx.fewer_names);
        let llmod = llvm::LLVMRustParseBitcodeForThinLTO(
            llcx,
            self.data().as_ptr(),
            self.data().len(),
            self.shared.module_names[self.idx].as_ptr(),
        );
        if llmod.is_null() {
            let msg = format!("failed to parse bitcode for thin LTO module");
            return Err(write::llvm_err(&diag_handler, msg));
        }
        let mtrans = ModuleTranslation {
            source: ModuleSource::Translated(ModuleLlvm {
                llmod,
                llcx,
                tm,
            }),
            llmod_id: self.name().to_string(),
            name: self.name().to_string(),
            kind: ModuleKind::Regular,
        };
        cgcx.save_temp_bitcode(&mtrans, "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() {
            let msg = format!("multiple source DICompileUnits found");
            return Err(write::llvm_err(&diag_handler, msg))
        }

        // Like with "fat" LTO, get some better optimizations if landing pads
        // are disabled by removing all landing pads.
        if cgcx.no_landing_pads {
            llvm::LLVMRustMarkAllFunctionsNounwind(llmod);
            cgcx.save_temp_bitcode(&mtrans, "thin-lto-after-nounwind");
            timeline.record("nounwind");
        }

        // 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
        // each of these would be supported by upstream LLVM but that's perhaps
        // a patch for another day!
        //
        // You can find some more comments about these functions in the LLVM
        // bindings we've got (currently `PassWrapper.cpp`)
        if !llvm::LLVMRustPrepareThinLTORename(self.shared.data.0, llmod) {
            let msg = format!("failed to prepare thin LTO module");
            return Err(write::llvm_err(&diag_handler, msg))
        }
        cgcx.save_temp_bitcode(&mtrans, "thin-lto-after-rename");
        timeline.record("rename");
        if !llvm::LLVMRustPrepareThinLTOResolveWeak(self.shared.data.0, llmod) {
            let msg = format!("failed to prepare thin LTO module");
            return Err(write::llvm_err(&diag_handler, msg))
        }
        cgcx.save_temp_bitcode(&mtrans, "thin-lto-after-resolve");
        timeline.record("resolve");
        if !llvm::LLVMRustPrepareThinLTOInternalize(self.shared.data.0, llmod) {
            let msg = format!("failed to prepare thin LTO module");
            return Err(write::llvm_err(&diag_handler, msg))
        }
        cgcx.save_temp_bitcode(&mtrans, "thin-lto-after-internalize");
        timeline.record("internalize");
        if !llvm::LLVMRustPrepareThinLTOImport(self.shared.data.0, llmod) {
            let msg = format!("failed to prepare thin LTO module");
            return Err(write::llvm_err(&diag_handler, msg))
        }
        cgcx.save_temp_bitcode(&mtrans, "thin-lto-after-import");
        timeline.record("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 tihs 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.
        llvm::LLVMRustThinLTOPatchDICompileUnit(llmod, cu1);
        cgcx.save_temp_bitcode(&mtrans, "thin-lto-after-patch");
        timeline.record("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
        // populate a thin-specific pass manager, which presumably LLVM treats a
        // little differently.
        info!("running thin lto passes over {}", mtrans.name);
        let config = cgcx.config(mtrans.kind);
        run_pass_manager(cgcx, tm, llmod, config, true);
        cgcx.save_temp_bitcode(&mtrans, "thin-lto-after-pm");
        timeline.record("thin-done");

        // FIXME: this is a hack around a bug in LLVM right now. Discovered in
        // #46910 it was found out that on 32-bit MSVC LLVM will hit a codegen
        // error if there's an available_externally function in the LLVM module.
        // Typically we don't actually use these functions but ThinLTO makes
        // heavy use of them when inlining across modules.
        //
        // Tracked upstream at https://bugs.llvm.org/show_bug.cgi?id=35736 this
        // function call (and its definition on the C++ side of things)
        // shouldn't be necessary eventually and we can safetly delete these few
        // lines.
        llvm::LLVMRustThinLTORemoveAvailableExternally(llmod);
        cgcx.save_temp_bitcode(&mtrans, "thin-lto-after-rm-ae");
        timeline.record("no-ae");

        Ok(mtrans)
    }
}