6 files changed, 44 insertions, 21 deletions
diff --git a/compiler/rustc_codegen_llvm/src/back/lto.rs b/compiler/rustc_codegen_llvm/src/back/lto.rs
index a8b49e9552c..925898d8173 100644
--- a/compiler/rustc_codegen_llvm/src/back/lto.rs
+++ b/compiler/rustc_codegen_llvm/src/back/lto.rs
@@ -584,12 +584,10 @@ fn thin_lto(
     }
 }
 
-fn enable_autodiff_settings(ad: &[config::AutoDiff], module: &mut ModuleCodegen<ModuleLlvm>) {
+fn enable_autodiff_settings(ad: &[config::AutoDiff]) {
     for &val in ad {
+        // We intentionally don't use a wildcard, to not forget handling anything new.
         match val {
-            config::AutoDiff::PrintModBefore => {
-                unsafe { llvm::LLVMDumpModule(module.module_llvm.llmod()) };
-            }
             config::AutoDiff::PrintPerf => {
                 llvm::set_print_perf(true);
             }
@@ -603,17 +601,23 @@ fn enable_autodiff_settings(ad: &[config::AutoDiff], module: &mut ModuleCodegen<
                 llvm::set_inline(true);
             }
             config::AutoDiff::LooseTypes => {
-                llvm::set_loose_types(false);
+                llvm::set_loose_types(true);
             }
             config::AutoDiff::PrintSteps => {
                 llvm::set_print(true);
             }
-            // We handle this below
+            // We handle this in the PassWrapper.cpp
+            config::AutoDiff::PrintPasses => {}
+            // We handle this in the PassWrapper.cpp
+            config::AutoDiff::PrintModBefore => {}
+            // We handle this in the PassWrapper.cpp
             config::AutoDiff::PrintModAfter => {}
-            // We handle this below
+            // We handle this in the PassWrapper.cpp
             config::AutoDiff::PrintModFinal => {}
             // This is required and already checked
             config::AutoDiff::Enable => {}
+            // We handle this below
+            config::AutoDiff::NoPostopt => {}
         }
     }
     // This helps with handling enums for now.
@@ -647,27 +651,27 @@ pub(crate) fn run_pass_manager(
     // We then run the llvm_optimize function a second time, to optimize the code which we generated
     // in the enzyme differentiation pass.
     let enable_ad = config.autodiff.contains(&config::AutoDiff::Enable);
-    let stage =
-        if enable_ad { write::AutodiffStage::DuringAD } else { write::AutodiffStage::PostAD };
+    let stage = if thin {
+        write::AutodiffStage::PreAD
+    } else {
+        if enable_ad { write::AutodiffStage::DuringAD } else { write::AutodiffStage::PostAD }
+    };
 
     if enable_ad {
-        enable_autodiff_settings(&config.autodiff, module);
+        enable_autodiff_settings(&config.autodiff);
     }
 
     unsafe {
         write::llvm_optimize(cgcx, dcx, module, None, config, opt_level, opt_stage, stage)?;
     }
 
-    if cfg!(llvm_enzyme) && enable_ad {
-        // This is the post-autodiff IR, mainly used for testing and educational purposes.
-        if config.autodiff.contains(&config::AutoDiff::PrintModAfter) {
-            unsafe { llvm::LLVMDumpModule(module.module_llvm.llmod()) };
-        }
-
+    if cfg!(llvm_enzyme) && enable_ad && !thin {
         let opt_stage = llvm::OptStage::FatLTO;
         let stage = write::AutodiffStage::PostAD;
-        unsafe {
-            write::llvm_optimize(cgcx, dcx, module, None, config, opt_level, opt_stage, stage)?;
+        if !config.autodiff.contains(&config::AutoDiff::NoPostopt) {
+            unsafe {
+                write::llvm_optimize(cgcx, dcx, module, None, config, opt_level, opt_stage, stage)?;
+            }
         }
 
         // This is the final IR, so people should be able to inspect the optimized autodiff output,
diff --git a/compiler/rustc_codegen_llvm/src/back/write.rs b/compiler/rustc_codegen_llvm/src/back/write.rs
index 18d221d232e..4ac77c8f7f1 100644
--- a/compiler/rustc_codegen_llvm/src/back/write.rs
+++ b/compiler/rustc_codegen_llvm/src/back/write.rs
@@ -572,6 +572,10 @@ pub(crate) unsafe fn llvm_optimize(
 
     let consider_ad = cfg!(llvm_enzyme) && config.autodiff.contains(&config::AutoDiff::Enable);
     let run_enzyme = autodiff_stage == AutodiffStage::DuringAD;
+    let print_before_enzyme = config.autodiff.contains(&config::AutoDiff::PrintModBefore);
+    let print_after_enzyme = config.autodiff.contains(&config::AutoDiff::PrintModAfter);
+    let print_passes = config.autodiff.contains(&config::AutoDiff::PrintPasses);
+    let merge_functions;
     let unroll_loops;
     let vectorize_slp;
     let vectorize_loop;
@@ -579,13 +583,20 @@ pub(crate) unsafe fn llvm_optimize(
     // When we build rustc with enzyme/autodiff support, we want to postpone size-increasing
     // optimizations until after differentiation. Our pipeline is thus: (opt + enzyme), (full opt).
     // We therefore have two calls to llvm_optimize, if autodiff is used.
+    //
+    // We also must disable merge_functions, since autodiff placeholder/dummy bodies tend to be
+    // identical. We run opts before AD, so there is a chance that LLVM will merge our dummies.
+    // In that case, we lack some dummy bodies and can't replace them with the real AD code anymore.
+    // We then would need to abort compilation. This was especially common in test cases.
     if consider_ad && autodiff_stage != AutodiffStage::PostAD {
+        merge_functions = false;
         unroll_loops = false;
         vectorize_slp = false;
         vectorize_loop = false;
     } else {
         unroll_loops =
             opt_level != config::OptLevel::Size && opt_level != config::OptLevel::SizeMin;
+        merge_functions = config.merge_functions;
         vectorize_slp = config.vectorize_slp;
         vectorize_loop = config.vectorize_loop;
     }
@@ -663,13 +674,16 @@ pub(crate) unsafe fn llvm_optimize(
             thin_lto_buffer,
             config.emit_thin_lto,
             config.emit_thin_lto_summary,
-            config.merge_functions,
+            merge_functions,
             unroll_loops,
             vectorize_slp,
             vectorize_loop,
             config.no_builtins,
             config.emit_lifetime_markers,
             run_enzyme,
+            print_before_enzyme,
+            print_after_enzyme,
+            print_passes,
             sanitizer_options.as_ref(),
             pgo_gen_path.as_ref().map_or(std::ptr::null(), |s| s.as_ptr()),
             pgo_use_path.as_ref().map_or(std::ptr::null(), |s| s.as_ptr()),
diff --git a/compiler/rustc_codegen_llvm/src/builder.rs b/compiler/rustc_codegen_llvm/src/builder.rs
index 27f7f95f100..04c8118b616 100644
--- a/compiler/rustc_codegen_llvm/src/builder.rs
+++ b/compiler/rustc_codegen_llvm/src/builder.rs
@@ -594,6 +594,7 @@ impl<'a, 'll, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'll, 'tcx> {
     fn load(&mut self, ty: &'ll Type, ptr: &'ll Value, align: Align) -> &'ll Value {
         unsafe {
             let load = llvm::LLVMBuildLoad2(self.llbuilder, ty, ptr, UNNAMED);
+            let align = align.min(self.cx().tcx.sess.target.max_reliable_alignment());
             llvm::LLVMSetAlignment(load, align.bytes() as c_uint);
             load
         }
@@ -807,6 +808,7 @@ impl<'a, 'll, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'll, 'tcx> {
         assert_eq!(self.cx.type_kind(self.cx.val_ty(ptr)), TypeKind::Pointer);
         unsafe {
             let store = llvm::LLVMBuildStore(self.llbuilder, val, ptr);
+            let align = align.min(self.cx().tcx.sess.target.max_reliable_alignment());
             let align =
                 if flags.contains(MemFlags::UNALIGNED) { 1 } else { align.bytes() as c_uint };
             llvm::LLVMSetAlignment(store, align);
diff --git a/compiler/rustc_codegen_llvm/src/builder/autodiff.rs b/compiler/rustc_codegen_llvm/src/builder/autodiff.rs
index e7c071d05aa..0147bd5a665 100644
--- a/compiler/rustc_codegen_llvm/src/builder/autodiff.rs
+++ b/compiler/rustc_codegen_llvm/src/builder/autodiff.rs
@@ -473,7 +473,7 @@ pub(crate) fn differentiate<'ll>(
         return Err(diag_handler.handle().emit_almost_fatal(AutoDiffWithoutEnable));
     }
 
-    // Before dumping the module, we want all the TypeTrees to become part of the module.
+    // Here we replace the placeholder code with the actual autodiff code, which calls Enzyme.
     for item in diff_items.iter() {
         let name = item.source.clone();
         let fn_def: Option<&llvm::Value> = cx.get_function(&name);
diff --git a/compiler/rustc_codegen_llvm/src/lib.rs b/compiler/rustc_codegen_llvm/src/lib.rs
index 425381b0ffa..b2feeacdb46 100644
--- a/compiler/rustc_codegen_llvm/src/lib.rs
+++ b/compiler/rustc_codegen_llvm/src/lib.rs
@@ -6,6 +6,7 @@
 
 // tidy-alphabetical-start
 #![allow(internal_features)]
+#![cfg_attr(bootstrap, feature(let_chains))]
 #![doc(html_root_url = "https://doc.rust-lang.org/nightly/nightly-rustc/")]
 #![doc(rust_logo)]
 #![feature(assert_matches)]
@@ -15,7 +16,6 @@
 #![feature(if_let_guard)]
 #![feature(impl_trait_in_assoc_type)]
 #![feature(iter_intersperse)]
-#![feature(let_chains)]
 #![feature(rustdoc_internals)]
 #![feature(slice_as_array)]
 #![feature(try_blocks)]
diff --git a/compiler/rustc_codegen_llvm/src/llvm/ffi.rs b/compiler/rustc_codegen_llvm/src/llvm/ffi.rs
index 9ff04f72903..ffb490dcdc2 100644
--- a/compiler/rustc_codegen_llvm/src/llvm/ffi.rs
+++ b/compiler/rustc_codegen_llvm/src/llvm/ffi.rs
@@ -2454,6 +2454,9 @@ unsafe extern "C" {
         DisableSimplifyLibCalls: bool,
         EmitLifetimeMarkers: bool,
         RunEnzyme: bool,
+        PrintBeforeEnzyme: bool,
+        PrintAfterEnzyme: bool,
+        PrintPasses: bool,
         SanitizerOptions: Option<&SanitizerOptions>,
         PGOGenPath: *const c_char,
         PGOUsePath: *const c_char,