Rollup merge of #144197 - KMJ-007:type-tree, r=ZuseZ4

TypeTree support in autodiff

# TypeTrees for Autodiff

## What are TypeTrees?
Memory layout descriptors for Enzyme. Tell Enzyme exactly how types are structured in memory so it can compute derivatives efficiently.

## Structure
```rust
TypeTree(Vec<Type>)

Type {
    offset: isize,  // byte offset (-1 = everywhere)
    size: usize,    // size in bytes
    kind: Kind,     // Float, Integer, Pointer, etc.
    child: TypeTree // nested structure
}
```

## Example: `fn compute(x: &f32, data: &[f32]) -> f32`

**Input 0: `x: &f32`**
```rust
TypeTree(vec![Type {
    offset: -1, size: 8, kind: Pointer,
    child: TypeTree(vec![Type {
        offset: -1, size: 4, kind: Float,
        child: TypeTree::new()
    }])
}])
```

**Input 1: `data: &[f32]`**
```rust
TypeTree(vec![Type {
    offset: -1, size: 8, kind: Pointer,
    child: TypeTree(vec![Type {
        offset: -1, size: 4, kind: Float,  // -1 = all elements
        child: TypeTree::new()
    }])
}])
```

**Output: `f32`**
```rust
TypeTree(vec![Type {
    offset: -1, size: 4, kind: Float,
    child: TypeTree::new()
}])
```

## Why Needed?
- Enzyme can't deduce complex type layouts from LLVM IR
- Prevents slow memory pattern analysis
- Enables correct derivative computation for nested structures
- Tells Enzyme which bytes are differentiable vs metadata

## What Enzyme Does With This Information:

Without TypeTrees (current state):
```llvm
; Enzyme sees generic LLVM IR:
define float ``@distance(ptr*`` %p1, ptr* %p2) {
; Has to guess what these pointers point to
; Slow analysis of all memory operations
; May miss optimization opportunities
}
```

With TypeTrees (our implementation):
```llvm
define "enzyme_type"="{[]:Float@float}" float ``@distance(``
    ptr "enzyme_type"="{[]:Pointer}" %p1,
    ptr "enzyme_type"="{[]:Pointer}" %p2
) {
; Enzyme knows exact type layout
; Can generate efficient derivative code directly
}
```

# TypeTrees - Offset and -1 Explained

## Type Structure

```rust
Type {
    offset: isize, // WHERE this type starts
    size: usize,   // HOW BIG this type is
    kind: Kind,    // WHAT KIND of data (Float, Int, Pointer)
    child: TypeTree // WHAT'S INSIDE (for pointers/containers)
}
```

## Offset Values

### Regular Offset (0, 4, 8, etc.)
**Specific byte position within a structure**

```rust
struct Point {
    x: f32, // offset 0, size 4
    y: f32, // offset 4, size 4
    id: i32, // offset 8, size 4
}
```

TypeTree for `&Point` (internal representation):
```rust
TypeTree(vec![
    Type { offset: 0, size: 4, kind: Float },   // x at byte 0
    Type { offset: 4, size: 4, kind: Float },   // y at byte 4
    Type { offset: 8, size: 4, kind: Integer }  // id at byte 8
])
```

Generates LLVM:
```llvm
"enzyme_type"="{[]:Float@float}"
```

### Offset -1 (Special: "Everywhere")
**Means "this pattern repeats for ALL elements"**

#### Example 1: Array `[f32; 100]`
```rust
TypeTree(vec![Type {
    offset: -1, // ALL positions
    size: 4,    // each f32 is 4 bytes
    kind: Float, // every element is float
}])
```

Instead of listing 100 separate Types with offsets `0,4,8,12...396`

#### Example 2: Slice `&[i32]`
```rust
// Pointer to slice data
TypeTree(vec![Type {
    offset: -1, size: 8, kind: Pointer,
    child: TypeTree(vec![Type {
        offset: -1, // ALL slice elements
        size: 4,    // each i32 is 4 bytes
        kind: Integer
    }])
}])
```

#### Example 3: Mixed Structure
```rust
struct Container {
    header: i64,        // offset 0
    data: [f32; 1000],  // offset 8, but elements use -1
}
```

```rust
TypeTree(vec![
    Type { offset: 0, size: 8, kind: Integer }, // header
    Type { offset: 8, size: 4000, kind: Pointer,
        child: TypeTree(vec![Type {
            offset: -1, size: 4, kind: Float // ALL array elements
        }])
    }
])
```

Author: matthiaskrgr

compiler/rustc_ast/src/expand/autodiff_attrs.rs

@@ -6,6 +6,7 @@
 use std::fmt::{self, Display, Formatter};
 use std::str::FromStr;
 
+use crate::expand::typetree::TypeTree;
 use crate::expand::{Decodable, Encodable, HashStable_Generic};
 use crate::{Ty, TyKind};
 
@@ -84,6 +85,8 @@ pub struct AutoDiffItem {
     /// The name of the function being generated
     pub target: String,
     pub attrs: AutoDiffAttrs,
+    pub inputs: Vec<TypeTree>,
+    pub output: TypeTree,
 }
 
 #[derive(Clone, Eq, PartialEq, Encodable, Decodable, Debug, HashStable_Generic)]
@@ -275,14 +278,22 @@ impl AutoDiffAttrs {
         !matches!(self.mode, DiffMode::Error | DiffMode::Source)
     }
 
-    pub fn into_item(self, source: String, target: String) -> AutoDiffItem {
-        AutoDiffItem { source, target, attrs: self }
+    pub fn into_item(
+        self,
+        source: String,
+        target: String,
+        inputs: Vec<TypeTree>,
+        output: TypeTree,
+    ) -> AutoDiffItem {
+        AutoDiffItem { source, target, inputs, output, attrs: self }
     }
 }
 
 impl fmt::Display for AutoDiffItem {
     fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         write!(f, "Differentiating {} -> {}", self.source, self.target)?;
-        write!(f, " with attributes: {:?}", self.attrs)
+        write!(f, " with attributes: {:?}", self.attrs)?;
+        write!(f, " with inputs: {:?}", self.inputs)?;
+        write!(f, " with output: {:?}", self.output)
     }
 }

compiler/rustc_ast/src/expand/typetree.rs

@@ -31,6 +31,7 @@ pub enum Kind {
     Half,
     Float,
     Double,
+    F128,
     Unknown,
 }
 

compiler/rustc_codegen_gcc/src/builder.rs

@@ -1383,6 +1383,7 @@ impl<'a, 'gcc, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'gcc, 'tcx> {
         _src_align: Align,
         size: RValue<'gcc>,
         flags: MemFlags,
+        _tt: Option<rustc_ast::expand::typetree::FncTree>, // Autodiff TypeTrees are LLVM-only, ignored in GCC backend
     ) {
         assert!(!flags.contains(MemFlags::NONTEMPORAL), "non-temporal memcpy not supported");
         let size = self.intcast(size, self.type_size_t(), false);

compiler/rustc_codegen_gcc/src/intrinsic/mod.rs

@@ -770,6 +770,7 @@ impl<'gcc, 'tcx> ArgAbiExt<'gcc, 'tcx> for ArgAbi<'tcx, Ty<'tcx>> {
                     scratch_align,
                     bx.const_usize(self.layout.size.bytes()),
                     MemFlags::empty(),
+                    None,
                 );
 
                 bx.lifetime_end(scratch, scratch_size);

compiler/rustc_codegen_llvm/src/abi.rs

@@ -246,6 +246,7 @@ impl<'ll, 'tcx> ArgAbiExt<'ll, 'tcx> for ArgAbi<'tcx, Ty<'tcx>> {
                     scratch_align,
                     bx.const_usize(copy_bytes),
                     MemFlags::empty(),
+                    None,
                 );
                 bx.lifetime_end(llscratch, scratch_size);
             }

compiler/rustc_codegen_llvm/src/back/lto.rs

@@ -563,6 +563,8 @@ fn enable_autodiff_settings(ad: &[config::AutoDiff]) {
             config::AutoDiff::Enable => {}
             // We handle this below
             config::AutoDiff::NoPostopt => {}
+            // Disables TypeTree generation
+            config::AutoDiff::NoTT => {}
         }
     }
     // This helps with handling enums for now.

compiler/rustc_codegen_llvm/src/builder.rs

@@ -2,6 +2,7 @@ use std::borrow::{Borrow, Cow};
 use std::ops::Deref;
 use std::{iter, ptr};
 
+use rustc_ast::expand::typetree::FncTree;
 pub(crate) mod autodiff;
 pub(crate) mod gpu_offload;
 
@@ -1107,11 +1108,12 @@ impl<'a, 'll, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'll, 'tcx> {
         src_align: Align,
         size: &'ll Value,
         flags: MemFlags,
+        tt: Option<FncTree>,
     ) {
         assert!(!flags.contains(MemFlags::NONTEMPORAL), "non-temporal memcpy not supported");
         let size = self.intcast(size, self.type_isize(), false);
         let is_volatile = flags.contains(MemFlags::VOLATILE);
-        unsafe {
+        let memcpy = unsafe {
             llvm::LLVMRustBuildMemCpy(
                 self.llbuilder,
                 dst,
@@ -1120,7 +1122,16 @@ impl<'a, 'll, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'll, 'tcx> {
                 src_align.bytes() as c_uint,
                 size,
                 is_volatile,
-            );
+            )
+        };
+
+        // TypeTree metadata for memcpy is especially important: when Enzyme encounters
+        // a memcpy during autodiff, it needs to know the structure of the data being
+        // copied to properly track derivatives. For example, copying an array of floats
+        // vs. copying a struct with mixed types requires different derivative handling.
+        // The TypeTree tells Enzyme exactly what memory layout to expect.
+        if let Some(tt) = tt {
+            crate::typetree::add_tt(self.cx().llmod, self.cx().llcx, memcpy, tt);
         }
     }
 

compiler/rustc_codegen_llvm/src/builder/autodiff.rs

@@ -1,6 +1,7 @@
 use std::ptr;
 
 use rustc_ast::expand::autodiff_attrs::{AutoDiffAttrs, DiffActivity, DiffMode};
+use rustc_ast::expand::typetree::FncTree;
 use rustc_codegen_ssa::common::TypeKind;
 use rustc_codegen_ssa::traits::{BaseTypeCodegenMethods, BuilderMethods};
 use rustc_middle::ty::{Instance, PseudoCanonicalInput, TyCtxt, TypingEnv};
@@ -294,6 +295,7 @@ pub(crate) fn generate_enzyme_call<'ll, 'tcx>(
     fn_args: &[&'ll Value],
     attrs: AutoDiffAttrs,
     dest: PlaceRef<'tcx, &'ll Value>,
+    fnc_tree: FncTree,
 ) {
     // We have to pick the name depending on whether we want forward or reverse mode autodiff.
     let mut ad_name: String = match attrs.mode {
@@ -370,6 +372,10 @@ pub(crate) fn generate_enzyme_call<'ll, 'tcx>(
         fn_args,
     );
 
+    if !fnc_tree.args.is_empty() || !fnc_tree.ret.0.is_empty() {
+        crate::typetree::add_tt(cx.llmod, cx.llcx, fn_to_diff, fnc_tree);
+    }
+
     let call = builder.call(enzyme_ty, None, None, ad_fn, &args, None, None);
 
     builder.store_to_place(call, dest.val);

compiler/rustc_codegen_llvm/src/intrinsic.rs

@@ -1212,6 +1212,9 @@ fn codegen_autodiff<'ll, 'tcx>(
         &mut diff_attrs.input_activity,
     );
 
+    let fnc_tree =
+        rustc_middle::ty::fnc_typetrees(tcx, fn_source.ty(tcx, TypingEnv::fully_monomorphized()));
+
     // Build body
     generate_enzyme_call(
         bx,
@@ -1222,6 +1225,7 @@ fn codegen_autodiff<'ll, 'tcx>(
         &val_arr,
         diff_attrs.clone(),
         result,
+        fnc_tree,
     );
 }
 

compiler/rustc_codegen_llvm/src/lib.rs

@@ -68,6 +68,7 @@ mod llvm_util;
 mod mono_item;
 mod type_;
 mod type_of;
+mod typetree;
 mod va_arg;
 mod value;