llvm-context: modularize compiler builtin functions (#234)

- Add the revive runtime function interface to minimize boiler plate
code.
- Outline heavily repeated code into dedicated functions to bring down
code size.
- The code size tests builds optimized for size.
- Function attributes are passed as slices.

This significantly brings down the code size for all OpenZeppelin wizard
contracts (using all possible features) compiled against OpenZeppelin
`v5.0.0` with size optimizations.

|contract|| `-Oz` main | `-Oz` PR || `-O3` main | `-O3` PR |
|-|-|-|-|-|-|-|
|erc1155.sol||100K|67K||114K|147K|
|erc20.sol||120K|90K||160K|191K|
|erc721.sol||128K|101K||178K|214K|
|governor.sol||226K|165K||293K|349K|
|rwa.sol||116K|85K||154K|185K|
|stable.sol||116K|86K||155K|192K|

On the flip side this introduces a heavy penalty for cycle optimized
builds. Setting the no-inline attributes for cycle optimized builds
helps a lot but heavily penalizes runtime speed (LLVM does not yet
inline everything properly - to be investigated later on).

Next steps:
- Modularize more functions
- Refactor the YUL function arguments to use pointers instead of values
- Afterwards check if LLVM still has trouble inline-ing properly on O3
or set the no-inline attribute if it does not penalize runtime
performance too bad.

crates/llvm-context/src/polkavm/context/pointer/heap.rs

@@ -0,0 +1,110 @@
+//! The revive simulated EVM linear memory pointer functions.
+
+use inkwell::values::BasicValueEnum;
+
+use crate::polkavm::context::runtime::RuntimeFunction;
+use crate::polkavm::context::Context;
+use crate::polkavm::Dependency;
+use crate::polkavm::WriteLLVM;
+
+/// Load a word size value from a heap pointer.
+pub struct LoadWord;
+
+impl<D> RuntimeFunction<D> for LoadWord
+where
+    D: Dependency + Clone,
+{
+    const NAME: &'static str = "__revive_load_heap_word";
+
+    fn r#type<'ctx>(context: &Context<'ctx, D>) -> inkwell::types::FunctionType<'ctx> {
+        context
+            .word_type()
+            .fn_type(&[context.xlen_type().into()], false)
+    }
+
+    fn emit_body<'ctx>(
+        &self,
+        context: &mut Context<'ctx, D>,
+    ) -> anyhow::Result<Option<BasicValueEnum<'ctx>>> {
+        let offset = Self::paramater(context, 0).into_int_value();
+        let length = context
+            .xlen_type()
+            .const_int(revive_common::BYTE_LENGTH_WORD as u64, false);
+        let pointer = context.build_heap_gep(offset, length)?;
+        let value = context
+            .builder()
+            .build_load(context.word_type(), pointer.value, "value")?;
+        context
+            .basic_block()
+            .get_last_instruction()
+            .expect("Always exists")
+            .set_alignment(revive_common::BYTE_LENGTH_BYTE as u32)
+            .expect("Alignment is valid");
+
+        let swapped_value = context.build_byte_swap(value)?;
+        Ok(Some(swapped_value))
+    }
+}
+
+impl<D> WriteLLVM<D> for LoadWord
+where
+    D: Dependency + Clone,
+{
+    fn declare(&mut self, context: &mut Context<D>) -> anyhow::Result<()> {
+        <Self as RuntimeFunction<_>>::declare(self, context)
+    }
+
+    fn into_llvm(self, context: &mut Context<D>) -> anyhow::Result<()> {
+        <Self as RuntimeFunction<_>>::emit(&self, context)
+    }
+}
+
+/// Store a word size value through a heap pointer.
+pub struct StoreWord;
+
+impl<D> RuntimeFunction<D> for StoreWord
+where
+    D: Dependency + Clone,
+{
+    const NAME: &'static str = "__revive_store_heap_word";
+
+    fn r#type<'ctx>(context: &Context<'ctx, D>) -> inkwell::types::FunctionType<'ctx> {
+        context.void_type().fn_type(
+            &[context.xlen_type().into(), context.word_type().into()],
+            false,
+        )
+    }
+
+    fn emit_body<'ctx>(
+        &self,
+        context: &mut Context<'ctx, D>,
+    ) -> anyhow::Result<Option<BasicValueEnum<'ctx>>> {
+        let offset = Self::paramater(context, 0).into_int_value();
+        let length = context
+            .xlen_type()
+            .const_int(revive_common::BYTE_LENGTH_WORD as u64, false);
+        let pointer = context.build_heap_gep(offset, length)?;
+
+        let value = context.build_byte_swap(Self::paramater(context, 1))?;
+
+        context
+            .builder()
+            .build_store(pointer.value, value)?
+            .set_alignment(revive_common::BYTE_LENGTH_BYTE as u32)
+            .expect("Alignment is valid");
+        Ok(None)
+    }
+}
+
+impl<D> WriteLLVM<D> for StoreWord
+where
+    D: Dependency + Clone,
+{
+    fn declare(&mut self, context: &mut Context<D>) -> anyhow::Result<()> {
+        <Self as RuntimeFunction<_>>::declare(self, context)
+    }
+
+    fn into_llvm(self, context: &mut Context<D>) -> anyhow::Result<()> {
+        <Self as RuntimeFunction<_>>::emit(&self, context)
+    }
+}

crates/llvm-context/src/polkavm/context/pointer/mod.rs

@@ -0,0 +1,136 @@
+//! The LLVM pointer.
+
+use inkwell::types::BasicType;
+
+use crate::polkavm::context::address_space::AddressSpace;
+use crate::polkavm::context::global::Global;
+use crate::polkavm::context::Context;
+use crate::polkavm::Dependency;
+
+pub mod heap;
+pub mod storage;
+
+/// The LLVM pointer.
+#[derive(Debug, Clone, Copy)]
+pub struct Pointer<'ctx> {
+    /// The pointee type.
+    pub r#type: inkwell::types::BasicTypeEnum<'ctx>,
+    /// The address space.
+    pub address_space: AddressSpace,
+    /// The pointer value.
+    pub value: inkwell::values::PointerValue<'ctx>,
+}
+
+impl<'ctx> Pointer<'ctx> {
+    /// A shortcut constructor.
+    pub fn new<T>(
+        r#type: T,
+        address_space: AddressSpace,
+        value: inkwell::values::PointerValue<'ctx>,
+    ) -> Self
+    where
+        T: BasicType<'ctx>,
+    {
+        Self {
+            r#type: r#type.as_basic_type_enum(),
+            address_space,
+            value,
+        }
+    }
+
+    /// Wraps a 256-bit primitive type pointer.
+    pub fn new_stack_field<D>(
+        context: &Context<'ctx, D>,
+        value: inkwell::values::PointerValue<'ctx>,
+    ) -> Self
+    where
+        D: Dependency + Clone,
+    {
+        Self {
+            r#type: context.word_type().as_basic_type_enum(),
+            address_space: AddressSpace::Stack,
+            value,
+        }
+    }
+
+    /// Creates a new pointer with the specified `offset`.
+    pub fn new_with_offset<D, T>(
+        context: &Context<'ctx, D>,
+        address_space: AddressSpace,
+        r#type: T,
+        offset: inkwell::values::IntValue<'ctx>,
+        name: &str,
+    ) -> Self
+    where
+        D: Dependency + Clone,
+        T: BasicType<'ctx>,
+    {
+        assert_ne!(
+            address_space,
+            AddressSpace::Stack,
+            "Stack pointers cannot be addressed"
+        );
+
+        let offset = context.safe_truncate_int_to_xlen(offset).unwrap();
+        let value = context
+            .builder
+            .build_int_to_ptr(offset, context.llvm().ptr_type(address_space.into()), name)
+            .unwrap();
+        Self::new(r#type, address_space, value)
+    }
+
+    /// Casts the pointer into another type.
+    pub fn cast<T>(self, r#type: T) -> Self
+    where
+        T: BasicType<'ctx>,
+    {
+        Self {
+            r#type: r#type.as_basic_type_enum(),
+            address_space: self.address_space,
+            value: self.value,
+        }
+    }
+
+    /// Cast this pointer to a register sized integer value.
+    pub fn to_int<D>(&self, context: &Context<'ctx, D>) -> inkwell::values::IntValue<'ctx>
+    where
+        D: Dependency + Clone,
+    {
+        context
+            .builder()
+            .build_ptr_to_int(self.value, context.xlen_type(), "ptr_to_xlen")
+            .expect("we should be positioned")
+    }
+
+    pub fn address_space_cast<D>(
+        self,
+        context: &Context<'ctx, D>,
+        address_space: AddressSpace,
+        name: &str,
+    ) -> anyhow::Result<Self>
+    where
+        D: Dependency + Clone,
+    {
+        let value = context.builder().build_address_space_cast(
+            self.value,
+            context.llvm().ptr_type(address_space.into()),
+            name,
+        )?;
+
+        Ok(Self {
+            address_space,
+            value,
+            ..self
+        })
+    }
+}
+
+impl<'ctx> From<Global<'ctx>> for Pointer<'ctx> {
+    fn from(global: Global<'ctx>) -> Self {
+        Self {
+            r#type: global.r#type,
+            address_space: AddressSpace::Stack,
+            value: global.value.as_pointer_value(),
+        }
+    }
+}

crates/llvm-context/src/polkavm/context/pointer/storage.rs

@@ -0,0 +1,135 @@
+//! The revive storage pointer functions.
+
+use inkwell::values::BasicValueEnum;
+
+use crate::polkavm::context::runtime::RuntimeFunction;
+use crate::polkavm::context::Context;
+use crate::polkavm::Dependency;
+use crate::polkavm::WriteLLVM;
+
+/// Load a word size value from a storage pointer.
+pub struct LoadWord;
+
+impl<D> RuntimeFunction<D> for LoadWord
+where
+    D: Dependency + Clone,
+{
+    const NAME: &'static str = "__revive_load_storage_word";
+
+    fn r#type<'ctx>(context: &Context<'ctx, D>) -> inkwell::types::FunctionType<'ctx> {
+        context.word_type().fn_type(
+            &[context.xlen_type().into(), context.word_type().into()],
+            false,
+        )
+    }
+
+    fn emit_body<'ctx>(
+        &self,
+        context: &mut Context<'ctx, D>,
+    ) -> anyhow::Result<Option<BasicValueEnum<'ctx>>> {
+        let is_transient = Self::paramater(context, 0);
+        let key_value = Self::paramater(context, 1);
+
+        let key_pointer = context.build_alloca_at_entry(context.word_type(), "key_pointer");
+        let value_pointer = context.build_alloca_at_entry(context.word_type(), "value_pointer");
+        let length_pointer = context.build_alloca_at_entry(context.xlen_type(), "length_pointer");
+
+        context
+            .builder()
+            .build_store(key_pointer.value, key_value)?;
+        context.build_store(value_pointer, context.word_const(0))?;
+        context.build_store(
+            length_pointer,
+            context
+                .xlen_type()
+                .const_int(revive_common::BYTE_LENGTH_WORD as u64, false),
+        )?;
+
+        let arguments = [
+            is_transient,
+            key_pointer.to_int(context).into(),
+            context.xlen_type().const_all_ones().into(),
+            value_pointer.to_int(context).into(),
+            length_pointer.to_int(context).into(),
+        ];
+        context.build_runtime_call(revive_runtime_api::polkavm_imports::GET_STORAGE, &arguments);
+
+        // We do not to check the return value: Solidity assumes infallible loads.
+        // If a key doesn't exist the "zero" value is returned (ensured by above write).
+
+        Ok(Some(context.build_load(value_pointer, "storage_value")?))
+    }
+}
+
+impl<D> WriteLLVM<D> for LoadWord
+where
+    D: Dependency + Clone,
+{
+    fn declare(&mut self, context: &mut Context<D>) -> anyhow::Result<()> {
+        <Self as RuntimeFunction<_>>::declare(self, context)
+    }
+
+    fn into_llvm(self, context: &mut Context<D>) -> anyhow::Result<()> {
+        <Self as RuntimeFunction<_>>::emit(&self, context)
+    }
+}
+
+/// Store a word size value through a storage pointer.
+pub struct StoreWord;
+
+impl<D> RuntimeFunction<D> for StoreWord
+where
+    D: Dependency + Clone,
+{
+    const NAME: &'static str = "__revive_store_storage_word";
+
+    fn r#type<'ctx>(context: &Context<'ctx, D>) -> inkwell::types::FunctionType<'ctx> {
+        context.void_type().fn_type(
+            &[
+                context.xlen_type().into(),
+                context.word_type().into(),
+                context.word_type().into(),
+            ],
+            false,
+        )
+    }
+
+    fn emit_body<'ctx>(
+        &self,
+        context: &mut Context<'ctx, D>,
+    ) -> anyhow::Result<Option<BasicValueEnum<'ctx>>> {
+        let is_transient = Self::paramater(context, 0);
+        let key = Self::paramater(context, 1);
+        let value = Self::paramater(context, 2);
+
+        let key_pointer = context.build_alloca_at_entry(context.word_type(), "key_pointer");
+        let value_pointer = context.build_alloca_at_entry(context.word_type(), "value_pointer");
+
+        context.build_store(key_pointer, key)?;
+        context.build_store(value_pointer, value)?;
+
+        let arguments = [
+            is_transient,
+            key_pointer.to_int(context).into(),
+            context.xlen_type().const_all_ones().into(),
+            value_pointer.to_int(context).into(),
+            context.integer_const(crate::polkavm::XLEN, 32).into(),
+        ];
+        context.build_runtime_call(revive_runtime_api::polkavm_imports::SET_STORAGE, &arguments);
+
+        Ok(None)
+    }
+}
+
+impl<D> WriteLLVM<D> for StoreWord
+where
+    D: Dependency + Clone,
+{
+    fn declare(&mut self, context: &mut Context<D>) -> anyhow::Result<()> {
+        <Self as RuntimeFunction<_>>::declare(self, context)
+    }
+
+    fn into_llvm(self, context: &mut Context<D>) -> anyhow::Result<()> {
+        <Self as RuntimeFunction<_>>::emit(&self, context)
+    }
+}