chore: builtin wrapping shift left (#3270)

Co-authored-by: jfecher <[email protected]>
Co-authored-by: Tom French <[email protected]>
Co-authored-by: kek kek kek <[email protected]>
Co-authored-by: jfecher <[email protected]>
Co-authored-by: José Pedro Sousa <[email protected]>
Co-authored-by: Álvaro Rodríguez <[email protected]>
Co-authored-by: kevaundray <[email protected]>
Co-authored-by: josh crites <[email protected]>
Co-authored-by: Gustavo Giráldez <[email protected]>
Co-authored-by: Alex Gherghisan <[email protected]>
Co-authored-by: Savio <[email protected]>
Co-authored-by: github-merge-queue[bot] <github-merge-queue[bot]@users.noreply.github.com>
Co-authored-by: Martin Verzilli <[email protected]>
Co-authored-by: vezenovm <[email protected]>
Co-authored-by: Jan Beneš <[email protected]>
Co-authored-by: Maddiaa <[email protected]>
Co-authored-by: Tom French <[email protected]>

compiler/noirc_evaluator/src/ssa/function_builder/mod.rs

@@ -1,4 +1,4 @@
-use std::borrow::Cow;
+use std::{borrow::Cow, rc::Rc};
 
 use acvm::FieldElement;
 use noirc_errors::Location;
@@ -16,7 +16,8 @@ use super::{
         basic_block::BasicBlock,
         dfg::{CallStack, InsertInstructionResult},
         function::RuntimeType,
-        instruction::{InstructionId, Intrinsic},
+        instruction::{Endian, InstructionId, Intrinsic},
+        types::NumericType,
     },
     ssa_gen::Ssa,
 };
@@ -259,9 +260,126 @@ impl FunctionBuilder {
         arguments: Vec<ValueId>,
         result_types: Vec<Type>,
     ) -> Cow<[ValueId]> {
+        if let Value::Intrinsic(intrinsic) = &self.current_function.dfg[func] {
+            if intrinsic == &Intrinsic::WrappingShiftLeft {
+                let result_type = self.current_function.dfg.type_of_value(arguments[0]);
+                let bit_size = match result_type {
+                    Type::Numeric(NumericType::Signed { bit_size })
+                    | Type::Numeric(NumericType::Unsigned { bit_size }) => bit_size,
+                    _ => {
+                        unreachable!("ICE: Truncation attempted on non-integer");
+                    }
+                };
+                return self
+                    .insert_wrapping_shift_left(arguments[0], arguments[1], bit_size)
+                    .results();
+            }
+        }
+
         self.insert_instruction(Instruction::Call { func, arguments }, Some(result_types)).results()
     }
 
+    /// Insert ssa instructions which computes lhs << rhs by doing lhs*2^rhs
+    pub(crate) fn insert_shift_left(&mut self, lhs: ValueId, rhs: ValueId) -> ValueId {
+        let base = self.field_constant(FieldElement::from(2_u128));
+        let pow = self.pow(base, rhs);
+        let typ = self.current_function.dfg.type_of_value(lhs);
+        let pow = self.insert_cast(pow, typ);
+        self.insert_binary(lhs, BinaryOp::Mul, pow)
+    }
+
+    /// Insert ssa instructions which computes lhs << rhs by doing lhs*2^rhs
+    /// and truncate the result to bit_size
+    fn insert_wrapping_shift_left(
+        &mut self,
+        lhs: ValueId,
+        rhs: ValueId,
+        bit_size: u32,
+    ) -> InsertInstructionResult {
+        let base = self.field_constant(FieldElement::from(2_u128));
+        let typ = self.current_function.dfg.type_of_value(lhs);
+        let (max_bit, pow) = if let Some(rhs_constant) =
+            self.current_function.dfg.get_numeric_constant(rhs)
+        {
+            // Happy case is that we know precisely by how many bits the the integer will
+            // increase: lhs_bit_size + rhs
+            let (rhs_bit_size_pow_2, overflows) =
+                2_u32.overflowing_pow(rhs_constant.to_u128() as u32);
+            if overflows {
+                let zero = self.numeric_constant(FieldElement::zero(), typ);
+                return InsertInstructionResult::SimplifiedTo(zero);
+            }
+            let pow = self.numeric_constant(FieldElement::from(rhs_bit_size_pow_2 as u128), typ);
+            (bit_size + (rhs_constant.to_u128() as u32), pow)
+        } else {
+            // we use a predicate to nullify the result in case of overflow
+            let bit_size_var =
+                self.numeric_constant(FieldElement::from(bit_size as u128), typ.clone());
+            let overflow = self.insert_binary(rhs, BinaryOp::Lt, bit_size_var);
+            let one = self.numeric_constant(FieldElement::one(), Type::unsigned(1));
+            let predicate = self.insert_binary(overflow, BinaryOp::Eq, one);
+            let predicate = self.insert_cast(predicate, typ.clone());
+
+            let pow = self.pow(base, rhs);
+            let pow = self.insert_cast(pow, typ);
+            (FieldElement::max_num_bits(), self.insert_binary(predicate, BinaryOp::Mul, pow))
+        };
+
+        let instruction = Instruction::Binary(Binary { lhs, rhs: pow, operator: BinaryOp::Mul });
+        if max_bit <= bit_size {
+            self.insert_instruction(instruction, None)
+        } else {
+            let result = self.insert_instruction(instruction, None).first();
+            self.insert_instruction(
+                Instruction::Truncate { value: result, bit_size, max_bit_size: max_bit },
+                None,
+            )
+        }
+    }
+
+    /// Insert ssa instructions which computes lhs >> rhs by doing lhs/2^rhs
+    pub(crate) fn insert_shift_right(&mut self, lhs: ValueId, rhs: ValueId) -> ValueId {
+        let base = self.field_constant(FieldElement::from(2_u128));
+        let pow = self.pow(base, rhs);
+        self.insert_binary(lhs, BinaryOp::Div, pow)
+    }
+
+    /// Computes lhs^rhs via square&multiply, using the bits decomposition of rhs
+    /// Pseudo-code of the computation:
+    /// let mut r = 1;
+    /// let rhs_bits = to_bits(rhs);
+    /// for i in 1 .. bit_size + 1 {
+    ///     let r_squared = r * r;
+    ///     let b = rhs_bits[bit_size - i];
+    ///     r = (r_squared * lhs * b) + (1 - b) * r_squared;
+    /// }
+    pub(crate) fn pow(&mut self, lhs: ValueId, rhs: ValueId) -> ValueId {
+        let typ = self.current_function.dfg.type_of_value(rhs);
+        if let Type::Numeric(NumericType::Unsigned { bit_size }) = typ {
+            let to_bits = self.import_intrinsic_id(Intrinsic::ToBits(Endian::Little));
+            let length = self.field_constant(FieldElement::from(bit_size as i128));
+            let result_types =
+                vec![Type::field(), Type::Array(Rc::new(vec![Type::bool()]), bit_size as usize)];
+            let rhs_bits = self.insert_call(to_bits, vec![rhs, length], result_types);
+            let rhs_bits = rhs_bits[1];
+            let one = self.field_constant(FieldElement::one());
+            let mut r = one;
+            for i in 1..bit_size + 1 {
+                let r_squared = self.insert_binary(r, BinaryOp::Mul, r);
+                let a = self.insert_binary(r_squared, BinaryOp::Mul, lhs);
+                let idx = self.field_constant(FieldElement::from((bit_size - i) as i128));
+                let b = self.insert_array_get(rhs_bits, idx, Type::field());
+                let r1 = self.insert_binary(a, BinaryOp::Mul, b);
+                let c = self.insert_binary(one, BinaryOp::Sub, b);
+                let r2 = self.insert_binary(c, BinaryOp::Mul, r_squared);
+                r = self.insert_binary(r1, BinaryOp::Add, r2);
+            }
+            r
+        } else {
+            unreachable!("Value must be unsigned in power operation");
+        }
+    }
+
     /// Insert an instruction to extract an element from an array
     pub(crate) fn insert_array_get(
         &mut self,

compiler/noirc_evaluator/src/ssa/ir/instruction.rs

@@ -46,6 +46,7 @@ pub(crate) enum Intrinsic {
     BlackBox(BlackBoxFunc),
     FromField,
     AsField,
+    WrappingShiftLeft,
 }
 
 impl std::fmt::Display for Intrinsic {
@@ -68,6 +69,7 @@ impl std::fmt::Display for Intrinsic {
             Intrinsic::BlackBox(function) => write!(f, "{function}"),
             Intrinsic::FromField => write!(f, "from_field"),
             Intrinsic::AsField => write!(f, "as_field"),
+            Intrinsic::WrappingShiftLeft => write!(f, "wrapping_shift_left"),
         }
     }
 }
@@ -92,7 +94,8 @@ impl Intrinsic {
             | Intrinsic::ToBits(_)
             | Intrinsic::ToRadix(_)
             | Intrinsic::FromField
-            | Intrinsic::AsField => false,
+            | Intrinsic::AsField
+            | Intrinsic::WrappingShiftLeft => false,
 
             // Some black box functions have side-effects
             Intrinsic::BlackBox(func) => matches!(func, BlackBoxFunc::RecursiveAggregation),
@@ -119,6 +122,7 @@ impl Intrinsic {
             "to_be_bits" => Some(Intrinsic::ToBits(Endian::Big)),
             "from_field" => Some(Intrinsic::FromField),
             "as_field" => Some(Intrinsic::AsField),
+            "wrapping_shift_left" => Some(Intrinsic::WrappingShiftLeft),
             other => BlackBoxFunc::lookup(other).map(Intrinsic::BlackBox),
         }
     }

compiler/noirc_evaluator/src/ssa/ir/instruction/call.rs

@@ -245,6 +245,9 @@ pub(super) fn simplify_call(
             let instruction = Instruction::Cast(arguments[0], ctrl_typevars.unwrap().remove(0));
             SimplifyResult::SimplifiedToInstruction(instruction)
         }
+        Intrinsic::WrappingShiftLeft => {
+            unreachable!("ICE - wrapping shift left should have been proccessed before")
+        }
     }
 }
 

compiler/noirc_evaluator/src/ssa/ssa_gen/context.rs

@@ -13,7 +13,7 @@ use crate::ssa::function_builder::FunctionBuilder;
 use crate::ssa::ir::dfg::DataFlowGraph;
 use crate::ssa::ir::function::FunctionId as IrFunctionId;
 use crate::ssa::ir::function::{Function, RuntimeType};
-use crate::ssa::ir::instruction::{BinaryOp, Endian, Intrinsic};
+use crate::ssa::ir::instruction::BinaryOp;
 use crate::ssa::ir::map::AtomicCounter;
 use crate::ssa::ir::types::{NumericType, Type};
 use crate::ssa::ir::value::ValueId;
@@ -265,50 +265,6 @@ impl<'a> FunctionContext<'a> {
         Ok(self.builder.numeric_constant(value, typ))
     }
 
-    /// Insert ssa instructions which computes lhs << rhs by doing lhs*2^rhs
-    fn insert_shift_left(&mut self, lhs: ValueId, rhs: ValueId) -> ValueId {
-        let base = self.builder.field_constant(FieldElement::from(2_u128));
-        let pow = self.pow(base, rhs);
-        let typ = self.builder.current_function.dfg.type_of_value(lhs);
-        let pow = self.builder.insert_cast(pow, typ);
-        self.builder.insert_binary(lhs, BinaryOp::Mul, pow)
-    }
-
-    /// Insert ssa instructions which computes lhs >> rhs by doing lhs/2^rhs
-    fn insert_shift_right(&mut self, lhs: ValueId, rhs: ValueId) -> ValueId {
-        let base = self.builder.field_constant(FieldElement::from(2_u128));
-        let pow = self.pow(base, rhs);
-        self.builder.insert_binary(lhs, BinaryOp::Div, pow)
-    }
-
-    /// Computes lhs^rhs via square&multiply, using the bits decomposition of rhs
-    fn pow(&mut self, lhs: ValueId, rhs: ValueId) -> ValueId {
-        let typ = self.builder.current_function.dfg.type_of_value(rhs);
-        if let Type::Numeric(NumericType::Unsigned { bit_size }) = typ {
-            let to_bits = self.builder.import_intrinsic_id(Intrinsic::ToBits(Endian::Little));
-            let length = self.builder.field_constant(FieldElement::from(bit_size as i128));
-            let result_types =
-                vec![Type::field(), Type::Array(Rc::new(vec![Type::bool()]), bit_size as usize)];
-            let rhs_bits = self.builder.insert_call(to_bits, vec![rhs, length], result_types);
-            let rhs_bits = rhs_bits[1];
-            let one = self.builder.field_constant(FieldElement::one());
-            let mut r = one;
-            for i in 1..bit_size + 1 {
-                let r1 = self.builder.insert_binary(r, BinaryOp::Mul, r);
-                let a = self.builder.insert_binary(r1, BinaryOp::Mul, lhs);
-                let idx = self.builder.field_constant(FieldElement::from((bit_size - i) as i128));
-                let b = self.builder.insert_array_get(rhs_bits, idx, Type::field());
-                let r2 = self.builder.insert_binary(a, BinaryOp::Mul, b);
-                let c = self.builder.insert_binary(one, BinaryOp::Sub, b);
-                let r3 = self.builder.insert_binary(c, BinaryOp::Mul, r1);
-                r = self.builder.insert_binary(r2, BinaryOp::Add, r3);
-            }
-            r
-        } else {
-            unreachable!("Value must be unsigned in power operation");
-        }
-    }
-
     /// Insert a binary instruction at the end of the current block.
     /// Converts the form of the binary instruction as necessary
     /// (e.g. swapping arguments, inserting a not) to represent it in the IR.
@@ -321,8 +277,8 @@ impl<'a> FunctionContext<'a> {
         location: Location,
     ) -> Values {
         let mut result = match operator {
-            BinaryOpKind::ShiftLeft => self.insert_shift_left(lhs, rhs),
-            BinaryOpKind::ShiftRight => self.insert_shift_right(lhs, rhs),
+            BinaryOpKind::ShiftLeft => self.builder.insert_shift_left(lhs, rhs),
+            BinaryOpKind::ShiftRight => self.builder.insert_shift_right(lhs, rhs),
             BinaryOpKind::Equal | BinaryOpKind::NotEqual
                 if matches!(self.builder.type_of_value(lhs), Type::Array(..)) =>
             {

noir_stdlib/src/lib.nr

@@ -60,6 +60,7 @@ pub fn wrapping_mul<T>(x : T, y: T) -> T {
     crate::from_field(crate::as_field(x) * crate::as_field(y))
 }
 
-pub fn wrapping_shift_left<T>(x : T, y: T) -> T {
-    crate::from_field(crate::as_field(x) * 2.pow_32(crate::as_field(y)))
-}
+/// Shift-left x by y bits
+/// If the result overflow the bitsize; it does not fail and returns 0 instead
+#[builtin(wrapping_shift_left)]
+pub fn wrapping_shift_left<T>(x : T, y: T) -> T {}

noir_stdlib/src/sha256.nr

@@ -7,7 +7,7 @@ fn rotr32(a: u32, b: u32) -> u32 // 32-bit right rotation
 {
     // None of the bits overlap between `(a >> b)` and `(a << (32 - b))`
     // Addition is then equivalent to OR, with fewer constraints.
-    (a >> b) + (a << (32 - b))
+    (a >> b) + (crate::wrapping_shift_left(a, 32 - b))
 }
 
 fn ch(x: u32, y: u32, z: u32) -> u32

noir_stdlib/src/sha512.nr

@@ -7,7 +7,7 @@ fn rotr64(a: u64, b: u64) -> u64 // 64-bit right rotation
 {
     // None of the bits overlap between `(a >> b)` and `(a << (64 - b))`
     // Addition is then equivalent to OR, with fewer constraints.
-    (a >> b) + (a << (64 - b))
+    (a >> b) + (crate::wrapping_shift_left(a, 64 - b))
 }
 
 fn sha_ch(x: u64, y: u64, z: u64) -> u64