feat!: Bit shift is restricted to u8 right operand (#4907)

# Description

## Problem\*

Resolves #4841

## Summary\*
bit shift now requires the number of bits to be u8.


## Additional Context



## Documentation\*

Check one:
- [ ] No documentation needed.
- [X] Documentation included in this PR.
- [ ] **[For Experimental Features]** Documentation to be submitted in a
separate PR.

# PR Checklist\*

- [ ] I have tested the changes locally.
- [ ] I have formatted the changes with [Prettier](https://prettier.io/)
and/or `cargo fmt` on default settings.

---------

Co-authored-by: Tom French <tom@tomfren.ch>

.github/workflows/publish-nargo.yml

@@ -23,7 +23,7 @@ permissions:
 
 jobs:
   build-apple-darwin:
-    runs-on: macos-latest
+    runs-on: macos-12
     env:
       CROSS_CONFIG: ${{ github.workspace }}/.github/Cross.toml
       NIGHTLY_RELEASE: ${{ inputs.tag == '' }}

acvm-repo/brillig_vm/src/arithmetic.rs

@@ -71,7 +71,9 @@ pub(crate) fn evaluate_binary_int_op(
             }
         }
     })?;
-    let rhs = rhs.expect_integer_with_bit_size(bit_size).map_err(|err| match err {
+    let rhs_bit_size =
+        if op == &BinaryIntOp::Shl || op == &BinaryIntOp::Shr { 8 } else { bit_size };
+    let rhs = rhs.expect_integer_with_bit_size(rhs_bit_size).map_err(|err| match err {
         MemoryTypeError::MismatchedBitSize { value_bit_size, expected_bit_size } => {
             BrilligArithmeticError::MismatchedRhsBitSize {
                 rhs_bit_size: value_bit_size,

compiler/noirc_evaluator/src/brillig/brillig_gen/brillig_block.rs

@@ -1279,8 +1279,11 @@ impl<'block> BrilligBlock<'block> {
         dfg: &DataFlowGraph,
         result_variable: SingleAddrVariable,
     ) {
-        let binary_type =
-            type_of_binary_operation(dfg[binary.lhs].get_type(), dfg[binary.rhs].get_type());
+        let binary_type = type_of_binary_operation(
+            dfg[binary.lhs].get_type(),
+            dfg[binary.rhs].get_type(),
+            binary.operator,
+        );
 
         let left = self.convert_ssa_single_addr_value(binary.lhs, dfg);
         let right = self.convert_ssa_single_addr_value(binary.rhs, dfg);
@@ -1766,7 +1769,7 @@ impl<'block> BrilligBlock<'block> {
 }
 
 /// Returns the type of the operation considering the types of the operands
-pub(crate) fn type_of_binary_operation(lhs_type: &Type, rhs_type: &Type) -> Type {
+pub(crate) fn type_of_binary_operation(lhs_type: &Type, rhs_type: &Type, op: BinaryOp) -> Type {
     match (lhs_type, rhs_type) {
         (_, Type::Function) | (Type::Function, _) => {
             unreachable!("Functions are invalid in binary operations")
@@ -1782,12 +1785,15 @@ pub(crate) fn type_of_binary_operation(lhs_type: &Type, rhs_type: &Type) -> Type
         }
         // If both sides are numeric type, then we expect their types to be
         // the same.
-        (Type::Numeric(lhs_type), Type::Numeric(rhs_type)) => {
+        (Type::Numeric(lhs_type), Type::Numeric(rhs_type))
+            if op != BinaryOp::Shl && op != BinaryOp::Shr =>
+        {
             assert_eq!(
                 lhs_type, rhs_type,
                 "lhs and rhs types in a binary operation are always the same but got {lhs_type} and {rhs_type}"
             );
             Type::Numeric(*lhs_type)
         }
+        _ => lhs_type.clone(),
     }
 }

compiler/noirc_evaluator/src/brillig/brillig_ir/instructions.rs

@@ -75,12 +75,6 @@ impl BrilligContext {
         result: SingleAddrVariable,
         operation: BrilligBinaryOp,
     ) {
-        assert!(
-            lhs.bit_size == rhs.bit_size,
-            "Not equal bit size for lhs and rhs: lhs {}, rhs {}",
-            lhs.bit_size,
-            rhs.bit_size
-        );
         let is_field_op = lhs.bit_size == FieldElement::max_num_bits();
         let expected_result_bit_size =
             BrilligContext::binary_result_bit_size(operation, lhs.bit_size);

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

@@ -230,7 +230,12 @@ impl FunctionBuilder {
     ) -> ValueId {
         let lhs_type = self.type_of_value(lhs);
         let rhs_type = self.type_of_value(rhs);
-        assert_eq!(lhs_type, rhs_type, "ICE - Binary instruction operands must have the same type");
+        if operator != BinaryOp::Shl && operator != BinaryOp::Shr {
+            assert_eq!(
+                lhs_type, rhs_type,
+                "ICE - Binary instruction operands must have the same type"
+            );
+        }
         let instruction = Instruction::Binary(Binary { lhs, rhs, operator });
         self.insert_instruction(instruction, None).first()
     }

compiler/noirc_evaluator/src/ssa/opt/remove_bit_shifts.rs

@@ -117,7 +117,7 @@ impl Context<'_> {
         } 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());
+                self.numeric_constant(FieldElement::from(bit_size as u128), Type::unsigned(8));
             let overflow = self.insert_binary(rhs, BinaryOp::Lt, bit_size_var);
             let predicate = self.insert_cast(overflow, typ.clone());
             // we can safely cast to unsigned because overflow_checks prevent bit-shift with a negative value

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

@@ -344,7 +344,7 @@ impl<'a> FunctionContext<'a> {
                         self.insert_safe_cast(result, result_type, location)
                     }
                     BinaryOpKind::ShiftLeft | BinaryOpKind::ShiftRight => {
-                        self.check_shift_overflow(result, rhs, bit_size, location, true)
+                        self.check_shift_overflow(result, rhs, bit_size, location)
                     }
                     _ => unreachable!("operator {} should not overflow", operator),
                 }
@@ -408,7 +408,7 @@ impl<'a> FunctionContext<'a> {
                             }
                         }
 
-                        self.check_shift_overflow(result, rhs, bit_size, location, false);
+                        self.check_shift_overflow(result, rhs, bit_size, location);
                     }
 
                     _ => unreachable!("operator {} should not overflow", operator),
@@ -430,32 +430,12 @@ impl<'a> FunctionContext<'a> {
         rhs: ValueId,
         bit_size: u32,
         location: Location,
-        is_signed: bool,
     ) -> ValueId {
         let one = self.builder.numeric_constant(FieldElement::one(), Type::bool());
-        let rhs = if is_signed {
-            self.insert_safe_cast(rhs, Type::unsigned(bit_size), location)
-        } else {
-            rhs
-        };
-        // Bit-shift with a negative number is an overflow
-        if is_signed {
-            // We compute the sign of rhs.
-            let half_width = self.builder.numeric_constant(
-                FieldElement::from(2_i128.pow(bit_size - 1)),
-                Type::unsigned(bit_size),
-            );
-            let sign = self.builder.insert_binary(rhs, BinaryOp::Lt, half_width);
-            self.builder.set_location(location).insert_constrain(
-                sign,
-                one,
-                Some("attempt to bit-shift with overflow".to_owned().into()),
-            );
-        }
+        assert!(self.builder.current_function.dfg.type_of_value(rhs) == Type::unsigned(8));
 
-        let max = self
-            .builder
-            .numeric_constant(FieldElement::from(bit_size as i128), Type::unsigned(bit_size));
+        let max =
+            self.builder.numeric_constant(FieldElement::from(bit_size as i128), Type::unsigned(8));
         let overflow = self.builder.insert_binary(rhs, BinaryOp::Lt, max);
         self.builder.set_location(location).insert_constrain(
             overflow,

compiler/noirc_frontend/src/hir/type_check/errors.rs

@@ -80,6 +80,8 @@ pub enum TypeCheckError {
     FieldModulo { span: Span },
     #[error("Fields cannot be compared, try casting to an integer first")]
     FieldComparison { span: Span },
+    #[error("The bit count in a bit-shift operation must fit in a u8, try casting the right hand side into a u8 first")]
+    InvalidShiftSize { span: Span },
     #[error("The number of bits to use for this bitwise operation is ambiguous. Either the operand's type or return type should be specified")]
     AmbiguousBitWidth { span: Span },
     #[error("Error with additional context")]
@@ -234,7 +236,8 @@ impl From<TypeCheckError> for Diagnostic {
             | TypeCheckError::UnconstrainedReferenceToConstrained { span }
             | TypeCheckError::UnconstrainedSliceReturnToConstrained { span }
             | TypeCheckError::NonConstantSliceLength { span }
-            | TypeCheckError::StringIndexAssign { span } => {
+            | TypeCheckError::StringIndexAssign { span }
+            | TypeCheckError::InvalidShiftSize { span } => {
                 Diagnostic::simple_error(error.to_string(), String::new(), span)
             }
             TypeCheckError::PublicReturnType { typ, span } => Diagnostic::simple_error(

compiler/noirc_frontend/src/hir/type_check/expr.rs

@@ -1,7 +1,8 @@
 use iter_extended::vecmap;
 use noirc_errors::Span;
 
-use crate::ast::{BinaryOpKind, UnaryOp};
+use crate::ast::{BinaryOpKind, IntegerBitSize, UnaryOp};
+use crate::macros_api::Signedness;
 use crate::{
     hir::{resolution::resolver::verify_mutable_reference, type_check::errors::Source},
     hir_def::{
@@ -1129,11 +1130,30 @@ impl<'interner> TypeChecker<'interner> {
                 if let TypeBinding::Bound(binding) = &*int.borrow() {
                     return self.infix_operand_type_rules(binding, op, other, span);
                 }
-
+                if op.kind == BinaryOpKind::ShiftLeft || op.kind == BinaryOpKind::ShiftRight {
+                    self.unify(
+                        rhs_type,
+                        &Type::Integer(Signedness::Unsigned, IntegerBitSize::Eight),
+                        || TypeCheckError::InvalidShiftSize { span },
+                    );
+                    let use_impl = if lhs_type.is_numeric() {
+                        let integer_type = Type::polymorphic_integer(self.interner);
+                        self.bind_type_variables_for_infix(lhs_type, op, &integer_type, span)
+                    } else {
+                        true
+                    };
+                    return Ok((lhs_type.clone(), use_impl));
+                }
                 let use_impl = self.bind_type_variables_for_infix(lhs_type, op, rhs_type, span);
                 Ok((other.clone(), use_impl))
             }
             (Integer(sign_x, bit_width_x), Integer(sign_y, bit_width_y)) => {
+                if op.kind == BinaryOpKind::ShiftLeft || op.kind == BinaryOpKind::ShiftRight {
+                    if *sign_y != Signedness::Unsigned || *bit_width_y != IntegerBitSize::Eight {
+                        return Err(TypeCheckError::InvalidShiftSize { span });
+                    }
+                    return Ok((Integer(*sign_x, *bit_width_x), false));
+                }
                 if sign_x != sign_y {
                     return Err(TypeCheckError::IntegerSignedness {
                         sign_x: *sign_x,
@@ -1165,6 +1185,12 @@ impl<'interner> TypeChecker<'interner> {
             (Bool, Bool) => Ok((Bool, false)),
 
             (lhs, rhs) => {
+                if op.kind == BinaryOpKind::ShiftLeft || op.kind == BinaryOpKind::ShiftRight {
+                    if rhs == &Type::Integer(Signedness::Unsigned, IntegerBitSize::Eight) {
+                        return Ok((lhs.clone(), true));
+                    }
+                    return Err(TypeCheckError::InvalidShiftSize { span });
+                }
                 self.unify(lhs, rhs, || TypeCheckError::TypeMismatchWithSource {
                     expected: lhs.clone(),
                     actual: rhs.clone(),

docs/docs/noir/concepts/ops.md

@@ -30,8 +30,8 @@ sidebar_position: 3
 | ^         |              XOR two private input types together              |                  Types must be integer |
 | &         |              AND two private input types together              |                  Types must be integer |
 | \|        |              OR two private input types together               |                  Types must be integer |
-| \<\<        |        Left shift an integer by another integer amount         |                  Types must be integer |
-| >>        |        Right shift an integer by another integer amount        |                  Types must be integer |
+| \<\<        |        Left shift an integer by another integer amount         | Types must be integer, shift must be u8 |
+| >>        |        Right shift an integer by another integer amount        | Types must be integer, shift must be u8 |
 | !         |                     Bitwise not of a value                     |        Type must be integer or boolean |
 | \<         |       returns a bool if one value is less than the other       | Upper bound must have a known bit size |
 | \<=        | returns a bool if one value is less than or equal to the other | Upper bound must have a known bit size |

noir_stdlib/src/ops.nr

@@ -126,30 +126,30 @@ impl BitXor for i64 { fn bitxor(self, other: i64) -> i64 { self ^ other } }
 
 // docs:start:shl-trait
 trait Shl {
-    fn shl(self, other: Self) -> Self;
+    fn shl(self, other: u8) -> Self;
 }
 // docs:end:shl-trait
 
-impl Shl for u32 { fn shl(self, other: u32) -> u32 { self << other } }
-impl Shl for u64 { fn shl(self, other: u64) -> u64 { self << other } }
+impl Shl for u32 { fn shl(self, other: u8) -> u32 { self << other } }
+impl Shl for u64 { fn shl(self, other: u8) -> u64 { self << other } }
 impl Shl for u8 { fn shl(self, other: u8) -> u8 { self << other } }
-impl Shl for u1 { fn shl(self, other: u1) -> u1 { self << other } }
+impl Shl for u1 { fn shl(self, other: u8) -> u1 { self << other } }
 
-impl Shl for i8 { fn shl(self, other: i8) -> i8 { self << other } }
-impl Shl for i32 { fn shl(self, other: i32) -> i32 { self << other } }
-impl Shl for i64 { fn shl(self, other: i64) -> i64 { self << other } }
+impl Shl for i8 { fn shl(self, other: u8) -> i8 { self << other } }
+impl Shl for i32 { fn shl(self, other: u8) -> i32 { self << other } }
+impl Shl for i64 { fn shl(self, other: u8) -> i64 { self << other } }
 
 // docs:start:shr-trait
 trait Shr {
-    fn shr(self, other: Self) -> Self;
+    fn shr(self, other: u8) -> Self;
 }
 // docs:end:shr-trait
 
-impl Shr for u64 { fn shr(self, other: u64) -> u64 { self >> other } }
-impl Shr for u32 { fn shr(self, other: u32) -> u32 { self >> other } }
+impl Shr for u64 { fn shr(self, other: u8) -> u64 { self >> other } }
+impl Shr for u32 { fn shr(self, other: u8) -> u32 { self >> other } }
 impl Shr for u8 { fn shr(self, other: u8) -> u8 { self >> other } }
-impl Shr for u1 { fn shr(self, other: u1) -> u1 { self >> other } }
+impl Shr for u1 { fn shr(self, other: u8) -> u1 { self >> other } }
 
-impl Shr for i8 { fn shr(self, other: i8) -> i8 { self >> other } }
-impl Shr for i32 { fn shr(self, other: i32) -> i32 { self >> other } }
-impl Shr for i64 { fn shr(self, other: i64) -> i64 { self >> other } }
+impl Shr for i8 { fn shr(self, other: u8) -> i8 { self >> other } }
+impl Shr for i32 { fn shr(self, other: u8) -> i32 { self >> other } }
+impl Shr for i64 { fn shr(self, other: u8) -> i64 { self >> other } }

noir_stdlib/src/sha512.nr

@@ -2,7 +2,7 @@
 // 64 bytes.
 // Internal functions act on 64-bit unsigned integers for simplicity.
 // Auxiliary mappings; names as in FIPS PUB 180-4
-fn rotr64(a: u64, b: u64) -> u64 // 64-bit right rotation
+fn rotr64(a: u64, b: u8) -> 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.

noir_stdlib/src/uint128.nr

@@ -256,9 +256,9 @@ impl BitXor for U128 {
 }
 
 impl Shl for U128 { 
-    fn shl(self, other: U128) -> U128 { 
-        assert(other < U128::from_u64s_le(128,0), "attempt to shift left with overflow");
-        let exp_bits = other.lo.to_be_bits(7);
+    fn shl(self, other: u8) -> U128 { 
+        assert(other < 128, "attempt to shift left with overflow");
+        let exp_bits = (other as Field).to_be_bits(7);
 
         let mut r: Field = 2;
         let mut y: Field = 1;
@@ -271,9 +271,9 @@ impl Shl for U128 {
 }
 
 impl Shr for U128 { 
-    fn shr(self, other: U128) -> U128 { 
-        assert(other < U128::from_u64s_le(128,0), "attempt to shift right with overflow");
-        let exp_bits = other.lo.to_be_bits(7);
+    fn shr(self, other: u8) -> U128 { 
+        assert(other < 128, "attempt to shift right with overflow");
+        let exp_bits = (other as Field).to_be_bits(7);
 
         let mut r: Field = 2;
         let mut y: Field = 1;

test_programs/compile_success_empty/brillig_integer_binary_operations/src/main.nr

@@ -23,8 +23,8 @@ fn main() {
     assert((x | y) == or(x, y));
     // TODO SSA => ACIR has some issues with xor ops
     assert(check_xor(x, y, 4));
-    assert((x >> y) == shr(x, y));
-    assert((x << y) == shl(x, y));
+    assert((x >> y as u8) == shr(x, y as u8));
+    assert((x << y as u8) == shl(x, y as u8));
 }
 
 unconstrained fn add(x: u32, y: u32) -> u32 {
@@ -67,11 +67,11 @@ unconstrained fn check_xor(x: u32, y: u32, result: u32) -> bool {
     (x ^ y) == result
 }
 
-unconstrained fn shr(x: u32, y: u32) -> u32 {
+unconstrained fn shr(x: u32, y: u8) -> u32 {
     x >> y
 }
 
-unconstrained fn shl(x: u32, y: u32) -> u32 {
+unconstrained fn shl(x: u32, y: u8) -> u32 {
     x << y
 }
 

test_programs/execution_success/bit_shifts_comptime/src/main.nr

@@ -14,7 +14,7 @@ fn main(x: u64) {
     //regression for 3481
     assert(x << 63 == 0);
 
-    assert_eq((1 as u64) << (32 as u64), 0x0100000000);
+    assert_eq((1 as u64) << 32, 0x0100000000);
 }
 
 fn regression_2250() {

test_programs/execution_success/bit_shifts_runtime/src/main.nr

@@ -1,4 +1,4 @@
-fn main(x: u64, y: u64) {
+fn main(x: u64, y: u8) {
     // runtime shifts on compile-time known values
     assert(64 << y == 128);
     assert(64 >> y == 32);
@@ -11,10 +11,10 @@ fn main(x: u64, y: u64) {
     let mut b: i8 = x as i8;
     assert(b << 1 == -128);
     assert(b >> 2 == 16);
-    assert(b >> a == 32);
+    assert(b >> y == 32);
     a = -a;
     assert(a << 7 == -128);
-    assert(a << -a == -2);
+    assert(a << y == -2);
 
-    assert(x >> x == 0);
+    assert(x >> (x as u8) == 0);
 }

test_programs/execution_success/brillig_bit_shifts_runtime/src/main.nr

@@ -1,7 +1,7 @@
-unconstrained fn main(x: u64, y: u64) {
+unconstrained fn main(x: u64, y: u8) {
     // runtime shifts on compile-time known values
-    assert(64 << y == 128);
-    assert(64 >> y == 32);
+    assert(64 as u32 << y == 128);
+    assert(64 as u32 >> y == 32);
     // runtime shifts on runtime values
     assert(x << y == 128);
     assert(x >> y == 32);
@@ -11,10 +11,10 @@ unconstrained fn main(x: u64, y: u64) {
     let mut b: i8 = x as i8;
     assert(b << 1 == -128);
     assert(b >> 2 == 16);
-    assert(b >> a == 32);
+    assert(b >> y == 32);
     a = -a;
     assert(a << 7 == -128);
-    assert(a << -a == -2);
+    assert(a << y == -2);
 
-    assert(x >> x == 0);
+    assert(x >> (x as u8) == 0);
 }