Determinism corpus

Stores eight f32 values in order: NaN, positive infinity, negative infinity, positive zero, negative zero, positive one, negative one, and a generated NaN. The expectation shows the exact bytes of each array element in that order.

Stores eight f64 values in order: NaN, positive infinity, negative infinity, positive zero, negative zero, positive one, negative one, and a generated NaN. The expectation shows the exact bytes of each array element in that order.

Hashes the mixed f32 byte image to check the stdlib hash path used by snapshots and other raw memory observers.

Hashes the mixed f64 byte image to check the stdlib hash path used by snapshots and other raw memory observers.

Reads a stored f32 NaN through the generic byte conversion path and checks its canonical memory representation.

Reads a stored f64 NaN through the generic byte conversion path and checks its canonical memory representation.

Observes an f32 NaN folded at compile time from zero divided by zero and requires the canonical bit pattern.

Observes an f32 NaN folded at compile time from infinity multiplied by zero and requires the canonical bit pattern.

Observes an f32 NaN folded at compile time from infinity minus itself and requires the canonical bit pattern.

Observes an f32 NaN folded at compile time from the square root of a negative value and requires the canonical bit pattern.

Observes an f64 NaN folded at compile time from zero divided by zero and requires the canonical bit pattern.

Observes an f64 NaN folded at compile time from infinity multiplied by zero and requires the canonical bit pattern.

Observes an f64 NaN folded at compile time from infinity minus itself and requires the canonical bit pattern.

Observes an f64 NaN folded at compile time from the square root of a negative value and requires the canonical bit pattern.

Demotes an f64 NaN produced at runtime and checks that the observed result uses the canonical f32 pattern.

Observes an f32 NaN produced at runtime from zero divided by zero and requires the canonical bit pattern.

Observes an f32 NaN produced at runtime from infinity multiplied by zero and requires the canonical bit pattern.

Observes an f32 NaN produced at runtime from infinity minus itself and requires the canonical bit pattern.

Observes an f32 NaN produced at runtime from the square root of a negative value and requires the canonical bit pattern.

Observes an f64 NaN produced at runtime from zero divided by zero and requires the canonical bit pattern.

Observes an f64 NaN produced at runtime from infinity multiplied by zero and requires the canonical bit pattern.

Observes an f64 NaN produced at runtime from infinity minus itself and requires the canonical bit pattern.

Observes an f64 NaN produced at runtime from the square root of a negative value and requires the canonical bit pattern.

Promotes an f32 NaN produced at runtime and checks that the observed result uses the canonical f64 pattern.

Stores several independently produced f32 NaNs and checks their raw heap bytes.

Stores several independently produced f64 NaNs and checks their raw heap bytes.

Stores a vector containing a NaN produced by division and finite lanes, checking canonicalization without disturbing the other lanes.

Stores a vector containing a NaN produced by square root, signed zero, and finite lanes, checking every observed bit.

Converts f32 the first exactly representable value above the destination maximum to i32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f32 NaN to i32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f32 negative infinity to i32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f32 positive infinity to i32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f32 the first exactly representable value above the destination maximum to u32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f32 NaN to u32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f32 negative infinity to u32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f32 positive infinity to u32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 NaN to i32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 negative infinity to i32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 positive infinity to i32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 NaN to i64 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 negative infinity to i64 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 positive infinity to i64 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 NaN to u32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 negative infinity to u32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 positive infinity to u32 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 NaN to u64 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 negative infinity to u64 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Converts f64 positive infinity to u64 and checks Wax's saturating result instead of the host CPU's conversion behavior.

Passes an array of mixed width float structs from the host and checks that float leaves canonicalize without changing integers or padding.

Carries the host array payload through an untyped byte crossing, proving which noncanonical bytes the typed array probe received.

Carries the host value payload through an untyped byte crossing, proving which noncanonical bytes the typed value probe received.

Passes a mixed width float struct by value from the host and checks canonical float leaves plus exact byte assignment and padding.

Applies an i32 arithmetic right shift by 33 and checks Wax's masked shift count rule.

Applies an i32 arithmetic right shift by -1 and checks Wax's masked shift count rule.

Applies an i32 left shift by 32 and checks Wax's masked shift count rule.

Applies an i32 left shift by 33 and checks Wax's masked shift count rule.

Applies an i32 left shift by -1 and checks Wax's masked shift count rule.

Applies an i64 arithmetic right shift by 65 and checks Wax's masked shift count rule.

Applies an i64 arithmetic right shift by -1 and checks Wax's masked shift count rule.

Applies an i64 left shift by 64 and checks Wax's masked shift count rule.

Applies an i64 left shift by 65 and checks Wax's masked shift count rule.

Applies an i64 left shift by -1 and checks Wax's masked shift count rule.

Applies an u32 logical right shift by 33 and checks Wax's masked shift count rule.

Applies an u64 logical right shift by 65 and checks Wax's masked shift count rule.

Runs i32 division with a zero divisor and requires the same named panic on every backend.

Runs i32 division with the signed minimum value and negative one and requires the same named panic on every backend.

Runs i64 division with a zero divisor and requires the same named panic on every backend.

Runs i64 division with the signed minimum value and negative one and requires the same named panic on every backend.

Runs i32 remainder with a zero divisor and requires the same named panic on every backend.

Runs i32 remainder with the signed minimum value and negative one and requires the same named panic on every backend.

Runs i64 remainder with a zero divisor and requires the same named panic on every backend.

Runs i64 remainder with the signed minimum value and negative one and requires the same named panic on every backend.

Promotes the minimum f32 subnormal to f64 and checks its exact widened representation.

Demotes the exact f64 representation of the minimum f32 subnormal and checks that the smallest f32 value survives.

Demotes the minimum f64 subnormal to f32 and checks that it underflows to zero.

For f32, adds the minimum subnormal to the maximum subnormal, carrying into the minimum normal.

For f32, halves the minimum subnormal, testing ties to even underflow to zero.

For f32, halves the negative minimum subnormal, testing the sign of underflowed zero.

For f32, subtracts the minimum subnormal from the minimum normal.

For f32, halves the minimum normal into the subnormal range.

For f32, halves three minimum subnormals, testing ties to even rounding upward.

For f64, adds the minimum subnormal to the maximum subnormal, carrying into the minimum normal.

For f64, halves the minimum subnormal, testing ties to even underflow to zero.

For f64, halves the negative minimum subnormal, testing the sign of underflowed zero.

For f64, subtracts the minimum subnormal from the minimum normal.

For f64, halves the minimum normal into the subnormal range.

For f64, halves three minimum subnormals, testing ties to even rounding upward.

Hashes hundreds of scalar reinterpretations plus NaN, subnormal, signed underflow, and contraction sensitive operations shared by every executable backend.