mirror/zig
1
0
Fork 0
mirror of https://github.com/ziglang/zig.git synced 2026-02-21 16:54:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'shawnl-lessThan'

Browse Source 
* always allow integer comparison operations no matter the
   bit width, signedness, or comptime-ness of operands.
   closes #2133
 * implement comparisons for vectors, which returns vector of
   bools.

closes #3001
This commit is contained in:
Andrew Kelley 2019-12-16 11:39:13 -05:00
commit 496f271d17
No known key found for this signature in database
GPG Key ID: 7C5F548F728501A9
8 changed files with 741 additions and 214 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

55
lib/std/math.zig
View File

@ -943,3 +943,58 @@ test "math.mulWide" {
testing.expect(mulWide(i8, 5, -5) == -25);
testing.expect(mulWide(u8, 100, 100) == 10000);
}
/// Not to be confused with `std.mem.Compare`.
pub const CompareOperator = enum {
/// Less than (`<`)
lt,
/// Less than or equal (`<=`)
lte,
/// Equal (`==`)
eq,
/// Greater than or equal (`>=`)
gte,
/// Greater than (`>`)
gt,
/// Not equal (`!=`)
neq,
};
/// This function does the same thing as comparison operators, however the
/// operator is a runtime-known enum value. Works on any operands that
/// support comparison operators.
pub fn compare(a: var, op: CompareOperator, b: var) bool {
return switch (op) {
.lt => a < b,
.lte => a <= b,
.eq => a == b,
.neq => a != b,
.gt => a > b,
.gte => a >= b,
};
}
test "math.lt, et al < <= > >= between signed and unsigned" {
testing.expect(compare(@as(i8, -1), .lt, @as(u8, 255)));
testing.expect(compare(@as(i8, 2), .gt, @as(u8, 1)));
testing.expect(!compare(@as(i8, -1), .gte, @as(u8, 255)));
testing.expect(compare(@as(u8, 255), .gt, @as(i8, -1)));
testing.expect(!compare(@as(u8, 255), .lte, @as(i8, -1)));
testing.expect(compare(@as(i8, -1), .lt, @as(u9, 255)));
testing.expect(!compare(@as(i8, -1), .gte, @as(u9, 255)));
testing.expect(compare(@as(u9, 255), .gt, @as(i8, -1)));
testing.expect(!compare(@as(u9, 255), .lte, @as(i8, -1)));
testing.expect(compare(@as(i9, -1), .lt, @as(u8, 255)));
testing.expect(!compare(@as(i9, -1), .gte, @as(u8, 255)));
testing.expect(compare(@as(u8, 255), .gt, @as(i9, -1)));
testing.expect(!compare(@as(u8, 255), .lte, @as(i9, -1)));
testing.expect(compare(@as(u8, 1), .lt, @as(u8, 2)));
testing.expect(@bitCast(u8, @as(i8, -1)) == @as(u8, 255));
testing.expect(!compare(@as(u8, 255), .eq, @as(i8, -1)));
testing.expect(compare(@as(u8, 1), .eq, @as(u8, 1)));
}

38
src/analyze.cpp
View File

@ -9219,3 +9219,41 @@ void copy_const_val(ZigValue *dest, ZigValue *src) {
dest->data.x_optional->parent.data.p_optional_payload.optional_val = dest;
}
}
bool type_is_numeric(ZigType *ty) {
switch (ty->id) {
case ZigTypeIdInvalid:
zig_unreachable();
case ZigTypeIdComptimeFloat:
case ZigTypeIdComptimeInt:
case ZigTypeIdInt:
case ZigTypeIdFloat:
return true;
case ZigTypeIdVector:
return type_is_numeric(ty->data.vector.elem_type);
case ZigTypeIdMetaType:
case ZigTypeIdVoid:
case ZigTypeIdBool:
case ZigTypeIdUnreachable:
case ZigTypeIdPointer:
case ZigTypeIdArray:
case ZigTypeIdStruct:
case ZigTypeIdUndefined:
case ZigTypeIdNull:
case ZigTypeIdOptional:
case ZigTypeIdErrorUnion:
case ZigTypeIdErrorSet:
case ZigTypeIdEnum:
case ZigTypeIdUnion:
case ZigTypeIdFn:
case ZigTypeIdBoundFn:
case ZigTypeIdOpaque:
case ZigTypeIdFnFrame:
case ZigTypeIdAnyFrame:
case ZigTypeIdEnumLiteral:
return false;
}
zig_unreachable();
}

1
src/analyze.hpp
View File

@ -279,4 +279,5 @@ bool is_anon_container(ZigType *ty);
void copy_const_val(ZigValue *dest, ZigValue *src);
bool type_has_optional_repr(ZigType *ty);
bool is_opt_err_set(ZigType *ty);
bool type_is_numeric(ZigType *ty);
#endif

24
src/bigint.cpp
View File

@ -1759,3 +1759,27 @@ void bigint_incr(BigInt *x) {
bigint_add(x, &copy, &one);
}
void bigint_decr(BigInt *x) {
if (x->digit_count == 0) {
bigint_init_signed(x, -1);
return;
}
if (x->digit_count == 1) {
if (x->is_negative && x->data.digit != UINT64_MAX) {
x->data.digit += 1;
return;
} else if (!x->is_negative && x->data.digit != 0) {
x->data.digit -= 1;
return;
}
}
BigInt copy;
bigint_init_bigint(&copy, x);
BigInt neg_one;
bigint_init_signed(&neg_one, -1);
bigint_add(x, &copy, &neg_one);
}

1
src/bigint.hpp
View File

@ -95,6 +95,7 @@ size_t bigint_bits_needed(const BigInt *op);
Cmp bigint_cmp_zero(const BigInt *op);
void bigint_incr(BigInt *value);
void bigint_decr(BigInt *value);
bool mul_u64_overflow(uint64_t op1, uint64_t op2, uint64_t *result);

760
src/ir.cpp
View File

@ -1792,6 +1792,24 @@ static IrInstruction *ir_build_bin_op(IrBuilder *irb, Scope *scope, AstNode *sou
return &bin_op_instruction->base;
}
static IrInstruction *ir_build_bin_op_gen(IrAnalyze *ira, IrInstruction *source_instr, ZigType *res_type,
IrBinOp op_id, IrInstruction *op1, IrInstruction *op2, bool safety_check_on)
{
IrInstructionBinOp *bin_op_instruction = ir_build_instruction<IrInstructionBinOp>(&ira->new_irb,
source_instr->scope, source_instr->source_node);
bin_op_instruction->base.value->type = res_type;
bin_op_instruction->op_id = op_id;
bin_op_instruction->op1 = op1;
bin_op_instruction->op2 = op2;
bin_op_instruction->safety_check_on = safety_check_on;
ir_ref_instruction(op1, ira->new_irb.current_basic_block);
ir_ref_instruction(op2, ira->new_irb.current_basic_block);
return &bin_op_instruction->base;
}
static IrInstruction *ir_build_merge_err_sets(IrBuilder *irb, Scope *scope, AstNode *source_node,
IrInstruction *op1, IrInstruction *op2, Buf *type_name)
{
@ -9591,51 +9609,58 @@ static bool float_is_nan(ZigValue *op) {
}
static Cmp float_cmp(ZigValue *op1, ZigValue *op2) {
assert(op1->type == op2->type);
if (op1->type->id == ZigTypeIdComptimeFloat) {
return bigfloat_cmp(&op1->data.x_bigfloat, &op2->data.x_bigfloat);
} else if (op1->type->id == ZigTypeIdFloat) {
switch (op1->type->data.floating.bit_count) {
case 16:
if (f16_lt(op1->data.x_f16, op2->data.x_f16)) {
return CmpLT;
} else if (f16_lt(op2->data.x_f16, op1->data.x_f16)) {
return CmpGT;
} else {
return CmpEQ;
}
case 32:
if (op1->data.x_f32 > op2->data.x_f32) {
return CmpGT;
} else if (op1->data.x_f32 < op2->data.x_f32) {
return CmpLT;
} else {
return CmpEQ;
}
case 64:
if (op1->data.x_f64 > op2->data.x_f64) {
return CmpGT;
} else if (op1->data.x_f64 < op2->data.x_f64) {
return CmpLT;
} else {
return CmpEQ;
}
case 128:
if (f128M_lt(&op1->data.x_f128, &op2->data.x_f128)) {
return CmpLT;
} else if (f128M_eq(&op1->data.x_f128, &op2->data.x_f128)) {
return CmpEQ;
} else {
return CmpGT;
}
default:
zig_unreachable();
if (op1->type == op2->type) {
if (op1->type->id == ZigTypeIdComptimeFloat) {
return bigfloat_cmp(&op1->data.x_bigfloat, &op2->data.x_bigfloat);
} else if (op1->type->id == ZigTypeIdFloat) {
switch (op1->type->data.floating.bit_count) {
case 16:
if (f16_lt(op1->data.x_f16, op2->data.x_f16)) {
return CmpLT;
} else if (f16_lt(op2->data.x_f16, op1->data.x_f16)) {
return CmpGT;
} else {
return CmpEQ;
}
case 32:
if (op1->data.x_f32 > op2->data.x_f32) {
return CmpGT;
} else if (op1->data.x_f32 < op2->data.x_f32) {
return CmpLT;
} else {
return CmpEQ;
}
case 64:
if (op1->data.x_f64 > op2->data.x_f64) {
return CmpGT;
} else if (op1->data.x_f64 < op2->data.x_f64) {
return CmpLT;
} else {
return CmpEQ;
}
case 128:
if (f128M_lt(&op1->data.x_f128, &op2->data.x_f128)) {
return CmpLT;
} else if (f128M_eq(&op1->data.x_f128, &op2->data.x_f128)) {
return CmpEQ;
} else {
return CmpGT;
}
default:
zig_unreachable();
}
} else {
zig_unreachable();
}
} else {
zig_unreachable();
}
BigFloat op1_big;
BigFloat op2_big;
float_init_bigfloat(op1, &op1_big);
float_init_bigfloat(op2, &op2_big);
return bigfloat_cmp(&op1_big, &op2_big);
}
// This function cannot handle NaN
static Cmp float_cmp_zero(ZigValue *op) {
if (op->type->id == ZigTypeIdComptimeFloat) {
return bigfloat_cmp_zero(&op->data.x_bigfloat);
@ -14421,18 +14446,7 @@ static IrInstruction *ir_evaluate_bin_op_cmp(IrAnalyze *ira, ZigType *resolved_t
if (op1_val->special == ConstValSpecialUndef ||
op2_val->special == ConstValSpecialUndef)
return ir_const_undef(ira, &bin_op_instruction->base, resolved_type);
if (resolved_type->id == ZigTypeIdComptimeFloat || resolved_type->id == ZigTypeIdFloat) {
if (float_is_nan(op1_val) || float_is_nan(op2_val)) {
return ir_const_bool(ira, &bin_op_instruction->base, op_id == IrBinOpCmpNotEq);
}
Cmp cmp_result = float_cmp(op1_val, op2_val);
bool answer = resolve_cmp_op_id(op_id, cmp_result);
return ir_const_bool(ira, &bin_op_instruction->base, answer);
} else if (resolved_type->id == ZigTypeIdComptimeInt || resolved_type->id == ZigTypeIdInt) {
Cmp cmp_result = bigint_cmp(&op1_val->data.x_bigint, &op2_val->data.x_bigint);
bool answer = resolve_cmp_op_id(op_id, cmp_result);
return ir_const_bool(ira, &bin_op_instruction->base, answer);
} else if (resolved_type->id == ZigTypeIdPointer && op_id != IrBinOpCmpEq && op_id != IrBinOpCmpNotEq) {
if (resolved_type->id == ZigTypeIdPointer && op_id != IrBinOpCmpEq && op_id != IrBinOpCmpNotEq) {
if ((op1_val->data.x_ptr.special == ConstPtrSpecialHardCodedAddr ||
op1_val->data.x_ptr.special == ConstPtrSpecialNull) &&
(op2_val->data.x_ptr.special == ConstPtrSpecialHardCodedAddr ||
@ -14482,6 +14496,12 @@ static Error lazy_cmp_zero(AstNode *source_node, ZigValue *val, Cmp *result) {
case ZigTypeIdInt:
*result = bigint_cmp_zero(&val->data.x_bigint);
return ErrorNone;
case ZigTypeIdComptimeFloat:
case ZigTypeIdFloat:
if (float_is_nan(val))
return ErrorNotLazy;
*result = float_cmp_zero(val);
return ErrorNone;
default:
return ErrorNotLazy;
}
@ -14511,9 +14531,450 @@ static Error lazy_cmp_zero(AstNode *source_node, ZigValue *val, Cmp *result) {
zig_unreachable();
}
static IrInstruction *ir_analyze_bin_op_cmp(IrAnalyze *ira, IrInstructionBinOp *bin_op_instruction) {
static ErrorMsg *ir_eval_bin_op_cmp_scalar(IrAnalyze *ira, IrInstruction *source_instr,
ZigValue *op1_val, IrBinOp op_id, ZigValue *op2_val, ZigValue *out_val)
{
Error err;
{
// Before resolving the values, we special case comparisons against zero. These can often
// be done without resolving lazy values, preventing potential dependency loops.
Cmp op1_cmp_zero;
if ((err = lazy_cmp_zero(source_instr->source_node, op1_val, &op1_cmp_zero))) {
if (err == ErrorNotLazy) goto never_mind_just_calculate_it_normally;
return ira->codegen->trace_err;
}
Cmp op2_cmp_zero;
if ((err = lazy_cmp_zero(source_instr->source_node, op2_val, &op2_cmp_zero))) {
if (err == ErrorNotLazy) goto never_mind_just_calculate_it_normally;
return ira->codegen->trace_err;
}
bool can_cmp_zero = false;
Cmp cmp_result;
if (op1_cmp_zero == CmpEQ && op2_cmp_zero == CmpEQ) {
can_cmp_zero = true;
cmp_result = CmpEQ;
} else if (op1_cmp_zero == CmpGT && op2_cmp_zero == CmpEQ) {
can_cmp_zero = true;
cmp_result = CmpGT;
} else if (op1_cmp_zero == CmpEQ && op2_cmp_zero == CmpGT) {
can_cmp_zero = true;
cmp_result = CmpLT;
} else if (op1_cmp_zero == CmpLT && op2_cmp_zero == CmpEQ) {
can_cmp_zero = true;
cmp_result = CmpLT;
} else if (op1_cmp_zero == CmpEQ && op2_cmp_zero == CmpLT) {
can_cmp_zero = true;
cmp_result = CmpGT;
} else if (op1_cmp_zero == CmpLT && op2_cmp_zero == CmpGT) {
can_cmp_zero = true;
cmp_result = CmpLT;
} else if (op1_cmp_zero == CmpGT && op2_cmp_zero == CmpLT) {
can_cmp_zero = true;
cmp_result = CmpGT;
}
if (can_cmp_zero) {
bool answer = resolve_cmp_op_id(op_id, cmp_result);
out_val->special = ConstValSpecialStatic;
out_val->data.x_bool = answer;
return nullptr;
}
}
never_mind_just_calculate_it_normally:
if ((err = ir_resolve_const_val(ira->codegen, ira->new_irb.exec, source_instr->source_node,
op1_val, UndefOk)))
{
return ira->codegen->trace_err;
}
if ((err = ir_resolve_const_val(ira->codegen, ira->new_irb.exec, source_instr->source_node,
op2_val, UndefOk)))
{
return ira->codegen->trace_err;
}
if (op1_val->special == ConstValSpecialUndef || op2_val->special == ConstValSpecialUndef) {
out_val->special = ConstValSpecialUndef;
return nullptr;
}
bool op1_is_float = op1_val->type->id == ZigTypeIdFloat || op1_val->type->id == ZigTypeIdComptimeFloat;
bool op2_is_float = op2_val->type->id == ZigTypeIdFloat || op2_val->type->id == ZigTypeIdComptimeFloat;
if (op1_is_float && op2_is_float) {
if (float_is_nan(op1_val) || float_is_nan(op2_val)) {
out_val->special = ConstValSpecialStatic;
out_val->data.x_bool = op_id == IrBinOpCmpNotEq;
return nullptr;
}
if (op1_val->type->id == ZigTypeIdComptimeFloat) {
IrInstruction *tmp = ir_const_noval(ira, source_instr);
tmp->value = op1_val;
IrInstruction *casted = ir_implicit_cast(ira, tmp, op2_val->type);
op1_val = casted->value;
} else if (op2_val->type->id == ZigTypeIdComptimeFloat) {
IrInstruction *tmp = ir_const_noval(ira, source_instr);
tmp->value = op2_val;
IrInstruction *casted = ir_implicit_cast(ira, tmp, op1_val->type);
op2_val = casted->value;
}
Cmp cmp_result = float_cmp(op1_val, op2_val);
out_val->special = ConstValSpecialStatic;
out_val->data.x_bool = resolve_cmp_op_id(op_id, cmp_result);
return nullptr;
}
bool op1_is_int = op1_val->type->id == ZigTypeIdInt || op1_val->type->id == ZigTypeIdComptimeInt;
bool op2_is_int = op2_val->type->id == ZigTypeIdInt || op2_val->type->id == ZigTypeIdComptimeInt;
BigInt *op1_bigint;
BigInt *op2_bigint;
bool need_to_free_op1_bigint = false;
bool need_to_free_op2_bigint = false;
if (op1_is_float) {
op1_bigint = allocate<BigInt>(1, "BigInt");
need_to_free_op1_bigint = true;
float_init_bigint(op1_bigint, op1_val);
} else {
assert(op1_is_int);
op1_bigint = &op1_val->data.x_bigint;
}
if (op2_is_float) {
op2_bigint = allocate<BigInt>(1, "BigInt");
need_to_free_op2_bigint = true;
float_init_bigint(op2_bigint, op2_val);
} else {
assert(op2_is_int);
op2_bigint = &op2_val->data.x_bigint;
}
Cmp cmp_result = bigint_cmp(op1_bigint, op2_bigint);
out_val->special = ConstValSpecialStatic;
out_val->data.x_bool = resolve_cmp_op_id(op_id, cmp_result);
if (need_to_free_op1_bigint) destroy(op1_bigint, "BigInt");
if (need_to_free_op2_bigint) destroy(op2_bigint, "BigInt");
return nullptr;
}
static IrInstruction *ir_analyze_bin_op_cmp_numeric(IrAnalyze *ira, IrInstruction *source_instr,
IrInstruction *op1, IrInstruction *op2, IrBinOp op_id)
{
Error err;
ZigType *scalar_result_type = ira->codegen->builtin_types.entry_bool;
ZigType *result_type = scalar_result_type;
ZigType *op1_scalar_type = op1->value->type;
ZigType *op2_scalar_type = op2->value->type;
if (op1->value->type->id == ZigTypeIdVector && op2->value->type->id == ZigTypeIdVector) {
if (op1->value->type->data.vector.len != op2->value->type->data.vector.len) {
ir_add_error(ira, source_instr,
buf_sprintf("vector length mismatch: %" PRIu32 " and %" PRIu32,
op1->value->type->data.vector.len, op2->value->type->data.vector.len));
return ira->codegen->invalid_instruction;
}
result_type = get_vector_type(ira->codegen, op1->value->type->data.vector.len, scalar_result_type);
op1_scalar_type = op1->value->type->data.vector.elem_type;
op2_scalar_type = op2->value->type->data.vector.elem_type;
} else if (op1->value->type->id == ZigTypeIdVector || op2->value->type->id == ZigTypeIdVector) {
ir_add_error(ira, source_instr,
buf_sprintf("mixed scalar and vector operands to comparison operator: '%s' and '%s'",
buf_ptr(&op1->value->type->name), buf_ptr(&op2->value->type->name)));
return ira->codegen->invalid_instruction;
}
bool opv_op1;
switch (type_has_one_possible_value(ira->codegen, op1->value->type)) {
case OnePossibleValueInvalid:
return ira->codegen->invalid_instruction;
case OnePossibleValueYes:
opv_op1 = true;
break;
case OnePossibleValueNo:
opv_op1 = false;
break;
}
bool opv_op2;
switch (type_has_one_possible_value(ira->codegen, op2->value->type)) {
case OnePossibleValueInvalid:
return ira->codegen->invalid_instruction;
case OnePossibleValueYes:
opv_op2 = true;
break;
case OnePossibleValueNo:
opv_op2 = false;
break;
}
Cmp op1_cmp_zero;
bool have_op1_cmp_zero = false;
if ((err = lazy_cmp_zero(source_instr->source_node, op1->value, &op1_cmp_zero))) {
if (err != ErrorNotLazy) return ira->codegen->invalid_instruction;
} else {
have_op1_cmp_zero = true;
}
Cmp op2_cmp_zero;
bool have_op2_cmp_zero = false;
if ((err = lazy_cmp_zero(source_instr->source_node, op2->value, &op2_cmp_zero))) {
if (err != ErrorNotLazy) return ira->codegen->invalid_instruction;
} else {
have_op2_cmp_zero = true;
}
if (((opv_op1 || instr_is_comptime(op1)) && (opv_op2 || instr_is_comptime(op2))) ||
(have_op1_cmp_zero && have_op2_cmp_zero))
{
IrInstruction *result_instruction = ir_const(ira, source_instr, result_type);
ZigValue *out_val = result_instruction->value;
if (result_type->id == ZigTypeIdVector) {
size_t len = result_type->data.vector.len;
expand_undef_array(ira->codegen, op1->value);
expand_undef_array(ira->codegen, op2->value);
out_val->special = ConstValSpecialUndef;
expand_undef_array(ira->codegen, out_val);
for (size_t i = 0; i < len; i += 1) {
ZigValue *scalar_op1_val = &op1->value->data.x_array.data.s_none.elements[i];
ZigValue *scalar_op2_val = &op2->value->data.x_array.data.s_none.elements[i];
ZigValue *scalar_out_val = &out_val->data.x_array.data.s_none.elements[i];
assert(scalar_out_val->type == scalar_result_type);
ErrorMsg *msg = ir_eval_bin_op_cmp_scalar(ira, source_instr,
scalar_op1_val, op_id, scalar_op2_val, scalar_out_val);
if (msg != nullptr) {
add_error_note(ira->codegen, msg, source_instr->source_node,
buf_sprintf("when computing vector element at index %" ZIG_PRI_usize, i));
return ira->codegen->invalid_instruction;
}
}
out_val->type = result_type;
out_val->special = ConstValSpecialStatic;
} else {
if (ir_eval_bin_op_cmp_scalar(ira, source_instr, op1->value, op_id,
op2->value, out_val) != nullptr)
{
return ira->codegen->invalid_instruction;
}
}
return result_instruction;
}
// If one operand has a comptime-known comparison with 0, and the other operand is unsigned, we might
// know the answer, depending on the operator.
// TODO make this work with vectors
if (have_op1_cmp_zero && op2_scalar_type->id == ZigTypeIdInt && !op2_scalar_type->data.integral.is_signed) {
if (op1_cmp_zero == CmpEQ) {
// 0 <= unsigned_x // true
// 0 > unsigned_x // false
switch (op_id) {
case IrBinOpCmpLessOrEq:
return ir_const_bool(ira, source_instr, true);
case IrBinOpCmpGreaterThan:
return ir_const_bool(ira, source_instr, false);
default:
break;
}
} else if (op1_cmp_zero == CmpLT) {
// -1 != unsigned_x // true
// -1 <= unsigned_x // true
// -1 < unsigned_x // true
// -1 == unsigned_x // false
// -1 >= unsigned_x // false
// -1 > unsigned_x // false
switch (op_id) {
case IrBinOpCmpNotEq:
case IrBinOpCmpLessOrEq:
case IrBinOpCmpLessThan:
return ir_const_bool(ira, source_instr, true);
case IrBinOpCmpEq:
case IrBinOpCmpGreaterOrEq:
case IrBinOpCmpGreaterThan:
return ir_const_bool(ira, source_instr, false);
default:
break;
}
}
}
if (have_op2_cmp_zero && op1_scalar_type->id == ZigTypeIdInt && !op1_scalar_type->data.integral.is_signed) {
if (op2_cmp_zero == CmpEQ) {
// unsigned_x < 0 // false
// unsigned_x >= 0 // true
switch (op_id) {
case IrBinOpCmpLessThan:
return ir_const_bool(ira, source_instr, false);
case IrBinOpCmpGreaterOrEq:
return ir_const_bool(ira, source_instr, true);
default:
break;
}
} else if (op2_cmp_zero == CmpLT) {
// unsigned_x != -1 // true
// unsigned_x >= -1 // true
// unsigned_x > -1 // true
// unsigned_x == -1 // false
// unsigned_x < -1 // false
// unsigned_x <= -1 // false
switch (op_id) {
case IrBinOpCmpNotEq:
case IrBinOpCmpGreaterOrEq:
case IrBinOpCmpGreaterThan:
return ir_const_bool(ira, source_instr, true);
case IrBinOpCmpEq:
case IrBinOpCmpLessThan:
case IrBinOpCmpLessOrEq:
return ir_const_bool(ira, source_instr, false);
default:
break;
}
}
}
// It must be a runtime comparison.
// For floats, emit a float comparison instruction.
bool op1_is_float = op1_scalar_type->id == ZigTypeIdFloat || op1_scalar_type->id == ZigTypeIdComptimeFloat;
bool op2_is_float = op2_scalar_type->id == ZigTypeIdFloat || op2_scalar_type->id == ZigTypeIdComptimeFloat;
if (op1_is_float && op2_is_float) {
// Implicit cast the smaller one to the larger one.
ZigType *dest_scalar_type;
if (op1_scalar_type->id == ZigTypeIdComptimeFloat) {
dest_scalar_type = op2_scalar_type;
} else if (op2_scalar_type->id == ZigTypeIdComptimeFloat) {
dest_scalar_type = op1_scalar_type;
} else if (op1_scalar_type->data.floating.bit_count >= op2_scalar_type->data.floating.bit_count) {
dest_scalar_type = op1_scalar_type;
} else {
dest_scalar_type = op2_scalar_type;
}
ZigType *dest_type = (result_type->id == ZigTypeIdVector) ?
get_vector_type(ira->codegen, result_type->data.vector.len, dest_scalar_type) : dest_scalar_type;
IrInstruction *casted_op1 = ir_implicit_cast(ira, op1, dest_type);
IrInstruction *casted_op2 = ir_implicit_cast(ira, op2, dest_type);
if (type_is_invalid(casted_op1->value->type) || type_is_invalid(casted_op2->value->type))
return ira->codegen->invalid_instruction;
return ir_build_bin_op_gen(ira, source_instr, result_type, op_id, casted_op1, casted_op2, true);
}
// For mixed unsigned integer sizes, implicit cast both operands to the larger integer.
// For mixed signed and unsigned integers, implicit cast both operands to a signed
// integer with + 1 bit.
// For mixed floats and integers, extract the integer part from the float, cast that to
// a signed integer with mantissa bits + 1, and if there was any non-integral part of the float,
// add/subtract 1.
bool dest_int_is_signed = false;
if (have_op1_cmp_zero) {
if (op1_cmp_zero == CmpLT) dest_int_is_signed = true;
} else if (op1_is_float) {
dest_int_is_signed = true;
} else if (op1_scalar_type->id == ZigTypeIdInt && op1_scalar_type->data.integral.is_signed) {
dest_int_is_signed = true;
}
if (have_op2_cmp_zero) {
if (op2_cmp_zero == CmpLT) dest_int_is_signed = true;
} else if (op2_is_float) {
dest_int_is_signed = true;
} else if (op2->value->type->id == ZigTypeIdInt && op2->value->type->data.integral.is_signed) {
dest_int_is_signed = true;
}
ZigType *dest_float_type = nullptr;
uint32_t op1_bits;
if (instr_is_comptime(op1)) {
ZigValue *op1_val = ir_resolve_const(ira, op1, UndefOk);
if (op1_val == nullptr)
return ira->codegen->invalid_instruction;
if (op1_val->special == ConstValSpecialUndef)
return ir_const_undef(ira, source_instr, ira->codegen->builtin_types.entry_bool);
if (result_type->id == ZigTypeIdVector) {
ir_add_error(ira, op1, buf_sprintf("compiler bug: TODO: support comptime vector here"));
return ira->codegen->invalid_instruction;
}
bool is_unsigned;
if (op1_is_float) {
BigInt bigint = {};
float_init_bigint(&bigint, op1_val);
Cmp zcmp = float_cmp_zero(op1_val);
if (float_has_fraction(op1_val)) {
if (op_id == IrBinOpCmpEq || op_id == IrBinOpCmpNotEq) {
return ir_const_bool(ira, source_instr, op_id == IrBinOpCmpNotEq);
}
if (zcmp == CmpLT) {
bigint_decr(&bigint);
} else {
bigint_incr(&bigint);
}
}
op1_bits = bigint_bits_needed(&bigint);
is_unsigned = zcmp != CmpLT;
} else {
op1_bits = bigint_bits_needed(&op1_val->data.x_bigint);
is_unsigned = bigint_cmp_zero(&op1_val->data.x_bigint) != CmpLT;
}
if (is_unsigned && dest_int_is_signed) {
op1_bits += 1;
}
} else if (op1_is_float) {
dest_float_type = op1_scalar_type;
} else {
ir_assert(op1_scalar_type->id == ZigTypeIdInt, source_instr);
op1_bits = op1_scalar_type->data.integral.bit_count;
if (!op1_scalar_type->data.integral.is_signed && dest_int_is_signed) {
op1_bits += 1;
}
}
uint32_t op2_bits;
if (instr_is_comptime(op2)) {
ZigValue *op2_val = ir_resolve_const(ira, op2, UndefOk);
if (op2_val == nullptr)
return ira->codegen->invalid_instruction;
if (op2_val->special == ConstValSpecialUndef)
return ir_const_undef(ira, source_instr, ira->codegen->builtin_types.entry_bool);
if (result_type->id == ZigTypeIdVector) {
ir_add_error(ira, op2, buf_sprintf("compiler bug: TODO: support comptime vector here"));
return ira->codegen->invalid_instruction;
}
bool is_unsigned;
if (op2_is_float) {
BigInt bigint = {};
float_init_bigint(&bigint, op2_val);
Cmp zcmp = float_cmp_zero(op2_val);
if (float_has_fraction(op2_val)) {
if (op_id == IrBinOpCmpEq || op_id == IrBinOpCmpNotEq) {
return ir_const_bool(ira, source_instr, op_id == IrBinOpCmpNotEq);
}
if (zcmp == CmpLT) {
bigint_decr(&bigint);
} else {
bigint_incr(&bigint);
}
}
op2_bits = bigint_bits_needed(&bigint);
is_unsigned = zcmp != CmpLT;
} else {
op2_bits = bigint_bits_needed(&op2_val->data.x_bigint);
is_unsigned = bigint_cmp_zero(&op2_val->data.x_bigint) != CmpLT;
}
if (is_unsigned && dest_int_is_signed) {
op2_bits += 1;
}
} else if (op2_is_float) {
dest_float_type = op2_scalar_type;
} else {
ir_assert(op2_scalar_type->id == ZigTypeIdInt, source_instr);
op2_bits = op2_scalar_type->data.integral.bit_count;
if (!op2_scalar_type->data.integral.is_signed && dest_int_is_signed) {
op2_bits += 1;
}
}
uint32_t dest_int_bits = (op1_bits > op2_bits) ? op1_bits : op2_bits;
ZigType *dest_scalar_type = (dest_float_type == nullptr) ?
get_int_type(ira->codegen, dest_int_is_signed, dest_int_bits) : dest_float_type;
ZigType *dest_type = (result_type->id == ZigTypeIdVector) ?
get_vector_type(ira->codegen, result_type->data.vector.len, dest_scalar_type) : dest_scalar_type;
IrInstruction *casted_op1 = ir_implicit_cast(ira, op1, dest_type);
if (type_is_invalid(casted_op1->value->type))
return ira->codegen->invalid_instruction;
IrInstruction *casted_op2 = ir_implicit_cast(ira, op2, dest_type);
if (type_is_invalid(casted_op2->value->type))
return ira->codegen->invalid_instruction;
return ir_build_bin_op_gen(ira, source_instr, result_type, op_id, casted_op1, casted_op2, true);
}
static IrInstruction *ir_analyze_bin_op_cmp(IrAnalyze *ira, IrInstructionBinOp *bin_op_instruction) {
IrInstruction *op1 = bin_op_instruction->op1->child;
if (type_is_invalid(op1->value->type))
return ira->codegen->invalid_instruction;
@ -14726,6 +15187,13 @@ static IrInstruction *ir_analyze_bin_op_cmp(IrAnalyze *ira, IrInstructionBinOp *
return result;
}
if (type_is_numeric(op1->value->type) && type_is_numeric(op2->value->type)) {
// This operation allows any combination of integer and float types, regardless of the
// signed-ness, comptime-ness, and bit-width. So peer type resolution is incorrect for
// numeric types.
return ir_analyze_bin_op_cmp_numeric(ira, &bin_op_instruction->base, op1, op2, op_id);
}
IrInstruction *instructions[] = {op1, op2};
ZigType *resolved_type = ir_resolve_peer_types(ira, source_node, nullptr, instructions, 2);
if (type_is_invalid(resolved_type))
@ -14741,8 +15209,7 @@ static IrInstruction *ir_analyze_bin_op_cmp(IrAnalyze *ira, IrInstructionBinOp *
case ZigTypeIdInt:
case ZigTypeIdFloat:
case ZigTypeIdVector:
operator_allowed = true;
break;
zig_unreachable(); // handled with the type_is_numeric checks above
case ZigTypeIdBool:
case ZigTypeIdMetaType:
@ -14802,50 +15269,6 @@ static IrInstruction *ir_analyze_bin_op_cmp(IrAnalyze *ira, IrInstructionBinOp *
}
if (one_possible_value || (instr_is_comptime(casted_op1) && instr_is_comptime(casted_op2))) {
{
// Before resolving the values, we special case comparisons against zero. These can often be done
// without resolving lazy values, preventing potential dependency loops.
Cmp op1_cmp_zero;
if ((err = lazy_cmp_zero(bin_op_instruction->base.source_node, casted_op1->value, &op1_cmp_zero))) {
if (err == ErrorNotLazy) goto never_mind_just_calculate_it_normally;
return ira->codegen->invalid_instruction;
}
Cmp op2_cmp_zero;
if ((err = lazy_cmp_zero(bin_op_instruction->base.source_node, casted_op2->value, &op2_cmp_zero))) {
if (err == ErrorNotLazy) goto never_mind_just_calculate_it_normally;
return ira->codegen->invalid_instruction;
}
bool can_cmp_zero = false;
Cmp cmp_result;
if (op1_cmp_zero == CmpEQ && op2_cmp_zero == CmpEQ) {
can_cmp_zero = true;
cmp_result = CmpEQ;
} else if (op1_cmp_zero == CmpGT && op2_cmp_zero == CmpEQ) {
can_cmp_zero = true;
cmp_result = CmpGT;
} else if (op1_cmp_zero == CmpEQ && op2_cmp_zero == CmpGT) {
can_cmp_zero = true;
cmp_result = CmpLT;
} else if (op1_cmp_zero == CmpLT && op2_cmp_zero == CmpEQ) {
can_cmp_zero = true;
cmp_result = CmpLT;
} else if (op1_cmp_zero == CmpEQ && op2_cmp_zero == CmpLT) {
can_cmp_zero = true;
cmp_result = CmpGT;
} else if (op1_cmp_zero == CmpLT && op2_cmp_zero == CmpGT) {
can_cmp_zero = true;
cmp_result = CmpLT;
} else if (op1_cmp_zero == CmpGT && op2_cmp_zero == CmpLT) {
can_cmp_zero = true;
cmp_result = CmpGT;
}
if (can_cmp_zero) {
bool answer = resolve_cmp_op_id(op_id, cmp_result);
return ir_const_bool(ira, &bin_op_instruction->base, answer);
}
}
never_mind_just_calculate_it_normally:
ZigValue *op1_val = one_possible_value ? casted_op1->value : ir_resolve_const(ira, casted_op1, UndefBad);
if (op1_val == nullptr)
return ira->codegen->invalid_instruction;
@ -14870,43 +15293,6 @@ never_mind_just_calculate_it_normally:
return result;
}
// some comparisons with unsigned numbers can be evaluated
if (resolved_type->id == ZigTypeIdInt && !resolved_type->data.integral.is_signed) {
ZigValue *known_left_val;
IrBinOp flipped_op_id;
if (instr_is_comptime(casted_op1)) {
known_left_val = ir_resolve_const(ira, casted_op1, UndefBad);
if (known_left_val == nullptr)
return ira->codegen->invalid_instruction;
flipped_op_id = op_id;
} else if (instr_is_comptime(casted_op2)) {
known_left_val = ir_resolve_const(ira, casted_op2, UndefBad);
if (known_left_val == nullptr)
return ira->codegen->invalid_instruction;
if (op_id == IrBinOpCmpLessThan) {
flipped_op_id = IrBinOpCmpGreaterThan;
} else if (op_id == IrBinOpCmpGreaterThan) {
flipped_op_id = IrBinOpCmpLessThan;
} else if (op_id == IrBinOpCmpLessOrEq) {
flipped_op_id = IrBinOpCmpGreaterOrEq;
} else if (op_id == IrBinOpCmpGreaterOrEq) {
flipped_op_id = IrBinOpCmpLessOrEq;
} else {
flipped_op_id = op_id;
}
} else {
known_left_val = nullptr;
}
if (known_left_val != nullptr && bigint_cmp_zero(&known_left_val->data.x_bigint) == CmpEQ &&
(flipped_op_id == IrBinOpCmpLessOrEq || flipped_op_id == IrBinOpCmpGreaterThan))
{
bool answer = (flipped_op_id == IrBinOpCmpLessOrEq);
return ir_const_bool(ira, &bin_op_instruction->base, answer);
}
}
IrInstruction *result = ir_build_bin_op(&ira->new_irb,
bin_op_instruction->base.scope, bin_op_instruction->base.source_node,
op_id, casted_op1, casted_op2, bin_op_instruction->safety_check_on);
@ -23959,26 +24345,32 @@ static IrInstruction *ir_analyze_instruction_int_to_float(IrAnalyze *ira, IrInst
return ir_resolve_cast(ira, &instruction->base, target, dest_type, CastOpIntToFloat);
}
static IrInstruction *ir_analyze_float_to_int(IrAnalyze *ira, IrInstruction *source_instr,
ZigType *dest_type, IrInstruction *operand, AstNode *operand_source_node)
{
if (operand->value->type->id == ZigTypeIdComptimeInt) {
return ir_implicit_cast(ira, operand, dest_type);
}
if (operand->value->type->id != ZigTypeIdFloat && operand->value->type->id != ZigTypeIdComptimeFloat) {
ir_add_error_node(ira, operand_source_node, buf_sprintf("expected float type, found '%s'",
buf_ptr(&operand->value->type->name)));
return ira->codegen->invalid_instruction;
}
return ir_resolve_cast(ira, source_instr, operand, dest_type, CastOpFloatToInt);
}
static IrInstruction *ir_analyze_instruction_float_to_int(IrAnalyze *ira, IrInstructionFloatToInt *instruction) {
ZigType *dest_type = ir_resolve_type(ira, instruction->dest_type->child);
if (type_is_invalid(dest_type))
return ira->codegen->invalid_instruction;
IrInstruction *target = instruction->target->child;
if (type_is_invalid(target->value->type))
IrInstruction *operand = instruction->target->child;
if (type_is_invalid(operand->value->type))
return ira->codegen->invalid_instruction;
if (target->value->type->id == ZigTypeIdComptimeInt) {
return ir_implicit_cast(ira, target, dest_type);
}
if (target->value->type->id != ZigTypeIdFloat && target->value->type->id != ZigTypeIdComptimeFloat) {
ir_add_error(ira, instruction->target, buf_sprintf("expected float type, found '%s'",
buf_ptr(&target->value->type->name)));
return ira->codegen->invalid_instruction;
}
return ir_resolve_cast(ira, &instruction->base, target, dest_type, CastOpFloatToInt);
return ir_analyze_float_to_int(ira, &instruction->base, dest_type, operand, instruction->target->source_node);
}
static IrInstruction *ir_analyze_instruction_err_to_int(IrAnalyze *ira, IrInstructionErrToInt *instruction) {
@ -27111,13 +27503,13 @@ static IrInstruction *ir_analyze_instruction_save_err_ret_addr(IrAnalyze *ira, I
return result;
}
static void ir_eval_float_op(IrAnalyze *ira, IrInstructionFloatOp *source_instr, ZigType *float_type,
ZigValue *op, ZigValue *out_val) {
static void ir_eval_float_op(IrAnalyze *ira, IrInstruction *source_instr, BuiltinFnId fop, ZigType *float_type,
ZigValue *op, ZigValue *out_val)
{
assert(ira && source_instr && float_type && out_val && op);
assert(float_type->id == ZigTypeIdFloat ||
float_type->id == ZigTypeIdComptimeFloat);
BuiltinFnId fop = source_instr->op;
unsigned bits;
switch (float_type->id) {
@ -27291,45 +27683,39 @@ static void ir_eval_float_op(IrAnalyze *ira, IrInstructionFloatOp *source_instr,
}
}
static IrInstruction *ir_analyze_instruction_float_op(IrAnalyze *ira, IrInstructionFloatOp *instruction) {
IrInstruction *type = instruction->type->child;
if (type_is_invalid(type->value->type))
return ira->codegen->invalid_instruction;
ZigType *expr_type = ir_resolve_type(ira, type);
if (type_is_invalid(expr_type))
return ira->codegen->invalid_instruction;
static IrInstruction *ir_analyze_float_op(IrAnalyze *ira, IrInstruction *source_instr,
ZigType *expr_type, AstNode *expr_type_src_node, IrInstruction *operand, BuiltinFnId op)
{
// Only allow float types, and vectors of floats.
ZigType *float_type = (expr_type->id == ZigTypeIdVector) ? expr_type->data.vector.elem_type : expr_type;
if (float_type->id != ZigTypeIdFloat && float_type->id != ZigTypeIdComptimeFloat) {
ir_add_error(ira, instruction->type, buf_sprintf("@%s does not support type '%s'", float_op_to_name(instruction->op, false), buf_ptr(&float_type->name)));
ir_add_error_node(ira, expr_type_src_node,
buf_sprintf("@%s does not support type '%s'",
float_op_to_name(op, false), buf_ptr(&float_type->name)));
return ira->codegen->invalid_instruction;
}
IrInstruction *op1 = instruction->op1->child;
if (type_is_invalid(op1->value->type))
IrInstruction *casted_op = ir_implicit_cast(ira, operand, float_type);
if (type_is_invalid(casted_op->value->type))
return ira->codegen->invalid_instruction;
IrInstruction *casted_op1 = ir_implicit_cast(ira, op1, float_type);
if (type_is_invalid(casted_op1->value->type))
return ira->codegen->invalid_instruction;
if (instr_is_comptime(casted_op1)) {
// Our comptime 16-bit and 128-bit support is quite limited.
if (instr_is_comptime(casted_op)) {
if ((float_type->id == ZigTypeIdComptimeFloat ||
float_type->data.floating.bit_count == 16 ||
float_type->data.floating.bit_count == 128) &&
instruction->op != BuiltinFnIdSqrt) {
ir_add_error(ira, instruction->type, buf_sprintf("@%s does not support type '%s'", float_op_to_name(instruction->op, false), buf_ptr(&float_type->name)));
op != BuiltinFnIdSqrt)
{
ir_add_error(ira, source_instr,
buf_sprintf("compiler bug: TODO make @%s support type '%s'",
float_op_to_name(op, false), buf_ptr(&float_type->name)));
return ira->codegen->invalid_instruction;
}
ZigValue *op1_const = ir_resolve_const(ira, casted_op1, UndefBad);
ZigValue *op1_const = ir_resolve_const(ira, casted_op, UndefBad);
if (!op1_const)
return ira->codegen->invalid_instruction;
IrInstruction *result = ir_const(ira, &instruction->base, expr_type);
IrInstruction *result = ir_const(ira, source_instr, expr_type);
ZigValue *out_val = result->value;
if (expr_type->id == ZigTypeIdVector) {
@ -27342,26 +27728,38 @@ static IrInstruction *ir_analyze_instruction_float_op(IrAnalyze *ira, IrInstruct
ZigValue *float_out_val = &out_val->data.x_array.data.s_none.elements[i];
assert(float_operand_op1->type == float_type);
assert(float_out_val->type == float_type);
ir_eval_float_op(ira, instruction, float_type,
op1_const, float_out_val);
ir_eval_float_op(ira, source_instr, op, float_type, op1_const, float_out_val);
float_out_val->type = float_type;
}
out_val->type = expr_type;
out_val->special = ConstValSpecialStatic;
} else {
ir_eval_float_op(ira, instruction, float_type, op1_const, out_val);
ir_eval_float_op(ira, source_instr, op, float_type, op1_const, out_val);
}
return result;
}
ir_assert(float_type->id == ZigTypeIdFloat, &instruction->base);
ir_assert(float_type->id == ZigTypeIdFloat, source_instr);
IrInstruction *result = ir_build_float_op(&ira->new_irb, instruction->base.scope,
instruction->base.source_node, nullptr, casted_op1, instruction->op);
IrInstruction *result = ir_build_float_op(&ira->new_irb, source_instr->scope,
source_instr->source_node, nullptr, casted_op, op);
result->value->type = expr_type;
return result;
}
static IrInstruction *ir_analyze_instruction_float_op(IrAnalyze *ira, IrInstructionFloatOp *instruction) {
ZigType *expr_type = ir_resolve_type(ira, instruction->type->child);
if (type_is_invalid(expr_type))
return ira->codegen->invalid_instruction;
IrInstruction *operand = instruction->op1->child;
if (type_is_invalid(operand->value->type))
return ira->codegen->invalid_instruction;
return ir_analyze_float_op(ira, &instruction->base, expr_type, instruction->type->source_node,
operand, instruction->op);
}
static IrInstruction *ir_analyze_instruction_bswap(IrAnalyze *ira, IrInstructionBswap *instruction) {
Error err;

58
test/compile_errors.zig
View File

@ -4174,58 +4174,50 @@ pub fn addCases(cases: *tests.CompileErrorContext) void {
"tmp.zig:3:5: error: use of undefined value here causes undefined behavior",
});
cases.add("equal on undefined value",
cases.add("comparison operators with undefined value",
\\// operator ==
\\comptime {
\\ var a: i64 = undefined;
\\ _ = a == a;
\\ var x: i32 = 0;
\\ if (a == a) x += 1;
\\}
, &[_][]const u8{
"tmp.zig:3:9: error: use of undefined value here causes undefined behavior",
});
cases.add("not equal on undefined value",
\\// operator !=
\\comptime {
\\ var a: i64 = undefined;
\\ _ = a != a;
\\ var x: i32 = 0;
\\ if (a != a) x += 1;
\\}
, &[_][]const u8{
"tmp.zig:3:9: error: use of undefined value here causes undefined behavior",
});
cases.add("greater than on undefined value",
\\// operator >
\\comptime {
\\ var a: i64 = undefined;
\\ _ = a > a;
\\ var x: i32 = 0;
\\ if (a > a) x += 1;
\\}
, &[_][]const u8{
"tmp.zig:3:9: error: use of undefined value here causes undefined behavior",
});
cases.add("greater than equal on undefined value",
\\// operator <
\\comptime {
\\ var a: i64 = undefined;
\\ _ = a >= a;
\\ var x: i32 = 0;
\\ if (a < a) x += 1;
\\}
, &[_][]const u8{
"tmp.zig:3:9: error: use of undefined value here causes undefined behavior",
});
cases.add("less than on undefined value",
\\// operator >=
\\comptime {
\\ var a: i64 = undefined;
\\ _ = a < a;
\\ var x: i32 = 0;
\\ if (a >= a) x += 1;
\\}
, &[_][]const u8{
"tmp.zig:3:9: error: use of undefined value here causes undefined behavior",
});
cases.add("less than equal on undefined value",
\\// operator <=
\\comptime {
\\ var a: i64 = undefined;
\\ _ = a <= a;
\\ var x: i32 = 0;
\\ if (a <= a) x += 1;
\\}
, &[_][]const u8{
"tmp.zig:3:9: error: use of undefined value here causes undefined behavior",
"tmp.zig:5:11: error: use of undefined value here causes undefined behavior",
"tmp.zig:11:11: error: use of undefined value here causes undefined behavior",
"tmp.zig:17:11: error: use of undefined value here causes undefined behavior",
"tmp.zig:23:11: error: use of undefined value here causes undefined behavior",
"tmp.zig:29:11: error: use of undefined value here causes undefined behavior",
"tmp.zig:35:11: error: use of undefined value here causes undefined behavior",
});
cases.add("and on undefined value",

18
test/stage1/behavior/math.zig
View File

@ -4,6 +4,7 @@ const expectEqual = std.testing.expectEqual;
const expectEqualSlices = std.testing.expectEqualSlices;
const maxInt = std.math.maxInt;
const minInt = std.math.minInt;
const mem = std.mem;
test "division" {
if (@import("builtin").arch == .riscv64) {
@ -653,3 +654,20 @@ test "128-bit multiplication" {
var c = a * b;
expect(c == 6);
}
test "vector comparison" {
const S = struct {
fn doTheTest() void {
var a: @Vector(6, i32) = [_]i32{1, 3, -1, 5, 7, 9};
var b: @Vector(6, i32) = [_]i32{-1, 3, 0, 6, 10, -10};
expect(mem.eql(bool, &@as([6]bool, a < b), &[_]bool{false, false, true, true, true, false}));
expect(mem.eql(bool, &@as([6]bool, a <= b), &[_]bool{false, true, true, true, true, false}));
expect(mem.eql(bool, &@as([6]bool, a == b), &[_]bool{false, true, false, false, false, false}));
expect(mem.eql(bool, &@as([6]bool, a != b), &[_]bool{true, false, true, true, true, true}));
expect(mem.eql(bool, &@as([6]bool, a > b), &[_]bool{true, false, false, false, false, true}));
expect(mem.eql(bool, &@as([6]bool, a >= b), &[_]bool{true, true, false, false, false, true}));
}
};
S.doTheTest();
comptime S.doTheTest();
}