architeuthis/nac3
1
0
Fork 0
forked from M-Labs/nac3
Code Pull Requests Activity

Merge pull request 'Fix float**int with negative power' (#254) from neg_powi_fix into master

Browse Source 
Reviewed-on: M-Labs/nac3#254
This commit is contained in:
pca006132 2022-04-04 22:43:20 +08:00
commit 0d10044d66
2 changed files with 30 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
nac3core/src/codegen/expr.rs
View File

@ -948,15 +948,12 @@ pub fn gen_binop_expr<'ctx, 'a, G: CodeGenerator>(
} else if ty1 == ctx.primitives.float && ty2 == ctx.primitives.int32 { } else if ty1 == ctx.primitives.float && ty2 == ctx.primitives.int32 {
// Pow is the only operator that would pass typecheck between float and int // Pow is the only operator that would pass typecheck between float and int
assert!(*op == Operator::Pow); assert!(*op == Operator::Pow);
// TODO: throw exception when rhs is out of i16 bound let i32_t = ctx.ctx.i32_type();
// since llvm intrinsic only support to i16 for f64 let pow_intr = ctx.module.get_function("llvm.powi.f64.i32").unwrap_or_else(|| {
let i16_t = ctx.ctx.i16_type();
let pow_intr = ctx.module.get_function("llvm.powi.f64.i16").unwrap_or_else(|| {
let f64_t = ctx.ctx.f64_type(); let f64_t = ctx.ctx.f64_type();
let ty = f64_t.fn_type(&[f64_t.into(), i16_t.into()], false); let ty = f64_t.fn_type(&[f64_t.into(), i32_t.into()], false);
ctx.module.add_function("llvm.powi.f64.i16", ty, None) ctx.module.add_function("llvm.powi.f64.i32", ty, None)
}); });
let right = ctx.builder.build_int_truncate(right.into_int_value(), i16_t, "r_pow");
ctx.builder ctx.builder
.build_call(pow_intr, &[left.into(), right.into()], "f_pow_i") .build_call(pow_intr, &[left.into(), right.into()], "f_pow_i")
.try_as_basic_value() .try_as_basic_value()

26
nac3standalone/demo/src/pow.py Normal file
View File

@ -0,0 +1,26 @@
@extern
def output_float64(f: float):
...
def run() -> int32:
output_float64(float(3 ** 1))
output_float64(float(3 ** 0))
output_float64(float(3 ** 19))
output_float64(1.0 ** -100)
output_float64(1.0 ** -2)
output_float64(1.0 ** 0)
output_float64(1.0 ** 1)
output_float64(1.0 ** 100)
output_float64(3.0 ** 0)
output_float64(3.0 ** 1)
output_float64(3.0 ** 2)
output_float64(3.0 ** -1)
output_float64(3.0 ** -2)
output_float64(3.0 ** -32767)
output_float64(3.0 ** -3.0)
output_float64(3.0 ** -0.0)
output_float64(3.0 ** 0.0)
output_float64(4.0 ** 0.5)
output_float64(4.0 ** -0.5)
return 0