Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Remove now-redundant range check on u128 -> f32 casts #67328

Merged
merged 1 commit into from
Dec 20, 2019
Merged

Remove now-redundant range check on u128 -> f32 casts #67328

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
43 changes: 7 additions & 36 deletions src/librustc_codegen_ssa/mir/rvalue.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -341,6 +341,13 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
llval
}
}
(CastTy::Int(_), CastTy::Float) => {
if signed {
bx.sitofp(llval, ll_t_out)
} else {
bx.uitofp(llval, ll_t_out)
}
}
(CastTy::Ptr(_), CastTy::Ptr(_)) |
(CastTy::FnPtr, CastTy::Ptr(_)) |
(CastTy::RPtr(_), CastTy::Ptr(_)) =>
Expand All @@ -352,8 +359,6 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
let usize_llval = bx.intcast(llval, bx.cx().type_isize(), signed);
bx.inttoptr(usize_llval, ll_t_out)
}
(CastTy::Int(_), CastTy::Float) =>
cast_int_to_float(&mut bx, signed, llval, ll_t_in, ll_t_out),
(CastTy::Float, CastTy::Int(IntTy::I)) =>
cast_float_to_int(&mut bx, true, llval, ll_t_in, ll_t_out),
(CastTy::Float, CastTy::Int(_)) =>
Expand Down Expand Up @@ -720,40 +725,6 @@ impl<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>> FunctionCx<'a, 'tcx, Bx> {
}
}

fn cast_int_to_float<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>>(
bx: &mut Bx,
signed: bool,
x: Bx::Value,
int_ty: Bx::Type,
float_ty: Bx::Type
) -> Bx::Value {
// Most integer types, even i128, fit into [-f32::MAX, f32::MAX] after rounding.
// It's only u128 -> f32 that can cause overflows (i.e., should yield infinity).
// LLVM's uitofp produces undef in those cases, so we manually check for that case.
let is_u128_to_f32 = !signed &&
bx.cx().int_width(int_ty) == 128 &&
bx.cx().float_width(float_ty) == 32;
if is_u128_to_f32 {
// All inputs greater or equal to (f32::MAX + 0.5 ULP) are rounded to infinity,
// and for everything else LLVM's uitofp works just fine.
use rustc_apfloat::ieee::Single;
const MAX_F32_PLUS_HALF_ULP: u128 = ((1 << (Single::PRECISION + 1)) - 1)
<< (Single::MAX_EXP - Single::PRECISION as i16);
let max = bx.cx().const_uint_big(int_ty, MAX_F32_PLUS_HALF_ULP);
let overflow = bx.icmp(IntPredicate::IntUGE, x, max);
let infinity_bits = bx.cx().const_u32(ieee::Single::INFINITY.to_bits() as u32);
let infinity = bx.bitcast(infinity_bits, float_ty);
let fp = bx.uitofp(x, float_ty);
bx.select(overflow, infinity, fp)
} else {
if signed {
bx.sitofp(x, float_ty)
} else {
bx.uitofp(x, float_ty)
}
}
}

fn cast_float_to_int<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>>(
bx: &mut Bx,
signed: bool,
Expand Down