Auto merge of #112516 - erikdesjardins:loop, r=davidtwco
cg_llvm: use index-based loop in write_operand_repeatedly This should be easier for LLVM to analyze. Fixes #111603 This needs a perf run. [cc](https://github.com/rust-lang/rust/issues/111603#issuecomment-1567531178) `@caojoshua`
This commit is contained in:
commit
3c554f5cb4
@ -572,8 +572,6 @@ fn write_operand_repeatedly(
|
||||
) {
|
||||
let zero = self.const_usize(0);
|
||||
let count = self.const_usize(count);
|
||||
let start = dest.project_index(self, zero).llval;
|
||||
let end = dest.project_index(self, count).llval;
|
||||
|
||||
let header_bb = self.append_sibling_block("repeat_loop_header");
|
||||
let body_bb = self.append_sibling_block("repeat_loop_body");
|
||||
@ -582,24 +580,18 @@ fn write_operand_repeatedly(
|
||||
self.br(header_bb);
|
||||
|
||||
let mut header_bx = Self::build(self.cx, header_bb);
|
||||
let current = header_bx.phi(self.val_ty(start), &[start], &[self.llbb()]);
|
||||
let i = header_bx.phi(self.val_ty(zero), &[zero], &[self.llbb()]);
|
||||
|
||||
let keep_going = header_bx.icmp(IntPredicate::IntNE, current, end);
|
||||
let keep_going = header_bx.icmp(IntPredicate::IntULT, i, count);
|
||||
header_bx.cond_br(keep_going, body_bb, next_bb);
|
||||
|
||||
let mut body_bx = Self::build(self.cx, body_bb);
|
||||
let align = dest.align.restrict_for_offset(dest.layout.field(self.cx(), 0).size);
|
||||
cg_elem
|
||||
.val
|
||||
.store(&mut body_bx, PlaceRef::new_sized_aligned(current, cg_elem.layout, align));
|
||||
let dest_elem = dest.project_index(&mut body_bx, i);
|
||||
cg_elem.val.store(&mut body_bx, dest_elem);
|
||||
|
||||
let next = body_bx.inbounds_gep(
|
||||
self.backend_type(cg_elem.layout),
|
||||
current,
|
||||
&[self.const_usize(1)],
|
||||
);
|
||||
let next = body_bx.unchecked_uadd(i, self.const_usize(1));
|
||||
body_bx.br(header_bb);
|
||||
header_bx.add_incoming_to_phi(current, next, body_bb);
|
||||
header_bx.add_incoming_to_phi(i, next, body_bb);
|
||||
|
||||
*self = Self::build(self.cx, next_bb);
|
||||
}
|
||||
|
@ -5,6 +5,18 @@
|
||||
|
||||
use std::sync::Arc;
|
||||
|
||||
// CHECK-LABEL: @new_from_array
|
||||
#[no_mangle]
|
||||
pub fn new_from_array(x: u64) -> Arc<[u64]> {
|
||||
// Ensure that we only generate one alloca for the array.
|
||||
|
||||
// CHECK: alloca
|
||||
// CHECK-SAME: [1000 x i64]
|
||||
// CHECK-NOT: alloca
|
||||
let array = [x; 1000];
|
||||
Arc::new(array)
|
||||
}
|
||||
|
||||
// CHECK-LABEL: @new_uninit
|
||||
#[no_mangle]
|
||||
pub fn new_uninit(x: u64) -> Arc<[u64; 1000]> {
|
||||
|
Loading…
Reference in New Issue
Block a user