user0/rust
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

rustc: Use memmove in copy_ty. 45% LLVM codegen speed improvement.

Browse source
This commit is contained in:
Patrick Walton 2011-08-04 16:13:25 -07:00
parent e0985c1060
commit ae3312002a
1 changed files with 7 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

9
src/comp/middle/trans.rs
View file

@ -2380,7 +2380,7 @@ fn call_memmove(cx: &@block_ctxt, dst: ValueRef, src: ValueRef,
let src_ptr = cx.build.PointerCast(src, T_ptr(T_i8()));
let dst_ptr = cx.build.PointerCast(dst, T_ptr(T_i8()));
let size = cx.build.IntCast(n_bytes, T_i32());
let align = C_int(0);
let align = C_int(1);
let volatile = C_bool(false);
ret rslt(cx,
cx.build.Call(memmove,
@ -2411,7 +2411,12 @@ fn memmove_ty(cx: &@block_ctxt, dst: ValueRef, src: ValueRef, t: &ty::t) ->
if ty::type_has_dynamic_size(bcx_tcx(cx), t) {
let llsz = size_of(cx, t);
ret call_memmove(llsz.bcx, dst, src, llsz.val);
} else { ret rslt(cx, cx.build.Store(cx.build.Load(src), dst)); }
} else if ty::type_is_structural(bcx_tcx(cx), t) {
let llsz = llsize_of(type_of(bcx_ccx(cx), cx.sp, t));
ret call_memmove(cx, dst, src, llsz);
} else {
ret rslt(cx, cx.build.Store(cx.build.Load(src), dst));
}
}
// Duplicates any heap-owned memory owned by a value of the given type.