llvm-6502/test/Linker/type-unique-src-type.ll
Rafael Espindola 55d5ea3be3 Split the set of identified struct types into opaque and non-opaque ones.
The non-opaque part can be structurally uniqued. To keep this to just
a hash lookup, we don't try to unique cyclic types.

Also change the type mapping algorithm to be optimistic about a type
not being recursive and only create a new type when proven to be wrong.
This is not as strong as trying to speculate that we can keep the source
type, but is simpler (no speculation to revert) and more powerfull
than what we had before (we don't copy non-recursive types at least).

I initially wrote this to try to replace the name based type merging.
It is not strong enough to replace it, but is is a useful addition.

With this patch the number of named struct types is a clang lto bootstrap goes
from 49674 to 15986.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223278 91177308-0d34-0410-b5e6-96231b3b80d8
2014-12-03 22:36:37 +00:00

25 lines
687 B
LLVM

; RUN: llvm-as %s -o %t.bc
; RUN: llvm-link -S %t.bc -o - | FileCheck %s
; RUN: llvm-link -S %s -o - | FileCheck %s
; Test that we don't try to map %C.0 and C and then try to map %C to a new type.
; This used to happen when lazy loading since we wouldn't then identify %C
; as a destination type until it was too late.
; CHECK: %C.0 = type { %B }
; CHECK-NEXT: %B = type { %A }
; CHECK-NEXT: %A = type { i8 }
; CHECK: @g1 = external global %C.0
; CHECK: getelementptr %C.0* null, i64 0, i32 0, i32 0
%A = type { i8 }
%B = type { %A }
%C = type { %B }
%C.0 = type { %B }
define void @f1() {
getelementptr %C* null, i64 0, i32 0, i32 0
ret void
}
@g1 = external global %C.0