mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-08-21 18:29:46 +00:00
c9ad3ab624
The pass used to be enabled by default with CodeGenOpt::Less (-O1). This is too aggressive, considering the pass indiscriminately merges all globals together. Currently, performance doesn't always improve, and, on code that uses few globals (e.g., the odd file- or function- static), more often than not is degraded by the optimization. Lengthy discussion can be found on llvmdev (AArch64-focused; ARM has similar problems): http://lists.cs.uiuc.edu/pipermail/llvmdev/2015-February/082800.html Also, it makes tooling and debuggers less useful when dealing with globals and data sections. GlobalMerge needs to better identify those cases that benefit, and this will be done separately. In the meantime, move the pass to run with -O3 rather than -O1, on both ARM and AArch64. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@233024 91177308-0d34-0410-b5e6-96231b3b80d8
75 lines
3.0 KiB
LLVM
75 lines
3.0 KiB
LLVM
; RUN: llc < %s -mtriple=thumb-apple-darwin -O3 -global-merge-on-const=true | FileCheck %s
|
|
; Test the ARMGlobalMerge pass. Use -march=thumb because it has a small
|
|
; value for the maximum offset (127).
|
|
|
|
; A local array that exceeds the maximum offset should not be merged.
|
|
; CHECK: g0:
|
|
@g0 = internal global [32 x i32] [ i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 1, i32 2 ]
|
|
|
|
; Global variables marked with "used" attribute must be kept
|
|
; CHECK: g8
|
|
@g8 = internal global i32 0
|
|
@llvm.used = appending global [1 x i8*] [i8* bitcast (i32* @g8 to i8*)], section "llvm.metadata"
|
|
|
|
; Global used in landing pad instruction must be kept
|
|
; CHECK: ZTIi
|
|
@_ZTIi = internal global i8* null
|
|
|
|
define i32 @_Z9exceptioni(i32 %arg) {
|
|
bb:
|
|
%tmp = invoke i32 @_Z14throwSomethingi(i32 %arg)
|
|
to label %bb9 unwind label %bb1
|
|
|
|
bb1: ; preds = %bb
|
|
%tmp2 = landingpad { i8*, i32 } personality i8* bitcast (i32 (...)* @__gxx_personality_sj0 to i8*)
|
|
catch i8* bitcast (i8** @_ZTIi to i8*)
|
|
%tmp3 = extractvalue { i8*, i32 } %tmp2, 1
|
|
%tmp4 = tail call i32 @llvm.eh.typeid.for(i8* bitcast (i8** @_ZTIi to i8*))
|
|
%tmp5 = icmp eq i32 %tmp3, %tmp4
|
|
br i1 %tmp5, label %bb6, label %bb10
|
|
|
|
bb6: ; preds = %bb1
|
|
%tmp7 = extractvalue { i8*, i32 } %tmp2, 0
|
|
%tmp8 = tail call i8* @__cxa_begin_catch(i8* %tmp7)
|
|
tail call void @__cxa_end_catch()
|
|
br label %bb9
|
|
|
|
bb9: ; preds = %bb6, %bb
|
|
%res.0 = phi i32 [ 0, %bb6 ], [ %tmp, %bb ]
|
|
ret i32 %res.0
|
|
|
|
bb10: ; preds = %bb1
|
|
resume { i8*, i32 } %tmp2
|
|
}
|
|
|
|
declare i32 @_Z14throwSomethingi(i32)
|
|
|
|
declare i32 @__gxx_personality_sj0(...)
|
|
|
|
declare i32 @llvm.eh.typeid.for(i8*)
|
|
|
|
declare i8* @__cxa_begin_catch(i8*)
|
|
|
|
declare void @__cxa_end_catch()
|
|
|
|
; CHECK: _MergedGlobals:
|
|
@g1 = internal global i32 1
|
|
@g2 = internal global i32 2
|
|
|
|
; Make sure that the complete variable fits within the range of the maximum
|
|
; offset. Having the starting offset in range is not sufficient.
|
|
; When this works properly, @g3 is placed in a separate chunk of merged globals.
|
|
; CHECK: _MergedGlobals1:
|
|
@g3 = internal global [30 x i32] [ i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10 ], align 4
|
|
|
|
; Global variables that can be placed in BSS should be kept together in a
|
|
; separate pool of merged globals.
|
|
; CHECK: _MergedGlobals2
|
|
@g4 = internal global i32 0
|
|
@g5 = internal global i32 0
|
|
|
|
; Global variables that are constant can be merged together
|
|
; CHECK: _MergedGlobals3
|
|
@g6 = internal constant [12 x i32] zeroinitializer, align 4
|
|
@g7 = internal constant [12 x i32] zeroinitializer, align 4
|