1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-24 19:52:54 +01:00
llvm-mirror/test/CodeGen/AArch64/global-merge-4.ll
Ahmed Bougacha dda2ff1737 [AArch64, ARM] Enable GlobalMerge with -O3 rather than -O1.
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.

llvm-svn: 233024
2015-03-23 21:17:36 +00:00

74 lines
4.1 KiB
LLVM

; RUN: llc %s -mtriple=aarch64-linux-gnuabi -O3 -enable-global-merge -o - | FileCheck %s
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-n32:64-S128"
target triple = "arm64-apple-ios7.0.0"
@bar = internal global [5 x i32] zeroinitializer, align 4
@baz = internal global [5 x i32] zeroinitializer, align 4
@foo = internal global [5 x i32] zeroinitializer, align 4
; Function Attrs: nounwind ssp
define internal void @initialize() #0 {
%1 = tail call i32 bitcast (i32 (...)* @calc to i32 ()*)() #2
store i32 %1, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @bar, i64 0, i64 0), align 4
%2 = tail call i32 bitcast (i32 (...)* @calc to i32 ()*)() #2
store i32 %2, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @baz, i64 0, i64 0), align 4
%3 = tail call i32 bitcast (i32 (...)* @calc to i32 ()*)() #2
store i32 %3, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @bar, i64 0, i64 1), align 4
%4 = tail call i32 bitcast (i32 (...)* @calc to i32 ()*)() #2
store i32 %4, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @baz, i64 0, i64 1), align 4
%5 = tail call i32 bitcast (i32 (...)* @calc to i32 ()*)() #2
store i32 %5, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @bar, i64 0, i64 2), align 4
%6 = tail call i32 bitcast (i32 (...)* @calc to i32 ()*)() #2
store i32 %6, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @baz, i64 0, i64 2), align 4
%7 = tail call i32 bitcast (i32 (...)* @calc to i32 ()*)() #2
store i32 %7, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @bar, i64 0, i64 3), align 4
%8 = tail call i32 bitcast (i32 (...)* @calc to i32 ()*)() #2
store i32 %8, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @baz, i64 0, i64 3), align 4
%9 = tail call i32 bitcast (i32 (...)* @calc to i32 ()*)() #2
store i32 %9, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @bar, i64 0, i64 4), align 4
%10 = tail call i32 bitcast (i32 (...)* @calc to i32 ()*)() #2
store i32 %10, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @baz, i64 0, i64 4), align 4
ret void
}
declare i32 @calc(...)
; Function Attrs: nounwind ssp
define internal void @calculate() #0 {
%1 = load i32, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @bar, i64 0, i64 0), align 4
%2 = load i32, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @baz, i64 0, i64 0), align 4
%3 = mul nsw i32 %2, %1
store i32 %3, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @foo, i64 0, i64 0), align 4
%4 = load i32, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @bar, i64 0, i64 1), align 4
%5 = load i32, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @baz, i64 0, i64 1), align 4
%6 = mul nsw i32 %5, %4
store i32 %6, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @foo, i64 0, i64 1), align 4
%7 = load i32, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @bar, i64 0, i64 2), align 4
%8 = load i32, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @baz, i64 0, i64 2), align 4
%9 = mul nsw i32 %8, %7
store i32 %9, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @foo, i64 0, i64 2), align 4
%10 = load i32, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @bar, i64 0, i64 3), align 4
%11 = load i32, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @baz, i64 0, i64 3), align 4
%12 = mul nsw i32 %11, %10
store i32 %12, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @foo, i64 0, i64 3), align 4
%13 = load i32, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @bar, i64 0, i64 4), align 4
%14 = load i32, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @baz, i64 0, i64 4), align 4
%15 = mul nsw i32 %14, %13
store i32 %15, i32* getelementptr inbounds ([5 x i32], [5 x i32]* @foo, i64 0, i64 4), align 4
ret void
}
; Function Attrs: nounwind readnone ssp
define internal i32* @returnFoo() #1 {
ret i32* getelementptr inbounds ([5 x i32], [5 x i32]* @foo, i64 0, i64 0)
}
;CHECK: .type _MergedGlobals,@object // @_MergedGlobals
;CHECK: .local _MergedGlobals
;CHECK: .comm _MergedGlobals,60,16
attributes #0 = { nounwind ssp }
attributes #1 = { nounwind readnone ssp }
attributes #2 = { nounwind }