llvm/test/CodeGen/AArch64/large-frame.ll
Tim Northover 148ac534fc AArch64: refactor frame handling to use movz/movk for overlarge offsets.
In the near future litpools will be in a different section, which means that
any access to them is at least two instructions. This makes the case for a
movz/movk pair (if total offset <= 32-bits) even more compelling.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175257 91177308-0d34-0410-b5e6-96231b3b80d8
2013-02-15 09:33:26 +00:00

115 lines
3.1 KiB
LLVM

; RUN: llc -verify-machineinstrs -mtriple=aarch64-none-linux-gnu < %s | FileCheck %s
declare void @use_addr(i8*)
@addr = global i8* null
define void @test_bigframe() {
; CHECK: test_bigframe:
%var1 = alloca i8, i32 20000000
%var2 = alloca i8, i32 16
%var3 = alloca i8, i32 20000000
; CHECK: sub sp, sp, #496
; CHECK: str x30, [sp, #488]
; Total adjust is 39999536
; CHECK: movz [[SUBCONST:x[0-9]+]], #22576
; CHECK: movk [[SUBCONST]], #610, lsl #16
; CHECK: sub sp, sp, [[SUBCONST]]
; Total offset is 20000024
; CHECK: movz [[VAR1OFFSET:x[0-9]+]], #11544
; CHECK: movk [[VAR1OFFSET]], #305, lsl #16
; CHECK: add {{x[0-9]+}}, sp, [[VAR1OFFSET]]
store volatile i8* %var1, i8** @addr
%var1plus2 = getelementptr i8* %var1, i32 2
store volatile i8* %var1plus2, i8** @addr
; CHECK: movz [[VAR2OFFSET:x[0-9]+]], #11528
; CHECK: movk [[VAR2OFFSET]], #305, lsl #16
; CHECK: add {{x[0-9]+}}, sp, [[VAR2OFFSET]]
store volatile i8* %var2, i8** @addr
%var2plus2 = getelementptr i8* %var2, i32 2
store volatile i8* %var2plus2, i8** @addr
store volatile i8* %var3, i8** @addr
%var3plus2 = getelementptr i8* %var3, i32 2
store volatile i8* %var3plus2, i8** @addr
; CHECK: movz [[ADDCONST:x[0-9]+]], #22576
; CHECK: movk [[ADDCONST]], #610, lsl #16
; CHECK: add sp, sp, [[ADDCONST]]
ret void
}
define void @test_mediumframe() {
; CHECK: test_mediumframe:
%var1 = alloca i8, i32 1000000
%var2 = alloca i8, i32 16
%var3 = alloca i8, i32 1000000
; CHECK: sub sp, sp, #496
; CHECK: str x30, [sp, #488]
; CHECK: sub sp, sp, #688
; CHECK-NEXT: sub sp, sp, #488, lsl #12
store volatile i8* %var1, i8** @addr
; CHECK: add [[VAR1ADDR:x[0-9]+]], sp, #600
; CHECK: add [[VAR1ADDR]], [[VAR1ADDR]], #244, lsl #12
%var1plus2 = getelementptr i8* %var1, i32 2
store volatile i8* %var1plus2, i8** @addr
; CHECK: add [[VAR1PLUS2:x[0-9]+]], {{x[0-9]+}}, #2
store volatile i8* %var2, i8** @addr
; CHECK: add [[VAR2ADDR:x[0-9]+]], sp, #584
; CHECK: add [[VAR2ADDR]], [[VAR2ADDR]], #244, lsl #12
%var2plus2 = getelementptr i8* %var2, i32 2
store volatile i8* %var2plus2, i8** @addr
; CHECK: add [[VAR2PLUS2:x[0-9]+]], {{x[0-9]+}}, #2
store volatile i8* %var3, i8** @addr
%var3plus2 = getelementptr i8* %var3, i32 2
store volatile i8* %var3plus2, i8** @addr
; CHECK: add sp, sp, #688
; CHECK: add sp, sp, #488, lsl #12
; CHECK: ldr x30, [sp, #488]
; CHECK: add sp, sp, #496
ret void
}
@bigspace = global [8 x i64] zeroinitializer
; If temporary registers are allocated for adjustment, they should *not* clobber
; argument registers.
define void @test_tempallocation([8 x i64] %val) nounwind {
; CHECK: test_tempallocation:
%var = alloca i8, i32 1000000
; CHECK: sub sp, sp,
; Make sure the prologue is reasonably efficient
; CHECK-NEXT: stp x29, x30, [sp,
; CHECK-NEXT: stp x25, x26, [sp,
; CHECK-NEXT: stp x23, x24, [sp,
; CHECK-NEXT: stp x21, x22, [sp,
; CHECK-NEXT: stp x19, x20, [sp,
; Make sure we don't trash an argument register
; CHECK-NOT: movz {{x[0-7],}}
; CHECK: sub sp, sp,
; CHECK-NOT: movz {{x[0-7],}}
; CHECK: bl use_addr
call void @use_addr(i8* %var)
store [8 x i64] %val, [8 x i64]* @bigspace
ret void
; CHECK: ret
}