Files
clang-p2996/llvm/test/CodeGen/SystemZ/cond-move-08.mir
Jonas Paulsson 16b7cc69ef [SystemZ] Eliminate call sequence instructions early. (#77812)
On SystemZ, the outgoing argument area which is big enough for all calls
in the function is created once during the prolog, as opposed to
adjusting the stack around each call. The call-sequence instructions are
therefore not really useful any more than to compute the maximum call
frame size, which has so far been done by PEI, but can just as well be
done at an earlier point.

This patch removes the mapping of the CallFrameSetupOpcode and
CallFrameDestroyOpcode and instead computes the MaxCallFrameSize
directly after instruction selection and then removes the ADJCALLSTACK
pseudos. This removes the confusing pseudos and also avoids the problem
of having to keep the call frame size accurate when creating new MBBs.

This fixes #76618 which exposed the need to maintain the call frame size
when splitting blocks (which was not done).
2024-03-28 18:26:38 +01:00

179 lines
5.4 KiB
YAML

# RUN: llc -mtriple=s390x-linux-gnu -mcpu=z15 -start-before=greedy %s -o - \
# RUN: | FileCheck %s
#
# Test that regalloc manages (via regalloc hints) to avoid a LOCRMux jump
# sequence expansion, and a SELR/LOCR instuction is emitted.
--- |
; ModuleID = 'tc.ll'
source_filename = "tc.ll"
target datalayout = "E-m:e-i1:8:16-i8:8:16-i64:64-f128:64-v128:64-a:8:16-n32:64"
@globvar = external global i32
declare void @fun() #0
define void @fun1() #0 {
bb5:
br label %bb6
bb6: ; preds = %bb33, %bb5
%tmp = phi i1 [ %tmp34, %bb33 ], [ undef, %bb5 ]
br label %bb7
bb7: ; preds = %bb7, %bb6
%lsr.iv1 = phi ptr [ %0, %bb7 ], [ undef, %bb6 ]
%tmp8 = phi i32 [ %tmp27, %bb7 ], [ -1000000, %bb6 ]
%tmp9 = phi i64 [ %tmp28, %bb7 ], [ 0, %bb6 ]
%lsr3 = trunc i64 %tmp9 to i32
%lsr.iv12 = bitcast ptr %lsr.iv1 to ptr
%tmp11 = load i32, ptr %lsr.iv12
%tmp12 = icmp sgt i32 %tmp11, undef
%tmp13 = trunc i64 %tmp9 to i32
%tmp14 = select i1 %tmp12, i32 %lsr3, i32 0
%tmp15 = select i1 %tmp12, i32 %tmp13, i32 %tmp8
%tmp16 = load i32, ptr undef
%tmp17 = select i1 false, i32 undef, i32 %tmp14
%tmp18 = select i1 false, i32 undef, i32 %tmp15
%tmp19 = select i1 false, i32 %tmp16, i32 undef
%tmp20 = select i1 undef, i32 undef, i32 %tmp17
%tmp21 = select i1 undef, i32 undef, i32 %tmp18
%tmp22 = select i1 undef, i32 undef, i32 %tmp19
%tmp23 = or i64 %tmp9, 3
%tmp24 = icmp sgt i32 undef, %tmp22
%tmp25 = trunc i64 %tmp23 to i32
%tmp26 = select i1 %tmp24, i32 %tmp25, i32 %tmp20
%tmp27 = select i1 %tmp24, i32 %tmp25, i32 %tmp21
%tmp28 = add nuw nsw i64 %tmp9, 4
%tmp29 = icmp eq i64 undef, 0
%scevgep = getelementptr [512 x i32], ptr %lsr.iv1, i64 0, i64 4
%0 = bitcast ptr %scevgep to ptr
br i1 %tmp29, label %bb30, label %bb7
bb30: ; preds = %bb7
%tmp32 = icmp sgt i32 %tmp27, -1000000
br i1 %tmp32, label %bb33, label %bb35
bb33: ; preds = %bb30
call void @fun()
store i32 %tmp26, ptr @globvar
%tmp34 = icmp ugt i32 undef, 1
br label %bb6
bb35: ; preds = %bb30
br i1 %tmp, label %bb37, label %bb38
bb37: ; preds = %bb35
unreachable
bb38: ; preds = %bb35
unreachable
}
; Function Attrs: nounwind
declare void @llvm.stackprotector(ptr, ptr) #1
attributes #0 = { "target-cpu"="z15" }
attributes #1 = { nounwind }
...
# CHECK: locr
# CHECK-NOT: risblg
---
name: fun1
alignment: 16
tracksRegLiveness: true
registers:
- { id: 0, class: grx32bit }
- { id: 1, class: addr64bit }
- { id: 2, class: grx32bit }
- { id: 3, class: addr64bit }
- { id: 4, class: gr32bit }
- { id: 5, class: grx32bit }
- { id: 6, class: gr64bit }
- { id: 7, class: gr64bit }
- { id: 8, class: grx32bit }
- { id: 9, class: grx32bit }
- { id: 10, class: gr64bit }
- { id: 11, class: grx32bit }
- { id: 12, class: gr64bit }
- { id: 13, class: grx32bit }
- { id: 14, class: gr32bit }
- { id: 15, class: gr32bit }
- { id: 16, class: grx32bit }
- { id: 17, class: grx32bit }
- { id: 18, class: gr32bit }
- { id: 19, class: addr64bit }
- { id: 20, class: grx32bit }
- { id: 21, class: gr32bit }
- { id: 22, class: gr64bit }
- { id: 23, class: grx32bit }
- { id: 24, class: grx32bit }
- { id: 25, class: grx32bit }
- { id: 26, class: addr64bit }
- { id: 27, class: grx32bit }
- { id: 28, class: addr64bit }
frameInfo:
adjustsStack: true
hasCalls: true
body: |
bb.0.bb5:
%25:grx32bit = IMPLICIT_DEF
bb.1.bb6:
%28:addr64bit = LGHI 0
%27:grx32bit = IIFMux 4293967296
%26:addr64bit = IMPLICIT_DEF
bb.2.bb7:
successors: %bb.3(0x04000000), %bb.2(0x7c000000)
%14:gr32bit = LMux %26, 0, $noreg :: (load (s32) from %ir.lsr.iv12)
CR %14, undef %15:gr32bit, implicit-def $cc
%16:grx32bit = COPY %28.subreg_l32
%16:grx32bit = LOCHIMux %16, 0, 14, 12, implicit $cc
%17:grx32bit = SELRMux %27, %28.subreg_l32, 14, 2, implicit killed $cc
%18:gr32bit = LMux undef %19:addr64bit, 0, $noreg :: (load (s32) from `ptr undef`)
%20:grx32bit = COPY %28.subreg_l32
%20:grx32bit = OILMux %20, 3, implicit-def dead $cc
CR undef %21:gr32bit, %18, implicit-def $cc
%4:gr32bit = SELRMux %16, %20, 14, 2, implicit $cc
%27:grx32bit = SELRMux %17, %20, 14, 2, implicit killed $cc
%28:addr64bit = nuw nsw LA %28, 4, $noreg
%26:addr64bit = LA %26, 16, $noreg
CGHI undef %22:gr64bit, 0, implicit-def $cc
BRC 14, 6, %bb.2, implicit killed $cc
J %bb.3
bb.3.bb30:
successors: %bb.4(0x7fffffff), %bb.5(0x00000001)
CFIMux %27, -999999, implicit-def $cc
BRC 14, 4, %bb.5, implicit killed $cc
J %bb.4
bb.4.bb33:
CallBRASL @fun, csr_systemz_elf, implicit-def dead $r14d, implicit-def dead $cc
STRL %4, @globvar :: (store (s32) into @globvar)
CLFIMux undef %23:grx32bit, 1, implicit-def $cc
%25:grx32bit = LHIMux 0
%25:grx32bit = LOCHIMux %25, 1, 14, 2, implicit killed $cc
J %bb.1
bb.5.bb35:
successors: %bb.6, %bb.7
TMLMux %25, 1, implicit-def $cc
BRC 15, 8, %bb.7, implicit killed $cc
J %bb.6
bb.6.bb37:
successors:
bb.7.bb38:
...