1 # RUN: llc -run-pass=prologepilog -reverse-csr-restore-seq -o - -mtriple=aarch64-- %s | FileCheck %s --check-prefixes=CHECK,BEFORELDSTOPT 2 # RUN: llc -start-before=prologepilog -stop-after=aarch64-ldst-opt -reverse-csr-restore-seq -o - -mtriple=aarch64-- %s | FileCheck %s --check-prefixes=CHECK,AFTERLDSTOPT 3 # 4 --- | 5 6 define void @foo() nounwind { entry: unreachable } 7 8 define void @bar() nounwind { entry: unreachable } 9 10 define void @baz() nounwind { entry: unreachable } 11 12 ... 13 --- 14 name: foo 15 # CHECK-LABEL: name: foo 16 tracksRegLiveness: true 17 body: | 18 bb.0: 19 $x19 = IMPLICIT_DEF 20 $x20 = IMPLICIT_DEF 21 $x21 = IMPLICIT_DEF 22 $x22 = IMPLICIT_DEF 23 $x23 = IMPLICIT_DEF 24 $x24 = IMPLICIT_DEF 25 $x25 = IMPLICIT_DEF 26 $x26 = IMPLICIT_DEF 27 28 ; The local stack size is 0, so the last ldp in the sequence will also 29 ; restore the stack. 30 ; CHECK: $x24, $x23 = frame-destroy LDPXi $sp, 2 31 ; CHECK-NEXT: $x22, $x21 = frame-destroy LDPXi $sp, 4 32 ; CHECK-NEXT: $x20, $x19 = frame-destroy LDPXi $sp, 6 33 34 ; Before running the load-store optimizer, we emit a ldp and an add. 35 ; BEFORELDSTOPT-NEXT: $x26, $x25 = frame-destroy LDPXi $sp, 0 36 ; BEFORELDSTOPT-NEXT: $sp = frame-destroy ADDXri $sp, 64, 0 37 38 ; We want to make sure that after running the load-store optimizer, the ldp 39 ; and the add get merged into a post-index ldp. 40 ; AFTERLDSTOPT-NEXT: early-clobber $sp, $x26, $x25 = frame-destroy LDPXpost $sp, 8 41 42 RET_ReallyLR 43 ... 44 --- 45 name: bar 46 # CHECK-LABEL: name: bar 47 tracksRegLiveness: true 48 stack: 49 - { id : 0, size: 8, alignment: 4, 50 stack-id: 0, callee-saved-register: '', callee-saved-restored: true, 51 local-offset: -4, debug-info-variable: '', debug-info-expression: '', 52 debug-info-location: '' } 53 54 body: | 55 bb.0: 56 $x19 = IMPLICIT_DEF 57 $x20 = IMPLICIT_DEF 58 $x21 = IMPLICIT_DEF 59 $x22 = IMPLICIT_DEF 60 $x23 = IMPLICIT_DEF 61 $x24 = IMPLICIT_DEF 62 $x25 = IMPLICIT_DEF 63 $x26 = IMPLICIT_DEF 64 65 ; The local stack size is not 0, and we can combine the CSR stack size with 66 ; the local stack size. This results in rewriting the offsets for all the 67 ; save/restores and forbids us to merge the stack adjustment and the last pop. 68 ; In this case, there is no point of moving the first CSR pair at the end. 69 ; CHECK: $x26, $x25 = frame-destroy LDPXi $sp, 2 70 ; CHECK-NEXT: $x24, $x23 = frame-destroy LDPXi $sp, 4 71 ; CHECK-NEXT: $x22, $x21 = frame-destroy LDPXi $sp, 6 72 ; CHECK-NEXT: $x20, $x19 = frame-destroy LDPXi $sp, 8 73 ; CHECK-NEXT: $sp = frame-destroy ADDXri $sp, 80, 0 74 RET_ReallyLR 75 ... 76 --- 77 # Check that the load from the offset 0 is moved at the end even when hasFP is 78 # false. 79 name: baz 80 # CHECK-LABEL: name: baz 81 alignment: 2 82 tracksRegLiveness: true 83 frameInfo: 84 adjustsStack: true 85 hasCalls: true 86 body: | 87 bb.0: 88 successors: %bb.1 89 90 $x0 = IMPLICIT_DEF 91 $x20 = IMPLICIT_DEF 92 $x21 = IMPLICIT_DEF 93 94 ADJCALLSTACKDOWN 0, 0, implicit-def dead $sp, implicit $sp 95 BL @foo, csr_aarch64_aapcs, implicit-def dead $lr, implicit $sp, implicit $x0, implicit-def $sp, implicit-def $x0 96 ADJCALLSTACKUP 0, 0, implicit-def dead $sp, implicit $sp 97 B %bb.1 98 99 bb.1: 100 ; CHECK: $x20, $lr = frame-destroy LDPXi $sp, 2 101 ; BEFORELDSTOPT-NEXT: $x21 = frame-destroy LDRXui $sp, 0 102 ; BEFORELDSTOPT-NEXT: $sp = frame-destroy ADDXri $sp, 32, 0 103 104 ; AFTERLDSTOPT-NEXT: early-clobber $sp, $x21 = frame-destroy LDRXpost $sp, 32 105 RET_ReallyLR 106 ... 107
