derek/TheRock
1
0 Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
Files
1a182e2dac58879f215c0fad0f5f78c76dcafe56
TheRock/compiler/amd-llvm/llvm/test/CodeGen/AArch64/ldst-opt.ll
T
derek.holloway 7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
Details
CodeQL / Analyze (python) (push) Canceled after 0s
Details
pre-commit / pre-commit (push) Canceled after 0s
Details
Multi-Arch CI / setup (push) Canceled after 0s
Details
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Details
Setup (Multi-Arch) / setup (push) Canceled after 0s
Details
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Details
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
Details
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Details
Multi-Arch CI / CI Summary (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Details
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Details
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
Details
Flatten subprojects into a single base monorepo
2026-09-02 16:34:47 -07:00

2270 lines
72 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
; RUN: llc -mtriple=aarch64-linux-gnu -aarch64-enable-atomic-cfg-tidy=0 -disable-lsr -verify-machineinstrs -o - %s | FileCheck --check-prefixes=CHECK,NOSTRICTALIGN %s
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+strict-align -aarch64-enable-atomic-cfg-tidy=0 -disable-lsr -verify-machineinstrs -o - %s | FileCheck --check-prefixes=CHECK,STRICTALIGN %s
; This file contains tests for the AArch64 load/store optimizer.
%padding = type { ptr, ptr, ptr, ptr }
%s.byte = type { i8, i8 }
%s.halfword = type { i16, i16 }
%s.word = type { i32, i32 }
%s.doubleword = type { i64, i32 }
%s.quadword = type { fp128, i32 }
%s.float = type { float, i32 }
%s.double = type { double, i32 }
%struct.byte = type { %padding, %s.byte }
%struct.halfword = type { %padding, %s.halfword }
%struct.word = type { %padding, %s.word }
%struct.doubleword = type { %padding, %s.doubleword }
%struct.quadword = type { %padding, %s.quadword }
%struct.float = type { %padding, %s.float }
%struct.double = type { %padding, %s.double }
; Check the following transform:
;
; (ldr|str) X, [x0, #32]
; ...
; add x0, x0, #32
; ->
; (ldr|str) X, [x0, #32]!
;
; with X being either w1, x1, s0, d0 or q0.
declare void @bar_byte(ptr, i8)
define void @load_pre_indexed_byte(ptr %ptr) nounwind {
; CHECK-LABEL: load_pre_indexed_byte:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldrb w1, [x0, #32]!
; CHECK-NEXT: b bar_byte
entry:
%a = getelementptr inbounds %struct.byte, ptr %ptr, i64 0, i32 1, i32 0
%add = load i8, ptr %a, align 4
br label %bar
bar:
%c = getelementptr inbounds %struct.byte, ptr %ptr, i64 0, i32 1
tail call void @bar_byte(ptr %c, i8 %add)
ret void
}
define void @store_pre_indexed_byte(ptr %ptr, i8 %val) nounwind {
; CHECK-LABEL: store_pre_indexed_byte:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: strb w1, [x0, #32]!
; CHECK-NEXT: b bar_byte
entry:
%a = getelementptr inbounds %struct.byte, ptr %ptr, i64 0, i32 1, i32 0
store i8 %val, ptr %a, align 4
br label %bar
bar:
%c = getelementptr inbounds %struct.byte, ptr %ptr, i64 0, i32 1
tail call void @bar_byte(ptr %c, i8 %val)
ret void
}
declare void @bar_halfword(ptr, i16)
define void @load_pre_indexed_halfword(ptr %ptr) nounwind {
; CHECK-LABEL: load_pre_indexed_halfword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldrh w1, [x0, #32]!
; CHECK-NEXT: b bar_halfword
entry:
%a = getelementptr inbounds %struct.halfword, ptr %ptr, i64 0, i32 1, i32 0
%add = load i16, ptr %a, align 4
br label %bar
bar:
%c = getelementptr inbounds %struct.halfword, ptr %ptr, i64 0, i32 1
tail call void @bar_halfword(ptr %c, i16 %add)
ret void
}
define void @store_pre_indexed_halfword(ptr %ptr, i16 %val) nounwind {
; CHECK-LABEL: store_pre_indexed_halfword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: strh w1, [x0, #32]!
; CHECK-NEXT: b bar_halfword
entry:
%a = getelementptr inbounds %struct.halfword, ptr %ptr, i64 0, i32 1, i32 0
store i16 %val, ptr %a, align 4
br label %bar
bar:
%c = getelementptr inbounds %struct.halfword, ptr %ptr, i64 0, i32 1
tail call void @bar_halfword(ptr %c, i16 %val)
ret void
}
declare void @bar_word(ptr, i32)
define void @load_pre_indexed_word(ptr %ptr) nounwind {
; CHECK-LABEL: load_pre_indexed_word:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr w1, [x0, #32]!
; CHECK-NEXT: b bar_word
entry:
%a = getelementptr inbounds %struct.word, ptr %ptr, i64 0, i32 1, i32 0
%add = load i32, ptr %a, align 4
br label %bar
bar:
%c = getelementptr inbounds %struct.word, ptr %ptr, i64 0, i32 1
tail call void @bar_word(ptr %c, i32 %add)
ret void
}
define void @store_pre_indexed_word(ptr %ptr, i32 %val) nounwind {
; CHECK-LABEL: store_pre_indexed_word:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str w1, [x0, #32]!
; CHECK-NEXT: b bar_word
entry:
%a = getelementptr inbounds %struct.word, ptr %ptr, i64 0, i32 1, i32 0
store i32 %val, ptr %a, align 4
br label %bar
bar:
%c = getelementptr inbounds %struct.word, ptr %ptr, i64 0, i32 1
tail call void @bar_word(ptr %c, i32 %val)
ret void
}
declare void @bar_doubleword(ptr, i64)
define void @load_pre_indexed_doubleword(ptr %ptr) nounwind {
; CHECK-LABEL: load_pre_indexed_doubleword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr x1, [x0, #32]!
; CHECK-NEXT: b bar_doubleword
entry:
%a = getelementptr inbounds %struct.doubleword, ptr %ptr, i64 0, i32 1, i32 0
%add = load i64, ptr %a, align 8
br label %bar
bar:
%c = getelementptr inbounds %struct.doubleword, ptr %ptr, i64 0, i32 1
tail call void @bar_doubleword(ptr %c, i64 %add)
ret void
}
define void @store_pre_indexed_doubleword(ptr %ptr, i64 %val) nounwind {
; CHECK-LABEL: store_pre_indexed_doubleword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x1, [x0, #32]!
; CHECK-NEXT: b bar_doubleword
entry:
%a = getelementptr inbounds %struct.doubleword, ptr %ptr, i64 0, i32 1, i32 0
store i64 %val, ptr %a, align 8
br label %bar
bar:
%c = getelementptr inbounds %struct.doubleword, ptr %ptr, i64 0, i32 1
tail call void @bar_doubleword(ptr %c, i64 %val)
ret void
}
declare void @bar_quadword(ptr, fp128)
define void @load_pre_indexed_quadword(ptr %ptr) nounwind {
; CHECK-LABEL: load_pre_indexed_quadword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr q0, [x0, #32]!
; CHECK-NEXT: b bar_quadword
entry:
%a = getelementptr inbounds %struct.quadword, ptr %ptr, i64 0, i32 1, i32 0
%add = load fp128, ptr %a, align 16
br label %bar
bar:
%c = getelementptr inbounds %struct.quadword, ptr %ptr, i64 0, i32 1
tail call void @bar_quadword(ptr %c, fp128 %add)
ret void
}
define void @store_pre_indexed_quadword(ptr %ptr, fp128 %val) nounwind {
; CHECK-LABEL: store_pre_indexed_quadword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str q0, [x0, #32]!
; CHECK-NEXT: b bar_quadword
entry:
%a = getelementptr inbounds %struct.quadword, ptr %ptr, i64 0, i32 1, i32 0
store fp128 %val, ptr %a, align 16
br label %bar
bar:
%c = getelementptr inbounds %struct.quadword, ptr %ptr, i64 0, i32 1
tail call void @bar_quadword(ptr %c, fp128 %val)
ret void
}
declare void @bar_float(ptr, float)
define void @load_pre_indexed_float(ptr %ptr) nounwind {
; CHECK-LABEL: load_pre_indexed_float:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr s0, [x0, #32]!
; CHECK-NEXT: b bar_float
entry:
%a = getelementptr inbounds %struct.float, ptr %ptr, i64 0, i32 1, i32 0
%add = load float, ptr %a, align 4
br label %bar
bar:
%c = getelementptr inbounds %struct.float, ptr %ptr, i64 0, i32 1
tail call void @bar_float(ptr %c, float %add)
ret void
}
define void @store_pre_indexed_float(ptr %ptr, float %val) nounwind {
; CHECK-LABEL: store_pre_indexed_float:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str s0, [x0, #32]!
; CHECK-NEXT: b bar_float
entry:
%a = getelementptr inbounds %struct.float, ptr %ptr, i64 0, i32 1, i32 0
store float %val, ptr %a, align 4
br label %bar
bar:
%c = getelementptr inbounds %struct.float, ptr %ptr, i64 0, i32 1
tail call void @bar_float(ptr %c, float %val)
ret void
}
declare void @bar_double(ptr, double)
define void @load_pre_indexed_double(ptr %ptr) nounwind {
; CHECK-LABEL: load_pre_indexed_double:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr d0, [x0, #32]!
; CHECK-NEXT: b bar_double
entry:
%a = getelementptr inbounds %struct.double, ptr %ptr, i64 0, i32 1, i32 0
%add = load double, ptr %a, align 8
br label %bar
bar:
%c = getelementptr inbounds %struct.double, ptr %ptr, i64 0, i32 1
tail call void @bar_double(ptr %c, double %add)
ret void
}
define void @store_pre_indexed_double(ptr %ptr, double %val) nounwind {
; CHECK-LABEL: store_pre_indexed_double:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str d0, [x0, #32]!
; CHECK-NEXT: b bar_double
entry:
%a = getelementptr inbounds %struct.double, ptr %ptr, i64 0, i32 1, i32 0
store double %val, ptr %a, align 8
br label %bar
bar:
%c = getelementptr inbounds %struct.double, ptr %ptr, i64 0, i32 1
tail call void @bar_double(ptr %c, double %val)
ret void
}
; Check the following transform:
;
; (ldp|stp) w1, w2 [x0, #32]
; ...
; add x0, x0, #32
; ->
; (ldp|stp) w1, w2, [x0, #32]!
;
define void @load_pair_pre_indexed_word(ptr %ptr) nounwind {
; CHECK-LABEL: load_pair_pre_indexed_word:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldp w8, w9, [x0, #32]!
; CHECK-NEXT: add w1, w8, w9
; CHECK-NEXT: b bar_word
entry:
%a = getelementptr inbounds %struct.word, ptr %ptr, i64 0, i32 1, i32 0
%a1 = load i32, ptr %a, align 4
%b = getelementptr inbounds %struct.word, ptr %ptr, i64 0, i32 1, i32 1
%b1 = load i32, ptr %b, align 4
%add = add i32 %a1, %b1
br label %bar
bar:
%c = getelementptr inbounds %struct.word, ptr %ptr, i64 0, i32 1
tail call void @bar_word(ptr %c, i32 %add)
ret void
}
define void @store_pair_pre_indexed_word(ptr %ptr, i32 %val) nounwind {
; CHECK-LABEL: store_pair_pre_indexed_word:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: stp w1, w1, [x0, #32]!
; CHECK-NEXT: b bar_word
entry:
%a = getelementptr inbounds %struct.word, ptr %ptr, i64 0, i32 1, i32 0
store i32 %val, ptr %a, align 4
%b = getelementptr inbounds %struct.word, ptr %ptr, i64 0, i32 1, i32 1
store i32 %val, ptr %b, align 4
br label %bar
bar:
%c = getelementptr inbounds %struct.word, ptr %ptr, i64 0, i32 1
tail call void @bar_word(ptr %c, i32 %val)
ret void
}
; Check the following transform:
;
; add x8, x8, #16
; ...
; ldr X, [x8]
; ->
; ldr X, [x8, #16]!
;
; with X being either w0, x0, s0, d0 or q0.
%pre.struct.i32 = type { i32, i32, i32, i32, i32}
%pre.struct.i64 = type { i32, i64, i64, i64, i64}
%pre.struct.i128 = type { i32, <2 x i64>, <2 x i64>, <2 x i64>}
%pre.struct.float = type { i32, float, float, float}
%pre.struct.double = type { i32, double, double, double}
define i32 @load_pre_indexed_word2(ptr %this, i1 %cond, ptr %load2) nounwind {
; CHECK-LABEL: load_pre_indexed_word2:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB16_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: ldr w0, [x8, #4]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB16_2: // %if.end
; CHECK-NEXT: add x8, x2, #8
; CHECK-NEXT: ldr w0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i32, ptr %load1, i64 0, i32 1
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i32, ptr %load2, i64 0, i32 2
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
%ret = load i32, ptr %retptr
ret i32 %ret
}
define i64 @load_pre_indexed_doubleword2(ptr %this, i1 %cond, ptr %load2) nounwind {
; CHECK-LABEL: load_pre_indexed_doubleword2:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB17_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: ldr x0, [x8, #8]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB17_2: // %if.end
; CHECK-NEXT: add x8, x2, #16
; CHECK-NEXT: ldr x0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i64, ptr %load1, i64 0, i32 1
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i64, ptr %load2, i64 0, i32 2
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
%ret = load i64, ptr %retptr
ret i64 %ret
}
define <2 x i64> @load_pre_indexed_quadword2(ptr %this, i1 %cond, ptr %load2) nounwind {
; CHECK-LABEL: load_pre_indexed_quadword2:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB18_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: ldr q0, [x8, #16]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB18_2: // %if.end
; CHECK-NEXT: add x8, x2, #32
; CHECK-NEXT: ldr q0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i128, ptr %load1, i64 0, i32 1
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i128, ptr %load2, i64 0, i32 2
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
%ret = load <2 x i64>, ptr %retptr
ret <2 x i64> %ret
}
define float @load_pre_indexed_float2(ptr %this, i1 %cond, ptr %load2) nounwind {
; CHECK-LABEL: load_pre_indexed_float2:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB19_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: ldr s0, [x8, #4]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB19_2: // %if.end
; CHECK-NEXT: add x8, x2, #8
; CHECK-NEXT: ldr s0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.float, ptr %load1, i64 0, i32 1
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.float, ptr %load2, i64 0, i32 2
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
%ret = load float, ptr %retptr
ret float %ret
}
define double @load_pre_indexed_double2(ptr %this, i1 %cond, ptr %load2) nounwind {
; CHECK-LABEL: load_pre_indexed_double2:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB20_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: ldr d0, [x8, #8]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB20_2: // %if.end
; CHECK-NEXT: add x8, x2, #16
; CHECK-NEXT: ldr d0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.double, ptr %load1, i64 0, i32 1
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.double, ptr %load2, i64 0, i32 2
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
%ret = load double, ptr %retptr
ret double %ret
}
define i32 @load_pre_indexed_word3(ptr %this, i1 %cond, ptr %load2) nounwind {
; CHECK-LABEL: load_pre_indexed_word3:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB21_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: ldr w0, [x8, #12]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB21_2: // %if.end
; CHECK-NEXT: add x8, x2, #16
; CHECK-NEXT: ldr w0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i32, ptr %load1, i64 0, i32 3
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i32, ptr %load2, i64 0, i32 4
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
%ret = load i32, ptr %retptr
ret i32 %ret
}
define i64 @load_pre_indexed_doubleword3(ptr %this, i1 %cond, ptr %load2) nounwind {
; CHECK-LABEL: load_pre_indexed_doubleword3:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB22_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: ldr x0, [x8, #16]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB22_2: // %if.end
; CHECK-NEXT: add x8, x2, #24
; CHECK-NEXT: ldr x0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i64, ptr %load1, i64 0, i32 2
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i64, ptr %load2, i64 0, i32 3
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
%ret = load i64, ptr %retptr
ret i64 %ret
}
define <2 x i64> @load_pre_indexed_quadword3(ptr %this, i1 %cond, ptr %load2) nounwind {
; CHECK-LABEL: load_pre_indexed_quadword3:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB23_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: ldr q0, [x8, #32]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB23_2: // %if.end
; CHECK-NEXT: add x8, x2, #48
; CHECK-NEXT: ldr q0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i128, ptr %load1, i64 0, i32 2
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i128, ptr %load2, i64 0, i32 3
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
%ret = load <2 x i64>, ptr %retptr
ret <2 x i64> %ret
}
define float @load_pre_indexed_float3(ptr %this, i1 %cond, ptr %load2) nounwind {
; CHECK-LABEL: load_pre_indexed_float3:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB24_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: ldr s0, [x8, #8]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB24_2: // %if.end
; CHECK-NEXT: add x8, x2, #12
; CHECK-NEXT: ldr s0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.float, ptr %load1, i64 0, i32 2
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.float, ptr %load2, i64 0, i32 3
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
%ret = load float, ptr %retptr
ret float %ret
}
define double @load_pre_indexed_double3(ptr %this, i1 %cond, ptr %load2) nounwind {
; CHECK-LABEL: load_pre_indexed_double3:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB25_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: ldr d0, [x8, #16]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB25_2: // %if.end
; CHECK-NEXT: add x8, x2, #24
; CHECK-NEXT: ldr d0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.double, ptr %load1, i64 0, i32 2
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.double, ptr %load2, i64 0, i32 3
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
%ret = load double, ptr %retptr
ret double %ret
}
; Check the following transform:
;
; add x8, x8, #16
; ...
; str X, [x8]
; ->
; str X, [x8, #16]!
;
; with X being either w0, x0, s0, d0 or q0.
define void @store_pre_indexed_word2(ptr %this, i1 %cond, ptr %load2, i32 %val) nounwind {
; CHECK-LABEL: store_pre_indexed_word2:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB26_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: str w3, [x8, #4]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB26_2: // %if.end
; CHECK-NEXT: add x8, x2, #8
; CHECK-NEXT: str w3, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i32, ptr %load1, i64 0, i32 1
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i32, ptr %load2, i64 0, i32 2
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
store i32 %val, ptr %retptr
ret void
}
define void @store_pre_indexed_doubleword2(ptr %this, i1 %cond, ptr %load2, i64 %val) nounwind {
; CHECK-LABEL: store_pre_indexed_doubleword2:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB27_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: str x3, [x8, #8]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB27_2: // %if.end
; CHECK-NEXT: add x8, x2, #16
; CHECK-NEXT: str x3, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i64, ptr %load1, i64 0, i32 1
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i64, ptr %load2, i64 0, i32 2
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
store i64 %val, ptr %retptr
ret void
}
define void @store_pre_indexed_quadword2(ptr %this, i1 %cond, ptr %load2, <2 x i64> %val) nounwind {
; CHECK-LABEL: store_pre_indexed_quadword2:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB28_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: str q0, [x8, #16]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB28_2: // %if.end
; CHECK-NEXT: add x8, x2, #32
; CHECK-NEXT: str q0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i128, ptr %load1, i64 0, i32 1
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i128, ptr %load2, i64 0, i32 2
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
store <2 x i64> %val, ptr %retptr
ret void
}
define void @store_pre_indexed_float2(ptr %this, i1 %cond, ptr %load2, float %val) nounwind {
; CHECK-LABEL: store_pre_indexed_float2:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB29_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: str s0, [x8, #4]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB29_2: // %if.end
; CHECK-NEXT: add x8, x2, #8
; CHECK-NEXT: str s0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.float, ptr %load1, i64 0, i32 1
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.float, ptr %load2, i64 0, i32 2
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
store float %val, ptr %retptr
ret void
}
define void @store_pre_indexed_double2(ptr %this, i1 %cond, ptr %load2, double %val) nounwind {
; CHECK-LABEL: store_pre_indexed_double2:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB30_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: str d0, [x8, #8]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB30_2: // %if.end
; CHECK-NEXT: add x8, x2, #16
; CHECK-NEXT: str d0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.double, ptr %load1, i64 0, i32 1
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.double, ptr %load2, i64 0, i32 2
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
store double %val, ptr %retptr
ret void
}
define void @store_pre_indexed_word3(ptr %this, i1 %cond, ptr %load2, i32 %val) nounwind {
; CHECK-LABEL: store_pre_indexed_word3:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB31_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: str w3, [x8, #12]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB31_2: // %if.end
; CHECK-NEXT: add x8, x2, #16
; CHECK-NEXT: str w3, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i32, ptr %load1, i64 0, i32 3
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i32, ptr %load2, i64 0, i32 4
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
store i32 %val, ptr %retptr
ret void
}
define void @store_pre_indexed_doubleword3(ptr %this, i1 %cond, ptr %load2, i64 %val) nounwind {
; CHECK-LABEL: store_pre_indexed_doubleword3:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB32_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: str x3, [x8, #24]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB32_2: // %if.end
; CHECK-NEXT: add x8, x2, #32
; CHECK-NEXT: str x3, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i64, ptr %load1, i64 0, i32 3
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i64, ptr %load2, i64 0, i32 4
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
store i64 %val, ptr %retptr
ret void
}
define void @store_pre_indexed_quadword3(ptr %this, i1 %cond, ptr %load2, <2 x i64> %val) nounwind {
; CHECK-LABEL: store_pre_indexed_quadword3:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB33_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: str q0, [x8, #32]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB33_2: // %if.end
; CHECK-NEXT: add x8, x2, #48
; CHECK-NEXT: str q0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.i128, ptr %load1, i64 0, i32 2
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.i128, ptr %load2, i64 0, i32 3
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
store <2 x i64> %val, ptr %retptr
ret void
}
define void @store_pre_indexed_float3(ptr %this, i1 %cond, ptr %load2, float %val) nounwind {
; CHECK-LABEL: store_pre_indexed_float3:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB34_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: str s0, [x8, #8]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB34_2: // %if.end
; CHECK-NEXT: add x8, x2, #12
; CHECK-NEXT: str s0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.float, ptr %load1, i64 0, i32 2
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.float, ptr %load2, i64 0, i32 3
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
store float %val, ptr %retptr
ret void
}
define void @store_pre_indexed_double3(ptr %this, i1 %cond, ptr %load2, double %val) nounwind {
; CHECK-LABEL: store_pre_indexed_double3:
; CHECK: // %bb.0:
; CHECK-NEXT: tbz w1, #0, .LBB35_2
; CHECK-NEXT: // %bb.1: // %if.then
; CHECK-NEXT: ldr x8, [x0]
; CHECK-NEXT: str d0, [x8, #16]!
; CHECK-NEXT: ret
; CHECK-NEXT: .LBB35_2: // %if.end
; CHECK-NEXT: add x8, x2, #24
; CHECK-NEXT: str d0, [x8]
; CHECK-NEXT: ret
br i1 %cond, label %if.then, label %if.end
if.then:
%load1 = load ptr, ptr %this
%gep1 = getelementptr inbounds %pre.struct.double, ptr %load1, i64 0, i32 2
br label %return
if.end:
%gep2 = getelementptr inbounds %pre.struct.double, ptr %load2, i64 0, i32 3
br label %return
return:
%retptr = phi ptr [ %gep1, %if.then ], [ %gep2, %if.end ]
store double %val, ptr %retptr
ret void
}
; Check the following transform:
;
; ldr X, [x20]
; ...
; add x20, x20, #32
; ->
; ldr X, [x20], #32
;
; with X being either w0, x0, s0, d0 or q0.
define void @load_post_indexed_byte(ptr %array, i64 %count) nounwind {
; CHECK-LABEL: load_post_indexed_byte:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: add x20, x0, #2
; CHECK-NEXT: .LBB36_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldurb w0, [x20, #-1]
; CHECK-NEXT: bl "use-byte"
; CHECK-NEXT: ldrb w0, [x20], #4
; CHECK-NEXT: bl "use-byte"
; CHECK-NEXT: subs x19, x19, #4
; CHECK-NEXT: b.ne .LBB36_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr i8, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr i8, ptr %iv2, i64 -1
%load = load i8, ptr %gep2
call void @use-byte(i8 %load)
%load2 = load i8, ptr %iv2
call void @use-byte(i8 %load2)
%iv.next = add i64 %iv, -4
%gep3 = getelementptr i8, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @load_post_indexed_halfword(ptr %array, i64 %count) nounwind {
; CHECK-LABEL: load_post_indexed_halfword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: add x20, x0, #4
; CHECK-NEXT: .LBB37_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldurh w0, [x20, #-2]
; CHECK-NEXT: bl "use-halfword"
; CHECK-NEXT: ldrh w0, [x20], #8
; CHECK-NEXT: bl "use-halfword"
; CHECK-NEXT: subs x19, x19, #4
; CHECK-NEXT: b.ne .LBB37_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr i16, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr i16, ptr %iv2, i64 -1
%load = load i16, ptr %gep2
call void @use-halfword(i16 %load)
%load2 = load i16, ptr %iv2
call void @use-halfword(i16 %load2)
%iv.next = add i64 %iv, -4
%gep3 = getelementptr i16, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @load_post_indexed_word(ptr %array, i64 %count) nounwind {
; CHECK-LABEL: load_post_indexed_word:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: add x20, x0, #8
; CHECK-NEXT: .LBB38_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur w0, [x20, #-4]
; CHECK-NEXT: bl "use-word"
; CHECK-NEXT: ldr w0, [x20], #16
; CHECK-NEXT: bl "use-word"
; CHECK-NEXT: subs x19, x19, #4
; CHECK-NEXT: b.ne .LBB38_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr i32, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr i32, ptr %iv2, i64 -1
%load = load i32, ptr %gep2
call void @use-word(i32 %load)
%load2 = load i32, ptr %iv2
call void @use-word(i32 %load2)
%iv.next = add i64 %iv, -4
%gep3 = getelementptr i32, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @load_post_indexed_doubleword(ptr %array, i64 %count) nounwind {
; CHECK-LABEL: load_post_indexed_doubleword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: add x20, x0, #16
; CHECK-NEXT: .LBB39_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur x0, [x20, #-8]
; CHECK-NEXT: bl "use-doubleword"
; CHECK-NEXT: ldr x0, [x20], #32
; CHECK-NEXT: bl "use-doubleword"
; CHECK-NEXT: subs x19, x19, #4
; CHECK-NEXT: b.ne .LBB39_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr i64, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr i64, ptr %iv2, i64 -1
%load = load i64, ptr %gep2
call void @use-doubleword(i64 %load)
%load2 = load i64, ptr %iv2
call void @use-doubleword(i64 %load2)
%iv.next = add i64 %iv, -4
%gep3 = getelementptr i64, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @load_post_indexed_quadword(ptr %array, i64 %count) nounwind {
; CHECK-LABEL: load_post_indexed_quadword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: add x20, x0, #32
; CHECK-NEXT: .LBB40_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur q0, [x20, #-16]
; CHECK-NEXT: bl "use-quadword"
; CHECK-NEXT: ldr q0, [x20], #64
; CHECK-NEXT: bl "use-quadword"
; CHECK-NEXT: subs x19, x19, #4
; CHECK-NEXT: b.ne .LBB40_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr <2 x i64>, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr <2 x i64>, ptr %iv2, i64 -1
%load = load <2 x i64>, ptr %gep2
call void @use-quadword(<2 x i64> %load)
%load2 = load <2 x i64>, ptr %iv2
call void @use-quadword(<2 x i64> %load2)
%iv.next = add i64 %iv, -4
%gep3 = getelementptr <2 x i64>, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @load_post_indexed_float(ptr %array, i64 %count) nounwind {
; CHECK-LABEL: load_post_indexed_float:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: add x20, x0, #8
; CHECK-NEXT: .LBB41_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur s0, [x20, #-4]
; CHECK-NEXT: bl "use-float"
; CHECK-NEXT: ldr s0, [x20], #16
; CHECK-NEXT: bl "use-float"
; CHECK-NEXT: subs x19, x19, #4
; CHECK-NEXT: b.ne .LBB41_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr float, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr float, ptr %iv2, i64 -1
%load = load float, ptr %gep2
call void @use-float(float %load)
%load2 = load float, ptr %iv2
call void @use-float(float %load2)
%iv.next = add i64 %iv, -4
%gep3 = getelementptr float, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @load_post_indexed_double(ptr %array, i64 %count) nounwind {
; CHECK-LABEL: load_post_indexed_double:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: add x20, x0, #16
; CHECK-NEXT: .LBB42_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur d0, [x20, #-8]
; CHECK-NEXT: bl "use-double"
; CHECK-NEXT: ldr d0, [x20], #32
; CHECK-NEXT: bl "use-double"
; CHECK-NEXT: subs x19, x19, #4
; CHECK-NEXT: b.ne .LBB42_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr double, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr double, ptr %iv2, i64 -1
%load = load double, ptr %gep2
call void @use-double(double %load)
%load2 = load double, ptr %iv2
call void @use-double(double %load2)
%iv.next = add i64 %iv, -4
%gep3 = getelementptr double, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
; Check the following transform:
;
; str X, [x20]
; ...
; add x20, x20, #32
; ->
; str X, [x20], #32
;
; with X being either w0, x0, s0, d0 or q0.
define void @store_post_indexed_byte(ptr %array, i64 %count, i8 %val) nounwind {
; CHECK-LABEL: store_post_indexed_byte:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: stp x30, x21, [sp, #-32]! // 16-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov w19, w2
; CHECK-NEXT: mov x20, x1
; CHECK-NEXT: add x21, x0, #2
; CHECK-NEXT: .LBB43_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldurb w0, [x21, #-1]
; CHECK-NEXT: bl "use-byte"
; CHECK-NEXT: subs x20, x20, #4
; CHECK-NEXT: strb w19, [x21], #4
; CHECK-NEXT: b.ne .LBB43_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldp x30, x21, [sp], #32 // 16-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr i8, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr i8, ptr %iv2, i64 -1
%load = load i8, ptr %gep2
call void @use-byte(i8 %load)
store i8 %val, ptr %iv2
%iv.next = add i64 %iv, -4
%gep3 = getelementptr i8, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @store_post_indexed_halfword(ptr %array, i64 %count, i16 %val) nounwind {
; CHECK-LABEL: store_post_indexed_halfword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: stp x30, x21, [sp, #-32]! // 16-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov w19, w2
; CHECK-NEXT: mov x20, x1
; CHECK-NEXT: add x21, x0, #4
; CHECK-NEXT: .LBB44_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldurh w0, [x21, #-2]
; CHECK-NEXT: bl "use-halfword"
; CHECK-NEXT: subs x20, x20, #4
; CHECK-NEXT: strh w19, [x21], #8
; CHECK-NEXT: b.ne .LBB44_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldp x30, x21, [sp], #32 // 16-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr i16, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr i16, ptr %iv2, i64 -1
%load = load i16, ptr %gep2
call void @use-halfword(i16 %load)
store i16 %val, ptr %iv2
%iv.next = add i64 %iv, -4
%gep3 = getelementptr i16, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @store_post_indexed_word(ptr %array, i64 %count, i32 %val) nounwind {
; CHECK-LABEL: store_post_indexed_word:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: stp x30, x21, [sp, #-32]! // 16-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov w19, w2
; CHECK-NEXT: mov x20, x1
; CHECK-NEXT: add x21, x0, #8
; CHECK-NEXT: .LBB45_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur w0, [x21, #-4]
; CHECK-NEXT: bl "use-word"
; CHECK-NEXT: subs x20, x20, #4
; CHECK-NEXT: str w19, [x21], #16
; CHECK-NEXT: b.ne .LBB45_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldp x30, x21, [sp], #32 // 16-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr i32, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr i32, ptr %iv2, i64 -1
%load = load i32, ptr %gep2
call void @use-word(i32 %load)
store i32 %val, ptr %iv2
%iv.next = add i64 %iv, -4
%gep3 = getelementptr i32, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @store_post_indexed_doubleword(ptr %array, i64 %count, i64 %val) nounwind {
; CHECK-LABEL: store_post_indexed_doubleword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: stp x30, x21, [sp, #-32]! // 16-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov x19, x2
; CHECK-NEXT: mov x20, x1
; CHECK-NEXT: add x21, x0, #16
; CHECK-NEXT: .LBB46_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur x0, [x21, #-8]
; CHECK-NEXT: bl "use-doubleword"
; CHECK-NEXT: subs x20, x20, #4
; CHECK-NEXT: str x19, [x21], #32
; CHECK-NEXT: b.ne .LBB46_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldp x30, x21, [sp], #32 // 16-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr i64, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr i64, ptr %iv2, i64 -1
%load = load i64, ptr %gep2
call void @use-doubleword(i64 %load)
store i64 %val, ptr %iv2
%iv.next = add i64 %iv, -4
%gep3 = getelementptr i64, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @store_post_indexed_quadword(ptr %array, i64 %count, <2 x i64> %val) nounwind {
; CHECK-LABEL: store_post_indexed_quadword:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: sub sp, sp, #48
; CHECK-NEXT: stp x20, x19, [sp, #32] // 16-byte Folded Spill
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: add x20, x0, #32
; CHECK-NEXT: str x30, [sp, #16] // 8-byte Spill
; CHECK-NEXT: str q0, [sp] // 16-byte Spill
; CHECK-NEXT: .LBB47_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur q0, [x20, #-16]
; CHECK-NEXT: bl "use-quadword"
; CHECK-NEXT: ldr q0, [sp] // 16-byte Reload
; CHECK-NEXT: subs x19, x19, #4
; CHECK-NEXT: str q0, [x20], #64
; CHECK-NEXT: b.ne .LBB47_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #32] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp, #16] // 8-byte Reload
; CHECK-NEXT: add sp, sp, #48
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr <2 x i64>, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr <2 x i64>, ptr %iv2, i64 -1
%load = load <2 x i64>, ptr %gep2
call void @use-quadword(<2 x i64> %load)
store <2 x i64> %val, ptr %iv2
%iv.next = add i64 %iv, -4
%gep3 = getelementptr <2 x i64>, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @store_post_indexed_float(ptr %array, i64 %count, float %val) nounwind {
; CHECK-LABEL: store_post_indexed_float:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str d8, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: fmov s8, s0
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: add x20, x0, #8
; CHECK-NEXT: str x30, [sp, #8] // 8-byte Spill
; CHECK-NEXT: .LBB48_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur s0, [x20, #-4]
; CHECK-NEXT: bl "use-float"
; CHECK-NEXT: subs x19, x19, #4
; CHECK-NEXT: str s8, [x20], #16
; CHECK-NEXT: b.ne .LBB48_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp, #8] // 8-byte Reload
; CHECK-NEXT: ldr d8, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr float, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr float, ptr %iv2, i64 -1
%load = load float, ptr %gep2
call void @use-float(float %load)
store float %val, ptr %iv2
%iv.next = add i64 %iv, -4
%gep3 = getelementptr float, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
define void @store_post_indexed_double(ptr %array, i64 %count, double %val) nounwind {
; CHECK-LABEL: store_post_indexed_double:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str d8, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: fmov d8, d0
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: add x20, x0, #16
; CHECK-NEXT: str x30, [sp, #8] // 8-byte Spill
; CHECK-NEXT: .LBB49_1: // %body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur d0, [x20, #-8]
; CHECK-NEXT: bl "use-double"
; CHECK-NEXT: subs x19, x19, #4
; CHECK-NEXT: str d8, [x20], #32
; CHECK-NEXT: b.ne .LBB49_1
; CHECK-NEXT: // %bb.2: // %exit
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp, #8] // 8-byte Reload
; CHECK-NEXT: ldr d8, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%gep1 = getelementptr double, ptr %array, i64 2
br label %body
body:
%iv2 = phi ptr [ %gep3, %body ], [ %gep1, %entry ]
%iv = phi i64 [ %iv.next, %body ], [ %count, %entry ]
%gep2 = getelementptr double, ptr %iv2, i64 -1
%load = load double, ptr %gep2
call void @use-double(double %load)
store double %val, ptr %iv2
%iv.next = add i64 %iv, -4
%gep3 = getelementptr double, ptr %iv2, i64 4
%cond = icmp eq i64 %iv.next, 0
br i1 %cond, label %exit, label %body
exit:
ret void
}
declare void @use-byte(i8)
declare void @use-halfword(i16)
declare void @use-word(i32)
declare void @use-doubleword(i64)
declare void @use-quadword(<2 x i64>)
declare void @use-float(float)
declare void @use-double(double)
; Check the following transform:
;
; stp w0, [x20]
; ...
; add x20, x20, #32
; ->
; stp w0, [x20], #32
define void @store_pair_post_indexed_word() nounwind {
; CHECK-LABEL: store_pair_post_indexed_word:
; CHECK: // %bb.0:
; CHECK-NEXT: sub sp, sp, #16
; CHECK-NEXT: ldp w8, w9, [sp, #8]
; CHECK-NEXT: stp w8, w9, [sp], #16
; CHECK-NEXT: ret
%src = alloca { i32, i32 }, align 8
%dst = alloca { i32, i32 }, align 8
%src.realp = getelementptr inbounds { i32, i32 }, ptr %src, i32 0, i32 0
%src.real = load i32, ptr %src.realp
%src.imagp = getelementptr inbounds { i32, i32 }, ptr %src, i32 0, i32 1
%src.imag = load i32, ptr %src.imagp
%dst.realp = getelementptr inbounds { i32, i32 }, ptr %dst, i32 0, i32 0
%dst.imagp = getelementptr inbounds { i32, i32 }, ptr %dst, i32 0, i32 1
store i32 %src.real, ptr %dst.realp
store i32 %src.imag, ptr %dst.imagp
ret void
}
define void @store_pair_post_indexed_doubleword() nounwind {
; CHECK-LABEL: store_pair_post_indexed_doubleword:
; CHECK: // %bb.0:
; CHECK-NEXT: sub sp, sp, #32
; CHECK-NEXT: ldp x8, x9, [sp, #16]
; CHECK-NEXT: stp x8, x9, [sp], #32
; CHECK-NEXT: ret
%src = alloca { i64, i64 }, align 8
%dst = alloca { i64, i64 }, align 8
%src.realp = getelementptr inbounds { i64, i64 }, ptr %src, i32 0, i32 0
%src.real = load i64, ptr %src.realp
%src.imagp = getelementptr inbounds { i64, i64 }, ptr %src, i32 0, i32 1
%src.imag = load i64, ptr %src.imagp
%dst.realp = getelementptr inbounds { i64, i64 }, ptr %dst, i32 0, i32 0
%dst.imagp = getelementptr inbounds { i64, i64 }, ptr %dst, i32 0, i32 1
store i64 %src.real, ptr %dst.realp
store i64 %src.imag, ptr %dst.imagp
ret void
}
define void @store_pair_post_indexed_float() nounwind {
; CHECK-LABEL: store_pair_post_indexed_float:
; CHECK: // %bb.0:
; CHECK-NEXT: sub sp, sp, #16
; CHECK-NEXT: ldp s0, s1, [sp, #8]
; CHECK-NEXT: stp s0, s1, [sp], #16
; CHECK-NEXT: ret
%src = alloca { float, float }, align 8
%dst = alloca { float, float }, align 8
%src.realp = getelementptr inbounds { float, float }, ptr %src, i32 0, i32 0
%src.real = load float, ptr %src.realp
%src.imagp = getelementptr inbounds { float, float }, ptr %src, i32 0, i32 1
%src.imag = load float, ptr %src.imagp
%dst.realp = getelementptr inbounds { float, float }, ptr %dst, i32 0, i32 0
%dst.imagp = getelementptr inbounds { float, float }, ptr %dst, i32 0, i32 1
store float %src.real, ptr %dst.realp
store float %src.imag, ptr %dst.imagp
ret void
}
define void @store_pair_post_indexed_double() nounwind {
; CHECK-LABEL: store_pair_post_indexed_double:
; CHECK: // %bb.0:
; CHECK-NEXT: sub sp, sp, #32
; CHECK-NEXT: ldp d0, d1, [sp, #16]
; CHECK-NEXT: stp d0, d1, [sp], #32
; CHECK-NEXT: ret
%src = alloca { double, double }, align 8
%dst = alloca { double, double }, align 8
%src.realp = getelementptr inbounds { double, double }, ptr %src, i32 0, i32 0
%src.real = load double, ptr %src.realp
%src.imagp = getelementptr inbounds { double, double }, ptr %src, i32 0, i32 1
%src.imag = load double, ptr %src.imagp
%dst.realp = getelementptr inbounds { double, double }, ptr %dst, i32 0, i32 0
%dst.imagp = getelementptr inbounds { double, double }, ptr %dst, i32 0, i32 1
store double %src.real, ptr %dst.realp
store double %src.imag, ptr %dst.imagp
ret void
}
; Check the following transform:
;
; (ldr|str) X, [x20]
; ...
; sub x20, x20, #16
; ->
; (ldr|str) X, [x20], #-16
;
; with X being either w0, x0, s0, d0 or q0.
define void @post_indexed_sub_word(ptr %a, ptr %b, i64 %count) nounwind {
; CHECK-LABEL: post_indexed_sub_word:
; CHECK: // %bb.0:
; CHECK-NEXT: .LBB54_1: // %for.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur w8, [x1, #-4]
; CHECK-NEXT: subs x2, x2, #1
; CHECK-NEXT: stur w8, [x0, #-4]
; CHECK-NEXT: ldr w8, [x1], #-8
; CHECK-NEXT: str w8, [x0], #-8
; CHECK-NEXT: b.gt .LBB54_1
; CHECK-NEXT: // %bb.2: // %end
; CHECK-NEXT: ret
br label %for.body
for.body:
%phi1 = phi ptr [ %gep4, %for.body ], [ %b, %0 ]
%phi2 = phi ptr [ %gep3, %for.body ], [ %a, %0 ]
%i = phi i64 [ %dec.i, %for.body], [ %count, %0 ]
%gep1 = getelementptr i32, ptr %phi1, i64 -1
%load1 = load i32, ptr %gep1
%gep2 = getelementptr i32, ptr %phi2, i64 -1
store i32 %load1, ptr %gep2
%load2 = load i32, ptr %phi1
store i32 %load2, ptr %phi2
%dec.i = add nsw i64 %i, -1
%gep3 = getelementptr i32, ptr %phi2, i64 -2
%gep4 = getelementptr i32, ptr %phi1, i64 -2
%cond = icmp sgt i64 %dec.i, 0
br i1 %cond, label %for.body, label %end
end:
ret void
}
define void @post_indexed_sub_doubleword(ptr %a, ptr %b, i64 %count) nounwind {
; CHECK-LABEL: post_indexed_sub_doubleword:
; CHECK: // %bb.0:
; CHECK-NEXT: .LBB55_1: // %for.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur x8, [x1, #-8]
; CHECK-NEXT: subs x2, x2, #1
; CHECK-NEXT: stur x8, [x0, #-8]
; CHECK-NEXT: ldr x8, [x1], #-16
; CHECK-NEXT: str x8, [x0], #-16
; CHECK-NEXT: b.gt .LBB55_1
; CHECK-NEXT: // %bb.2: // %end
; CHECK-NEXT: ret
br label %for.body
for.body:
%phi1 = phi ptr [ %gep4, %for.body ], [ %b, %0 ]
%phi2 = phi ptr [ %gep3, %for.body ], [ %a, %0 ]
%i = phi i64 [ %dec.i, %for.body], [ %count, %0 ]
%gep1 = getelementptr i64, ptr %phi1, i64 -1
%load1 = load i64, ptr %gep1
%gep2 = getelementptr i64, ptr %phi2, i64 -1
store i64 %load1, ptr %gep2
%load2 = load i64, ptr %phi1
store i64 %load2, ptr %phi2
%dec.i = add nsw i64 %i, -1
%gep3 = getelementptr i64, ptr %phi2, i64 -2
%gep4 = getelementptr i64, ptr %phi1, i64 -2
%cond = icmp sgt i64 %dec.i, 0
br i1 %cond, label %for.body, label %end
end:
ret void
}
define void @post_indexed_sub_quadword(ptr %a, ptr %b, i64 %count) nounwind {
; CHECK-LABEL: post_indexed_sub_quadword:
; CHECK: // %bb.0:
; CHECK-NEXT: .LBB56_1: // %for.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur q0, [x1, #-16]
; CHECK-NEXT: subs x2, x2, #1
; CHECK-NEXT: stur q0, [x0, #-16]
; CHECK-NEXT: ldr q0, [x1], #-32
; CHECK-NEXT: str q0, [x0], #-32
; CHECK-NEXT: b.gt .LBB56_1
; CHECK-NEXT: // %bb.2: // %end
; CHECK-NEXT: ret
br label %for.body
for.body:
%phi1 = phi ptr [ %gep4, %for.body ], [ %b, %0 ]
%phi2 = phi ptr [ %gep3, %for.body ], [ %a, %0 ]
%i = phi i64 [ %dec.i, %for.body], [ %count, %0 ]
%gep1 = getelementptr <2 x i64>, ptr %phi1, i64 -1
%load1 = load <2 x i64>, ptr %gep1
%gep2 = getelementptr <2 x i64>, ptr %phi2, i64 -1
store <2 x i64> %load1, ptr %gep2
%load2 = load <2 x i64>, ptr %phi1
store <2 x i64> %load2, ptr %phi2
%dec.i = add nsw i64 %i, -1
%gep3 = getelementptr <2 x i64>, ptr %phi2, i64 -2
%gep4 = getelementptr <2 x i64>, ptr %phi1, i64 -2
%cond = icmp sgt i64 %dec.i, 0
br i1 %cond, label %for.body, label %end
end:
ret void
}
define void @post_indexed_sub_float(ptr %a, ptr %b, i64 %count) nounwind {
; CHECK-LABEL: post_indexed_sub_float:
; CHECK: // %bb.0:
; CHECK-NEXT: .LBB57_1: // %for.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur s0, [x1, #-4]
; CHECK-NEXT: subs x2, x2, #1
; CHECK-NEXT: stur s0, [x0, #-4]
; CHECK-NEXT: ldr s0, [x1], #-8
; CHECK-NEXT: str s0, [x0], #-8
; CHECK-NEXT: b.gt .LBB57_1
; CHECK-NEXT: // %bb.2: // %end
; CHECK-NEXT: ret
br label %for.body
for.body:
%phi1 = phi ptr [ %gep4, %for.body ], [ %b, %0 ]
%phi2 = phi ptr [ %gep3, %for.body ], [ %a, %0 ]
%i = phi i64 [ %dec.i, %for.body], [ %count, %0 ]
%gep1 = getelementptr float, ptr %phi1, i64 -1
%load1 = load float, ptr %gep1
%gep2 = getelementptr float, ptr %phi2, i64 -1
store float %load1, ptr %gep2
%load2 = load float, ptr %phi1
store float %load2, ptr %phi2
%dec.i = add nsw i64 %i, -1
%gep3 = getelementptr float, ptr %phi2, i64 -2
%gep4 = getelementptr float, ptr %phi1, i64 -2
%cond = icmp sgt i64 %dec.i, 0
br i1 %cond, label %for.body, label %end
end:
ret void
}
define void @post_indexed_sub_double(ptr %a, ptr %b, i64 %count) nounwind {
; CHECK-LABEL: post_indexed_sub_double:
; CHECK: // %bb.0:
; CHECK-NEXT: .LBB58_1: // %for.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldur d0, [x1, #-8]
; CHECK-NEXT: subs x2, x2, #1
; CHECK-NEXT: stur d0, [x0, #-8]
; CHECK-NEXT: ldr d0, [x1], #-16
; CHECK-NEXT: str d0, [x0], #-16
; CHECK-NEXT: b.gt .LBB58_1
; CHECK-NEXT: // %bb.2: // %end
; CHECK-NEXT: ret
br label %for.body
for.body:
%phi1 = phi ptr [ %gep4, %for.body ], [ %b, %0 ]
%phi2 = phi ptr [ %gep3, %for.body ], [ %a, %0 ]
%i = phi i64 [ %dec.i, %for.body], [ %count, %0 ]
%gep1 = getelementptr double, ptr %phi1, i64 -1
%load1 = load double, ptr %gep1
%gep2 = getelementptr double, ptr %phi2, i64 -1
store double %load1, ptr %gep2
%load2 = load double, ptr %phi1
store double %load2, ptr %phi2
%dec.i = add nsw i64 %i, -1
%gep3 = getelementptr double, ptr %phi2, i64 -2
%gep4 = getelementptr double, ptr %phi1, i64 -2
%cond = icmp sgt i64 %dec.i, 0
br i1 %cond, label %for.body, label %end
end:
ret void
}
define void @post_indexed_sub_doubleword_offset_min(ptr %a, ptr %b, i64 %count) nounwind {
; CHECK-LABEL: post_indexed_sub_doubleword_offset_min:
; CHECK: // %bb.0:
; CHECK-NEXT: .LBB59_1: // %for.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldr x8, [x1, #8]
; CHECK-NEXT: subs x2, x2, #1
; CHECK-NEXT: str x8, [x0, #8]
; CHECK-NEXT: ldr x8, [x1], #-256
; CHECK-NEXT: str x8, [x0], #-256
; CHECK-NEXT: b.gt .LBB59_1
; CHECK-NEXT: // %bb.2: // %end
; CHECK-NEXT: ret
br label %for.body
for.body:
%phi1 = phi ptr [ %gep4, %for.body ], [ %b, %0 ]
%phi2 = phi ptr [ %gep3, %for.body ], [ %a, %0 ]
%i = phi i64 [ %dec.i, %for.body], [ %count, %0 ]
%gep1 = getelementptr i64, ptr %phi1, i64 1
%load1 = load i64, ptr %gep1
%gep2 = getelementptr i64, ptr %phi2, i64 1
store i64 %load1, ptr %gep2
%load2 = load i64, ptr %phi1
store i64 %load2, ptr %phi2
%dec.i = add nsw i64 %i, -1
%gep3 = getelementptr i64, ptr %phi2, i64 -32
%gep4 = getelementptr i64, ptr %phi1, i64 -32
%cond = icmp sgt i64 %dec.i, 0
br i1 %cond, label %for.body, label %end
end:
ret void
}
define void @post_indexed_doubleword_offset_out_of_range(ptr %a, ptr %b, i64 %count) nounwind {
; CHECK-LABEL: post_indexed_doubleword_offset_out_of_range:
; CHECK: // %bb.0:
; CHECK-NEXT: .LBB60_1: // %for.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldr x8, [x1, #8]
; CHECK-NEXT: subs x2, x2, #1
; CHECK-NEXT: str x8, [x0, #8]
; CHECK-NEXT: ldr x8, [x1]
; CHECK-NEXT: add x1, x1, #256
; CHECK-NEXT: str x8, [x0]
; CHECK-NEXT: add x0, x0, #256
; CHECK-NEXT: b.gt .LBB60_1
; CHECK-NEXT: // %bb.2: // %end
; CHECK-NEXT: ret
br label %for.body
for.body:
%phi1 = phi ptr [ %gep4, %for.body ], [ %b, %0 ]
%phi2 = phi ptr [ %gep3, %for.body ], [ %a, %0 ]
%i = phi i64 [ %dec.i, %for.body], [ %count, %0 ]
%gep1 = getelementptr i64, ptr %phi1, i64 1
%load1 = load i64, ptr %gep1
%gep2 = getelementptr i64, ptr %phi2, i64 1
store i64 %load1, ptr %gep2
%load2 = load i64, ptr %phi1
store i64 %load2, ptr %phi2
%dec.i = add nsw i64 %i, -1
%gep3 = getelementptr i64, ptr %phi2, i64 32
%gep4 = getelementptr i64, ptr %phi1, i64 32
%cond = icmp sgt i64 %dec.i, 0
br i1 %cond, label %for.body, label %end
end:
ret void
}
define void @post_indexed_paired_min_offset(ptr %a, ptr %b, i64 %count) nounwind {
; CHECK-LABEL: post_indexed_paired_min_offset:
; CHECK: // %bb.0:
; CHECK-NEXT: .LBB61_1: // %for.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldp x9, x8, [x1], #-512
; CHECK-NEXT: subs x2, x2, #1
; CHECK-NEXT: stp x9, x8, [x0], #-512
; CHECK-NEXT: b.gt .LBB61_1
; CHECK-NEXT: // %bb.2: // %end
; CHECK-NEXT: ret
br label %for.body
for.body:
%phi1 = phi ptr [ %gep4, %for.body ], [ %b, %0 ]
%phi2 = phi ptr [ %gep3, %for.body ], [ %a, %0 ]
%i = phi i64 [ %dec.i, %for.body], [ %count, %0 ]
%gep1 = getelementptr i64, ptr %phi1, i64 1
%load1 = load i64, ptr %gep1
%gep2 = getelementptr i64, ptr %phi2, i64 1
%load2 = load i64, ptr %phi1
store i64 %load1, ptr %gep2
store i64 %load2, ptr %phi2
%dec.i = add nsw i64 %i, -1
%gep3 = getelementptr i64, ptr %phi2, i64 -64
%gep4 = getelementptr i64, ptr %phi1, i64 -64
%cond = icmp sgt i64 %dec.i, 0
br i1 %cond, label %for.body, label %end
end:
ret void
}
define void @post_indexed_paired_offset_out_of_range(ptr %a, ptr %b, i64 %count) nounwind {
; CHECK-LABEL: post_indexed_paired_offset_out_of_range:
; CHECK: // %bb.0:
; CHECK-NEXT: .LBB62_1: // %for.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldp x8, x9, [x1]
; CHECK-NEXT: subs x2, x2, #1
; CHECK-NEXT: add x1, x1, #512
; CHECK-NEXT: stp x9, x8, [x0]
; CHECK-NEXT: add x0, x0, #512
; CHECK-NEXT: b.gt .LBB62_1
; CHECK-NEXT: // %bb.2: // %end
; CHECK-NEXT: ret
br label %for.body
for.body:
%phi1 = phi ptr [ %gep4, %for.body ], [ %b, %0 ]
%phi2 = phi ptr [ %gep3, %for.body ], [ %a, %0 ]
%i = phi i64 [ %dec.i, %for.body], [ %count, %0 ]
%gep1 = getelementptr i64, ptr %phi1, i64 1
%load1 = load i64, ptr %phi1
%gep2 = getelementptr i64, ptr %phi2, i64 1
%load2 = load i64, ptr %gep1
store i64 %load1, ptr %gep2
store i64 %load2, ptr %phi2
%dec.i = add nsw i64 %i, -1
%gep3 = getelementptr i64, ptr %phi2, i64 64
%gep4 = getelementptr i64, ptr %phi1, i64 64
%cond = icmp sgt i64 %dec.i, 0
br i1 %cond, label %for.body, label %end
end:
ret void
}
; DAGCombiner::MergeConsecutiveStores merges this into a vector store,
; replaceZeroVectorStore should split the vector store back into
; scalar stores which should get merged by AArch64LoadStoreOptimizer.
define void @merge_zr32(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_zr32:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: str xzr, [x0]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_zr32:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: stp wzr, wzr, [x0]
; STRICTALIGN-NEXT: ret
entry:
store i32 0, ptr %p
%p1 = getelementptr i32, ptr %p, i32 1
store i32 0, ptr %p1
ret void
}
; Same as merge_zr32 but the merged stores should also get paried.
define void @merge_zr32_2(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_zr32_2:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: stp xzr, xzr, [x0]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_zr32_2:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: stp wzr, wzr, [x0]
; STRICTALIGN-NEXT: stp wzr, wzr, [x0, #8]
; STRICTALIGN-NEXT: ret
entry:
store i32 0, ptr %p
%p1 = getelementptr i32, ptr %p, i32 1
store i32 0, ptr %p1
%p2 = getelementptr i32, ptr %p, i64 2
store i32 0, ptr %p2
%p3 = getelementptr i32, ptr %p, i64 3
store i32 0, ptr %p3
ret void
}
; Like merge_zr32_2, but checking the largest allowed stp immediate offset.
define void @merge_zr32_2_offset(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_zr32_2_offset:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: stp xzr, xzr, [x0, #504]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_zr32_2_offset:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: str wzr, [x0, #504]
; STRICTALIGN-NEXT: str wzr, [x0, #508]
; STRICTALIGN-NEXT: str wzr, [x0, #512]
; STRICTALIGN-NEXT: str wzr, [x0, #516]
; STRICTALIGN-NEXT: ret
entry:
%p0 = getelementptr i32, ptr %p, i32 126
store i32 0, ptr %p0
%p1 = getelementptr i32, ptr %p, i32 127
store i32 0, ptr %p1
%p2 = getelementptr i32, ptr %p, i64 128
store i32 0, ptr %p2
%p3 = getelementptr i32, ptr %p, i64 129
store i32 0, ptr %p3
ret void
}
; Like merge_zr32, but replaceZeroVectorStore should not split this
; vector store since the address offset is too large for the stp
; instruction.
define void @no_merge_zr32_2_offset(ptr %p) {
; NOSTRICTALIGN-LABEL: no_merge_zr32_2_offset:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: movi v0.2d, #0000000000000000
; NOSTRICTALIGN-NEXT: str q0, [x0, #4096]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: no_merge_zr32_2_offset:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: str wzr, [x0, #4096]
; STRICTALIGN-NEXT: str wzr, [x0, #4100]
; STRICTALIGN-NEXT: str wzr, [x0, #4104]
; STRICTALIGN-NEXT: str wzr, [x0, #4108]
; STRICTALIGN-NEXT: ret
entry:
%p0 = getelementptr i32, ptr %p, i32 1024
store i32 0, ptr %p0
%p1 = getelementptr i32, ptr %p, i32 1025
store i32 0, ptr %p1
%p2 = getelementptr i32, ptr %p, i64 1026
store i32 0, ptr %p2
%p3 = getelementptr i32, ptr %p, i64 1027
store i32 0, ptr %p3
ret void
}
; Like merge_zr32, but replaceZeroVectorStore should not split the
; vector store since the zero constant vector has multiple uses, so we
; err on the side that allows for stp q instruction generation.
define void @merge_zr32_3(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_zr32_3:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: movi v0.2d, #0000000000000000
; NOSTRICTALIGN-NEXT: stp q0, q0, [x0]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_zr32_3:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: stp wzr, wzr, [x0]
; STRICTALIGN-NEXT: stp wzr, wzr, [x0, #8]
; STRICTALIGN-NEXT: stp wzr, wzr, [x0, #16]
; STRICTALIGN-NEXT: stp wzr, wzr, [x0, #24]
; STRICTALIGN-NEXT: ret
entry:
store i32 0, ptr %p
%p1 = getelementptr i32, ptr %p, i32 1
store i32 0, ptr %p1
%p2 = getelementptr i32, ptr %p, i64 2
store i32 0, ptr %p2
%p3 = getelementptr i32, ptr %p, i64 3
store i32 0, ptr %p3
%p4 = getelementptr i32, ptr %p, i64 4
store i32 0, ptr %p4
%p5 = getelementptr i32, ptr %p, i64 5
store i32 0, ptr %p5
%p6 = getelementptr i32, ptr %p, i64 6
store i32 0, ptr %p6
%p7 = getelementptr i32, ptr %p, i64 7
store i32 0, ptr %p7
ret void
}
; Like merge_zr32, but with 2-vector type.
define void @merge_zr32_2vec(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_zr32_2vec:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: str xzr, [x0]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_zr32_2vec:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: stp wzr, wzr, [x0]
; STRICTALIGN-NEXT: ret
entry:
store <2 x i32> zeroinitializer, ptr %p
ret void
}
; Like merge_zr32, but with 3-vector type.
define void @merge_zr32_3vec(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_zr32_3vec:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: str wzr, [x0, #8]
; NOSTRICTALIGN-NEXT: str xzr, [x0]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_zr32_3vec:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: stp wzr, wzr, [x0, #4]
; STRICTALIGN-NEXT: str wzr, [x0]
; STRICTALIGN-NEXT: ret
entry:
store <3 x i32> zeroinitializer, ptr %p
ret void
}
; Like merge_zr32, but with 4-vector type.
define void @merge_zr32_4vec(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_zr32_4vec:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: stp xzr, xzr, [x0]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_zr32_4vec:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: stp wzr, wzr, [x0, #8]
; STRICTALIGN-NEXT: stp wzr, wzr, [x0]
; STRICTALIGN-NEXT: ret
entry:
store <4 x i32> zeroinitializer, ptr %p
ret void
}
; Like merge_zr32, but with 2-vector float type.
define void @merge_zr32_2vecf(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_zr32_2vecf:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: str xzr, [x0]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_zr32_2vecf:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: stp wzr, wzr, [x0]
; STRICTALIGN-NEXT: ret
entry:
store <2 x float> zeroinitializer, ptr %p
ret void
}
; Like merge_zr32, but with 4-vector float type.
define void @merge_zr32_4vecf(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_zr32_4vecf:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: stp xzr, xzr, [x0]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_zr32_4vecf:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: stp wzr, wzr, [x0, #8]
; STRICTALIGN-NEXT: stp wzr, wzr, [x0]
; STRICTALIGN-NEXT: ret
entry:
store <4 x float> zeroinitializer, ptr %p
ret void
}
; Similar to merge_zr32, but for 64-bit values.
define void @merge_zr64(ptr %p) {
; CHECK-LABEL: merge_zr64:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: stp xzr, xzr, [x0]
; CHECK-NEXT: ret
entry:
store i64 0, ptr %p
%p1 = getelementptr i64, ptr %p, i64 1
store i64 0, ptr %p1
ret void
}
; Similar to merge_zr32, but for 64-bit values and with unaligned stores.
define void @merge_zr64_unalign(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_zr64_unalign:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: stp xzr, xzr, [x0]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_zr64_unalign:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: lsr x8, xzr, #40
; STRICTALIGN-NEXT: lsr x9, xzr, #24
; STRICTALIGN-NEXT: lsr x11, xzr, #16
; STRICTALIGN-NEXT: mov x10, x0
; STRICTALIGN-NEXT: lsr x12, xzr, #8
; STRICTALIGN-NEXT: strb wzr, [x0]
; STRICTALIGN-NEXT: strb w8, [x0, #13]
; STRICTALIGN-NEXT: strb wzr, [x10, #8]!
; STRICTALIGN-NEXT: strb w9, [x10, #3]
; STRICTALIGN-NEXT: strb w11, [x10, #2]
; STRICTALIGN-NEXT: strb w8, [x0, #5]
; STRICTALIGN-NEXT: lsr x8, xzr, #32
; STRICTALIGN-NEXT: strb w9, [x0, #3]
; STRICTALIGN-NEXT: lsr x9, xzr, #56
; STRICTALIGN-NEXT: strb w11, [x0, #2]
; STRICTALIGN-NEXT: lsr x11, xzr, #48
; STRICTALIGN-NEXT: strb w12, [x0, #9]
; STRICTALIGN-NEXT: strb w12, [x0, #1]
; STRICTALIGN-NEXT: strb w8, [x10, #4]!
; STRICTALIGN-NEXT: strb w9, [x10, #3]
; STRICTALIGN-NEXT: strb w11, [x10, #2]
; STRICTALIGN-NEXT: strb w8, [x0, #4]!
; STRICTALIGN-NEXT: strb w9, [x0, #3]
; STRICTALIGN-NEXT: strb w11, [x0, #2]
; STRICTALIGN-NEXT: ret
entry:
store <2 x i64> zeroinitializer, ptr %p, align 1
ret void
}
; Similar to merge_zr32_3, replaceZeroVectorStore should not split the
; vector store since the zero constant vector has multiple uses.
define void @merge_zr64_2(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_zr64_2:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: movi v0.2d, #0000000000000000
; NOSTRICTALIGN-NEXT: stp q0, q0, [x0]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_zr64_2:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: stp xzr, xzr, [x0]
; STRICTALIGN-NEXT: stp xzr, xzr, [x0, #16]
; STRICTALIGN-NEXT: ret
entry:
store i64 0, ptr %p
%p1 = getelementptr i64, ptr %p, i64 1
store i64 0, ptr %p1
%p2 = getelementptr i64, ptr %p, i64 2
store i64 0, ptr %p2
%p3 = getelementptr i64, ptr %p, i64 3
store i64 0, ptr %p3
ret void
}
; Like merge_zr64, but with 2-vector double type.
define void @merge_zr64_2vecd(ptr %p) {
; CHECK-LABEL: merge_zr64_2vecd:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: stp xzr, xzr, [x0]
; CHECK-NEXT: ret
entry:
store <2 x double> zeroinitializer, ptr %p
ret void
}
; Like merge_zr64, but with 3-vector i64 type.
define void @merge_zr64_3vec(ptr %p) {
; CHECK-LABEL: merge_zr64_3vec:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: stp xzr, xzr, [x0, #8]
; CHECK-NEXT: str xzr, [x0]
; CHECK-NEXT: ret
entry:
store <3 x i64> zeroinitializer, ptr %p
ret void
}
; Like merge_zr64_2, but with 4-vector double type.
define void @merge_zr64_4vecd(ptr %p) {
; CHECK-LABEL: merge_zr64_4vecd:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: movi v0.2d, #0000000000000000
; CHECK-NEXT: stp q0, q0, [x0]
; CHECK-NEXT: ret
entry:
store <4 x double> zeroinitializer, ptr %p
ret void
}
; Verify that non-consecutive merges do not generate q0
define void @merge_multiple_128bit_stores(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_multiple_128bit_stores:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: movi v0.2d, #0000000000000000
; NOSTRICTALIGN-NEXT: str q0, [x0]
; NOSTRICTALIGN-NEXT: stur q0, [x0, #24]
; NOSTRICTALIGN-NEXT: str q0, [x0, #48]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_multiple_128bit_stores:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: stp xzr, xzr, [x0]
; STRICTALIGN-NEXT: stp xzr, xzr, [x0, #24]
; STRICTALIGN-NEXT: stp xzr, xzr, [x0, #48]
; STRICTALIGN-NEXT: ret
entry:
store i64 0, ptr %p
%p1 = getelementptr i64, ptr %p, i64 1
store i64 0, ptr %p1
%p3 = getelementptr i64, ptr %p, i64 3
store i64 0, ptr %p3
%p4 = getelementptr i64, ptr %p, i64 4
store i64 0, ptr %p4
%p6 = getelementptr i64, ptr %p, i64 6
store i64 0, ptr %p6
%p7 = getelementptr i64, ptr %p, i64 7
store i64 0, ptr %p7
ret void
}
; Verify that large stores generate stp q
define void @merge_multiple_128bit_stores_consec(ptr %p) {
; NOSTRICTALIGN-LABEL: merge_multiple_128bit_stores_consec:
; NOSTRICTALIGN: // %bb.0: // %entry
; NOSTRICTALIGN-NEXT: movi v0.2d, #0000000000000000
; NOSTRICTALIGN-NEXT: stp q0, q0, [x0]
; NOSTRICTALIGN-NEXT: stp q0, q0, [x0, #32]
; NOSTRICTALIGN-NEXT: ret
;
; STRICTALIGN-LABEL: merge_multiple_128bit_stores_consec:
; STRICTALIGN: // %bb.0: // %entry
; STRICTALIGN-NEXT: stp xzr, xzr, [x0]
; STRICTALIGN-NEXT: stp xzr, xzr, [x0, #16]
; STRICTALIGN-NEXT: stp xzr, xzr, [x0, #32]
; STRICTALIGN-NEXT: stp xzr, xzr, [x0, #48]
; STRICTALIGN-NEXT: ret
entry:
store i64 0, ptr %p
%p1 = getelementptr i64, ptr %p, i64 1
store i64 0, ptr %p1
%p2 = getelementptr i64, ptr %p, i64 2
store i64 0, ptr %p2
%p3 = getelementptr i64, ptr %p, i64 3
store i64 0, ptr %p3
%p4 = getelementptr i64, ptr %p, i64 4
store i64 0, ptr %p4
%p5 = getelementptr i64, ptr %p, i64 5
store i64 0, ptr %p5
%p6 = getelementptr i64, ptr %p, i64 6
store i64 0, ptr %p6
%p7 = getelementptr i64, ptr %p, i64 7
store i64 0, ptr %p7
ret void
}
; Check for bug 34674 where invalid add of xzr was being generated.
define i64 @bug34674(ptr %p) {
; CHECK-LABEL: bug34674:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: mov x9, xzr
; CHECK-NEXT: mov x8, x0
; CHECK-NEXT: add x0, x9, #1
; CHECK-NEXT: stp xzr, xzr, [x8]
; CHECK-NEXT: ret
entry:
store <2 x i64> zeroinitializer, ptr %p
%ld = load i64, ptr %p
%add = add i64 %ld, 1
ret i64 %add
}
define void @trunc_splat_zero(ptr %ptr) {
; CHECK-LABEL: trunc_splat_zero:
; CHECK: // %bb.0:
; CHECK-NEXT: strh wzr, [x0]
; CHECK-NEXT: ret
store <2 x i8> zeroinitializer, ptr %ptr, align 2
ret void
}
define void @trunc_splat(ptr %ptr) {
; CHECK-LABEL: trunc_splat:
; CHECK: // %bb.0:
; CHECK-NEXT: mov w8, #42 // =0x2a
; CHECK-NEXT: movk w8, #42, lsl #16
; CHECK-NEXT: str w8, [x0]
; CHECK-NEXT: ret
store <2 x i16> <i16 42, i16 42>, ptr %ptr, align 4
ret void
}
Reference in new issue View Git Blame Copy Permalink