; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -fixup-byte-word-insts=1 < %s | FileCheck %s -check-prefix=BWON
; RUN: llc -fixup-byte-word-insts=0 < %s | FileCheck %s -check-prefix=BWOFF

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-s0:64:64-f80:128:128-n8:16:32:64-S128"
target triple = "x86_64-apple-macosx10.8.0"

%struct.A = type { i8, i8, i8, i8, i8, i8, i8, i8 }

; This has byte loads interspersed with byte stores, in a single
; basic-block loop.  The upper portion should be dead, so the movb loads
; should have been changed into movzbl instead.
define void @foo1(i32 %count, ptr noalias nocapture %q, ptr noalias nocapture %p) nounwind uwtable noinline ssp {
; BWON-LABEL: foo1:
; BWON:       ## %bb.0:
; BWON-NEXT:    testl %edi, %edi
; BWON-NEXT:    jle LBB0_2
; BWON-NEXT:    .p2align 4
; BWON-NEXT:  LBB0_1: ## %a4
; BWON-NEXT:    ## =>This Inner Loop Header: Depth=1
; BWON-NEXT:    movzbl (%rsi), %eax
; BWON-NEXT:    movb %al, (%rdx)
; BWON-NEXT:    movzbl 1(%rsi), %eax
; BWON-NEXT:    movb %al, 1(%rdx)
; BWON-NEXT:    addq $8, %rdx
; BWON-NEXT:    decl %edi
; BWON-NEXT:    jne LBB0_1
; BWON-NEXT:  LBB0_2: ## %._crit_edge
; BWON-NEXT:    retq
;
; BWOFF-LABEL: foo1:
; BWOFF:       ## %bb.0:
; BWOFF-NEXT:    testl %edi, %edi
; BWOFF-NEXT:    jle LBB0_2
; BWOFF-NEXT:    .p2align 4
; BWOFF-NEXT:  LBB0_1: ## %a4
; BWOFF-NEXT:    ## =>This Inner Loop Header: Depth=1
; BWOFF-NEXT:    movb (%rsi), %al
; BWOFF-NEXT:    movb %al, (%rdx)
; BWOFF-NEXT:    movb 1(%rsi), %al
; BWOFF-NEXT:    movb %al, 1(%rdx)
; BWOFF-NEXT:    addq $8, %rdx
; BWOFF-NEXT:    decl %edi
; BWOFF-NEXT:    jne LBB0_1
; BWOFF-NEXT:  LBB0_2: ## %._crit_edge
; BWOFF-NEXT:    retq
  %1 = icmp sgt i32 %count, 0
  br i1 %1, label %.lr.ph, label %._crit_edge

.lr.ph:                                           ; preds = %0
  %2 = getelementptr inbounds %struct.A, ptr %q, i64 0, i32 1
  br label %a4

a4:                                       ; preds = %3, %.lr.ph
  %i.02 = phi i32 [ 0, %.lr.ph ], [ %a9, %a4 ]
  %.01 = phi ptr [ %p, %.lr.ph ], [ %a10, %a4 ]
  %a5 = load i8, ptr %q, align 1
  store i8 %a5, ptr %.01, align 1
  %a8 = getelementptr inbounds %struct.A, ptr %.01, i64 0, i32 1
  %a6 = load i8, ptr %2, align 1
  store i8 %a6, ptr %a8, align 1
  %a9 = add nsw i32 %i.02, 1
  %a10 = getelementptr inbounds %struct.A, ptr %.01, i64 1
  %exitcond = icmp eq i32 %a9, %count
  br i1 %exitcond, label %._crit_edge, label %a4

._crit_edge:                                      ; preds = %3, %0
  ret void
}

%struct.B = type { i16, i16, i16, i16, i16, i16, i16, i16 }

; This has word loads interspersed with word stores.
; The upper portion should be dead, so the movw loads should have
; been changed into movzwl instead.
define void @foo2(i32 %count, ptr noalias nocapture %q, ptr noalias nocapture %p) nounwind uwtable noinline ssp {
; BWON-LABEL: foo2:
; BWON:       ## %bb.0:
; BWON-NEXT:    testl %edi, %edi
; BWON-NEXT:    jle LBB1_2
; BWON-NEXT:    .p2align 4
; BWON-NEXT:  LBB1_1: ## %a4
; BWON-NEXT:    ## =>This Inner Loop Header: Depth=1
; BWON-NEXT:    movzwl (%rsi), %eax
; BWON-NEXT:    movw %ax, (%rdx)
; BWON-NEXT:    movzwl 2(%rsi), %eax
; BWON-NEXT:    movw %ax, 2(%rdx)
; BWON-NEXT:    addq $16, %rdx
; BWON-NEXT:    decl %edi
; BWON-NEXT:    jne LBB1_1
; BWON-NEXT:  LBB1_2: ## %._crit_edge
; BWON-NEXT:    retq
;
; BWOFF-LABEL: foo2:
; BWOFF:       ## %bb.0:
; BWOFF-NEXT:    testl %edi, %edi
; BWOFF-NEXT:    jle LBB1_2
; BWOFF-NEXT:    .p2align 4
; BWOFF-NEXT:  LBB1_1: ## %a4
; BWOFF-NEXT:    ## =>This Inner Loop Header: Depth=1
; BWOFF-NEXT:    movw (%rsi), %ax
; BWOFF-NEXT:    movw %ax, (%rdx)
; BWOFF-NEXT:    movw 2(%rsi), %ax
; BWOFF-NEXT:    movw %ax, 2(%rdx)
; BWOFF-NEXT:    addq $16, %rdx
; BWOFF-NEXT:    decl %edi
; BWOFF-NEXT:    jne LBB1_1
; BWOFF-NEXT:  LBB1_2: ## %._crit_edge
; BWOFF-NEXT:    retq
  %1 = icmp sgt i32 %count, 0
  br i1 %1, label %.lr.ph, label %._crit_edge

.lr.ph:                                           ; preds = %0
  %2 = getelementptr inbounds %struct.B, ptr %q, i64 0, i32 1
  br label %a4

a4:                                       ; preds = %3, %.lr.ph
  %i.02 = phi i32 [ 0, %.lr.ph ], [ %a9, %a4 ]
  %.01 = phi ptr [ %p, %.lr.ph ], [ %a10, %a4 ]
  %a5 = load i16, ptr %q, align 2
  store i16 %a5, ptr %.01, align 2
  %a8 = getelementptr inbounds %struct.B, ptr %.01, i64 0, i32 1
  %a6 = load i16, ptr %2, align 2
  store i16 %a6, ptr %a8, align 2
  %a9 = add nsw i32 %i.02, 1
  %a10 = getelementptr inbounds %struct.B, ptr %.01, i64 1
  %exitcond = icmp eq i32 %a9, %count
  br i1 %exitcond, label %._crit_edge, label %a4

._crit_edge:                                      ; preds = %3, %0
  ret void
}

; This test contains nothing but a simple byte load and store.
; movb encodes smaller, but we use movzbl for the load for better perf.
define void @foo3(ptr%dst, ptr%src) {
; BWON-LABEL: foo3:
; BWON:       ## %bb.0:
; BWON-NEXT:    movzbl (%rsi), %eax
; BWON-NEXT:    movb %al, (%rdi)
; BWON-NEXT:    retq
;
; BWOFF-LABEL: foo3:
; BWOFF:       ## %bb.0:
; BWOFF-NEXT:    movb (%rsi), %al
; BWOFF-NEXT:    movb %al, (%rdi)
; BWOFF-NEXT:    retq
  %t0 = load i8, ptr%src, align 1
  store i8 %t0, ptr%dst, align 1
  ret void
}

; This test contains nothing but a simple word load and store.  Since
; movw and movzwl are the same size, we should always choose to use
; movzwl instead.
define void @foo4(ptr%dst, ptr%src) {
; BWON-LABEL: foo4:
; BWON:       ## %bb.0:
; BWON-NEXT:    movzwl (%rsi), %eax
; BWON-NEXT:    movw %ax, (%rdi)
; BWON-NEXT:    retq
;
; BWOFF-LABEL: foo4:
; BWOFF:       ## %bb.0:
; BWOFF-NEXT:    movw (%rsi), %ax
; BWOFF-NEXT:    movw %ax, (%rdi)
; BWOFF-NEXT:    retq
  %t0 = load i16, ptr%src, align 2
  store i16 %t0, ptr%dst, align 2
  ret void
}