Merging r308986 and r308963:

------------------------------------------------------------------------
r308963 | rksimon | 2017-07-25 03:33:36 -0700 (Tue, 25 Jul 2017) | 1 line

[X86] Add 24-byte memcmp tests (PR33914)
------------------------------------------------------------------------

------------------------------------------------------------------------
r308986 | rksimon | 2017-07-25 10:04:37 -0700 (Tue, 25 Jul 2017) | 9 lines

[X86][CGP] Reduce memcmp() expansion to 2 load pairs (PR33914)

D35067/rL308322 attempted to support up to 4 load pairs for memcmp inlining which resulted in regressions for some optimized libc memcmp implementations (PR33914).

Until we can match these more optimal cases, this patch reduces the memcmp expansion to a maximum of 2 load pairs (which matches what we do for -Os).

This patch should be considered for the 5.0.0 release branch as well

Differential Revision: https://reviews.llvm.org/D35830
------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_50@309127 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 4485fa277161493a1dfe900f68a162b72ab56fc7..ba8eb8656585a7864b182066ae4a939f4462cc3a 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -1672,8 +1672,8 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
 
   // TODO: These control memcmp expansion in CGP and could be raised higher, but
   // that needs to benchmarked and balanced with the potential use of vector
-  // load/store types (PR33329).
-  MaxLoadsPerMemcmp = 4;
+  // load/store types (PR33329, PR33914).
+  MaxLoadsPerMemcmp = 2;
   MaxLoadsPerMemcmpOptSize = 2;
 
   // Set loop alignment to 2^ExperimentalPrefLoopAlignment bytes (default: 2^4).

diff --git a/test/CodeGen/X86/memcmp-minsize.ll b/test/CodeGen/X86/memcmp-minsize.ll
index a7f42644ca2d5f58b895f168ff5e4ca2d9d276e0..a55c40f5bda821d5fbda07597006fbb23de19e0f 100644 (file)
--- a/test/CodeGen/X86/memcmp-minsize.ll
+++ b/test/CodeGen/X86/memcmp-minsize.ll
@@ -527,6 +527,93 @@ define i1 @length16_eq_const(i8* %X) nounwind minsize {
   ret i1 %c
 }
 
+; PR33914 - https://bugs.llvm.org/show_bug.cgi?id=33914
+
+define i32 @length24(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length24:
+; X86:       # BB#0:
+; X86-NEXT:    subl $16, %esp
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl %ecx, {{[0-9]+}}(%esp)
+; X86-NEXT:    movl %eax, (%esp)
+; X86-NEXT:    andl $0, {{[0-9]+}}(%esp)
+; X86-NEXT:    movl $24, {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    retl
+;
+; X64-LABEL: length24:
+; X64:       # BB#0:
+; X64-NEXT:    pushq $24
+; X64-NEXT:    popq %rdx
+; X64-NEXT:    jmp memcmp # TAILCALL
+  %m = tail call i32 @memcmp(i8* %X, i8* %Y, i64 24) nounwind
+  ret i32 %m
+}
+
+define i1 @length24_eq(i8* %x, i8* %y) nounwind minsize {
+; X86-LABEL: length24_eq:
+; X86:       # BB#0:
+; X86-NEXT:    subl $16, %esp
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl %ecx, {{[0-9]+}}(%esp)
+; X86-NEXT:    movl %eax, (%esp)
+; X86-NEXT:    andl $0, {{[0-9]+}}(%esp)
+; X86-NEXT:    movl $24, {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    testl %eax, %eax
+; X86-NEXT:    sete %al
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    retl
+;
+; X64-LABEL: length24_eq:
+; X64:       # BB#0:
+; X64-NEXT:    pushq %rax
+; X64-NEXT:    pushq $24
+; X64-NEXT:    popq %rdx
+; X64-NEXT:    callq memcmp
+; X64-NEXT:    testl %eax, %eax
+; X64-NEXT:    sete %al
+; X64-NEXT:    popq %rcx
+; X64-NEXT:    retq
+  %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 24) nounwind
+  %cmp = icmp eq i32 %call, 0
+  ret i1 %cmp
+}
+
+define i1 @length24_eq_const(i8* %X) nounwind minsize {
+; X86-LABEL: length24_eq_const:
+; X86:       # BB#0:
+; X86-NEXT:    subl $16, %esp
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl %eax, (%esp)
+; X86-NEXT:    andl $0, {{[0-9]+}}(%esp)
+; X86-NEXT:    movl $24, {{[0-9]+}}(%esp)
+; X86-NEXT:    movl $.L.str, {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    testl %eax, %eax
+; X86-NEXT:    setne %al
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    retl
+;
+; X64-LABEL: length24_eq_const:
+; X64:       # BB#0:
+; X64-NEXT:    pushq %rax
+; X64-NEXT:    pushq $24
+; X64-NEXT:    popq %rdx
+; X64-NEXT:    movl $.L.str, %esi
+; X64-NEXT:    callq memcmp
+; X64-NEXT:    testl %eax, %eax
+; X64-NEXT:    setne %al
+; X64-NEXT:    popq %rcx
+; X64-NEXT:    retq
+  %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i64 24) nounwind
+  %c = icmp ne i32 %m, 0
+  ret i1 %c
+}
+
 define i32 @length32(i8* %X, i8* %Y) nounwind minsize {
 ; X86-LABEL: length32:
 ; X86:       # BB#0:

diff --git a/test/CodeGen/X86/memcmp-optsize.ll b/test/CodeGen/X86/memcmp-optsize.ll
index 450205a966d2369cc96d96c5b9f5efe0546964df..4a5f30890513ea7d25f3928dae0421d591a8dbba 100644 (file)
--- a/test/CodeGen/X86/memcmp-optsize.ll
+++ b/test/CodeGen/X86/memcmp-optsize.ll
@@ -699,6 +699,82 @@ define i1 @length16_eq_const(i8* %X) nounwind optsize {
   ret i1 %c
 }
 
+; PR33914 - https://bugs.llvm.org/show_bug.cgi?id=33914
+
+define i32 @length24(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length24:
+; X86:       # BB#0:
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl $24
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    retl
+;
+; X64-LABEL: length24:
+; X64:       # BB#0:
+; X64-NEXT:    movl $24, %edx
+; X64-NEXT:    jmp memcmp # TAILCALL
+  %m = tail call i32 @memcmp(i8* %X, i8* %Y, i64 24) nounwind
+  ret i32 %m
+}
+
+define i1 @length24_eq(i8* %x, i8* %y) nounwind optsize {
+; X86-LABEL: length24_eq:
+; X86:       # BB#0:
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl $24
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    testl %eax, %eax
+; X86-NEXT:    sete %al
+; X86-NEXT:    retl
+;
+; X64-LABEL: length24_eq:
+; X64:       # BB#0:
+; X64-NEXT:    pushq %rax
+; X64-NEXT:    movl $24, %edx
+; X64-NEXT:    callq memcmp
+; X64-NEXT:    testl %eax, %eax
+; X64-NEXT:    sete %al
+; X64-NEXT:    popq %rcx
+; X64-NEXT:    retq
+  %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 24) nounwind
+  %cmp = icmp eq i32 %call, 0
+  ret i1 %cmp
+}
+
+define i1 @length24_eq_const(i8* %X) nounwind optsize {
+; X86-LABEL: length24_eq_const:
+; X86:       # BB#0:
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl $24
+; X86-NEXT:    pushl $.L.str
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    testl %eax, %eax
+; X86-NEXT:    setne %al
+; X86-NEXT:    retl
+;
+; X64-LABEL: length24_eq_const:
+; X64:       # BB#0:
+; X64-NEXT:    pushq %rax
+; X64-NEXT:    movl $.L.str, %esi
+; X64-NEXT:    movl $24, %edx
+; X64-NEXT:    callq memcmp
+; X64-NEXT:    testl %eax, %eax
+; X64-NEXT:    setne %al
+; X64-NEXT:    popq %rcx
+; X64-NEXT:    retq
+  %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i64 24) nounwind
+  %c = icmp ne i32 %m, 0
+  ret i1 %c
+}
+
 define i32 @length32(i8* %X, i8* %Y) nounwind optsize {
 ; X86-LABEL: length32:
 ; X86:       # BB#0:

diff --git a/test/CodeGen/X86/memcmp.ll b/test/CodeGen/X86/memcmp.ll
index 2e67827654624c9862061fde99bfdacefb921240..889f6a74bf7fe672fadac9fb2fbdfddbadc4478a 100644 (file)
--- a/test/CodeGen/X86/memcmp.ll
+++ b/test/CodeGen/X86/memcmp.ll
@@ -475,25 +475,14 @@ define i1 @length8_eq_const(i8* %X) nounwind {
 
 define i1 @length12_eq(i8* %X, i8* %Y) nounwind {
 ; X86-LABEL: length12_eq:
-; X86:       # BB#0: # %loadbb
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    movl (%ecx), %edx
-; X86-NEXT:    cmpl (%eax), %edx
-; X86-NEXT:    jne .LBB14_1
-; X86-NEXT:  # BB#2: # %loadbb1
-; X86-NEXT:    movl 4(%ecx), %edx
-; X86-NEXT:    cmpl 4(%eax), %edx
-; X86-NEXT:    jne .LBB14_1
-; X86-NEXT:  # BB#3: # %loadbb2
-; X86-NEXT:    movl 8(%ecx), %edx
-; X86-NEXT:    xorl %ecx, %ecx
-; X86-NEXT:    cmpl 8(%eax), %edx
-; X86-NEXT:    je .LBB14_4
-; X86-NEXT:  .LBB14_1: # %res_block
-; X86-NEXT:    movl $1, %ecx
-; X86-NEXT:  .LBB14_4: # %endblock
-; X86-NEXT:    testl %ecx, %ecx
+; X86:       # BB#0:
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl $12
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    testl %eax, %eax
 ; X86-NEXT:    setne %al
 ; X86-NEXT:    retl
 ;
@@ -520,40 +509,13 @@ define i1 @length12_eq(i8* %X, i8* %Y) nounwind {
 
 define i32 @length12(i8* %X, i8* %Y) nounwind {
 ; X86-LABEL: length12:
-; X86:       # BB#0: # %loadbb
-; X86-NEXT:    pushl %esi
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X86-NEXT:    movl (%esi), %ecx
-; X86-NEXT:    movl (%eax), %edx
-; X86-NEXT:    bswapl %ecx
-; X86-NEXT:    bswapl %edx
-; X86-NEXT:    cmpl %edx, %ecx
-; X86-NEXT:    jne .LBB15_1
-; X86-NEXT:  # BB#2: # %loadbb1
-; X86-NEXT:    movl 4(%esi), %ecx
-; X86-NEXT:    movl 4(%eax), %edx
-; X86-NEXT:    bswapl %ecx
-; X86-NEXT:    bswapl %edx
-; X86-NEXT:    cmpl %edx, %ecx
-; X86-NEXT:    jne .LBB15_1
-; X86-NEXT:  # BB#3: # %loadbb2
-; X86-NEXT:    movl 8(%esi), %ecx
-; X86-NEXT:    movl 8(%eax), %edx
-; X86-NEXT:    bswapl %ecx
-; X86-NEXT:    bswapl %edx
-; X86-NEXT:    xorl %eax, %eax
-; X86-NEXT:    cmpl %edx, %ecx
-; X86-NEXT:    jne .LBB15_1
-; X86-NEXT:  # BB#4: # %endblock
-; X86-NEXT:    popl %esi
-; X86-NEXT:    retl
-; X86-NEXT:  .LBB15_1: # %res_block
-; X86-NEXT:    cmpl %edx, %ecx
-; X86-NEXT:    movl $-1, %ecx
-; X86-NEXT:    movl $1, %eax
-; X86-NEXT:    cmovbl %ecx, %eax
-; X86-NEXT:    popl %esi
+; X86:       # BB#0:
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl $12
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    addl $16, %esp
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: length12:
@@ -588,47 +550,13 @@ define i32 @length12(i8* %X, i8* %Y) nounwind {
 
 define i32 @length16(i8* %X, i8* %Y) nounwind {
 ; X86-LABEL: length16:
-; X86:       # BB#0: # %loadbb
-; X86-NEXT:    pushl %esi
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X86-NEXT:    movl (%esi), %ecx
-; X86-NEXT:    movl (%eax), %edx
-; X86-NEXT:    bswapl %ecx
-; X86-NEXT:    bswapl %edx
-; X86-NEXT:    cmpl %edx, %ecx
-; X86-NEXT:    jne .LBB16_1
-; X86-NEXT:  # BB#2: # %loadbb1
-; X86-NEXT:    movl 4(%esi), %ecx
-; X86-NEXT:    movl 4(%eax), %edx
-; X86-NEXT:    bswapl %ecx
-; X86-NEXT:    bswapl %edx
-; X86-NEXT:    cmpl %edx, %ecx
-; X86-NEXT:    jne .LBB16_1
-; X86-NEXT:  # BB#3: # %loadbb2
-; X86-NEXT:    movl 8(%esi), %ecx
-; X86-NEXT:    movl 8(%eax), %edx
-; X86-NEXT:    bswapl %ecx
-; X86-NEXT:    bswapl %edx
-; X86-NEXT:    cmpl %edx, %ecx
-; X86-NEXT:    jne .LBB16_1
-; X86-NEXT:  # BB#4: # %loadbb3
-; X86-NEXT:    movl 12(%esi), %ecx
-; X86-NEXT:    movl 12(%eax), %edx
-; X86-NEXT:    bswapl %ecx
-; X86-NEXT:    bswapl %edx
-; X86-NEXT:    xorl %eax, %eax
-; X86-NEXT:    cmpl %edx, %ecx
-; X86-NEXT:    jne .LBB16_1
-; X86-NEXT:  # BB#5: # %endblock
-; X86-NEXT:    popl %esi
-; X86-NEXT:    retl
-; X86-NEXT:  .LBB16_1: # %res_block
-; X86-NEXT:    cmpl %edx, %ecx
-; X86-NEXT:    movl $-1, %ecx
-; X86-NEXT:    movl $1, %eax
-; X86-NEXT:    cmovbl %ecx, %eax
-; X86-NEXT:    popl %esi
+; X86:       # BB#0:
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl $16
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    addl $16, %esp
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: length16:
@@ -660,32 +588,29 @@ define i32 @length16(i8* %X, i8* %Y) nounwind {
 }
 
 define i1 @length16_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length16_eq:
-; X86:       # BB#0: # %loadbb
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    movl (%ecx), %edx
-; X86-NEXT:    cmpl (%eax), %edx
-; X86-NEXT:    jne .LBB17_1
-; X86-NEXT:  # BB#2: # %loadbb1
-; X86-NEXT:    movl 4(%ecx), %edx
-; X86-NEXT:    cmpl 4(%eax), %edx
-; X86-NEXT:    jne .LBB17_1
-; X86-NEXT:  # BB#3: # %loadbb2
-; X86-NEXT:    movl 8(%ecx), %edx
-; X86-NEXT:    cmpl 8(%eax), %edx
-; X86-NEXT:    jne .LBB17_1
-; X86-NEXT:  # BB#4: # %loadbb3
-; X86-NEXT:    movl 12(%ecx), %edx
-; X86-NEXT:    xorl %ecx, %ecx
-; X86-NEXT:    cmpl 12(%eax), %edx
-; X86-NEXT:    je .LBB17_5
-; X86-NEXT:  .LBB17_1: # %res_block
-; X86-NEXT:    movl $1, %ecx
-; X86-NEXT:  .LBB17_5: # %endblock
-; X86-NEXT:    testl %ecx, %ecx
-; X86-NEXT:    setne %al
-; X86-NEXT:    retl
+; X86-NOSSE-LABEL: length16_eq:
+; X86-NOSSE:       # BB#0:
+; X86-NOSSE-NEXT:    pushl $0
+; X86-NOSSE-NEXT:    pushl $16
+; X86-NOSSE-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT:    calll memcmp
+; X86-NOSSE-NEXT:    addl $16, %esp
+; X86-NOSSE-NEXT:    testl %eax, %eax
+; X86-NOSSE-NEXT:    setne %al
+; X86-NOSSE-NEXT:    retl
+;
+; X86-SSE2-LABEL: length16_eq:
+; X86-SSE2:       # BB#0:
+; X86-SSE2-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE2-NEXT:    movdqu (%ecx), %xmm0
+; X86-SSE2-NEXT:    movdqu (%eax), %xmm1
+; X86-SSE2-NEXT:    pcmpeqb %xmm0, %xmm1
+; X86-SSE2-NEXT:    pmovmskb %xmm1, %eax
+; X86-SSE2-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT:    setne %al
+; X86-SSE2-NEXT:    retl
 ;
 ; X64-LABEL: length16_eq:
 ; X64:       # BB#0: # %loadbb
@@ -709,27 +634,27 @@ define i1 @length16_eq(i8* %x, i8* %y) nounwind {
 }
 
 define i1 @length16_eq_const(i8* %X) nounwind {
-; X86-LABEL: length16_eq_const:
-; X86:       # BB#0: # %loadbb
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    cmpl $858927408, (%eax) # imm = 0x33323130
-; X86-NEXT:    jne .LBB18_1
-; X86-NEXT:  # BB#2: # %loadbb1
-; X86-NEXT:    cmpl $926299444, 4(%eax) # imm = 0x37363534
-; X86-NEXT:    jne .LBB18_1
-; X86-NEXT:  # BB#3: # %loadbb2
-; X86-NEXT:    cmpl $825243960, 8(%eax) # imm = 0x31303938
-; X86-NEXT:    jne .LBB18_1
-; X86-NEXT:  # BB#4: # %loadbb3
-; X86-NEXT:    xorl %ecx, %ecx
-; X86-NEXT:    cmpl $892613426, 12(%eax) # imm = 0x35343332
-; X86-NEXT:    je .LBB18_5
-; X86-NEXT:  .LBB18_1: # %res_block
-; X86-NEXT:    movl $1, %ecx
-; X86-NEXT:  .LBB18_5: # %endblock
-; X86-NEXT:    testl %ecx, %ecx
-; X86-NEXT:    sete %al
-; X86-NEXT:    retl
+; X86-NOSSE-LABEL: length16_eq_const:
+; X86-NOSSE:       # BB#0:
+; X86-NOSSE-NEXT:    pushl $0
+; X86-NOSSE-NEXT:    pushl $16
+; X86-NOSSE-NEXT:    pushl $.L.str
+; X86-NOSSE-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT:    calll memcmp
+; X86-NOSSE-NEXT:    addl $16, %esp
+; X86-NOSSE-NEXT:    testl %eax, %eax
+; X86-NOSSE-NEXT:    sete %al
+; X86-NOSSE-NEXT:    retl
+;
+; X86-SSE2-LABEL: length16_eq_const:
+; X86-SSE2:       # BB#0:
+; X86-SSE2-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT:    movdqu (%eax), %xmm0
+; X86-SSE2-NEXT:    pcmpeqb {{\.LCPI.*}}, %xmm0
+; X86-SSE2-NEXT:    pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT:    sete %al
+; X86-SSE2-NEXT:    retl
 ;
 ; X64-LABEL: length16_eq_const:
 ; X64:       # BB#0: # %loadbb
@@ -752,6 +677,82 @@ define i1 @length16_eq_const(i8* %X) nounwind {
   ret i1 %c
 }
 
+; PR33914 - https://bugs.llvm.org/show_bug.cgi?id=33914
+
+define i32 @length24(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length24:
+; X86:       # BB#0:
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl $24
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    retl
+;
+; X64-LABEL: length24:
+; X64:       # BB#0:
+; X64-NEXT:    movl $24, %edx
+; X64-NEXT:    jmp memcmp # TAILCALL
+  %m = tail call i32 @memcmp(i8* %X, i8* %Y, i64 24) nounwind
+  ret i32 %m
+}
+
+define i1 @length24_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length24_eq:
+; X86:       # BB#0:
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl $24
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    testl %eax, %eax
+; X86-NEXT:    sete %al
+; X86-NEXT:    retl
+;
+; X64-LABEL: length24_eq:
+; X64:       # BB#0:
+; X64-NEXT:    pushq %rax
+; X64-NEXT:    movl $24, %edx
+; X64-NEXT:    callq memcmp
+; X64-NEXT:    testl %eax, %eax
+; X64-NEXT:    sete %al
+; X64-NEXT:    popq %rcx
+; X64-NEXT:    retq
+  %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 24) nounwind
+  %cmp = icmp eq i32 %call, 0
+  ret i1 %cmp
+}
+
+define i1 @length24_eq_const(i8* %X) nounwind {
+; X86-LABEL: length24_eq_const:
+; X86:       # BB#0:
+; X86-NEXT:    pushl $0
+; X86-NEXT:    pushl $24
+; X86-NEXT:    pushl $.L.str
+; X86-NEXT:    pushl {{[0-9]+}}(%esp)
+; X86-NEXT:    calll memcmp
+; X86-NEXT:    addl $16, %esp
+; X86-NEXT:    testl %eax, %eax
+; X86-NEXT:    setne %al
+; X86-NEXT:    retl
+;
+; X64-LABEL: length24_eq_const:
+; X64:       # BB#0:
+; X64-NEXT:    pushq %rax
+; X64-NEXT:    movl $.L.str, %esi
+; X64-NEXT:    movl $24, %edx
+; X64-NEXT:    callq memcmp
+; X64-NEXT:    testl %eax, %eax
+; X64-NEXT:    setne %al
+; X64-NEXT:    popq %rcx
+; X64-NEXT:    retq
+  %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i64 24) nounwind
+  %c = icmp ne i32 %m, 0
+  ret i1 %c
+}
+
 define i32 @length32(i8* %X, i8* %Y) nounwind {
 ; X86-LABEL: length32:
 ; X86:       # BB#0:
@@ -764,43 +765,9 @@ define i32 @length32(i8* %X, i8* %Y) nounwind {
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: length32:
-; X64:       # BB#0: # %loadbb
-; X64-NEXT:    movq (%rdi), %rcx
-; X64-NEXT:    movq (%rsi), %rdx
-; X64-NEXT:    bswapq %rcx
-; X64-NEXT:    bswapq %rdx
-; X64-NEXT:    cmpq %rdx, %rcx
-; X64-NEXT:    jne .LBB19_1
-; X64-NEXT:  # BB#2: # %loadbb1
-; X64-NEXT:    movq 8(%rdi), %rcx
-; X64-NEXT:    movq 8(%rsi), %rdx
-; X64-NEXT:    bswapq %rcx
-; X64-NEXT:    bswapq %rdx
-; X64-NEXT:    cmpq %rdx, %rcx
-; X64-NEXT:    jne .LBB19_1
-; X64-NEXT:  # BB#3: # %loadbb2
-; X64-NEXT:    movq 16(%rdi), %rcx
-; X64-NEXT:    movq 16(%rsi), %rdx
-; X64-NEXT:    bswapq %rcx
-; X64-NEXT:    bswapq %rdx
-; X64-NEXT:    cmpq %rdx, %rcx
-; X64-NEXT:    jne .LBB19_1
-; X64-NEXT:  # BB#4: # %loadbb3
-; X64-NEXT:    movq 24(%rdi), %rcx
-; X64-NEXT:    movq 24(%rsi), %rdx
-; X64-NEXT:    bswapq %rcx
-; X64-NEXT:    bswapq %rdx
-; X64-NEXT:    xorl %eax, %eax
-; X64-NEXT:    cmpq %rdx, %rcx
-; X64-NEXT:    jne .LBB19_1
-; X64-NEXT:  # BB#5: # %endblock
-; X64-NEXT:    retq
-; X64-NEXT:  .LBB19_1: # %res_block
-; X64-NEXT:    cmpq %rdx, %rcx
-; X64-NEXT:    movl $-1, %ecx
-; X64-NEXT:    movl $1, %eax
-; X64-NEXT:    cmovbl %ecx, %eax
-; X64-NEXT:    retq
+; X64:       # BB#0:
+; X64-NEXT:    movl $32, %edx
+; X64-NEXT:    jmp memcmp # TAILCALL
   %m = tail call i32 @memcmp(i8* %X, i8* %Y, i64 32) nounwind
   ret i32 %m
 }
@@ -820,30 +787,25 @@ define i1 @length32_eq(i8* %x, i8* %y) nounwind {
 ; X86-NEXT:    sete %al
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: length32_eq:
-; X64:       # BB#0: # %loadbb
-; X64-NEXT:    movq (%rdi), %rax
-; X64-NEXT:    cmpq (%rsi), %rax
-; X64-NEXT:    jne .LBB20_1
-; X64-NEXT:  # BB#2: # %loadbb1
-; X64-NEXT:    movq 8(%rdi), %rax
-; X64-NEXT:    cmpq 8(%rsi), %rax
-; X64-NEXT:    jne .LBB20_1
-; X64-NEXT:  # BB#3: # %loadbb2
-; X64-NEXT:    movq 16(%rdi), %rax
-; X64-NEXT:    cmpq 16(%rsi), %rax
-; X64-NEXT:    jne .LBB20_1
-; X64-NEXT:  # BB#4: # %loadbb3
-; X64-NEXT:    movq 24(%rdi), %rcx
-; X64-NEXT:    xorl %eax, %eax
-; X64-NEXT:    cmpq 24(%rsi), %rcx
-; X64-NEXT:    je .LBB20_5
-; X64-NEXT:  .LBB20_1: # %res_block
-; X64-NEXT:    movl $1, %eax
-; X64-NEXT:  .LBB20_5: # %endblock
-; X64-NEXT:    testl %eax, %eax
-; X64-NEXT:    sete %al
-; X64-NEXT:    retq
+; X64-SSE2-LABEL: length32_eq:
+; X64-SSE2:       # BB#0:
+; X64-SSE2-NEXT:    pushq %rax
+; X64-SSE2-NEXT:    movl $32, %edx
+; X64-SSE2-NEXT:    callq memcmp
+; X64-SSE2-NEXT:    testl %eax, %eax
+; X64-SSE2-NEXT:    sete %al
+; X64-SSE2-NEXT:    popq %rcx
+; X64-SSE2-NEXT:    retq
+;
+; X64-AVX2-LABEL: length32_eq:
+; X64-AVX2:       # BB#0:
+; X64-AVX2-NEXT:    vmovdqu (%rdi), %ymm0
+; X64-AVX2-NEXT:    vpcmpeqb (%rsi), %ymm0, %ymm0
+; X64-AVX2-NEXT:    vpmovmskb %ymm0, %eax
+; X64-AVX2-NEXT:    cmpl $-1, %eax
+; X64-AVX2-NEXT:    sete %al
+; X64-AVX2-NEXT:    vzeroupper
+; X64-AVX2-NEXT:    retq
   %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 32) nounwind
   %cmp = icmp eq i32 %call, 0
   ret i1 %cmp
@@ -862,30 +824,26 @@ define i1 @length32_eq_const(i8* %X) nounwind {
 ; X86-NEXT:    setne %al
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: length32_eq_const:
-; X64:       # BB#0: # %loadbb
-; X64-NEXT:    movabsq $3978425819141910832, %rax # imm = 0x3736353433323130
-; X64-NEXT:    cmpq %rax, (%rdi)
-; X64-NEXT:    jne .LBB21_1
-; X64-NEXT:  # BB#2: # %loadbb1
-; X64-NEXT:    movabsq $3833745473465760056, %rax # imm = 0x3534333231303938
-; X64-NEXT:    cmpq %rax, 8(%rdi)
-; X64-NEXT:    jne .LBB21_1
-; X64-NEXT:  # BB#3: # %loadbb2
-; X64-NEXT:    movabsq $3689065127958034230, %rax # imm = 0x3332313039383736
-; X64-NEXT:    cmpq %rax, 16(%rdi)
-; X64-NEXT:    jne .LBB21_1
-; X64-NEXT:  # BB#4: # %loadbb3
-; X64-NEXT:    xorl %eax, %eax
-; X64-NEXT:    movabsq $3544395820347831604, %rcx # imm = 0x3130393837363534
-; X64-NEXT:    cmpq %rcx, 24(%rdi)
-; X64-NEXT:    je .LBB21_5
-; X64-NEXT:  .LBB21_1: # %res_block
-; X64-NEXT:    movl $1, %eax
-; X64-NEXT:  .LBB21_5: # %endblock
-; X64-NEXT:    testl %eax, %eax
-; X64-NEXT:    setne %al
-; X64-NEXT:    retq
+; X64-SSE2-LABEL: length32_eq_const:
+; X64-SSE2:       # BB#0:
+; X64-SSE2-NEXT:    pushq %rax
+; X64-SSE2-NEXT:    movl $.L.str, %esi
+; X64-SSE2-NEXT:    movl $32, %edx
+; X64-SSE2-NEXT:    callq memcmp
+; X64-SSE2-NEXT:    testl %eax, %eax
+; X64-SSE2-NEXT:    setne %al
+; X64-SSE2-NEXT:    popq %rcx
+; X64-SSE2-NEXT:    retq
+;
+; X64-AVX2-LABEL: length32_eq_const:
+; X64-AVX2:       # BB#0:
+; X64-AVX2-NEXT:    vmovdqu (%rdi), %ymm0
+; X64-AVX2-NEXT:    vpcmpeqb {{.*}}(%rip), %ymm0, %ymm0
+; X64-AVX2-NEXT:    vpmovmskb %ymm0, %eax
+; X64-AVX2-NEXT:    cmpl $-1, %eax
+; X64-AVX2-NEXT:    setne %al
+; X64-AVX2-NEXT:    vzeroupper
+; X64-AVX2-NEXT:    retq
   %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i64 32) nounwind
   %c = icmp ne i32 %m, 0
   ret i1 %c

diff --git a/test/Transforms/CodeGenPrepare/X86/memcmp.ll b/test/Transforms/CodeGenPrepare/X86/memcmp.ll
index 1dfc087619653d7f5c74cc857f4b56e6962b08dd..c5281a9e5733b58a0d48667a89a5a440d2c9de76 100644 (file)
--- a/test/Transforms/CodeGenPrepare/X86/memcmp.ll
+++ b/test/Transforms/CodeGenPrepare/X86/memcmp.ll
@@ -238,91 +238,9 @@ define i32 @cmp6(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 }
 
 define i32 @cmp7(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
-; X32-LABEL: @cmp7(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP2]])
-; X32-NEXT:    [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
-; X32-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[TMP4]], [[TMP5]]
-; X32-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X32:       res_block:
-; X32-NEXT:    [[PHI_SRC1:%.*]] = phi i32 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP17:%.*]], [[LOADBB1]] ]
-; X32-NEXT:    [[PHI_SRC2:%.*]] = phi i32 [ [[TMP5]], [[LOADBB]] ], [ [[TMP18:%.*]], [[LOADBB1]] ]
-; X32-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
-; X32-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i16*
-; X32-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i16*
-; X32-NEXT:    [[TMP11:%.*]] = getelementptr i16, i16* [[TMP9]], i16 2
-; X32-NEXT:    [[TMP12:%.*]] = getelementptr i16, i16* [[TMP10]], i16 2
-; X32-NEXT:    [[TMP13:%.*]] = load i16, i16* [[TMP11]]
-; X32-NEXT:    [[TMP14:%.*]] = load i16, i16* [[TMP12]]
-; X32-NEXT:    [[TMP15:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP13]])
-; X32-NEXT:    [[TMP16:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP14]])
-; X32-NEXT:    [[TMP17]] = zext i16 [[TMP15]] to i32
-; X32-NEXT:    [[TMP18]] = zext i16 [[TMP16]] to i32
-; X32-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[TMP17]], [[TMP18]]
-; X32-NEXT:    br i1 [[TMP19]], label [[LOADBB2:%.*]], label [[RES_BLOCK]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP20:%.*]] = getelementptr i8, i8* [[X]], i8 6
-; X32-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[Y]], i8 6
-; X32-NEXT:    [[TMP22:%.*]] = load i8, i8* [[TMP20]]
-; X32-NEXT:    [[TMP23:%.*]] = load i8, i8* [[TMP21]]
-; X32-NEXT:    [[TMP24:%.*]] = zext i8 [[TMP22]] to i32
-; X32-NEXT:    [[TMP25:%.*]] = zext i8 [[TMP23]] to i32
-; X32-NEXT:    [[TMP26:%.*]] = sub i32 [[TMP24]], [[TMP25]]
-; X32-NEXT:    br label [[ENDBLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ [[TMP26]], [[LOADBB2]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X32-NEXT:    ret i32 [[PHI_RES]]
-;
-; X64-LABEL: @cmp7(
-; X64-NEXT:  loadbb:
-; X64-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X64-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X64-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X64-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X64-NEXT:    [[TMP4:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP2]])
-; X64-NEXT:    [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
-; X64-NEXT:    [[TMP6:%.*]] = zext i32 [[TMP4]] to i64
-; X64-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP5]] to i64
-; X64-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[TMP6]], [[TMP7]]
-; X64-NEXT:    br i1 [[TMP8]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X64:       res_block:
-; X64-NEXT:    [[PHI_SRC1:%.*]] = phi i64 [ [[TMP6]], [[LOADBB:%.*]] ], [ [[TMP19:%.*]], [[LOADBB1]] ]
-; X64-NEXT:    [[PHI_SRC2:%.*]] = phi i64 [ [[TMP7]], [[LOADBB]] ], [ [[TMP20:%.*]], [[LOADBB1]] ]
-; X64-NEXT:    [[TMP9:%.*]] = icmp ult i64 [[PHI_SRC1]], [[PHI_SRC2]]
-; X64-NEXT:    [[TMP10:%.*]] = select i1 [[TMP9]], i32 -1, i32 1
-; X64-NEXT:    br label [[ENDBLOCK:%.*]]
-; X64:       loadbb1:
-; X64-NEXT:    [[TMP11:%.*]] = bitcast i8* [[X]] to i16*
-; X64-NEXT:    [[TMP12:%.*]] = bitcast i8* [[Y]] to i16*
-; X64-NEXT:    [[TMP13:%.*]] = getelementptr i16, i16* [[TMP11]], i16 2
-; X64-NEXT:    [[TMP14:%.*]] = getelementptr i16, i16* [[TMP12]], i16 2
-; X64-NEXT:    [[TMP15:%.*]] = load i16, i16* [[TMP13]]
-; X64-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP14]]
-; X64-NEXT:    [[TMP17:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP15]])
-; X64-NEXT:    [[TMP18:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP16]])
-; X64-NEXT:    [[TMP19]] = zext i16 [[TMP17]] to i64
-; X64-NEXT:    [[TMP20]] = zext i16 [[TMP18]] to i64
-; X64-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[TMP19]], [[TMP20]]
-; X64-NEXT:    br i1 [[TMP21]], label [[LOADBB2:%.*]], label [[RES_BLOCK]]
-; X64:       loadbb2:
-; X64-NEXT:    [[TMP22:%.*]] = getelementptr i8, i8* [[X]], i8 6
-; X64-NEXT:    [[TMP23:%.*]] = getelementptr i8, i8* [[Y]], i8 6
-; X64-NEXT:    [[TMP24:%.*]] = load i8, i8* [[TMP22]]
-; X64-NEXT:    [[TMP25:%.*]] = load i8, i8* [[TMP23]]
-; X64-NEXT:    [[TMP26:%.*]] = zext i8 [[TMP24]] to i32
-; X64-NEXT:    [[TMP27:%.*]] = zext i8 [[TMP25]] to i32
-; X64-NEXT:    [[TMP28:%.*]] = sub i32 [[TMP26]], [[TMP27]]
-; X64-NEXT:    br label [[ENDBLOCK]]
-; X64:       endblock:
-; X64-NEXT:    [[PHI_RES:%.*]] = phi i32 [ [[TMP28]], [[LOADBB2]] ], [ [[TMP10]], [[RES_BLOCK]] ]
-; X64-NEXT:    ret i32 [[PHI_RES]]
+; ALL-LABEL: @cmp7(
+; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 7)
+; ALL-NEXT:    ret i32 [[CALL]]
 ;
   %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 7)
   ret i32 %call
@@ -379,44 +297,8 @@ define i32 @cmp8(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 
 define i32 @cmp9(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 ; X32-LABEL: @cmp9(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP2]])
-; X32-NEXT:    [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
-; X32-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[TMP4]], [[TMP5]]
-; X32-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X32:       res_block:
-; X32-NEXT:    [[PHI_SRC1:%.*]] = phi i32 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP15:%.*]], [[LOADBB1]] ]
-; X32-NEXT:    [[PHI_SRC2:%.*]] = phi i32 [ [[TMP5]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1]] ]
-; X32-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
-; X32-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP9]], i32 1
-; X32-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP10]], i32 1
-; X32-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP11]]
-; X32-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP12]]
-; X32-NEXT:    [[TMP15]] = call i32 @llvm.bswap.i32(i32 [[TMP13]])
-; X32-NEXT:    [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; X32-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP15]], [[TMP16]]
-; X32-NEXT:    br i1 [[TMP17]], label [[LOADBB2:%.*]], label [[RES_BLOCK]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP18:%.*]] = getelementptr i8, i8* [[X]], i8 8
-; X32-NEXT:    [[TMP19:%.*]] = getelementptr i8, i8* [[Y]], i8 8
-; X32-NEXT:    [[TMP20:%.*]] = load i8, i8* [[TMP18]]
-; X32-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP19]]
-; X32-NEXT:    [[TMP22:%.*]] = zext i8 [[TMP20]] to i32
-; X32-NEXT:    [[TMP23:%.*]] = zext i8 [[TMP21]] to i32
-; X32-NEXT:    [[TMP24:%.*]] = sub i32 [[TMP22]], [[TMP23]]
-; X32-NEXT:    br label [[ENDBLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ [[TMP24]], [[LOADBB2]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X32-NEXT:    ret i32 [[PHI_RES]]
+; X32-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 9)
+; X32-NEXT:    ret i32 [[CALL]]
 ;
 ; X64-LABEL: @cmp9(
 ; X64-NEXT:  loadbb:
@@ -451,48 +333,8 @@ define i32 @cmp9(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 
 define i32 @cmp10(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 ; X32-LABEL: @cmp10(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP2]])
-; X32-NEXT:    [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
-; X32-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[TMP4]], [[TMP5]]
-; X32-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X32:       res_block:
-; X32-NEXT:    [[PHI_SRC1:%.*]] = phi i32 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP15:%.*]], [[LOADBB1]] ], [ [[TMP26:%.*]], [[LOADBB2:%.*]] ]
-; X32-NEXT:    [[PHI_SRC2:%.*]] = phi i32 [ [[TMP5]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1]] ], [ [[TMP27:%.*]], [[LOADBB2]] ]
-; X32-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
-; X32-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP9]], i32 1
-; X32-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP10]], i32 1
-; X32-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP11]]
-; X32-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP12]]
-; X32-NEXT:    [[TMP15]] = call i32 @llvm.bswap.i32(i32 [[TMP13]])
-; X32-NEXT:    [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; X32-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP15]], [[TMP16]]
-; X32-NEXT:    br i1 [[TMP17]], label [[LOADBB2]], label [[RES_BLOCK]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[X]] to i16*
-; X32-NEXT:    [[TMP19:%.*]] = bitcast i8* [[Y]] to i16*
-; X32-NEXT:    [[TMP20:%.*]] = getelementptr i16, i16* [[TMP18]], i16 4
-; X32-NEXT:    [[TMP21:%.*]] = getelementptr i16, i16* [[TMP19]], i16 4
-; X32-NEXT:    [[TMP22:%.*]] = load i16, i16* [[TMP20]]
-; X32-NEXT:    [[TMP23:%.*]] = load i16, i16* [[TMP21]]
-; X32-NEXT:    [[TMP24:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP22]])
-; X32-NEXT:    [[TMP25:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP23]])
-; X32-NEXT:    [[TMP26]] = zext i16 [[TMP24]] to i32
-; X32-NEXT:    [[TMP27]] = zext i16 [[TMP25]] to i32
-; X32-NEXT:    [[TMP28:%.*]] = icmp eq i32 [[TMP26]], [[TMP27]]
-; X32-NEXT:    br i1 [[TMP28]], label [[ENDBLOCK]], label [[RES_BLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB2]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X32-NEXT:    ret i32 [[PHI_RES]]
+; X32-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 10)
+; X32-NEXT:    ret i32 [[CALL]]
 ;
 ; X64-LABEL: @cmp10(
 ; X64-NEXT:  loadbb:
@@ -532,100 +374,9 @@ define i32 @cmp10(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 }
 
 define i32 @cmp11(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
-; X32-LABEL: @cmp11(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP2]])
-; X32-NEXT:    [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
-; X32-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[TMP4]], [[TMP5]]
-; X32-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X32:       res_block:
-; X32-NEXT:    [[PHI_SRC1:%.*]] = phi i32 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP15:%.*]], [[LOADBB1]] ], [ [[TMP26:%.*]], [[LOADBB2:%.*]] ]
-; X32-NEXT:    [[PHI_SRC2:%.*]] = phi i32 [ [[TMP5]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1]] ], [ [[TMP27:%.*]], [[LOADBB2]] ]
-; X32-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
-; X32-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP9]], i32 1
-; X32-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP10]], i32 1
-; X32-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP11]]
-; X32-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP12]]
-; X32-NEXT:    [[TMP15]] = call i32 @llvm.bswap.i32(i32 [[TMP13]])
-; X32-NEXT:    [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; X32-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP15]], [[TMP16]]
-; X32-NEXT:    br i1 [[TMP17]], label [[LOADBB2]], label [[RES_BLOCK]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[X]] to i16*
-; X32-NEXT:    [[TMP19:%.*]] = bitcast i8* [[Y]] to i16*
-; X32-NEXT:    [[TMP20:%.*]] = getelementptr i16, i16* [[TMP18]], i16 4
-; X32-NEXT:    [[TMP21:%.*]] = getelementptr i16, i16* [[TMP19]], i16 4
-; X32-NEXT:    [[TMP22:%.*]] = load i16, i16* [[TMP20]]
-; X32-NEXT:    [[TMP23:%.*]] = load i16, i16* [[TMP21]]
-; X32-NEXT:    [[TMP24:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP22]])
-; X32-NEXT:    [[TMP25:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP23]])
-; X32-NEXT:    [[TMP26]] = zext i16 [[TMP24]] to i32
-; X32-NEXT:    [[TMP27]] = zext i16 [[TMP25]] to i32
-; X32-NEXT:    [[TMP28:%.*]] = icmp eq i32 [[TMP26]], [[TMP27]]
-; X32-NEXT:    br i1 [[TMP28]], label [[LOADBB3:%.*]], label [[RES_BLOCK]]
-; X32:       loadbb3:
-; X32-NEXT:    [[TMP29:%.*]] = getelementptr i8, i8* [[X]], i8 10
-; X32-NEXT:    [[TMP30:%.*]] = getelementptr i8, i8* [[Y]], i8 10
-; X32-NEXT:    [[TMP31:%.*]] = load i8, i8* [[TMP29]]
-; X32-NEXT:    [[TMP32:%.*]] = load i8, i8* [[TMP30]]
-; X32-NEXT:    [[TMP33:%.*]] = zext i8 [[TMP31]] to i32
-; X32-NEXT:    [[TMP34:%.*]] = zext i8 [[TMP32]] to i32
-; X32-NEXT:    [[TMP35:%.*]] = sub i32 [[TMP33]], [[TMP34]]
-; X32-NEXT:    br label [[ENDBLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ [[TMP35]], [[LOADBB3]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X32-NEXT:    ret i32 [[PHI_RES]]
-;
-; X64-LABEL: @cmp11(
-; X64-NEXT:  loadbb:
-; X64-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i64*
-; X64-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP0]]
-; X64-NEXT:    [[TMP3:%.*]] = load i64, i64* [[TMP1]]
-; X64-NEXT:    [[TMP4:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP2]])
-; X64-NEXT:    [[TMP5:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP3]])
-; X64-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP4]], [[TMP5]]
-; X64-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X64:       res_block:
-; X64-NEXT:    [[PHI_SRC1:%.*]] = phi i64 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP17:%.*]], [[LOADBB1]] ]
-; X64-NEXT:    [[PHI_SRC2:%.*]] = phi i64 [ [[TMP5]], [[LOADBB]] ], [ [[TMP18:%.*]], [[LOADBB1]] ]
-; X64-NEXT:    [[TMP7:%.*]] = icmp ult i64 [[PHI_SRC1]], [[PHI_SRC2]]
-; X64-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X64-NEXT:    br label [[ENDBLOCK:%.*]]
-; X64:       loadbb1:
-; X64-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i16*
-; X64-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i16*
-; X64-NEXT:    [[TMP11:%.*]] = getelementptr i16, i16* [[TMP9]], i16 4
-; X64-NEXT:    [[TMP12:%.*]] = getelementptr i16, i16* [[TMP10]], i16 4
-; X64-NEXT:    [[TMP13:%.*]] = load i16, i16* [[TMP11]]
-; X64-NEXT:    [[TMP14:%.*]] = load i16, i16* [[TMP12]]
-; X64-NEXT:    [[TMP15:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP13]])
-; X64-NEXT:    [[TMP16:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP14]])
-; X64-NEXT:    [[TMP17]] = zext i16 [[TMP15]] to i64
-; X64-NEXT:    [[TMP18]] = zext i16 [[TMP16]] to i64
-; X64-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[TMP17]], [[TMP18]]
-; X64-NEXT:    br i1 [[TMP19]], label [[LOADBB2:%.*]], label [[RES_BLOCK]]
-; X64:       loadbb2:
-; X64-NEXT:    [[TMP20:%.*]] = getelementptr i8, i8* [[X]], i8 10
-; X64-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[Y]], i8 10
-; X64-NEXT:    [[TMP22:%.*]] = load i8, i8* [[TMP20]]
-; X64-NEXT:    [[TMP23:%.*]] = load i8, i8* [[TMP21]]
-; X64-NEXT:    [[TMP24:%.*]] = zext i8 [[TMP22]] to i32
-; X64-NEXT:    [[TMP25:%.*]] = zext i8 [[TMP23]] to i32
-; X64-NEXT:    [[TMP26:%.*]] = sub i32 [[TMP24]], [[TMP25]]
-; X64-NEXT:    br label [[ENDBLOCK]]
-; X64:       endblock:
-; X64-NEXT:    [[PHI_RES:%.*]] = phi i32 [ [[TMP26]], [[LOADBB2]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X64-NEXT:    ret i32 [[PHI_RES]]
+; ALL-LABEL: @cmp11(
+; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 11)
+; ALL-NEXT:    ret i32 [[CALL]]
 ;
   %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 11)
   ret i32 %call
@@ -633,46 +384,8 @@ define i32 @cmp11(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 
 define i32 @cmp12(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 ; X32-LABEL: @cmp12(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP2]])
-; X32-NEXT:    [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
-; X32-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[TMP4]], [[TMP5]]
-; X32-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X32:       res_block:
-; X32-NEXT:    [[PHI_SRC1:%.*]] = phi i32 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP15:%.*]], [[LOADBB1]] ], [ [[TMP24:%.*]], [[LOADBB2:%.*]] ]
-; X32-NEXT:    [[PHI_SRC2:%.*]] = phi i32 [ [[TMP5]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1]] ], [ [[TMP25:%.*]], [[LOADBB2]] ]
-; X32-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
-; X32-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP9]], i32 1
-; X32-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP10]], i32 1
-; X32-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP11]]
-; X32-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP12]]
-; X32-NEXT:    [[TMP15]] = call i32 @llvm.bswap.i32(i32 [[TMP13]])
-; X32-NEXT:    [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; X32-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP15]], [[TMP16]]
-; X32-NEXT:    br i1 [[TMP17]], label [[LOADBB2]], label [[RES_BLOCK]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP19:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP20:%.*]] = getelementptr i32, i32* [[TMP18]], i32 2
-; X32-NEXT:    [[TMP21:%.*]] = getelementptr i32, i32* [[TMP19]], i32 2
-; X32-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP20]]
-; X32-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP21]]
-; X32-NEXT:    [[TMP24]] = call i32 @llvm.bswap.i32(i32 [[TMP22]])
-; X32-NEXT:    [[TMP25]] = call i32 @llvm.bswap.i32(i32 [[TMP23]])
-; X32-NEXT:    [[TMP26:%.*]] = icmp eq i32 [[TMP24]], [[TMP25]]
-; X32-NEXT:    br i1 [[TMP26]], label [[ENDBLOCK]], label [[RES_BLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB2]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X32-NEXT:    ret i32 [[PHI_RES]]
+; X32-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 12)
+; X32-NEXT:    ret i32 [[CALL]]
 ;
 ; X64-LABEL: @cmp12(
 ; X64-NEXT:  loadbb:
@@ -712,268 +425,27 @@ define i32 @cmp12(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 }
 
 define i32 @cmp13(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
-; X32-LABEL: @cmp13(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP2]])
-; X32-NEXT:    [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
-; X32-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[TMP4]], [[TMP5]]
-; X32-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X32:       res_block:
-; X32-NEXT:    [[PHI_SRC1:%.*]] = phi i32 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP15:%.*]], [[LOADBB1]] ], [ [[TMP24:%.*]], [[LOADBB2:%.*]] ]
-; X32-NEXT:    [[PHI_SRC2:%.*]] = phi i32 [ [[TMP5]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1]] ], [ [[TMP25:%.*]], [[LOADBB2]] ]
-; X32-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
-; X32-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP9]], i32 1
-; X32-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP10]], i32 1
-; X32-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP11]]
-; X32-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP12]]
-; X32-NEXT:    [[TMP15]] = call i32 @llvm.bswap.i32(i32 [[TMP13]])
-; X32-NEXT:    [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; X32-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP15]], [[TMP16]]
-; X32-NEXT:    br i1 [[TMP17]], label [[LOADBB2]], label [[RES_BLOCK]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP19:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP20:%.*]] = getelementptr i32, i32* [[TMP18]], i32 2
-; X32-NEXT:    [[TMP21:%.*]] = getelementptr i32, i32* [[TMP19]], i32 2
-; X32-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP20]]
-; X32-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP21]]
-; X32-NEXT:    [[TMP24]] = call i32 @llvm.bswap.i32(i32 [[TMP22]])
-; X32-NEXT:    [[TMP25]] = call i32 @llvm.bswap.i32(i32 [[TMP23]])
-; X32-NEXT:    [[TMP26:%.*]] = icmp eq i32 [[TMP24]], [[TMP25]]
-; X32-NEXT:    br i1 [[TMP26]], label [[LOADBB3:%.*]], label [[RES_BLOCK]]
-; X32:       loadbb3:
-; X32-NEXT:    [[TMP27:%.*]] = getelementptr i8, i8* [[X]], i8 12
-; X32-NEXT:    [[TMP28:%.*]] = getelementptr i8, i8* [[Y]], i8 12
-; X32-NEXT:    [[TMP29:%.*]] = load i8, i8* [[TMP27]]
-; X32-NEXT:    [[TMP30:%.*]] = load i8, i8* [[TMP28]]
-; X32-NEXT:    [[TMP31:%.*]] = zext i8 [[TMP29]] to i32
-; X32-NEXT:    [[TMP32:%.*]] = zext i8 [[TMP30]] to i32
-; X32-NEXT:    [[TMP33:%.*]] = sub i32 [[TMP31]], [[TMP32]]
-; X32-NEXT:    br label [[ENDBLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ [[TMP33]], [[LOADBB3]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X32-NEXT:    ret i32 [[PHI_RES]]
-;
-; X64-LABEL: @cmp13(
-; X64-NEXT:  loadbb:
-; X64-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i64*
-; X64-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP0]]
-; X64-NEXT:    [[TMP3:%.*]] = load i64, i64* [[TMP1]]
-; X64-NEXT:    [[TMP4:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP2]])
-; X64-NEXT:    [[TMP5:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP3]])
-; X64-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP4]], [[TMP5]]
-; X64-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X64:       res_block:
-; X64-NEXT:    [[PHI_SRC1:%.*]] = phi i64 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP17:%.*]], [[LOADBB1]] ]
-; X64-NEXT:    [[PHI_SRC2:%.*]] = phi i64 [ [[TMP5]], [[LOADBB]] ], [ [[TMP18:%.*]], [[LOADBB1]] ]
-; X64-NEXT:    [[TMP7:%.*]] = icmp ult i64 [[PHI_SRC1]], [[PHI_SRC2]]
-; X64-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X64-NEXT:    br label [[ENDBLOCK:%.*]]
-; X64:       loadbb1:
-; X64-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i32*
-; X64-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i32*
-; X64-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP9]], i32 2
-; X64-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP10]], i32 2
-; X64-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP11]]
-; X64-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP12]]
-; X64-NEXT:    [[TMP15:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP13]])
-; X64-NEXT:    [[TMP16:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; X64-NEXT:    [[TMP17]] = zext i32 [[TMP15]] to i64
-; X64-NEXT:    [[TMP18]] = zext i32 [[TMP16]] to i64
-; X64-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[TMP17]], [[TMP18]]
-; X64-NEXT:    br i1 [[TMP19]], label [[LOADBB2:%.*]], label [[RES_BLOCK]]
-; X64:       loadbb2:
-; X64-NEXT:    [[TMP20:%.*]] = getelementptr i8, i8* [[X]], i8 12
-; X64-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[Y]], i8 12
-; X64-NEXT:    [[TMP22:%.*]] = load i8, i8* [[TMP20]]
-; X64-NEXT:    [[TMP23:%.*]] = load i8, i8* [[TMP21]]
-; X64-NEXT:    [[TMP24:%.*]] = zext i8 [[TMP22]] to i32
-; X64-NEXT:    [[TMP25:%.*]] = zext i8 [[TMP23]] to i32
-; X64-NEXT:    [[TMP26:%.*]] = sub i32 [[TMP24]], [[TMP25]]
-; X64-NEXT:    br label [[ENDBLOCK]]
-; X64:       endblock:
-; X64-NEXT:    [[PHI_RES:%.*]] = phi i32 [ [[TMP26]], [[LOADBB2]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X64-NEXT:    ret i32 [[PHI_RES]]
+; ALL-LABEL: @cmp13(
+; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 13)
+; ALL-NEXT:    ret i32 [[CALL]]
 ;
   %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 13)
   ret i32 %call
 }
 
 define i32 @cmp14(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
-; X32-LABEL: @cmp14(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP2]])
-; X32-NEXT:    [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
-; X32-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[TMP4]], [[TMP5]]
-; X32-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X32:       res_block:
-; X32-NEXT:    [[PHI_SRC1:%.*]] = phi i32 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP15:%.*]], [[LOADBB1]] ], [ [[TMP24:%.*]], [[LOADBB2:%.*]] ], [ [[TMP35:%.*]], [[LOADBB3:%.*]] ]
-; X32-NEXT:    [[PHI_SRC2:%.*]] = phi i32 [ [[TMP5]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1]] ], [ [[TMP25:%.*]], [[LOADBB2]] ], [ [[TMP36:%.*]], [[LOADBB3]] ]
-; X32-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
-; X32-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP9]], i32 1
-; X32-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP10]], i32 1
-; X32-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP11]]
-; X32-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP12]]
-; X32-NEXT:    [[TMP15]] = call i32 @llvm.bswap.i32(i32 [[TMP13]])
-; X32-NEXT:    [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; X32-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP15]], [[TMP16]]
-; X32-NEXT:    br i1 [[TMP17]], label [[LOADBB2]], label [[RES_BLOCK]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP19:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP20:%.*]] = getelementptr i32, i32* [[TMP18]], i32 2
-; X32-NEXT:    [[TMP21:%.*]] = getelementptr i32, i32* [[TMP19]], i32 2
-; X32-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP20]]
-; X32-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP21]]
-; X32-NEXT:    [[TMP24]] = call i32 @llvm.bswap.i32(i32 [[TMP22]])
-; X32-NEXT:    [[TMP25]] = call i32 @llvm.bswap.i32(i32 [[TMP23]])
-; X32-NEXT:    [[TMP26:%.*]] = icmp eq i32 [[TMP24]], [[TMP25]]
-; X32-NEXT:    br i1 [[TMP26]], label [[LOADBB3]], label [[RES_BLOCK]]
-; X32:       loadbb3:
-; X32-NEXT:    [[TMP27:%.*]] = bitcast i8* [[X]] to i16*
-; X32-NEXT:    [[TMP28:%.*]] = bitcast i8* [[Y]] to i16*
-; X32-NEXT:    [[TMP29:%.*]] = getelementptr i16, i16* [[TMP27]], i16 6
-; X32-NEXT:    [[TMP30:%.*]] = getelementptr i16, i16* [[TMP28]], i16 6
-; X32-NEXT:    [[TMP31:%.*]] = load i16, i16* [[TMP29]]
-; X32-NEXT:    [[TMP32:%.*]] = load i16, i16* [[TMP30]]
-; X32-NEXT:    [[TMP33:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP31]])
-; X32-NEXT:    [[TMP34:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP32]])
-; X32-NEXT:    [[TMP35]] = zext i16 [[TMP33]] to i32
-; X32-NEXT:    [[TMP36]] = zext i16 [[TMP34]] to i32
-; X32-NEXT:    [[TMP37:%.*]] = icmp eq i32 [[TMP35]], [[TMP36]]
-; X32-NEXT:    br i1 [[TMP37]], label [[ENDBLOCK]], label [[RES_BLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB3]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X32-NEXT:    ret i32 [[PHI_RES]]
-;
-; X64-LABEL: @cmp14(
-; X64-NEXT:  loadbb:
-; X64-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i64*
-; X64-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP0]]
-; X64-NEXT:    [[TMP3:%.*]] = load i64, i64* [[TMP1]]
-; X64-NEXT:    [[TMP4:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP2]])
-; X64-NEXT:    [[TMP5:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP3]])
-; X64-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP4]], [[TMP5]]
-; X64-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X64:       res_block:
-; X64-NEXT:    [[PHI_SRC1:%.*]] = phi i64 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP17:%.*]], [[LOADBB1]] ], [ [[TMP28:%.*]], [[LOADBB2:%.*]] ]
-; X64-NEXT:    [[PHI_SRC2:%.*]] = phi i64 [ [[TMP5]], [[LOADBB]] ], [ [[TMP18:%.*]], [[LOADBB1]] ], [ [[TMP29:%.*]], [[LOADBB2]] ]
-; X64-NEXT:    [[TMP7:%.*]] = icmp ult i64 [[PHI_SRC1]], [[PHI_SRC2]]
-; X64-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X64-NEXT:    br label [[ENDBLOCK:%.*]]
-; X64:       loadbb1:
-; X64-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i32*
-; X64-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i32*
-; X64-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP9]], i32 2
-; X64-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP10]], i32 2
-; X64-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP11]]
-; X64-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP12]]
-; X64-NEXT:    [[TMP15:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP13]])
-; X64-NEXT:    [[TMP16:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; X64-NEXT:    [[TMP17]] = zext i32 [[TMP15]] to i64
-; X64-NEXT:    [[TMP18]] = zext i32 [[TMP16]] to i64
-; X64-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[TMP17]], [[TMP18]]
-; X64-NEXT:    br i1 [[TMP19]], label [[LOADBB2]], label [[RES_BLOCK]]
-; X64:       loadbb2:
-; X64-NEXT:    [[TMP20:%.*]] = bitcast i8* [[X]] to i16*
-; X64-NEXT:    [[TMP21:%.*]] = bitcast i8* [[Y]] to i16*
-; X64-NEXT:    [[TMP22:%.*]] = getelementptr i16, i16* [[TMP20]], i16 6
-; X64-NEXT:    [[TMP23:%.*]] = getelementptr i16, i16* [[TMP21]], i16 6
-; X64-NEXT:    [[TMP24:%.*]] = load i16, i16* [[TMP22]]
-; X64-NEXT:    [[TMP25:%.*]] = load i16, i16* [[TMP23]]
-; X64-NEXT:    [[TMP26:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP24]])
-; X64-NEXT:    [[TMP27:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP25]])
-; X64-NEXT:    [[TMP28]] = zext i16 [[TMP26]] to i64
-; X64-NEXT:    [[TMP29]] = zext i16 [[TMP27]] to i64
-; X64-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[TMP28]], [[TMP29]]
-; X64-NEXT:    br i1 [[TMP30]], label [[ENDBLOCK]], label [[RES_BLOCK]]
-; X64:       endblock:
-; X64-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB2]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X64-NEXT:    ret i32 [[PHI_RES]]
+; ALL-LABEL: @cmp14(
+; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 14)
+; ALL-NEXT:    ret i32 [[CALL]]
 ;
   %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 14)
   ret i32 %call
 }
 
 define i32 @cmp15(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
-; X32-LABEL: @cmp15(
-; X32-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 15)
-; X32-NEXT:    ret i32 [[CALL]]
-;
-; X64-LABEL: @cmp15(
-; X64-NEXT:  loadbb:
-; X64-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i64*
-; X64-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP0]]
-; X64-NEXT:    [[TMP3:%.*]] = load i64, i64* [[TMP1]]
-; X64-NEXT:    [[TMP4:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP2]])
-; X64-NEXT:    [[TMP5:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP3]])
-; X64-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP4]], [[TMP5]]
-; X64-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X64:       res_block:
-; X64-NEXT:    [[PHI_SRC1:%.*]] = phi i64 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP17:%.*]], [[LOADBB1]] ], [ [[TMP28:%.*]], [[LOADBB2:%.*]] ]
-; X64-NEXT:    [[PHI_SRC2:%.*]] = phi i64 [ [[TMP5]], [[LOADBB]] ], [ [[TMP18:%.*]], [[LOADBB1]] ], [ [[TMP29:%.*]], [[LOADBB2]] ]
-; X64-NEXT:    [[TMP7:%.*]] = icmp ult i64 [[PHI_SRC1]], [[PHI_SRC2]]
-; X64-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X64-NEXT:    br label [[ENDBLOCK:%.*]]
-; X64:       loadbb1:
-; X64-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i32*
-; X64-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i32*
-; X64-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP9]], i32 2
-; X64-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP10]], i32 2
-; X64-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP11]]
-; X64-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP12]]
-; X64-NEXT:    [[TMP15:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP13]])
-; X64-NEXT:    [[TMP16:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; X64-NEXT:    [[TMP17]] = zext i32 [[TMP15]] to i64
-; X64-NEXT:    [[TMP18]] = zext i32 [[TMP16]] to i64
-; X64-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[TMP17]], [[TMP18]]
-; X64-NEXT:    br i1 [[TMP19]], label [[LOADBB2]], label [[RES_BLOCK]]
-; X64:       loadbb2:
-; X64-NEXT:    [[TMP20:%.*]] = bitcast i8* [[X]] to i16*
-; X64-NEXT:    [[TMP21:%.*]] = bitcast i8* [[Y]] to i16*
-; X64-NEXT:    [[TMP22:%.*]] = getelementptr i16, i16* [[TMP20]], i16 6
-; X64-NEXT:    [[TMP23:%.*]] = getelementptr i16, i16* [[TMP21]], i16 6
-; X64-NEXT:    [[TMP24:%.*]] = load i16, i16* [[TMP22]]
-; X64-NEXT:    [[TMP25:%.*]] = load i16, i16* [[TMP23]]
-; X64-NEXT:    [[TMP26:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP24]])
-; X64-NEXT:    [[TMP27:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP25]])
-; X64-NEXT:    [[TMP28]] = zext i16 [[TMP26]] to i64
-; X64-NEXT:    [[TMP29]] = zext i16 [[TMP27]] to i64
-; X64-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[TMP28]], [[TMP29]]
-; X64-NEXT:    br i1 [[TMP30]], label [[LOADBB3:%.*]], label [[RES_BLOCK]]
-; X64:       loadbb3:
-; X64-NEXT:    [[TMP31:%.*]] = getelementptr i8, i8* [[X]], i8 14
-; X64-NEXT:    [[TMP32:%.*]] = getelementptr i8, i8* [[Y]], i8 14
-; X64-NEXT:    [[TMP33:%.*]] = load i8, i8* [[TMP31]]
-; X64-NEXT:    [[TMP34:%.*]] = load i8, i8* [[TMP32]]
-; X64-NEXT:    [[TMP35:%.*]] = zext i8 [[TMP33]] to i32
-; X64-NEXT:    [[TMP36:%.*]] = zext i8 [[TMP34]] to i32
-; X64-NEXT:    [[TMP37:%.*]] = sub i32 [[TMP35]], [[TMP36]]
-; X64-NEXT:    br label [[ENDBLOCK]]
-; X64:       endblock:
-; X64-NEXT:    [[PHI_RES:%.*]] = phi i32 [ [[TMP37]], [[LOADBB3]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X64-NEXT:    ret i32 [[PHI_RES]]
+; ALL-LABEL: @cmp15(
+; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 15)
+; ALL-NEXT:    ret i32 [[CALL]]
 ;
   %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 15)
   ret i32 %call
@@ -981,57 +453,8 @@ define i32 @cmp15(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 
 define i32 @cmp16(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 ; X32-LABEL: @cmp16(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP2]])
-; X32-NEXT:    [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
-; X32-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[TMP4]], [[TMP5]]
-; X32-NEXT:    br i1 [[TMP6]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; X32:       res_block:
-; X32-NEXT:    [[PHI_SRC1:%.*]] = phi i32 [ [[TMP4]], [[LOADBB:%.*]] ], [ [[TMP15:%.*]], [[LOADBB1]] ], [ [[TMP24:%.*]], [[LOADBB2:%.*]] ], [ [[TMP33:%.*]], [[LOADBB3:%.*]] ]
-; X32-NEXT:    [[PHI_SRC2:%.*]] = phi i32 [ [[TMP5]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1]] ], [ [[TMP25:%.*]], [[LOADBB2]] ], [ [[TMP34:%.*]], [[LOADBB3]] ]
-; X32-NEXT:    [[TMP7:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
-; X32-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 -1, i32 1
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP9:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP10:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP9]], i32 1
-; X32-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP10]], i32 1
-; X32-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP11]]
-; X32-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP12]]
-; X32-NEXT:    [[TMP15]] = call i32 @llvm.bswap.i32(i32 [[TMP13]])
-; X32-NEXT:    [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; X32-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP15]], [[TMP16]]
-; X32-NEXT:    br i1 [[TMP17]], label [[LOADBB2]], label [[RES_BLOCK]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP18:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP19:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP20:%.*]] = getelementptr i32, i32* [[TMP18]], i32 2
-; X32-NEXT:    [[TMP21:%.*]] = getelementptr i32, i32* [[TMP19]], i32 2
-; X32-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP20]]
-; X32-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP21]]
-; X32-NEXT:    [[TMP24]] = call i32 @llvm.bswap.i32(i32 [[TMP22]])
-; X32-NEXT:    [[TMP25]] = call i32 @llvm.bswap.i32(i32 [[TMP23]])
-; X32-NEXT:    [[TMP26:%.*]] = icmp eq i32 [[TMP24]], [[TMP25]]
-; X32-NEXT:    br i1 [[TMP26]], label [[LOADBB3]], label [[RES_BLOCK]]
-; X32:       loadbb3:
-; X32-NEXT:    [[TMP27:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP28:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP29:%.*]] = getelementptr i32, i32* [[TMP27]], i32 3
-; X32-NEXT:    [[TMP30:%.*]] = getelementptr i32, i32* [[TMP28]], i32 3
-; X32-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP29]]
-; X32-NEXT:    [[TMP32:%.*]] = load i32, i32* [[TMP30]]
-; X32-NEXT:    [[TMP33]] = call i32 @llvm.bswap.i32(i32 [[TMP31]])
-; X32-NEXT:    [[TMP34]] = call i32 @llvm.bswap.i32(i32 [[TMP32]])
-; X32-NEXT:    [[TMP35:%.*]] = icmp eq i32 [[TMP33]], [[TMP34]]
-; X32-NEXT:    br i1 [[TMP35]], label [[ENDBLOCK]], label [[RES_BLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB3]] ], [ [[TMP8]], [[RES_BLOCK]] ]
-; X32-NEXT:    ret i32 [[PHI_RES]]
+; X32-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 16)
+; X32-NEXT:    ret i32 [[CALL]]
 ;
 ; X64-LABEL: @cmp16(
 ; X64-NEXT:  loadbb:
@@ -1189,43 +612,17 @@ define i32 @cmp_eq6(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 ; ALL-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
 ; ALL-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
 ; ALL-NEXT:    ret i32 [[CONV]]
-;
-  %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 6)
-  %cmp = icmp eq i32 %call, 0
-  %conv = zext i1 %cmp to i32
-  ret i32 %conv
-}
-
-define i32 @cmp_eq7(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
-; ALL-LABEL: @cmp_eq7(
-; ALL-NEXT:  loadbb:
-; ALL-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; ALL-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; ALL-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; ALL-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; ALL-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP2]], [[TMP3]]
-; ALL-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; ALL:       res_block:
-; ALL-NEXT:    br label [[ENDBLOCK:%.*]]
-; ALL:       loadbb1:
-; ALL-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i16*
-; ALL-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i16*
-; ALL-NEXT:    [[TMP7:%.*]] = getelementptr i16, i16* [[TMP5]], i16 2
-; ALL-NEXT:    [[TMP8:%.*]] = getelementptr i16, i16* [[TMP6]], i16 2
-; ALL-NEXT:    [[TMP9:%.*]] = load i16, i16* [[TMP7]]
-; ALL-NEXT:    [[TMP10:%.*]] = load i16, i16* [[TMP8]]
-; ALL-NEXT:    [[TMP11:%.*]] = icmp ne i16 [[TMP9]], [[TMP10]]
-; ALL-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; ALL:       loadbb2:
-; ALL-NEXT:    [[TMP12:%.*]] = getelementptr i8, i8* [[X]], i8 6
-; ALL-NEXT:    [[TMP13:%.*]] = getelementptr i8, i8* [[Y]], i8 6
-; ALL-NEXT:    [[TMP14:%.*]] = load i8, i8* [[TMP12]]
-; ALL-NEXT:    [[TMP15:%.*]] = load i8, i8* [[TMP13]]
-; ALL-NEXT:    [[TMP16:%.*]] = icmp ne i8 [[TMP14]], [[TMP15]]
-; ALL-NEXT:    br i1 [[TMP16]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; ALL:       endblock:
-; ALL-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB2]] ], [ 1, [[RES_BLOCK]] ]
-; ALL-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
+;
+  %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 6)
+  %cmp = icmp eq i32 %call, 0
+  %conv = zext i1 %cmp to i32
+  ret i32 %conv
+}
+
+define i32 @cmp_eq7(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
+; ALL-LABEL: @cmp_eq7(
+; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 7)
+; ALL-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
 ; ALL-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
 ; ALL-NEXT:    ret i32 [[CONV]]
 ;
@@ -1280,34 +677,8 @@ define i32 @cmp_eq8(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 
 define i32 @cmp_eq9(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 ; X32-LABEL: @cmp_eq9(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP2]], [[TMP3]]
-; X32-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; X32:       res_block:
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP5]], i32 1
-; X32-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[TMP6]], i32 1
-; X32-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]]
-; X32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP8]]
-; X32-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP9]], [[TMP10]]
-; X32-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP12:%.*]] = getelementptr i8, i8* [[X]], i8 8
-; X32-NEXT:    [[TMP13:%.*]] = getelementptr i8, i8* [[Y]], i8 8
-; X32-NEXT:    [[TMP14:%.*]] = load i8, i8* [[TMP12]]
-; X32-NEXT:    [[TMP15:%.*]] = load i8, i8* [[TMP13]]
-; X32-NEXT:    [[TMP16:%.*]] = icmp ne i8 [[TMP14]], [[TMP15]]
-; X32-NEXT:    br i1 [[TMP16]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB2]] ], [ 1, [[RES_BLOCK]] ]
-; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
+; X32-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 9)
+; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
 ; X32-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
 ; X32-NEXT:    ret i32 [[CONV]]
 ;
@@ -1342,36 +713,8 @@ define i32 @cmp_eq9(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 
 define i32 @cmp_eq10(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 ; X32-LABEL: @cmp_eq10(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP2]], [[TMP3]]
-; X32-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; X32:       res_block:
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP5]], i32 1
-; X32-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[TMP6]], i32 1
-; X32-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]]
-; X32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP8]]
-; X32-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP9]], [[TMP10]]
-; X32-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[X]] to i16*
-; X32-NEXT:    [[TMP13:%.*]] = bitcast i8* [[Y]] to i16*
-; X32-NEXT:    [[TMP14:%.*]] = getelementptr i16, i16* [[TMP12]], i16 4
-; X32-NEXT:    [[TMP15:%.*]] = getelementptr i16, i16* [[TMP13]], i16 4
-; X32-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP14]]
-; X32-NEXT:    [[TMP17:%.*]] = load i16, i16* [[TMP15]]
-; X32-NEXT:    [[TMP18:%.*]] = icmp ne i16 [[TMP16]], [[TMP17]]
-; X32-NEXT:    br i1 [[TMP18]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB2]] ], [ 1, [[RES_BLOCK]] ]
-; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
+; X32-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 10)
+; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
 ; X32-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
 ; X32-NEXT:    ret i32 [[CONV]]
 ;
@@ -1407,78 +750,11 @@ define i32 @cmp_eq10(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 }
 
 define i32 @cmp_eq11(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
-; X32-LABEL: @cmp_eq11(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP2]], [[TMP3]]
-; X32-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; X32:       res_block:
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP5]], i32 1
-; X32-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[TMP6]], i32 1
-; X32-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]]
-; X32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP8]]
-; X32-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP9]], [[TMP10]]
-; X32-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[X]] to i16*
-; X32-NEXT:    [[TMP13:%.*]] = bitcast i8* [[Y]] to i16*
-; X32-NEXT:    [[TMP14:%.*]] = getelementptr i16, i16* [[TMP12]], i16 4
-; X32-NEXT:    [[TMP15:%.*]] = getelementptr i16, i16* [[TMP13]], i16 4
-; X32-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP14]]
-; X32-NEXT:    [[TMP17:%.*]] = load i16, i16* [[TMP15]]
-; X32-NEXT:    [[TMP18:%.*]] = icmp ne i16 [[TMP16]], [[TMP17]]
-; X32-NEXT:    br i1 [[TMP18]], label [[RES_BLOCK]], label [[LOADBB3:%.*]]
-; X32:       loadbb3:
-; X32-NEXT:    [[TMP19:%.*]] = getelementptr i8, i8* [[X]], i8 10
-; X32-NEXT:    [[TMP20:%.*]] = getelementptr i8, i8* [[Y]], i8 10
-; X32-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP19]]
-; X32-NEXT:    [[TMP22:%.*]] = load i8, i8* [[TMP20]]
-; X32-NEXT:    [[TMP23:%.*]] = icmp ne i8 [[TMP21]], [[TMP22]]
-; X32-NEXT:    br i1 [[TMP23]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB3]] ], [ 1, [[RES_BLOCK]] ]
-; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
-; X32-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT:    ret i32 [[CONV]]
-;
-; X64-LABEL: @cmp_eq11(
-; X64-NEXT:  loadbb:
-; X64-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i64*
-; X64-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP0]]
-; X64-NEXT:    [[TMP3:%.*]] = load i64, i64* [[TMP1]]
-; X64-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP2]], [[TMP3]]
-; X64-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; X64:       res_block:
-; X64-NEXT:    br label [[ENDBLOCK:%.*]]
-; X64:       loadbb1:
-; X64-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i16*
-; X64-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i16*
-; X64-NEXT:    [[TMP7:%.*]] = getelementptr i16, i16* [[TMP5]], i16 4
-; X64-NEXT:    [[TMP8:%.*]] = getelementptr i16, i16* [[TMP6]], i16 4
-; X64-NEXT:    [[TMP9:%.*]] = load i16, i16* [[TMP7]]
-; X64-NEXT:    [[TMP10:%.*]] = load i16, i16* [[TMP8]]
-; X64-NEXT:    [[TMP11:%.*]] = icmp ne i16 [[TMP9]], [[TMP10]]
-; X64-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; X64:       loadbb2:
-; X64-NEXT:    [[TMP12:%.*]] = getelementptr i8, i8* [[X]], i8 10
-; X64-NEXT:    [[TMP13:%.*]] = getelementptr i8, i8* [[Y]], i8 10
-; X64-NEXT:    [[TMP14:%.*]] = load i8, i8* [[TMP12]]
-; X64-NEXT:    [[TMP15:%.*]] = load i8, i8* [[TMP13]]
-; X64-NEXT:    [[TMP16:%.*]] = icmp ne i8 [[TMP14]], [[TMP15]]
-; X64-NEXT:    br i1 [[TMP16]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; X64:       endblock:
-; X64-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB2]] ], [ 1, [[RES_BLOCK]] ]
-; X64-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
-; X64-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X64-NEXT:    ret i32 [[CONV]]
+; ALL-LABEL: @cmp_eq11(
+; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 11)
+; ALL-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; ALL-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; ALL-NEXT:    ret i32 [[CONV]]
 ;
   %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 11)
   %cmp = icmp eq i32 %call, 0
@@ -1488,36 +764,8 @@ define i32 @cmp_eq11(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 
 define i32 @cmp_eq12(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 ; X32-LABEL: @cmp_eq12(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP2]], [[TMP3]]
-; X32-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; X32:       res_block:
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP5]], i32 1
-; X32-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[TMP6]], i32 1
-; X32-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]]
-; X32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP8]]
-; X32-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP9]], [[TMP10]]
-; X32-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP13:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP14:%.*]] = getelementptr i32, i32* [[TMP12]], i32 2
-; X32-NEXT:    [[TMP15:%.*]] = getelementptr i32, i32* [[TMP13]], i32 2
-; X32-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP14]]
-; X32-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP15]]
-; X32-NEXT:    [[TMP18:%.*]] = icmp ne i32 [[TMP16]], [[TMP17]]
-; X32-NEXT:    br i1 [[TMP18]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB2]] ], [ 1, [[RES_BLOCK]] ]
-; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
+; X32-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 12)
+; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
 ; X32-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
 ; X32-NEXT:    ret i32 [[CONV]]
 ;
@@ -1553,78 +801,11 @@ define i32 @cmp_eq12(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 }
 
 define i32 @cmp_eq13(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
-; X32-LABEL: @cmp_eq13(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP2]], [[TMP3]]
-; X32-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; X32:       res_block:
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP5]], i32 1
-; X32-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[TMP6]], i32 1
-; X32-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]]
-; X32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP8]]
-; X32-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP9]], [[TMP10]]
-; X32-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP13:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP14:%.*]] = getelementptr i32, i32* [[TMP12]], i32 2
-; X32-NEXT:    [[TMP15:%.*]] = getelementptr i32, i32* [[TMP13]], i32 2
-; X32-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP14]]
-; X32-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP15]]
-; X32-NEXT:    [[TMP18:%.*]] = icmp ne i32 [[TMP16]], [[TMP17]]
-; X32-NEXT:    br i1 [[TMP18]], label [[RES_BLOCK]], label [[LOADBB3:%.*]]
-; X32:       loadbb3:
-; X32-NEXT:    [[TMP19:%.*]] = getelementptr i8, i8* [[X]], i8 12
-; X32-NEXT:    [[TMP20:%.*]] = getelementptr i8, i8* [[Y]], i8 12
-; X32-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP19]]
-; X32-NEXT:    [[TMP22:%.*]] = load i8, i8* [[TMP20]]
-; X32-NEXT:    [[TMP23:%.*]] = icmp ne i8 [[TMP21]], [[TMP22]]
-; X32-NEXT:    br i1 [[TMP23]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB3]] ], [ 1, [[RES_BLOCK]] ]
-; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
-; X32-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT:    ret i32 [[CONV]]
-;
-; X64-LABEL: @cmp_eq13(
-; X64-NEXT:  loadbb:
-; X64-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i64*
-; X64-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP0]]
-; X64-NEXT:    [[TMP3:%.*]] = load i64, i64* [[TMP1]]
-; X64-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP2]], [[TMP3]]
-; X64-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; X64:       res_block:
-; X64-NEXT:    br label [[ENDBLOCK:%.*]]
-; X64:       loadbb1:
-; X64-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i32*
-; X64-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i32*
-; X64-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP5]], i32 2
-; X64-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[TMP6]], i32 2
-; X64-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]]
-; X64-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP8]]
-; X64-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP9]], [[TMP10]]
-; X64-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; X64:       loadbb2:
-; X64-NEXT:    [[TMP12:%.*]] = getelementptr i8, i8* [[X]], i8 12
-; X64-NEXT:    [[TMP13:%.*]] = getelementptr i8, i8* [[Y]], i8 12
-; X64-NEXT:    [[TMP14:%.*]] = load i8, i8* [[TMP12]]
-; X64-NEXT:    [[TMP15:%.*]] = load i8, i8* [[TMP13]]
-; X64-NEXT:    [[TMP16:%.*]] = icmp ne i8 [[TMP14]], [[TMP15]]
-; X64-NEXT:    br i1 [[TMP16]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; X64:       endblock:
-; X64-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB2]] ], [ 1, [[RES_BLOCK]] ]
-; X64-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
-; X64-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X64-NEXT:    ret i32 [[CONV]]
+; ALL-LABEL: @cmp_eq13(
+; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 13)
+; ALL-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; ALL-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; ALL-NEXT:    ret i32 [[CONV]]
 ;
   %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 13)
   %cmp = icmp eq i32 %call, 0
@@ -1633,82 +814,11 @@ define i32 @cmp_eq13(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 }
 
 define i32 @cmp_eq14(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
-; X32-LABEL: @cmp_eq14(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP2]], [[TMP3]]
-; X32-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; X32:       res_block:
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP5]], i32 1
-; X32-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[TMP6]], i32 1
-; X32-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]]
-; X32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP8]]
-; X32-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP9]], [[TMP10]]
-; X32-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP13:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP14:%.*]] = getelementptr i32, i32* [[TMP12]], i32 2
-; X32-NEXT:    [[TMP15:%.*]] = getelementptr i32, i32* [[TMP13]], i32 2
-; X32-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP14]]
-; X32-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP15]]
-; X32-NEXT:    [[TMP18:%.*]] = icmp ne i32 [[TMP16]], [[TMP17]]
-; X32-NEXT:    br i1 [[TMP18]], label [[RES_BLOCK]], label [[LOADBB3:%.*]]
-; X32:       loadbb3:
-; X32-NEXT:    [[TMP19:%.*]] = bitcast i8* [[X]] to i16*
-; X32-NEXT:    [[TMP20:%.*]] = bitcast i8* [[Y]] to i16*
-; X32-NEXT:    [[TMP21:%.*]] = getelementptr i16, i16* [[TMP19]], i16 6
-; X32-NEXT:    [[TMP22:%.*]] = getelementptr i16, i16* [[TMP20]], i16 6
-; X32-NEXT:    [[TMP23:%.*]] = load i16, i16* [[TMP21]]
-; X32-NEXT:    [[TMP24:%.*]] = load i16, i16* [[TMP22]]
-; X32-NEXT:    [[TMP25:%.*]] = icmp ne i16 [[TMP23]], [[TMP24]]
-; X32-NEXT:    br i1 [[TMP25]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB3]] ], [ 1, [[RES_BLOCK]] ]
-; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
-; X32-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT:    ret i32 [[CONV]]
-;
-; X64-LABEL: @cmp_eq14(
-; X64-NEXT:  loadbb:
-; X64-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i64*
-; X64-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP0]]
-; X64-NEXT:    [[TMP3:%.*]] = load i64, i64* [[TMP1]]
-; X64-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP2]], [[TMP3]]
-; X64-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; X64:       res_block:
-; X64-NEXT:    br label [[ENDBLOCK:%.*]]
-; X64:       loadbb1:
-; X64-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i32*
-; X64-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i32*
-; X64-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP5]], i32 2
-; X64-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[TMP6]], i32 2
-; X64-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]]
-; X64-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP8]]
-; X64-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP9]], [[TMP10]]
-; X64-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; X64:       loadbb2:
-; X64-NEXT:    [[TMP12:%.*]] = bitcast i8* [[X]] to i16*
-; X64-NEXT:    [[TMP13:%.*]] = bitcast i8* [[Y]] to i16*
-; X64-NEXT:    [[TMP14:%.*]] = getelementptr i16, i16* [[TMP12]], i16 6
-; X64-NEXT:    [[TMP15:%.*]] = getelementptr i16, i16* [[TMP13]], i16 6
-; X64-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP14]]
-; X64-NEXT:    [[TMP17:%.*]] = load i16, i16* [[TMP15]]
-; X64-NEXT:    [[TMP18:%.*]] = icmp ne i16 [[TMP16]], [[TMP17]]
-; X64-NEXT:    br i1 [[TMP18]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; X64:       endblock:
-; X64-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB2]] ], [ 1, [[RES_BLOCK]] ]
-; X64-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
-; X64-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X64-NEXT:    ret i32 [[CONV]]
+; ALL-LABEL: @cmp_eq14(
+; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 14)
+; ALL-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; ALL-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; ALL-NEXT:    ret i32 [[CONV]]
 ;
   %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 14)
   %cmp = icmp eq i32 %call, 0
@@ -1717,52 +827,11 @@ define i32 @cmp_eq14(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 }
 
 define i32 @cmp_eq15(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
-; X32-LABEL: @cmp_eq15(
-; X32-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 15)
-; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
-; X32-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT:    ret i32 [[CONV]]
-;
-; X64-LABEL: @cmp_eq15(
-; X64-NEXT:  loadbb:
-; X64-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i64*
-; X64-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP0]]
-; X64-NEXT:    [[TMP3:%.*]] = load i64, i64* [[TMP1]]
-; X64-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP2]], [[TMP3]]
-; X64-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; X64:       res_block:
-; X64-NEXT:    br label [[ENDBLOCK:%.*]]
-; X64:       loadbb1:
-; X64-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i32*
-; X64-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i32*
-; X64-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP5]], i32 2
-; X64-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[TMP6]], i32 2
-; X64-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]]
-; X64-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP8]]
-; X64-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP9]], [[TMP10]]
-; X64-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; X64:       loadbb2:
-; X64-NEXT:    [[TMP12:%.*]] = bitcast i8* [[X]] to i16*
-; X64-NEXT:    [[TMP13:%.*]] = bitcast i8* [[Y]] to i16*
-; X64-NEXT:    [[TMP14:%.*]] = getelementptr i16, i16* [[TMP12]], i16 6
-; X64-NEXT:    [[TMP15:%.*]] = getelementptr i16, i16* [[TMP13]], i16 6
-; X64-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP14]]
-; X64-NEXT:    [[TMP17:%.*]] = load i16, i16* [[TMP15]]
-; X64-NEXT:    [[TMP18:%.*]] = icmp ne i16 [[TMP16]], [[TMP17]]
-; X64-NEXT:    br i1 [[TMP18]], label [[RES_BLOCK]], label [[LOADBB3:%.*]]
-; X64:       loadbb3:
-; X64-NEXT:    [[TMP19:%.*]] = getelementptr i8, i8* [[X]], i8 14
-; X64-NEXT:    [[TMP20:%.*]] = getelementptr i8, i8* [[Y]], i8 14
-; X64-NEXT:    [[TMP21:%.*]] = load i8, i8* [[TMP19]]
-; X64-NEXT:    [[TMP22:%.*]] = load i8, i8* [[TMP20]]
-; X64-NEXT:    [[TMP23:%.*]] = icmp ne i8 [[TMP21]], [[TMP22]]
-; X64-NEXT:    br i1 [[TMP23]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; X64:       endblock:
-; X64-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB3]] ], [ 1, [[RES_BLOCK]] ]
-; X64-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
-; X64-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X64-NEXT:    ret i32 [[CONV]]
+; ALL-LABEL: @cmp_eq15(
+; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 15)
+; ALL-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; ALL-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; ALL-NEXT:    ret i32 [[CONV]]
 ;
   %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 15)
   %cmp = icmp eq i32 %call, 0
@@ -1772,45 +841,8 @@ define i32 @cmp_eq15(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 
 define i32 @cmp_eq16(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 ; X32-LABEL: @cmp_eq16(
-; X32-NEXT:  loadbb:
-; X32-NEXT:    [[TMP0:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT:    [[TMP1:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]]
-; X32-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]]
-; X32-NEXT:    [[TMP4:%.*]] = icmp ne i32 [[TMP2]], [[TMP3]]
-; X32-NEXT:    br i1 [[TMP4]], label [[RES_BLOCK:%.*]], label [[LOADBB1:%.*]]
-; X32:       res_block:
-; X32-NEXT:    br label [[ENDBLOCK:%.*]]
-; X32:       loadbb1:
-; X32-NEXT:    [[TMP5:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP6:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[TMP5]], i32 1
-; X32-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[TMP6]], i32 1
-; X32-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP7]]
-; X32-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP8]]
-; X32-NEXT:    [[TMP11:%.*]] = icmp ne i32 [[TMP9]], [[TMP10]]
-; X32-NEXT:    br i1 [[TMP11]], label [[RES_BLOCK]], label [[LOADBB2:%.*]]
-; X32:       loadbb2:
-; X32-NEXT:    [[TMP12:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP13:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP14:%.*]] = getelementptr i32, i32* [[TMP12]], i32 2
-; X32-NEXT:    [[TMP15:%.*]] = getelementptr i32, i32* [[TMP13]], i32 2
-; X32-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP14]]
-; X32-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP15]]
-; X32-NEXT:    [[TMP18:%.*]] = icmp ne i32 [[TMP16]], [[TMP17]]
-; X32-NEXT:    br i1 [[TMP18]], label [[RES_BLOCK]], label [[LOADBB3:%.*]]
-; X32:       loadbb3:
-; X32-NEXT:    [[TMP19:%.*]] = bitcast i8* [[X]] to i32*
-; X32-NEXT:    [[TMP20:%.*]] = bitcast i8* [[Y]] to i32*
-; X32-NEXT:    [[TMP21:%.*]] = getelementptr i32, i32* [[TMP19]], i32 3
-; X32-NEXT:    [[TMP22:%.*]] = getelementptr i32, i32* [[TMP20]], i32 3
-; X32-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP21]]
-; X32-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP22]]
-; X32-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP23]], [[TMP24]]
-; X32-NEXT:    br i1 [[TMP25]], label [[RES_BLOCK]], label [[ENDBLOCK]]
-; X32:       endblock:
-; X32-NEXT:    [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB3]] ], [ 1, [[RES_BLOCK]] ]
-; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[PHI_RES]], 0
+; X32-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 16)
+; X32-NEXT:    [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
 ; X32-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
 ; X32-NEXT:    ret i32 [[CONV]]
 ;