[llvm] 8721490 - [x86] split memcmp tests for 32/64-bit targets; NFC
Sanjay Patel via llvm-commits
llvm-commits at lists.llvm.org
Sun Aug 15 11:06:22 PDT 2021
Author: Sanjay Patel
Date: 2021-08-15T13:51:18-04:00
New Revision: 8721490d3859b4d7fff633bdf3c71d888e75727c
URL: https://github.com/llvm/llvm-project/commit/8721490d3859b4d7fff633bdf3c71d888e75727c
DIFF: https://github.com/llvm/llvm-project/commit/8721490d3859b4d7fff633bdf3c71d888e75727c.diff
LOG: [x86] split memcmp tests for 32/64-bit targets; NFC
memcmp is defined as taking a size_t length arg,
so that differs depending on pointer size of the
target.
We casually matched non-compliant function signatures
as memcmp, but that can cause crashing as seen with
PR50850.
If we fix that bug, these tests would no longer be
testing the expected behavior for a 32-bit target,
so I have duplicated all tests and adjusted them
to match the stricter definition of memcmp/bcmp
by changing the length arg to i32 on a 32-bit target.
Added:
llvm/test/CodeGen/X86/memcmp-minsize-x32.ll
llvm/test/CodeGen/X86/memcmp-more-load-pairs-x32.ll
llvm/test/CodeGen/X86/memcmp-optsize-x32.ll
llvm/test/CodeGen/X86/memcmp-pgso-x32.ll
llvm/test/CodeGen/X86/memcmp-x32.ll
llvm/test/Transforms/ExpandMemCmp/X86/memcmp-x32.ll
Modified:
llvm/test/CodeGen/X86/memcmp-minsize.ll
llvm/test/CodeGen/X86/memcmp-more-load-pairs.ll
llvm/test/CodeGen/X86/memcmp-optsize.ll
llvm/test/CodeGen/X86/memcmp-pgso.ll
llvm/test/CodeGen/X86/memcmp.ll
llvm/test/Transforms/ExpandMemCmp/X86/memcmp.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/X86/memcmp-minsize-x32.ll b/llvm/test/CodeGen/X86/memcmp-minsize-x32.ll
new file mode 100644
index 0000000000000..9f56c38062f68
--- /dev/null
+++ b/llvm/test/CodeGen/X86/memcmp-minsize-x32.ll
@@ -0,0 +1,445 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=cmov | FileCheck %s --check-prefix=X86 --check-prefix=X86-NOSSE
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=X86 --check-prefix=X86-SSE2
+
+; This tests codegen time inlining/optimization of memcmp
+; rdar://6480398
+
+ at .str = private constant [65 x i8] c"0123456789012345678901234567890123456789012345678901234567890123\00", align 1
+
+declare dso_local i32 @memcmp(i8*, i8*, i32)
+
+define i32 @length2(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length2:
+; X86: # %bb.0:
+; X86-NEXT: pushl $2
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ ret i32 %m
+}
+
+define i1 @length2_eq(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length2_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: cmpw (%eax), %cx
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_eq_const(i8* %X) nounwind minsize {
+; X86-LABEL: length2_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: cmpw $12849, (%eax) # imm = 0x3231
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 1), i32 2) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length2_eq_nobuiltin_attr:
+; X86: # %bb.0:
+; X86-NEXT: pushl $2
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind nobuiltin
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length3(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length3:
+; X86: # %bb.0:
+; X86-NEXT: pushl $3
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 3) nounwind
+ ret i32 %m
+}
+
+define i1 @length3_eq(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length3_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $3
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 3) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length4(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length4:
+; X86: # %bb.0:
+; X86-NEXT: pushl $4
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ ret i32 %m
+}
+
+define i1 @length4_eq(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length4_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: cmpl (%eax), %ecx
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length4_eq_const(i8* %X) nounwind minsize {
+; X86-LABEL: length4_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: cmpl $875770417, (%eax) # imm = 0x34333231
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 1), i32 4) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length5(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length5:
+; X86: # %bb.0:
+; X86-NEXT: pushl $5
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ ret i32 %m
+}
+
+define i1 @length5_eq(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length5_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $5
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length8(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length8:
+; X86: # %bb.0:
+; X86-NEXT: pushl $8
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 8) nounwind
+ ret i32 %m
+}
+
+define i1 @length8_eq(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length8_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $8
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 8) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length8_eq_const(i8* %X) nounwind minsize {
+; X86-LABEL: length8_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $8
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 8) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length12_eq(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length12_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $12
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 12) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length12(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length12:
+; X86: # %bb.0:
+; X86-NEXT: pushl $12
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 12) nounwind
+ ret i32 %m
+}
+
+; PR33329 - https://bugs.llvm.org/show_bug.cgi?id=33329
+
+define i32 @length16(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length16:
+; X86: # %bb.0:
+; X86-NEXT: pushl $16
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 16) nounwind
+ ret i32 %m
+}
+
+define i1 @length16_eq(i8* %x, i8* %y) nounwind minsize {
+; X86-NOSSE-LABEL: length16_eq:
+; X86-NOSSE: # %bb.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 $12, %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
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 16) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length16_eq_const(i8* %X) nounwind minsize {
+; X86-NOSSE-LABEL: length16_eq_const:
+; X86-NOSSE: # %bb.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 $12, %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[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 16) nounwind
+ %c = icmp eq i32 %m, 0
+ 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: pushl $24
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 24) nounwind
+ ret i32 %m
+}
+
+define i1 @length24_eq(i8* %x, i8* %y) nounwind minsize {
+; X86-LABEL: length24_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $24
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 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: pushl $24
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 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:
+; X86-NEXT: pushl $32
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 32) nounwind
+ ret i32 %m
+}
+
+; PR33325 - https://bugs.llvm.org/show_bug.cgi?id=33325
+
+define i1 @length32_eq(i8* %x, i8* %y) nounwind minsize {
+; X86-LABEL: length32_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $32
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 32) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length32_eq_const(i8* %X) nounwind minsize {
+; X86-LABEL: length32_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $32
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 32) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length64(i8* %X, i8* %Y) nounwind minsize {
+; X86-LABEL: length64:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 64) nounwind
+ ret i32 %m
+}
+
+define i1 @length64_eq(i8* %x, i8* %y) nounwind minsize {
+; X86-LABEL: length64_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 64) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length64_eq_const(i8* %X) nounwind minsize {
+; X86-LABEL: length64_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 64) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
diff --git a/llvm/test/CodeGen/X86/memcmp-minsize.ll b/llvm/test/CodeGen/X86/memcmp-minsize.ll
index fec09f6b8d364..3bcd79c4c0cd4 100644
--- a/llvm/test/CodeGen/X86/memcmp-minsize.ll
+++ b/llvm/test/CodeGen/X86/memcmp-minsize.ll
@@ -1,6 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=cmov | FileCheck %s --check-prefix=X86 --check-prefix=X86-NOSSE
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=X86 --check-prefix=X86-SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefix=X64 --check-prefix=X64-SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx | FileCheck %s --check-prefix=X64 --check-prefix=X64-AVX --check-prefix=X64-AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx2 | FileCheck %s --check-prefix=X64 --check-prefix=X64-AVX --check-prefix=X64-AVX2
@@ -13,16 +11,6 @@
declare dso_local i32 @memcmp(i8*, i8*, i64)
define i32 @length2(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length2:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $2
-; 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: length2:
; X64: # %bb.0:
; X64-NEXT: pushq $2
@@ -33,15 +21,6 @@ define i32 @length2(i8* %X, i8* %Y) nounwind minsize {
}
define i1 @length2_eq(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length2_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: cmpw (%eax), %cx
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_eq:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -54,13 +33,6 @@ define i1 @length2_eq(i8* %X, i8* %Y) nounwind minsize {
}
define i1 @length2_eq_const(i8* %X) nounwind minsize {
-; X86-LABEL: length2_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: cmpw $12849, (%eax) # imm = 0x3231
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_eq_const:
; X64: # %bb.0:
; X64-NEXT: cmpw $12849, (%rdi) # imm = 0x3231
@@ -72,18 +44,6 @@ define i1 @length2_eq_const(i8* %X) nounwind minsize {
}
define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length2_eq_nobuiltin_attr:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $2
-; 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: length2_eq_nobuiltin_attr:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -100,16 +60,6 @@ define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind minsize {
}
define i32 @length3(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length3:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $3
-; 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: length3:
; X64: # %bb.0:
; X64-NEXT: pushq $3
@@ -120,18 +70,6 @@ define i32 @length3(i8* %X, i8* %Y) nounwind minsize {
}
define i1 @length3_eq(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length3_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $3
-; 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
-;
; X64-LABEL: length3_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -148,16 +86,6 @@ define i1 @length3_eq(i8* %X, i8* %Y) nounwind minsize {
}
define i32 @length4(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length4:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $4
-; 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: length4:
; X64: # %bb.0:
; X64-NEXT: pushq $4
@@ -168,15 +96,6 @@ define i32 @length4(i8* %X, i8* %Y) nounwind minsize {
}
define i1 @length4_eq(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length4_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: cmpl (%eax), %ecx
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_eq:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -189,13 +108,6 @@ define i1 @length4_eq(i8* %X, i8* %Y) nounwind minsize {
}
define i1 @length4_eq_const(i8* %X) nounwind minsize {
-; X86-LABEL: length4_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: cmpl $875770417, (%eax) # imm = 0x34333231
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_eq_const:
; X64: # %bb.0:
; X64-NEXT: cmpl $875770417, (%rdi) # imm = 0x34333231
@@ -207,16 +119,6 @@ define i1 @length4_eq_const(i8* %X) nounwind minsize {
}
define i32 @length5(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length5:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $5
-; 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: length5:
; X64: # %bb.0:
; X64-NEXT: pushq $5
@@ -227,18 +129,6 @@ define i32 @length5(i8* %X, i8* %Y) nounwind minsize {
}
define i1 @length5_eq(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length5_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $5
-; 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
-;
; X64-LABEL: length5_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -255,16 +145,6 @@ define i1 @length5_eq(i8* %X, i8* %Y) nounwind minsize {
}
define i32 @length8(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length8:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $8
-; 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: length8:
; X64: # %bb.0:
; X64-NEXT: pushq $8
@@ -275,18 +155,6 @@ define i32 @length8(i8* %X, i8* %Y) nounwind minsize {
}
define i1 @length8_eq(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length8_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $8
-; 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: length8_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -299,18 +167,6 @@ define i1 @length8_eq(i8* %X, i8* %Y) nounwind minsize {
}
define i1 @length8_eq_const(i8* %X) nounwind minsize {
-; X86-LABEL: length8_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $8
-; 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: length8_eq_const:
; X64: # %bb.0:
; X64-NEXT: movabsq $3978425819141910832, %rax # imm = 0x3736353433323130
@@ -323,18 +179,6 @@ define i1 @length8_eq_const(i8* %X) nounwind minsize {
}
define i1 @length12_eq(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length12_eq:
-; 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
-;
; X64-LABEL: length12_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -351,16 +195,6 @@ define i1 @length12_eq(i8* %X, i8* %Y) nounwind minsize {
}
define i32 @length12(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length12:
-; 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:
; X64: # %bb.0:
; X64-NEXT: pushq $12
@@ -373,15 +207,6 @@ define i32 @length12(i8* %X, i8* %Y) nounwind minsize {
; PR33329 - https://bugs.llvm.org/show_bug.cgi?id=33329
define i32 @length16(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length16:
-; 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:
; X64: # %bb.0:
@@ -393,30 +218,6 @@ define i32 @length16(i8* %X, i8* %Y) nounwind minsize {
}
define i1 @length16_eq(i8* %x, i8* %y) nounwind minsize {
-; 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-SSE2-LABEL: length16_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rsi), %xmm0
@@ -440,28 +241,6 @@ define i1 @length16_eq(i8* %x, i8* %y) nounwind minsize {
}
define i1 @length16_eq_const(i8* %X) nounwind minsize {
-; 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[0-9]+_[0-9]+}}, %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-SSE2-LABEL: length16_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -486,16 +265,6 @@ define i1 @length16_eq_const(i8* %X) nounwind minsize {
; 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: 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: pushq $24
@@ -506,18 +275,6 @@ define i32 @length24(i8* %X, i8* %Y) nounwind minsize {
}
define i1 @length24_eq(i8* %x, i8* %y) nounwind minsize {
-; 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
@@ -534,18 +291,6 @@ define i1 @length24_eq(i8* %x, i8* %y) nounwind minsize {
}
define i1 @length24_eq_const(i8* %X) nounwind minsize {
-; 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
@@ -563,16 +308,6 @@ define i1 @length24_eq_const(i8* %X) nounwind minsize {
}
define i32 @length32(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length32:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $32
-; 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: length32:
; X64: # %bb.0:
; X64-NEXT: pushq $32
@@ -585,18 +320,6 @@ define i32 @length32(i8* %X, i8* %Y) nounwind minsize {
; PR33325 - https://bugs.llvm.org/show_bug.cgi?id=33325
define i1 @length32_eq(i8* %x, i8* %y) nounwind minsize {
-; X86-LABEL: length32_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $32
-; 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-SSE2-LABEL: length32_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: pushq %rax
@@ -631,18 +354,6 @@ define i1 @length32_eq(i8* %x, i8* %y) nounwind minsize {
}
define i1 @length32_eq_const(i8* %X) nounwind minsize {
-; X86-LABEL: length32_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $32
-; 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-SSE2-LABEL: length32_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: pushq %rax
@@ -678,16 +389,6 @@ define i1 @length32_eq_const(i8* %X) nounwind minsize {
}
define i32 @length64(i8* %X, i8* %Y) nounwind minsize {
-; X86-LABEL: length64:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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: length64:
; X64: # %bb.0:
; X64-NEXT: pushq $64
@@ -698,18 +399,6 @@ define i32 @length64(i8* %X, i8* %Y) nounwind minsize {
}
define i1 @length64_eq(i8* %x, i8* %y) nounwind minsize {
-; X86-LABEL: length64_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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
-;
; X64-LABEL: length64_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -726,18 +415,6 @@ define i1 @length64_eq(i8* %x, i8* %y) nounwind minsize {
}
define i1 @length64_eq_const(i8* %X) nounwind minsize {
-; X86-LABEL: length64_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length64_eq_const:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
diff --git a/llvm/test/CodeGen/X86/memcmp-more-load-pairs-x32.ll b/llvm/test/CodeGen/X86/memcmp-more-load-pairs-x32.ll
new file mode 100644
index 0000000000000..edfaaaed7d849
--- /dev/null
+++ b/llvm/test/CodeGen/X86/memcmp-more-load-pairs-x32.ll
@@ -0,0 +1,2916 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; NOTE: This is a copy of llvm/test/CodeGen/X86/memcmp.ll with more load pairs. Please keep it that way.
+; RUN: llc -max-loads-per-memcmp=4 -memcmp-num-loads-per-block=4 < %s -mtriple=i686-unknown-unknown -mattr=cmov | FileCheck %s --check-prefixes=X86,X86-NOSSE
+; RUN: llc -max-loads-per-memcmp=4 -memcmp-num-loads-per-block=4 < %s -mtriple=i686-unknown-unknown -mattr=+sse | FileCheck %s --check-prefixes=X86,X86-SSE1
+; RUN: llc -max-loads-per-memcmp=4 -memcmp-num-loads-per-block=4 < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=X86,X86-SSE2
+; RUN: llc -max-loads-per-memcmp=4 -memcmp-num-loads-per-block=4 < %s -mtriple=i686-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=X86,X86-SSE41
+
+; This tests codegen time inlining/optimization of memcmp
+; rdar://6480398
+
+ at .str = private constant [513 x i8] c"01234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901\00", align 1
+
+declare dso_local i32 @memcmp(i8*, i8*, i32)
+
+define i32 @length0(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length0:
+; X86: # %bb.0:
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 0) nounwind
+ ret i32 %m
+ }
+
+define i1 @length0_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length0_eq:
+; X86: # %bb.0:
+; X86-NEXT: movb $1, %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 0) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length0_lt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length0_lt:
+; X86: # %bb.0:
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 0) nounwind
+ %c = icmp slt i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length2(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: rolw $8, %cx
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: movzwl %cx, %eax
+; X86-NEXT: movzwl %dx, %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ ret i32 %m
+}
+
+define i1 @length2_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: cmpw (%eax), %cx
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_lt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2_lt:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: rolw $8, %cx
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: movzwl %cx, %eax
+; X86-NEXT: movzwl %dx, %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ %c = icmp slt i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_gt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2_gt:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: movzwl (%eax), %eax
+; X86-NEXT: rolw $8, %cx
+; X86-NEXT: rolw $8, %ax
+; X86-NEXT: movzwl %cx, %ecx
+; X86-NEXT: movzwl %ax, %eax
+; X86-NEXT: subl %eax, %ecx
+; X86-NEXT: testl %ecx, %ecx
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ %c = icmp sgt i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_eq_const(i8* %X) nounwind {
+; X86-LABEL: length2_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl (%eax), %eax
+; X86-NEXT: cmpl $12849, %eax # imm = 0x3231
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 1), i32 2) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2_eq_nobuiltin_attr:
+; X86: # %bb.0:
+; X86-NEXT: pushl $2
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind nobuiltin
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length3(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length3:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: movzwl (%ecx), %esi
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: rolw $8, %si
+; X86-NEXT: cmpw %si, %dx
+; X86-NEXT: jne .LBB9_3
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movzbl 2(%eax), %eax
+; X86-NEXT: movzbl 2(%ecx), %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+; X86-NEXT: .LBB9_3: # %res_block
+; X86-NEXT: setae %al
+; X86-NEXT: movzbl %al, %eax
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 3) nounwind
+ ret i32 %m
+}
+
+define i1 @length3_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length3_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %edx
+; X86-NEXT: xorw (%eax), %dx
+; X86-NEXT: movb 2(%ecx), %cl
+; X86-NEXT: xorb 2(%eax), %cl
+; X86-NEXT: movzbl %cl, %eax
+; X86-NEXT: orw %dx, %ax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 3) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length4(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length4:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: seta %al
+; X86-NEXT: sbbl $0, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ ret i32 %m
+}
+
+define i1 @length4_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length4_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: cmpl (%eax), %ecx
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length4_lt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length4_lt:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: seta %al
+; X86-NEXT: sbbl $0, %eax
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ %c = icmp slt i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length4_gt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length4_gt:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: movl (%eax), %eax
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %eax
+; X86-NEXT: xorl %edx, %edx
+; X86-NEXT: cmpl %eax, %ecx
+; X86-NEXT: seta %dl
+; X86-NEXT: sbbl $0, %edx
+; X86-NEXT: testl %edx, %edx
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ %c = icmp sgt i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length4_eq_const(i8* %X) nounwind {
+; X86-LABEL: length4_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: cmpl $875770417, (%eax) # imm = 0x34333231
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 1), i32 4) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length5(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length5:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: movl (%ecx), %esi
+; X86-NEXT: bswapl %edx
+; X86-NEXT: bswapl %esi
+; X86-NEXT: cmpl %esi, %edx
+; X86-NEXT: jne .LBB16_3
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movzbl 4(%eax), %eax
+; X86-NEXT: movzbl 4(%ecx), %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+; X86-NEXT: .LBB16_3: # %res_block
+; X86-NEXT: setae %al
+; X86-NEXT: movzbl %al, %eax
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ ret i32 %m
+}
+
+define i1 @length5_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length5_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: movb 4(%ecx), %cl
+; X86-NEXT: xorb 4(%eax), %cl
+; X86-NEXT: movzbl %cl, %eax
+; X86-NEXT: orl %edx, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length5_lt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length5_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: movl (%ecx), %esi
+; X86-NEXT: bswapl %edx
+; X86-NEXT: bswapl %esi
+; X86-NEXT: cmpl %esi, %edx
+; X86-NEXT: jne .LBB18_3
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movzbl 4(%eax), %eax
+; X86-NEXT: movzbl 4(%ecx), %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: jmp .LBB18_2
+; X86-NEXT: .LBB18_3: # %res_block
+; X86-NEXT: setae %al
+; X86-NEXT: movzbl %al, %eax
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB18_2: # %endblock
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ %c = icmp slt i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length7(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length7:
+; X86: # %bb.0:
+; 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 .LBB19_2
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movl 3(%esi), %ecx
+; X86-NEXT: movl 3(%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: je .LBB19_3
+; X86-NEXT: .LBB19_2: # %res_block
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: setae %al
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB19_3: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 7) nounwind
+ ret i32 %m
+}
+
+define i1 @length7_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length7_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 3(%ecx), %ecx
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 3(%eax), %ecx
+; X86-NEXT: orl %edx, %ecx
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 7) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length7_lt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length7_lt:
+; X86: # %bb.0:
+; 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 .LBB21_2
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movl 3(%esi), %ecx
+; X86-NEXT: movl 3(%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: je .LBB21_3
+; X86-NEXT: .LBB21_2: # %res_block
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: setae %al
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB21_3: # %endblock
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 7) nounwind
+ %c = icmp slt i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length8(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length8:
+; X86: # %bb.0:
+; 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 .LBB22_2
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movl 4(%esi), %ecx
+; X86-NEXT: movl 4(%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: je .LBB22_3
+; X86-NEXT: .LBB22_2: # %res_block
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: setae %al
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB22_3: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 8) nounwind
+ ret i32 %m
+}
+
+define i1 @length8_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length8_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 4(%ecx), %ecx
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 4(%eax), %ecx
+; X86-NEXT: orl %edx, %ecx
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 8) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length8_eq_const(i8* %X) nounwind {
+; X86-LABEL: length8_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl $858927408, %ecx # imm = 0x33323130
+; X86-NEXT: xorl (%eax), %ecx
+; X86-NEXT: movl $926299444, %edx # imm = 0x37363534
+; X86-NEXT: xorl 4(%eax), %edx
+; X86-NEXT: orl %ecx, %edx
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 8) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length9_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length9_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 4(%ecx), %esi
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 4(%eax), %esi
+; X86-NEXT: orl %edx, %esi
+; X86-NEXT: movb 8(%ecx), %cl
+; X86-NEXT: xorb 8(%eax), %cl
+; X86-NEXT: movzbl %cl, %eax
+; X86-NEXT: orl %esi, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 9) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length10_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length10_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 4(%ecx), %esi
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 4(%eax), %esi
+; X86-NEXT: orl %edx, %esi
+; X86-NEXT: movzwl 8(%ecx), %ecx
+; X86-NEXT: xorw 8(%eax), %cx
+; X86-NEXT: movzwl %cx, %eax
+; X86-NEXT: orl %esi, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 10) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length11_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length11_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 4(%ecx), %esi
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 4(%eax), %esi
+; X86-NEXT: orl %edx, %esi
+; X86-NEXT: movl 7(%ecx), %ecx
+; X86-NEXT: xorl 7(%eax), %ecx
+; X86-NEXT: orl %esi, %ecx
+; X86-NEXT: sete %al
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 11) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length12_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length12_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 4(%ecx), %esi
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 4(%eax), %esi
+; X86-NEXT: orl %edx, %esi
+; X86-NEXT: movl 8(%ecx), %ecx
+; X86-NEXT: xorl 8(%eax), %ecx
+; X86-NEXT: orl %esi, %ecx
+; X86-NEXT: setne %al
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 12) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length12(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length12:
+; X86: # %bb.0:
+; 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 .LBB29_3
+; X86-NEXT: # %bb.1: # %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 .LBB29_3
+; X86-NEXT: # %bb.2: # %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: je .LBB29_4
+; X86-NEXT: .LBB29_3: # %res_block
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: setae %al
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB29_4: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 12) nounwind
+ ret i32 %m
+}
+
+define i1 @length13_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length13_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 4(%ecx), %esi
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 4(%eax), %esi
+; X86-NEXT: orl %edx, %esi
+; X86-NEXT: movl 8(%ecx), %edx
+; X86-NEXT: xorl 8(%eax), %edx
+; X86-NEXT: movb 12(%ecx), %cl
+; X86-NEXT: xorb 12(%eax), %cl
+; X86-NEXT: movzbl %cl, %eax
+; X86-NEXT: orl %edx, %eax
+; X86-NEXT: orl %esi, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 13) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length14_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length14_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 4(%ecx), %esi
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 4(%eax), %esi
+; X86-NEXT: orl %edx, %esi
+; X86-NEXT: movl 8(%ecx), %edx
+; X86-NEXT: xorl 8(%eax), %edx
+; X86-NEXT: movzwl 12(%ecx), %ecx
+; X86-NEXT: xorw 12(%eax), %cx
+; X86-NEXT: movzwl %cx, %eax
+; X86-NEXT: orl %edx, %eax
+; X86-NEXT: orl %esi, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 14) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length15_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length15_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 4(%ecx), %esi
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 4(%eax), %esi
+; X86-NEXT: orl %edx, %esi
+; X86-NEXT: movl 8(%ecx), %edx
+; X86-NEXT: xorl 8(%eax), %edx
+; X86-NEXT: movl 11(%ecx), %ecx
+; X86-NEXT: xorl 11(%eax), %ecx
+; X86-NEXT: orl %edx, %ecx
+; X86-NEXT: orl %esi, %ecx
+; X86-NEXT: sete %al
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 15) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+; PR33329 - https://bugs.llvm.org/show_bug.cgi?id=33329
+
+define i32 @length16(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length16:
+; X86: # %bb.0:
+; 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 .LBB33_4
+; X86-NEXT: # %bb.1: # %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 .LBB33_4
+; X86-NEXT: # %bb.2: # %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 .LBB33_4
+; X86-NEXT: # %bb.3: # %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: je .LBB33_5
+; X86-NEXT: .LBB33_4: # %res_block
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: setae %al
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB33_5: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 16) nounwind
+ ret i32 %m
+}
+
+define i1 @length16_eq(i8* %x, i8* %y) nounwind {
+; X86-NOSSE-LABEL: length16_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl %esi
+; X86-NOSSE-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NOSSE-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NOSSE-NEXT: movl (%ecx), %edx
+; X86-NOSSE-NEXT: movl 4(%ecx), %esi
+; X86-NOSSE-NEXT: xorl (%eax), %edx
+; X86-NOSSE-NEXT: xorl 4(%eax), %esi
+; X86-NOSSE-NEXT: orl %edx, %esi
+; X86-NOSSE-NEXT: movl 8(%ecx), %edx
+; X86-NOSSE-NEXT: xorl 8(%eax), %edx
+; X86-NOSSE-NEXT: movl 12(%ecx), %ecx
+; X86-NOSSE-NEXT: xorl 12(%eax), %ecx
+; X86-NOSSE-NEXT: orl %edx, %ecx
+; X86-NOSSE-NEXT: orl %esi, %ecx
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: popl %esi
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length16_eq:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl %esi
+; X86-SSE1-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE1-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE1-NEXT: movl (%ecx), %edx
+; X86-SSE1-NEXT: movl 4(%ecx), %esi
+; X86-SSE1-NEXT: xorl (%eax), %edx
+; X86-SSE1-NEXT: xorl 4(%eax), %esi
+; X86-SSE1-NEXT: orl %edx, %esi
+; X86-SSE1-NEXT: movl 8(%ecx), %edx
+; X86-SSE1-NEXT: xorl 8(%eax), %edx
+; X86-SSE1-NEXT: movl 12(%ecx), %ecx
+; X86-SSE1-NEXT: xorl 12(%eax), %ecx
+; X86-SSE1-NEXT: orl %edx, %ecx
+; X86-SSE1-NEXT: orl %esi, %ecx
+; X86-SSE1-NEXT: setne %al
+; X86-SSE1-NEXT: popl %esi
+; X86-SSE1-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
+;
+; X86-SSE41-LABEL: length16_eq:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu (%eax), %xmm1
+; X86-SSE41-NEXT: pxor %xmm0, %xmm1
+; X86-SSE41-NEXT: ptest %xmm1, %xmm1
+; X86-SSE41-NEXT: setne %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 16) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length16_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length16_lt:
+; X86: # %bb.0:
+; 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 .LBB35_4
+; X86-NEXT: # %bb.1: # %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 .LBB35_4
+; X86-NEXT: # %bb.2: # %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 .LBB35_4
+; X86-NEXT: # %bb.3: # %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: je .LBB35_5
+; X86-NEXT: .LBB35_4: # %res_block
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: setae %al
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB35_5: # %endblock
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 16) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length16_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length16_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT: movl (%esi), %eax
+; X86-NEXT: movl (%edx), %ecx
+; X86-NEXT: bswapl %eax
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: cmpl %ecx, %eax
+; X86-NEXT: jne .LBB36_4
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movl 4(%esi), %eax
+; X86-NEXT: movl 4(%edx), %ecx
+; X86-NEXT: bswapl %eax
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: cmpl %ecx, %eax
+; X86-NEXT: jne .LBB36_4
+; X86-NEXT: # %bb.2: # %loadbb2
+; X86-NEXT: movl 8(%esi), %eax
+; X86-NEXT: movl 8(%edx), %ecx
+; X86-NEXT: bswapl %eax
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: cmpl %ecx, %eax
+; X86-NEXT: jne .LBB36_4
+; X86-NEXT: # %bb.3: # %loadbb3
+; X86-NEXT: movl 12(%esi), %eax
+; X86-NEXT: movl 12(%edx), %ecx
+; X86-NEXT: bswapl %eax
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: xorl %edx, %edx
+; X86-NEXT: cmpl %ecx, %eax
+; X86-NEXT: je .LBB36_5
+; X86-NEXT: .LBB36_4: # %res_block
+; X86-NEXT: xorl %edx, %edx
+; X86-NEXT: cmpl %ecx, %eax
+; X86-NEXT: setae %dl
+; X86-NEXT: leal -1(%edx,%edx), %edx
+; X86-NEXT: .LBB36_5: # %endblock
+; X86-NEXT: testl %edx, %edx
+; X86-NEXT: setg %al
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 16) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length16_eq_const(i8* %X) nounwind {
+; X86-NOSSE-LABEL: length16_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl %esi
+; X86-NOSSE-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NOSSE-NEXT: movl $858927408, %ecx # imm = 0x33323130
+; X86-NOSSE-NEXT: xorl (%eax), %ecx
+; X86-NOSSE-NEXT: movl $926299444, %edx # imm = 0x37363534
+; X86-NOSSE-NEXT: xorl 4(%eax), %edx
+; X86-NOSSE-NEXT: orl %ecx, %edx
+; X86-NOSSE-NEXT: movl $825243960, %ecx # imm = 0x31303938
+; X86-NOSSE-NEXT: xorl 8(%eax), %ecx
+; X86-NOSSE-NEXT: movl $892613426, %esi # imm = 0x35343332
+; X86-NOSSE-NEXT: xorl 12(%eax), %esi
+; X86-NOSSE-NEXT: orl %ecx, %esi
+; X86-NOSSE-NEXT: orl %edx, %esi
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: popl %esi
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length16_eq_const:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl %esi
+; X86-SSE1-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE1-NEXT: movl $858927408, %ecx # imm = 0x33323130
+; X86-SSE1-NEXT: xorl (%eax), %ecx
+; X86-SSE1-NEXT: movl $926299444, %edx # imm = 0x37363534
+; X86-SSE1-NEXT: xorl 4(%eax), %edx
+; X86-SSE1-NEXT: orl %ecx, %edx
+; X86-SSE1-NEXT: movl $825243960, %ecx # imm = 0x31303938
+; X86-SSE1-NEXT: xorl 8(%eax), %ecx
+; X86-SSE1-NEXT: movl $892613426, %esi # imm = 0x35343332
+; X86-SSE1-NEXT: xorl 12(%eax), %esi
+; X86-SSE1-NEXT: orl %ecx, %esi
+; X86-SSE1-NEXT: orl %edx, %esi
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: popl %esi
+; X86-SSE1-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[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length16_eq_const:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movdqu (%eax), %xmm0
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 16) nounwind
+ %c = icmp eq i32 %m, 0
+ 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 $24
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 24) nounwind
+ ret i32 %m
+}
+
+define i1 @length24_eq(i8* %x, i8* %y) nounwind {
+; X86-NOSSE-LABEL: length24_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $24
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length24_eq:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $24
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length24_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 8(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 8(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length24_eq:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 8(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 8(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: por %xmm2, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 24) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length24_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length24_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $24
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 24) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length24_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length24_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $24
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 24) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length24_eq_const(i8* %X) nounwind {
+; X86-NOSSE-LABEL: length24_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $24
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length24_eq_const:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $24
+; X86-SSE1-NEXT: pushl $.L.str
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: setne %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length24_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 8(%eax), %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length24_eq_const:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movdqu (%eax), %xmm0
+; X86-SSE41-NEXT: movdqu 8(%eax), %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE41-NEXT: por %xmm1, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: setne %al
+; X86-SSE41-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 24) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length31(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length31:
+; X86: # %bb.0:
+; X86-NEXT: pushl $31
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 31) nounwind
+ ret i32 %m
+}
+
+define i1 @length31_eq(i8* %x, i8* %y) nounwind {
+; X86-NOSSE-LABEL: length31_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $31
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length31_eq:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $31
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length31_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 15(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 15(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length31_eq:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 15(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 15(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: por %xmm2, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 31) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length31_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length31_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $31
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 31) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length31_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length31_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $31
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 31) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length31_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
+; X86-NOSSE-LABEL: length31_eq_prefer128:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $31
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length31_eq_prefer128:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $31
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length31_eq_prefer128:
+; 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 15(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 15(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length31_eq_prefer128:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 15(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 15(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: por %xmm2, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 31) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length31_eq_const(i8* %X) nounwind {
+; X86-NOSSE-LABEL: length31_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $31
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length31_eq_const:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $31
+; X86-SSE1-NEXT: pushl $.L.str
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: setne %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length31_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 15(%eax), %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length31_eq_const:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movdqu (%eax), %xmm0
+; X86-SSE41-NEXT: movdqu 15(%eax), %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE41-NEXT: por %xmm1, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: setne %al
+; X86-SSE41-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 31) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length32(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length32:
+; X86: # %bb.0:
+; X86-NEXT: pushl $32
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 32) nounwind
+ ret i32 %m
+}
+
+; PR33325 - https://bugs.llvm.org/show_bug.cgi?id=33325
+
+define i1 @length32_eq(i8* %x, i8* %y) nounwind {
+; X86-NOSSE-LABEL: length32_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $32
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length32_eq:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $32
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length32_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 16(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length32_eq:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: por %xmm2, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 32) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length32_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length32_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $32
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 32) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length32_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length32_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $32
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 32) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length32_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
+; X86-NOSSE-LABEL: length32_eq_prefer128:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $32
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length32_eq_prefer128:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $32
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length32_eq_prefer128:
+; 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 16(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length32_eq_prefer128:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: por %xmm2, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 32) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length32_eq_const(i8* %X) nounwind {
+; X86-NOSSE-LABEL: length32_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $32
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length32_eq_const:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $32
+; X86-SSE1-NEXT: pushl $.L.str
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: setne %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length32_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length32_eq_const:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movdqu (%eax), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE41-NEXT: por %xmm1, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: setne %al
+; X86-SSE41-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 32) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length48(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length48:
+; X86: # %bb.0:
+; X86-NEXT: pushl $48
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 48) nounwind
+ ret i32 %m
+}
+
+define i1 @length48_eq(i8* %x, i8* %y) nounwind {
+; X86-NOSSE-LABEL: length48_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $48
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length48_eq:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $48
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length48_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 16(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: movdqu 32(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu 32(%eax), %xmm3
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm3
+; X86-SSE2-NEXT: pand %xmm0, %xmm3
+; X86-SSE2-NEXT: pand %xmm2, %xmm3
+; X86-SSE2-NEXT: pmovmskb %xmm3, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length48_eq:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: movdqu 32(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu 32(%eax), %xmm3
+; X86-SSE41-NEXT: pxor %xmm1, %xmm3
+; X86-SSE41-NEXT: por %xmm0, %xmm3
+; X86-SSE41-NEXT: por %xmm2, %xmm3
+; X86-SSE41-NEXT: ptest %xmm3, %xmm3
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 48) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length48_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length48_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $48
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 48) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length48_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length48_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $48
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 48) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length48_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
+; X86-NOSSE-LABEL: length48_eq_prefer128:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $48
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length48_eq_prefer128:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $48
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length48_eq_prefer128:
+; 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 16(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: movdqu 32(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu 32(%eax), %xmm3
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm3
+; X86-SSE2-NEXT: pand %xmm0, %xmm3
+; X86-SSE2-NEXT: pand %xmm2, %xmm3
+; X86-SSE2-NEXT: pmovmskb %xmm3, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length48_eq_prefer128:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: movdqu 32(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu 32(%eax), %xmm3
+; X86-SSE41-NEXT: pxor %xmm1, %xmm3
+; X86-SSE41-NEXT: por %xmm0, %xmm3
+; X86-SSE41-NEXT: por %xmm2, %xmm3
+; X86-SSE41-NEXT: ptest %xmm3, %xmm3
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 48) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length48_eq_const(i8* %X) nounwind {
+; X86-NOSSE-LABEL: length48_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $48
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length48_eq_const:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $48
+; X86-SSE1-NEXT: pushl $.L.str
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: setne %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length48_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE2-NEXT: movdqu 32(%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
+; X86-SSE2-NEXT: pand %xmm1, %xmm2
+; X86-SSE2-NEXT: pand %xmm0, %xmm2
+; X86-SSE2-NEXT: pmovmskb %xmm2, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length48_eq_const:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movdqu (%eax), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE41-NEXT: movdqu 32(%eax), %xmm2
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
+; X86-SSE41-NEXT: por %xmm1, %xmm2
+; X86-SSE41-NEXT: por %xmm0, %xmm2
+; X86-SSE41-NEXT: ptest %xmm2, %xmm2
+; X86-SSE41-NEXT: setne %al
+; X86-SSE41-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 48) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length63(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length63:
+; X86: # %bb.0:
+; X86-NEXT: pushl $63
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 63) nounwind
+ ret i32 %m
+}
+
+define i1 @length63_eq(i8* %x, i8* %y) nounwind {
+; X86-NOSSE-LABEL: length63_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $63
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length63_eq:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $63
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: setne %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length63_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 16(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: movdqu 32(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu 32(%eax), %xmm3
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm3
+; X86-SSE2-NEXT: movdqu 47(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu 47(%eax), %xmm4
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm4
+; X86-SSE2-NEXT: pand %xmm3, %xmm4
+; X86-SSE2-NEXT: pand %xmm0, %xmm4
+; X86-SSE2-NEXT: pand %xmm2, %xmm4
+; X86-SSE2-NEXT: pmovmskb %xmm4, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length63_eq:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: movdqu 32(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu 32(%eax), %xmm3
+; X86-SSE41-NEXT: pxor %xmm1, %xmm3
+; X86-SSE41-NEXT: movdqu 47(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu 47(%eax), %xmm4
+; X86-SSE41-NEXT: pxor %xmm1, %xmm4
+; X86-SSE41-NEXT: por %xmm3, %xmm4
+; X86-SSE41-NEXT: por %xmm0, %xmm4
+; X86-SSE41-NEXT: por %xmm2, %xmm4
+; X86-SSE41-NEXT: ptest %xmm4, %xmm4
+; X86-SSE41-NEXT: setne %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 63) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length63_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length63_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $63
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 63) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length63_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length63_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $63
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 63) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length63_eq_const(i8* %X) nounwind {
+; X86-NOSSE-LABEL: length63_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $63
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length63_eq_const:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $63
+; X86-SSE1-NEXT: pushl $.L.str
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length63_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE2-NEXT: movdqu 32(%eax), %xmm2
+; X86-SSE2-NEXT: movdqu 47(%eax), %xmm3
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm3
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
+; X86-SSE2-NEXT: pand %xmm3, %xmm2
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pand %xmm2, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length63_eq_const:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movdqu (%eax), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE41-NEXT: movdqu 32(%eax), %xmm2
+; X86-SSE41-NEXT: movdqu 47(%eax), %xmm3
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm3
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
+; X86-SSE41-NEXT: por %xmm3, %xmm2
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE41-NEXT: por %xmm2, %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE41-NEXT: por %xmm1, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 63) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length64(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length64:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 64) nounwind
+ ret i32 %m
+}
+
+define i1 @length64_eq(i8* %x, i8* %y) nounwind {
+; X86-NOSSE-LABEL: length64_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $64
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length64_eq:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $64
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: setne %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length64_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 16(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: movdqu 32(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu 32(%eax), %xmm3
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm3
+; X86-SSE2-NEXT: movdqu 48(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu 48(%eax), %xmm4
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm4
+; X86-SSE2-NEXT: pand %xmm3, %xmm4
+; X86-SSE2-NEXT: pand %xmm0, %xmm4
+; X86-SSE2-NEXT: pand %xmm2, %xmm4
+; X86-SSE2-NEXT: pmovmskb %xmm4, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length64_eq:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: movdqu 32(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu 32(%eax), %xmm3
+; X86-SSE41-NEXT: pxor %xmm1, %xmm3
+; X86-SSE41-NEXT: movdqu 48(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu 48(%eax), %xmm4
+; X86-SSE41-NEXT: pxor %xmm1, %xmm4
+; X86-SSE41-NEXT: por %xmm3, %xmm4
+; X86-SSE41-NEXT: por %xmm0, %xmm4
+; X86-SSE41-NEXT: por %xmm2, %xmm4
+; X86-SSE41-NEXT: ptest %xmm4, %xmm4
+; X86-SSE41-NEXT: setne %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 64) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length64_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length64_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 64) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length64_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length64_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 64) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length64_eq_const(i8* %X) nounwind {
+; X86-NOSSE-LABEL: length64_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $64
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length64_eq_const:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $64
+; X86-SSE1-NEXT: pushl $.L.str
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length64_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE2-NEXT: movdqu 32(%eax), %xmm2
+; X86-SSE2-NEXT: movdqu 48(%eax), %xmm3
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm3
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
+; X86-SSE2-NEXT: pand %xmm3, %xmm2
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pand %xmm2, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length64_eq_const:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movdqu (%eax), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE41-NEXT: movdqu 32(%eax), %xmm2
+; X86-SSE41-NEXT: movdqu 48(%eax), %xmm3
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm3
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
+; X86-SSE41-NEXT: por %xmm3, %xmm2
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE41-NEXT: por %xmm2, %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE41-NEXT: por %xmm1, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 64) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length96(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length96:
+; X86: # %bb.0:
+; X86-NEXT: pushl $96
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 96) nounwind
+ ret i32 %m
+}
+
+define i1 @length96_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length96_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $96
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 96) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length96_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length96_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $96
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 96) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length96_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length96_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $96
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 96) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length96_eq_const(i8* %X) nounwind {
+; X86-LABEL: length96_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $96
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 96) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length127(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length127:
+; X86: # %bb.0:
+; X86-NEXT: pushl $127
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 127) nounwind
+ ret i32 %m
+}
+
+define i1 @length127_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length127_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $127
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 127) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length127_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length127_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $127
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 127) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length127_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length127_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $127
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 127) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length127_eq_const(i8* %X) nounwind {
+; X86-LABEL: length127_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $127
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 127) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length128(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length128:
+; X86: # %bb.0:
+; X86-NEXT: pushl $128
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 128) nounwind
+ ret i32 %m
+}
+
+define i1 @length128_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length128_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $128
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 128) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length128_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length128_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $128
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 128) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length128_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length128_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $128
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 128) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length128_eq_const(i8* %X) nounwind {
+; X86-LABEL: length128_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $128
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 128) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length192(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length192:
+; X86: # %bb.0:
+; X86-NEXT: pushl $192
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 192) nounwind
+ ret i32 %m
+}
+
+define i1 @length192_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length192_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $192
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 192) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length192_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length192_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $192
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 192) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length192_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length192_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $192
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 192) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length192_eq_const(i8* %X) nounwind {
+; X86-LABEL: length192_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $192
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 192) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length255(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length255:
+; X86: # %bb.0:
+; X86-NEXT: pushl $255
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 255) nounwind
+ ret i32 %m
+}
+
+define i1 @length255_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length255_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $255
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 255) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length255_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length255_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $255
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 255) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length255_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length255_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $255
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 255) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length255_eq_const(i8* %X) nounwind {
+; X86-LABEL: length255_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $255
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 255) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length256(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length256:
+; X86: # %bb.0:
+; X86-NEXT: pushl $256 # imm = 0x100
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 256) nounwind
+ ret i32 %m
+}
+
+define i1 @length256_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length256_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $256 # imm = 0x100
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 256) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length256_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length256_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $256 # imm = 0x100
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 256) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length256_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length256_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $256 # imm = 0x100
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 256) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length256_eq_const(i8* %X) nounwind {
+; X86-LABEL: length256_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $256 # imm = 0x100
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 256) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length384(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length384:
+; X86: # %bb.0:
+; X86-NEXT: pushl $384 # imm = 0x180
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 384) nounwind
+ ret i32 %m
+}
+
+define i1 @length384_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length384_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $384 # imm = 0x180
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 384) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length384_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length384_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $384 # imm = 0x180
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 384) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length384_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length384_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $384 # imm = 0x180
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 384) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length384_eq_const(i8* %X) nounwind {
+; X86-LABEL: length384_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $384 # imm = 0x180
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 384) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length511(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length511:
+; X86: # %bb.0:
+; X86-NEXT: pushl $511 # imm = 0x1FF
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 511) nounwind
+ ret i32 %m
+}
+
+define i1 @length511_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length511_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $511 # imm = 0x1FF
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 511) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length511_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length511_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $511 # imm = 0x1FF
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 511) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length511_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length511_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $511 # imm = 0x1FF
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 511) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length511_eq_const(i8* %X) nounwind {
+; X86-LABEL: length511_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $511 # imm = 0x1FF
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 511) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length512(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length512:
+; X86: # %bb.0:
+; X86-NEXT: pushl $512 # imm = 0x200
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 512) nounwind
+ ret i32 %m
+}
+
+define i1 @length512_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length512_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $512 # imm = 0x200
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 512) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length512_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length512_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $512 # imm = 0x200
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 512) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length512_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length512_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $512 # imm = 0x200
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 512) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length512_eq_const(i8* %X) nounwind {
+; X86-LABEL: length512_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $512 # imm = 0x200
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 512) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+; This checks that we do not do stupid things with huge sizes.
+define i32 @huge_length(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: huge_length:
+; X86: # %bb.0:
+; X86-NEXT: pushl $-1
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 9223372036854775807) nounwind
+ ret i32 %m
+}
+
+define i1 @huge_length_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: huge_length_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $-1
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 9223372036854775807) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+; This checks non-constant sizes.
+define i32 @nonconst_length(i8* %X, i8* %Y, i32 %size) nounwind {
+; X86-LABEL: nonconst_length:
+; X86: # %bb.0:
+; X86-NEXT: jmp memcmp # TAILCALL
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 %size) nounwind
+ ret i32 %m
+}
+
+define i1 @nonconst_length_eq(i8* %X, i8* %Y, i32 %size) nounwind {
+; X86-LABEL: nonconst_length_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 %size) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
diff --git a/llvm/test/CodeGen/X86/memcmp-more-load-pairs.ll b/llvm/test/CodeGen/X86/memcmp-more-load-pairs.ll
index 84b454776c325..9fc03d421bea4 100644
--- a/llvm/test/CodeGen/X86/memcmp-more-load-pairs.ll
+++ b/llvm/test/CodeGen/X86/memcmp-more-load-pairs.ll
@@ -1,9 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; NOTE: This is a copy of llvm/test/CodeGen/X86/memcmp.ll with more load pairs. Please keep it that way.
-; RUN: llc -max-loads-per-memcmp=4 -memcmp-num-loads-per-block=4 < %s -mtriple=i686-unknown-unknown -mattr=cmov | FileCheck %s --check-prefixes=X86,X86-NOSSE
-; RUN: llc -max-loads-per-memcmp=4 -memcmp-num-loads-per-block=4 < %s -mtriple=i686-unknown-unknown -mattr=+sse | FileCheck %s --check-prefixes=X86,X86-SSE1
-; RUN: llc -max-loads-per-memcmp=4 -memcmp-num-loads-per-block=4 < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=X86,X86-SSE2
-; RUN: llc -max-loads-per-memcmp=4 -memcmp-num-loads-per-block=4 < %s -mtriple=i686-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=X86,X86-SSE41
; RUN: llc -max-loads-per-memcmp=4 -memcmp-num-loads-per-block=4 < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=X64,X64-SSE,X64-SSE2
; RUN: llc -max-loads-per-memcmp=4 -memcmp-num-loads-per-block=4 < %s -mtriple=x86_64-unknown-unknown -mattr=sse4.1 | FileCheck %s --check-prefixes=X64,X64-SSE,X64-SSE41
; RUN: llc -max-loads-per-memcmp=4 -memcmp-num-loads-per-block=4 < %s -mtriple=x86_64-unknown-unknown -mattr=avx | FileCheck %s --check-prefixes=X64,X64-AVX,X64-AVX1
@@ -23,11 +19,6 @@
declare dso_local i32 @memcmp(i8*, i8*, i64)
define i32 @length0(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length0:
-; X86: # %bb.0:
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length0:
; X64: # %bb.0:
; X64-NEXT: xorl %eax, %eax
@@ -37,11 +28,6 @@ define i32 @length0(i8* %X, i8* %Y) nounwind {
}
define i1 @length0_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length0_eq:
-; X86: # %bb.0:
-; X86-NEXT: movb $1, %al
-; X86-NEXT: retl
-;
; X64-LABEL: length0_eq:
; X64: # %bb.0:
; X64-NEXT: movb $1, %al
@@ -52,11 +38,6 @@ define i1 @length0_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length0_lt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length0_lt:
-; X86: # %bb.0:
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length0_lt:
; X64: # %bb.0:
; X64-NEXT: xorl %eax, %eax
@@ -67,19 +48,6 @@ define i1 @length0_lt(i8* %X, i8* %Y) nounwind {
}
define i32 @length2(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: rolw $8, %cx
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: movzwl %cx, %eax
-; X86-NEXT: movzwl %dx, %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length2:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -95,15 +63,6 @@ define i32 @length2(i8* %X, i8* %Y) nounwind {
}
define i1 @length2_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: cmpw (%eax), %cx
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_eq:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -116,21 +75,6 @@ define i1 @length2_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length2_lt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2_lt:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: rolw $8, %cx
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: movzwl %cx, %eax
-; X86-NEXT: movzwl %dx, %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length2_lt:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -149,21 +93,6 @@ define i1 @length2_lt(i8* %X, i8* %Y) nounwind {
}
define i1 @length2_gt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2_gt:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: movzwl (%eax), %eax
-; X86-NEXT: rolw $8, %cx
-; X86-NEXT: rolw $8, %ax
-; X86-NEXT: movzwl %cx, %ecx
-; X86-NEXT: movzwl %ax, %eax
-; X86-NEXT: subl %eax, %ecx
-; X86-NEXT: testl %ecx, %ecx
-; X86-NEXT: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_gt:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -182,14 +111,6 @@ define i1 @length2_gt(i8* %X, i8* %Y) nounwind {
}
define i1 @length2_eq_const(i8* %X) nounwind {
-; X86-LABEL: length2_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movzwl (%eax), %eax
-; X86-NEXT: cmpl $12849, %eax # imm = 0x3231
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_eq_const:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -202,18 +123,6 @@ define i1 @length2_eq_const(i8* %X) nounwind {
}
define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2_eq_nobuiltin_attr:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $2
-; 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: length2_eq_nobuiltin_attr:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -229,30 +138,6 @@ define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind {
}
define i32 @length3(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length3:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: movzwl (%ecx), %esi
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: rolw $8, %si
-; X86-NEXT: cmpw %si, %dx
-; X86-NEXT: jne .LBB9_3
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movzbl 2(%eax), %eax
-; X86-NEXT: movzbl 2(%ecx), %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-; X86-NEXT: .LBB9_3: # %res_block
-; X86-NEXT: setae %al
-; X86-NEXT: movzbl %al, %eax
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length3:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -276,19 +161,6 @@ define i32 @length3(i8* %X, i8* %Y) nounwind {
}
define i1 @length3_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length3_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %edx
-; X86-NEXT: xorw (%eax), %dx
-; X86-NEXT: movb 2(%ecx), %cl
-; X86-NEXT: xorb 2(%eax), %cl
-; X86-NEXT: movzbl %cl, %eax
-; X86-NEXT: orw %dx, %ax
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length3_eq:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -305,20 +177,6 @@ define i1 @length3_eq(i8* %X, i8* %Y) nounwind {
}
define i32 @length4(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length4:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: seta %al
-; X86-NEXT: sbbl $0, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length4:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %ecx
@@ -335,15 +193,6 @@ define i32 @length4(i8* %X, i8* %Y) nounwind {
}
define i1 @length4_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length4_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: cmpl (%eax), %ecx
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_eq:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -356,22 +205,6 @@ define i1 @length4_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length4_lt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length4_lt:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: seta %al
-; X86-NEXT: sbbl $0, %eax
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length4_lt:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %ecx
@@ -391,22 +224,6 @@ define i1 @length4_lt(i8* %X, i8* %Y) nounwind {
}
define i1 @length4_gt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length4_gt:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: movl (%eax), %eax
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %eax
-; X86-NEXT: xorl %edx, %edx
-; X86-NEXT: cmpl %eax, %ecx
-; X86-NEXT: seta %dl
-; X86-NEXT: sbbl $0, %edx
-; X86-NEXT: testl %edx, %edx
-; X86-NEXT: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_gt:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -426,13 +243,6 @@ define i1 @length4_gt(i8* %X, i8* %Y) nounwind {
}
define i1 @length4_eq_const(i8* %X) nounwind {
-; X86-LABEL: length4_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: cmpl $875770417, (%eax) # imm = 0x34333231
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_eq_const:
; X64: # %bb.0:
; X64-NEXT: cmpl $875770417, (%rdi) # imm = 0x34333231
@@ -444,30 +254,6 @@ define i1 @length4_eq_const(i8* %X) nounwind {
}
define i32 @length5(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length5:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: movl (%ecx), %esi
-; X86-NEXT: bswapl %edx
-; X86-NEXT: bswapl %esi
-; X86-NEXT: cmpl %esi, %edx
-; X86-NEXT: jne .LBB16_3
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movzbl 4(%eax), %eax
-; X86-NEXT: movzbl 4(%ecx), %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-; X86-NEXT: .LBB16_3: # %res_block
-; X86-NEXT: setae %al
-; X86-NEXT: movzbl %al, %eax
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length5:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -491,19 +277,6 @@ define i32 @length5(i8* %X, i8* %Y) nounwind {
}
define i1 @length5_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length5_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: movb 4(%ecx), %cl
-; X86-NEXT: xorb 4(%eax), %cl
-; X86-NEXT: movzbl %cl, %eax
-; X86-NEXT: orl %edx, %eax
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length5_eq:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -520,32 +293,6 @@ define i1 @length5_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length5_lt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length5_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: movl (%ecx), %esi
-; X86-NEXT: bswapl %edx
-; X86-NEXT: bswapl %esi
-; X86-NEXT: cmpl %esi, %edx
-; X86-NEXT: jne .LBB18_3
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movzbl 4(%eax), %eax
-; X86-NEXT: movzbl 4(%ecx), %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: jmp .LBB18_2
-; X86-NEXT: .LBB18_3: # %res_block
-; X86-NEXT: setae %al
-; X86-NEXT: movzbl %al, %eax
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB18_2: # %endblock
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length5_lt:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -574,34 +321,6 @@ define i1 @length5_lt(i8* %X, i8* %Y) nounwind {
}
define i32 @length7(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length7:
-; X86: # %bb.0:
-; 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 .LBB19_2
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movl 3(%esi), %ecx
-; X86-NEXT: movl 3(%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: je .LBB19_3
-; X86-NEXT: .LBB19_2: # %res_block
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: setae %al
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB19_3: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length7:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %ecx
@@ -630,18 +349,6 @@ define i32 @length7(i8* %X, i8* %Y) nounwind {
}
define i1 @length7_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length7_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 3(%ecx), %ecx
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 3(%eax), %ecx
-; X86-NEXT: orl %edx, %ecx
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length7_eq:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -657,36 +364,6 @@ define i1 @length7_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length7_lt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length7_lt:
-; X86: # %bb.0:
-; 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 .LBB21_2
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movl 3(%esi), %ecx
-; X86-NEXT: movl 3(%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: je .LBB21_3
-; X86-NEXT: .LBB21_2: # %res_block
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: setae %al
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB21_3: # %endblock
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length7_lt:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %ecx
@@ -718,34 +395,6 @@ define i1 @length7_lt(i8* %X, i8* %Y) nounwind {
}
define i32 @length8(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length8:
-; X86: # %bb.0:
-; 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 .LBB22_2
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movl 4(%esi), %ecx
-; X86-NEXT: movl 4(%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: je .LBB22_3
-; X86-NEXT: .LBB22_2: # %res_block
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: setae %al
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB22_3: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length8:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -762,18 +411,6 @@ define i32 @length8(i8* %X, i8* %Y) nounwind {
}
define i1 @length8_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length8_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 4(%ecx), %ecx
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 4(%eax), %ecx
-; X86-NEXT: orl %edx, %ecx
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length8_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -786,17 +423,6 @@ define i1 @length8_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length8_eq_const(i8* %X) nounwind {
-; X86-LABEL: length8_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl $858927408, %ecx # imm = 0x33323130
-; X86-NEXT: xorl (%eax), %ecx
-; X86-NEXT: movl $926299444, %edx # imm = 0x37363534
-; X86-NEXT: xorl 4(%eax), %edx
-; X86-NEXT: orl %ecx, %edx
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length8_eq_const:
; X64: # %bb.0:
; X64-NEXT: movabsq $3978425819141910832, %rax # imm = 0x3736353433323130
@@ -809,24 +435,6 @@ define i1 @length8_eq_const(i8* %X) nounwind {
}
define i1 @length9_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length9_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 4(%ecx), %esi
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 4(%eax), %esi
-; X86-NEXT: orl %edx, %esi
-; X86-NEXT: movb 8(%ecx), %cl
-; X86-NEXT: xorb 8(%eax), %cl
-; X86-NEXT: movzbl %cl, %eax
-; X86-NEXT: orl %esi, %eax
-; X86-NEXT: sete %al
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length9_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -843,24 +451,6 @@ define i1 @length9_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length10_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length10_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 4(%ecx), %esi
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 4(%eax), %esi
-; X86-NEXT: orl %edx, %esi
-; X86-NEXT: movzwl 8(%ecx), %ecx
-; X86-NEXT: xorw 8(%eax), %cx
-; X86-NEXT: movzwl %cx, %eax
-; X86-NEXT: orl %esi, %eax
-; X86-NEXT: sete %al
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length10_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -877,23 +467,6 @@ define i1 @length10_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length11_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length11_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 4(%ecx), %esi
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 4(%eax), %esi
-; X86-NEXT: orl %edx, %esi
-; X86-NEXT: movl 7(%ecx), %ecx
-; X86-NEXT: xorl 7(%eax), %ecx
-; X86-NEXT: orl %esi, %ecx
-; X86-NEXT: sete %al
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length11_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -909,23 +482,6 @@ define i1 @length11_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length12_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length12_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 4(%ecx), %esi
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 4(%eax), %esi
-; X86-NEXT: orl %edx, %esi
-; X86-NEXT: movl 8(%ecx), %ecx
-; X86-NEXT: xorl 8(%eax), %ecx
-; X86-NEXT: orl %esi, %ecx
-; X86-NEXT: setne %al
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length12_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -941,41 +497,6 @@ define i1 @length12_eq(i8* %X, i8* %Y) nounwind {
}
define i32 @length12(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length12:
-; X86: # %bb.0:
-; 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 .LBB29_3
-; X86-NEXT: # %bb.1: # %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 .LBB29_3
-; X86-NEXT: # %bb.2: # %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: je .LBB29_4
-; X86-NEXT: .LBB29_3: # %res_block
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: setae %al
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB29_4: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length12:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -1004,27 +525,6 @@ define i32 @length12(i8* %X, i8* %Y) nounwind {
}
define i1 @length13_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length13_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 4(%ecx), %esi
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 4(%eax), %esi
-; X86-NEXT: orl %edx, %esi
-; X86-NEXT: movl 8(%ecx), %edx
-; X86-NEXT: xorl 8(%eax), %edx
-; X86-NEXT: movb 12(%ecx), %cl
-; X86-NEXT: xorb 12(%eax), %cl
-; X86-NEXT: movzbl %cl, %eax
-; X86-NEXT: orl %edx, %eax
-; X86-NEXT: orl %esi, %eax
-; X86-NEXT: sete %al
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length13_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -1040,27 +540,6 @@ define i1 @length13_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length14_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length14_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 4(%ecx), %esi
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 4(%eax), %esi
-; X86-NEXT: orl %edx, %esi
-; X86-NEXT: movl 8(%ecx), %edx
-; X86-NEXT: xorl 8(%eax), %edx
-; X86-NEXT: movzwl 12(%ecx), %ecx
-; X86-NEXT: xorw 12(%eax), %cx
-; X86-NEXT: movzwl %cx, %eax
-; X86-NEXT: orl %edx, %eax
-; X86-NEXT: orl %esi, %eax
-; X86-NEXT: sete %al
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length14_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -1076,26 +555,6 @@ define i1 @length14_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length15_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length15_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 4(%ecx), %esi
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 4(%eax), %esi
-; X86-NEXT: orl %edx, %esi
-; X86-NEXT: movl 8(%ecx), %edx
-; X86-NEXT: xorl 8(%eax), %edx
-; X86-NEXT: movl 11(%ecx), %ecx
-; X86-NEXT: xorl 11(%eax), %ecx
-; X86-NEXT: orl %edx, %ecx
-; X86-NEXT: orl %esi, %ecx
-; X86-NEXT: sete %al
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length15_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -1113,48 +572,6 @@ define i1 @length15_eq(i8* %X, i8* %Y) nounwind {
; PR33329 - https://bugs.llvm.org/show_bug.cgi?id=33329
define i32 @length16(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length16:
-; X86: # %bb.0:
-; 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 .LBB33_4
-; X86-NEXT: # %bb.1: # %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 .LBB33_4
-; X86-NEXT: # %bb.2: # %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 .LBB33_4
-; X86-NEXT: # %bb.3: # %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: je .LBB33_5
-; X86-NEXT: .LBB33_4: # %res_block
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: setae %al
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB33_5: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length16:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -1183,69 +600,6 @@ define i32 @length16(i8* %X, i8* %Y) nounwind {
}
define i1 @length16_eq(i8* %x, i8* %y) nounwind {
-; X86-NOSSE-LABEL: length16_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl %esi
-; X86-NOSSE-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NOSSE-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NOSSE-NEXT: movl (%ecx), %edx
-; X86-NOSSE-NEXT: movl 4(%ecx), %esi
-; X86-NOSSE-NEXT: xorl (%eax), %edx
-; X86-NOSSE-NEXT: xorl 4(%eax), %esi
-; X86-NOSSE-NEXT: orl %edx, %esi
-; X86-NOSSE-NEXT: movl 8(%ecx), %edx
-; X86-NOSSE-NEXT: xorl 8(%eax), %edx
-; X86-NOSSE-NEXT: movl 12(%ecx), %ecx
-; X86-NOSSE-NEXT: xorl 12(%eax), %ecx
-; X86-NOSSE-NEXT: orl %edx, %ecx
-; X86-NOSSE-NEXT: orl %esi, %ecx
-; X86-NOSSE-NEXT: setne %al
-; X86-NOSSE-NEXT: popl %esi
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length16_eq:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl %esi
-; X86-SSE1-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE1-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE1-NEXT: movl (%ecx), %edx
-; X86-SSE1-NEXT: movl 4(%ecx), %esi
-; X86-SSE1-NEXT: xorl (%eax), %edx
-; X86-SSE1-NEXT: xorl 4(%eax), %esi
-; X86-SSE1-NEXT: orl %edx, %esi
-; X86-SSE1-NEXT: movl 8(%ecx), %edx
-; X86-SSE1-NEXT: xorl 8(%eax), %edx
-; X86-SSE1-NEXT: movl 12(%ecx), %ecx
-; X86-SSE1-NEXT: xorl 12(%eax), %ecx
-; X86-SSE1-NEXT: orl %edx, %ecx
-; X86-SSE1-NEXT: orl %esi, %ecx
-; X86-SSE1-NEXT: setne %al
-; X86-SSE1-NEXT: popl %esi
-; X86-SSE1-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
-;
-; X86-SSE41-LABEL: length16_eq:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu (%eax), %xmm1
-; X86-SSE41-NEXT: pxor %xmm0, %xmm1
-; X86-SSE41-NEXT: ptest %xmm1, %xmm1
-; X86-SSE41-NEXT: setne %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length16_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -1288,50 +642,6 @@ define i1 @length16_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length16_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length16_lt:
-; X86: # %bb.0:
-; 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 .LBB35_4
-; X86-NEXT: # %bb.1: # %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 .LBB35_4
-; X86-NEXT: # %bb.2: # %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 .LBB35_4
-; X86-NEXT: # %bb.3: # %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: je .LBB35_5
-; X86-NEXT: .LBB35_4: # %res_block
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: setae %al
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB35_5: # %endblock
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length16_lt:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -1363,50 +673,6 @@ define i1 @length16_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length16_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length16_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %edx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %esi
-; X86-NEXT: movl (%esi), %eax
-; X86-NEXT: movl (%edx), %ecx
-; X86-NEXT: bswapl %eax
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: cmpl %ecx, %eax
-; X86-NEXT: jne .LBB36_4
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movl 4(%esi), %eax
-; X86-NEXT: movl 4(%edx), %ecx
-; X86-NEXT: bswapl %eax
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: cmpl %ecx, %eax
-; X86-NEXT: jne .LBB36_4
-; X86-NEXT: # %bb.2: # %loadbb2
-; X86-NEXT: movl 8(%esi), %eax
-; X86-NEXT: movl 8(%edx), %ecx
-; X86-NEXT: bswapl %eax
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: cmpl %ecx, %eax
-; X86-NEXT: jne .LBB36_4
-; X86-NEXT: # %bb.3: # %loadbb3
-; X86-NEXT: movl 12(%esi), %eax
-; X86-NEXT: movl 12(%edx), %ecx
-; X86-NEXT: bswapl %eax
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: xorl %edx, %edx
-; X86-NEXT: cmpl %ecx, %eax
-; X86-NEXT: je .LBB36_5
-; X86-NEXT: .LBB36_4: # %res_block
-; X86-NEXT: xorl %edx, %edx
-; X86-NEXT: cmpl %ecx, %eax
-; X86-NEXT: setae %dl
-; X86-NEXT: leal -1(%edx,%edx), %edx
-; X86-NEXT: .LBB36_5: # %endblock
-; X86-NEXT: testl %edx, %edx
-; X86-NEXT: setg %al
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length16_gt:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -1438,63 +704,6 @@ define i1 @length16_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length16_eq_const(i8* %X) nounwind {
-; X86-NOSSE-LABEL: length16_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl %esi
-; X86-NOSSE-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NOSSE-NEXT: movl $858927408, %ecx # imm = 0x33323130
-; X86-NOSSE-NEXT: xorl (%eax), %ecx
-; X86-NOSSE-NEXT: movl $926299444, %edx # imm = 0x37363534
-; X86-NOSSE-NEXT: xorl 4(%eax), %edx
-; X86-NOSSE-NEXT: orl %ecx, %edx
-; X86-NOSSE-NEXT: movl $825243960, %ecx # imm = 0x31303938
-; X86-NOSSE-NEXT: xorl 8(%eax), %ecx
-; X86-NOSSE-NEXT: movl $892613426, %esi # imm = 0x35343332
-; X86-NOSSE-NEXT: xorl 12(%eax), %esi
-; X86-NOSSE-NEXT: orl %ecx, %esi
-; X86-NOSSE-NEXT: orl %edx, %esi
-; X86-NOSSE-NEXT: sete %al
-; X86-NOSSE-NEXT: popl %esi
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length16_eq_const:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl %esi
-; X86-SSE1-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE1-NEXT: movl $858927408, %ecx # imm = 0x33323130
-; X86-SSE1-NEXT: xorl (%eax), %ecx
-; X86-SSE1-NEXT: movl $926299444, %edx # imm = 0x37363534
-; X86-SSE1-NEXT: xorl 4(%eax), %edx
-; X86-SSE1-NEXT: orl %ecx, %edx
-; X86-SSE1-NEXT: movl $825243960, %ecx # imm = 0x31303938
-; X86-SSE1-NEXT: xorl 8(%eax), %ecx
-; X86-SSE1-NEXT: movl $892613426, %esi # imm = 0x35343332
-; X86-SSE1-NEXT: xorl 12(%eax), %esi
-; X86-SSE1-NEXT: orl %ecx, %esi
-; X86-SSE1-NEXT: orl %edx, %esi
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: popl %esi
-; X86-SSE1-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[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length16_eq_const:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movdqu (%eax), %xmm0
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length16_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -1537,16 +746,6 @@ define i1 @length16_eq_const(i8* %X) nounwind {
; 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: movq (%rdi), %rcx
@@ -1582,61 +781,6 @@ define i32 @length24(i8* %X, i8* %Y) nounwind {
}
define i1 @length24_eq(i8* %x, i8* %y) nounwind {
-; X86-NOSSE-LABEL: length24_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $24
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length24_eq:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $24
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length24_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 8(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 8(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length24_eq:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 8(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 8(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: por %xmm2, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length24_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -1694,18 +838,6 @@ define i1 @length24_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length24_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length24_lt:
-; 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: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length24_lt:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -1744,18 +876,6 @@ define i1 @length24_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length24_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length24_gt:
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length24_gt:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -1794,55 +914,6 @@ define i1 @length24_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length24_eq_const(i8* %X) nounwind {
-; X86-NOSSE-LABEL: length24_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $24
-; 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: setne %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length24_eq_const:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $24
-; X86-SSE1-NEXT: pushl $.L.str
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: setne %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length24_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 8(%eax), %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length24_eq_const:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movdqu (%eax), %xmm0
-; X86-SSE41-NEXT: movdqu 8(%eax), %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE41-NEXT: por %xmm1, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: setne %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length24_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -1895,16 +966,6 @@ define i1 @length24_eq_const(i8* %X) nounwind {
}
define i32 @length31(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length31:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $31
-; 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: length31:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -1947,61 +1008,6 @@ define i32 @length31(i8* %X, i8* %Y) nounwind {
}
define i1 @length31_eq(i8* %x, i8* %y) nounwind {
-; X86-NOSSE-LABEL: length31_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $31
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length31_eq:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $31
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length31_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 15(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 15(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length31_eq:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 15(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 15(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: por %xmm2, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length31_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -2058,18 +1064,6 @@ define i1 @length31_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length31_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length31_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $31
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length31_lt:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -2115,18 +1109,6 @@ define i1 @length31_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length31_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length31_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $31
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length31_gt:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -2172,61 +1154,6 @@ define i1 @length31_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length31_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
-; X86-NOSSE-LABEL: length31_eq_prefer128:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $31
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length31_eq_prefer128:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $31
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length31_eq_prefer128:
-; 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 15(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 15(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length31_eq_prefer128:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 15(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 15(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: por %xmm2, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length31_eq_prefer128:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -2283,55 +1210,6 @@ define i1 @length31_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"=
}
define i1 @length31_eq_const(i8* %X) nounwind {
-; X86-NOSSE-LABEL: length31_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $31
-; 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: setne %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length31_eq_const:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $31
-; X86-SSE1-NEXT: pushl $.L.str
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: setne %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length31_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 15(%eax), %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length31_eq_const:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movdqu (%eax), %xmm0
-; X86-SSE41-NEXT: movdqu 15(%eax), %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE41-NEXT: por %xmm1, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: setne %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length31_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -2384,16 +1262,6 @@ define i1 @length31_eq_const(i8* %X) nounwind {
}
define i32 @length32(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length32:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $32
-; 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: length32:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -2438,61 +1306,6 @@ define i32 @length32(i8* %X, i8* %Y) nounwind {
; PR33325 - https://bugs.llvm.org/show_bug.cgi?id=33325
define i1 @length32_eq(i8* %x, i8* %y) nounwind {
-; X86-NOSSE-LABEL: length32_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $32
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length32_eq:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $32
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length32_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 16(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length32_eq:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: por %xmm2, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length32_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -2562,18 +1375,6 @@ define i1 @length32_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length32_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length32_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $32
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length32_lt:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -2619,18 +1420,6 @@ define i1 @length32_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length32_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length32_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $32
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length32_gt:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -2676,61 +1465,6 @@ define i1 @length32_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length32_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
-; X86-NOSSE-LABEL: length32_eq_prefer128:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $32
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length32_eq_prefer128:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $32
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length32_eq_prefer128:
-; 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 16(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length32_eq_prefer128:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: por %xmm2, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length32_eq_prefer128:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -2787,55 +1521,6 @@ define i1 @length32_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"=
}
define i1 @length32_eq_const(i8* %X) nounwind {
-; X86-NOSSE-LABEL: length32_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $32
-; 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: setne %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length32_eq_const:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $32
-; X86-SSE1-NEXT: pushl $.L.str
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: setne %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length32_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length32_eq_const:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movdqu (%eax), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE41-NEXT: por %xmm1, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: setne %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length32_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -2901,16 +1586,6 @@ define i1 @length32_eq_const(i8* %X) nounwind {
}
define i32 @length48(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length48:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $48
-; 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: length48:
; X64: # %bb.0:
; X64-NEXT: movl $48, %edx
@@ -2920,69 +1595,6 @@ define i32 @length48(i8* %X, i8* %Y) nounwind {
}
define i1 @length48_eq(i8* %x, i8* %y) nounwind {
-; X86-NOSSE-LABEL: length48_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $48
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length48_eq:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $48
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length48_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 16(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: movdqu 32(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu 32(%eax), %xmm3
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm3
-; X86-SSE2-NEXT: pand %xmm0, %xmm3
-; X86-SSE2-NEXT: pand %xmm2, %xmm3
-; X86-SSE2-NEXT: pmovmskb %xmm3, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length48_eq:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: movdqu 32(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu 32(%eax), %xmm3
-; X86-SSE41-NEXT: pxor %xmm1, %xmm3
-; X86-SSE41-NEXT: por %xmm0, %xmm3
-; X86-SSE41-NEXT: por %xmm2, %xmm3
-; X86-SSE41-NEXT: ptest %xmm3, %xmm3
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length48_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -3075,18 +1687,6 @@ define i1 @length48_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length48_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length48_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $48
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length48_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3102,18 +1702,6 @@ define i1 @length48_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length48_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length48_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $48
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length48_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3129,69 +1717,6 @@ define i1 @length48_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length48_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
-; X86-NOSSE-LABEL: length48_eq_prefer128:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $48
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length48_eq_prefer128:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $48
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length48_eq_prefer128:
-; 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 16(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: movdqu 32(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu 32(%eax), %xmm3
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm3
-; X86-SSE2-NEXT: pand %xmm0, %xmm3
-; X86-SSE2-NEXT: pand %xmm2, %xmm3
-; X86-SSE2-NEXT: pmovmskb %xmm3, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length48_eq_prefer128:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: movdqu 32(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu 32(%eax), %xmm3
-; X86-SSE41-NEXT: pxor %xmm1, %xmm3
-; X86-SSE41-NEXT: por %xmm0, %xmm3
-; X86-SSE41-NEXT: por %xmm2, %xmm3
-; X86-SSE41-NEXT: ptest %xmm3, %xmm3
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length48_eq_prefer128:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -3263,61 +1788,6 @@ define i1 @length48_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"=
}
define i1 @length48_eq_const(i8* %X) nounwind {
-; X86-NOSSE-LABEL: length48_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $48
-; 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: setne %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length48_eq_const:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $48
-; X86-SSE1-NEXT: pushl $.L.str
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: setne %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length48_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE2-NEXT: movdqu 32(%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
-; X86-SSE2-NEXT: pand %xmm1, %xmm2
-; X86-SSE2-NEXT: pand %xmm0, %xmm2
-; X86-SSE2-NEXT: pmovmskb %xmm2, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length48_eq_const:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movdqu (%eax), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE41-NEXT: movdqu 32(%eax), %xmm2
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
-; X86-SSE41-NEXT: por %xmm1, %xmm2
-; X86-SSE41-NEXT: por %xmm0, %xmm2
-; X86-SSE41-NEXT: ptest %xmm2, %xmm2
-; X86-SSE41-NEXT: setne %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length48_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -3401,16 +1871,6 @@ define i1 @length48_eq_const(i8* %X) nounwind {
}
define i32 @length63(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length63:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $63
-; 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: length63:
; X64: # %bb.0:
; X64-NEXT: movl $63, %edx
@@ -3420,77 +1880,6 @@ define i32 @length63(i8* %X, i8* %Y) nounwind {
}
define i1 @length63_eq(i8* %x, i8* %y) nounwind {
-; X86-NOSSE-LABEL: length63_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $63
-; 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-SSE1-LABEL: length63_eq:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $63
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: setne %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length63_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 16(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: movdqu 32(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu 32(%eax), %xmm3
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm3
-; X86-SSE2-NEXT: movdqu 47(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu 47(%eax), %xmm4
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm4
-; X86-SSE2-NEXT: pand %xmm3, %xmm4
-; X86-SSE2-NEXT: pand %xmm0, %xmm4
-; X86-SSE2-NEXT: pand %xmm2, %xmm4
-; X86-SSE2-NEXT: pmovmskb %xmm4, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length63_eq:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: movdqu 32(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu 32(%eax), %xmm3
-; X86-SSE41-NEXT: pxor %xmm1, %xmm3
-; X86-SSE41-NEXT: movdqu 47(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu 47(%eax), %xmm4
-; X86-SSE41-NEXT: pxor %xmm1, %xmm4
-; X86-SSE41-NEXT: por %xmm3, %xmm4
-; X86-SSE41-NEXT: por %xmm0, %xmm4
-; X86-SSE41-NEXT: por %xmm2, %xmm4
-; X86-SSE41-NEXT: ptest %xmm4, %xmm4
-; X86-SSE41-NEXT: setne %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length63_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -3588,18 +1977,6 @@ define i1 @length63_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length63_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length63_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $63
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length63_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3615,18 +1992,6 @@ define i1 @length63_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length63_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length63_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $63
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length63_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3642,67 +2007,6 @@ define i1 @length63_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length63_eq_const(i8* %X) nounwind {
-; X86-NOSSE-LABEL: length63_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $63
-; 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-SSE1-LABEL: length63_eq_const:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $63
-; X86-SSE1-NEXT: pushl $.L.str
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length63_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE2-NEXT: movdqu 32(%eax), %xmm2
-; X86-SSE2-NEXT: movdqu 47(%eax), %xmm3
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm3
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
-; X86-SSE2-NEXT: pand %xmm3, %xmm2
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pand %xmm2, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length63_eq_const:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movdqu (%eax), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE41-NEXT: movdqu 32(%eax), %xmm2
-; X86-SSE41-NEXT: movdqu 47(%eax), %xmm3
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm3
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
-; X86-SSE41-NEXT: por %xmm3, %xmm2
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE41-NEXT: por %xmm2, %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE41-NEXT: por %xmm1, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length63_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -3792,16 +2096,6 @@ define i1 @length63_eq_const(i8* %X) nounwind {
}
define i32 @length64(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length64:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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: length64:
; X64: # %bb.0:
; X64-NEXT: movl $64, %edx
@@ -3811,77 +2105,6 @@ define i32 @length64(i8* %X, i8* %Y) nounwind {
}
define i1 @length64_eq(i8* %x, i8* %y) nounwind {
-; X86-NOSSE-LABEL: length64_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $64
-; 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-SSE1-LABEL: length64_eq:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $64
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: setne %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length64_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 16(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: movdqu 32(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu 32(%eax), %xmm3
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm3
-; X86-SSE2-NEXT: movdqu 48(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu 48(%eax), %xmm4
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm4
-; X86-SSE2-NEXT: pand %xmm3, %xmm4
-; X86-SSE2-NEXT: pand %xmm0, %xmm4
-; X86-SSE2-NEXT: pand %xmm2, %xmm4
-; X86-SSE2-NEXT: pmovmskb %xmm4, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length64_eq:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: movdqu 32(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu 32(%eax), %xmm3
-; X86-SSE41-NEXT: pxor %xmm1, %xmm3
-; X86-SSE41-NEXT: movdqu 48(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu 48(%eax), %xmm4
-; X86-SSE41-NEXT: pxor %xmm1, %xmm4
-; X86-SSE41-NEXT: por %xmm3, %xmm4
-; X86-SSE41-NEXT: por %xmm0, %xmm4
-; X86-SSE41-NEXT: por %xmm2, %xmm4
-; X86-SSE41-NEXT: ptest %xmm4, %xmm4
-; X86-SSE41-NEXT: setne %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length64_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -3994,18 +2217,6 @@ define i1 @length64_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length64_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length64_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length64_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4021,18 +2232,6 @@ define i1 @length64_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length64_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length64_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length64_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4048,67 +2247,6 @@ define i1 @length64_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length64_eq_const(i8* %X) nounwind {
-; X86-NOSSE-LABEL: length64_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $64
-; 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-SSE1-LABEL: length64_eq_const:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $64
-; X86-SSE1-NEXT: pushl $.L.str
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length64_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE2-NEXT: movdqu 32(%eax), %xmm2
-; X86-SSE2-NEXT: movdqu 48(%eax), %xmm3
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm3
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
-; X86-SSE2-NEXT: pand %xmm3, %xmm2
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pand %xmm2, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length64_eq_const:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movdqu (%eax), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE41-NEXT: movdqu 32(%eax), %xmm2
-; X86-SSE41-NEXT: movdqu 48(%eax), %xmm3
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm3
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm2
-; X86-SSE41-NEXT: por %xmm3, %xmm2
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE41-NEXT: por %xmm2, %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE41-NEXT: por %xmm1, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length64_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -4213,16 +2351,6 @@ define i1 @length64_eq_const(i8* %X) nounwind {
}
define i32 @length96(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length96:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $96
-; 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: length96:
; X64: # %bb.0:
; X64-NEXT: movl $96, %edx
@@ -4232,18 +2360,6 @@ define i32 @length96(i8* %X, i8* %Y) nounwind {
}
define i1 @length96_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length96_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $96
-; 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
-;
; X64-SSE-LABEL: length96_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -4342,18 +2458,6 @@ define i1 @length96_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length96_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length96_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $96
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length96_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4369,18 +2473,6 @@ define i1 @length96_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length96_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length96_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $96
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length96_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4396,18 +2488,6 @@ define i1 @length96_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length96_eq_const(i8* %X) nounwind {
-; X86-LABEL: length96_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $96
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE-LABEL: length96_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -4504,16 +2584,6 @@ define i1 @length96_eq_const(i8* %X) nounwind {
}
define i32 @length127(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length127:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $127
-; 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: length127:
; X64: # %bb.0:
; X64-NEXT: movl $127, %edx
@@ -4523,18 +2593,6 @@ define i32 @length127(i8* %X, i8* %Y) nounwind {
}
define i1 @length127_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length127_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $127
-; 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
-;
; X64-SSE-LABEL: length127_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -4640,18 +2698,6 @@ define i1 @length127_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length127_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length127_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $127
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length127_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4667,18 +2713,6 @@ define i1 @length127_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length127_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length127_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $127
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length127_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4694,18 +2728,6 @@ define i1 @length127_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length127_eq_const(i8* %X) nounwind {
-; X86-LABEL: length127_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $127
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE-LABEL: length127_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -4812,16 +2834,6 @@ define i1 @length127_eq_const(i8* %X) nounwind {
}
define i32 @length128(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length128:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $128
-; 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: length128:
; X64: # %bb.0:
; X64-NEXT: movl $128, %edx
@@ -4831,18 +2843,6 @@ define i32 @length128(i8* %X, i8* %Y) nounwind {
}
define i1 @length128_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length128_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $128
-; 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
-;
; X64-SSE-LABEL: length128_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -4948,18 +2948,6 @@ define i1 @length128_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length128_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length128_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $128
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length128_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4975,18 +2963,6 @@ define i1 @length128_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length128_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length128_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $128
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length128_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -5002,18 +2978,6 @@ define i1 @length128_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length128_eq_const(i8* %X) nounwind {
-; X86-LABEL: length128_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $128
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE-LABEL: length128_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -5120,16 +3084,6 @@ define i1 @length128_eq_const(i8* %X) nounwind {
}
define i32 @length192(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length192:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $192
-; 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: length192:
; X64: # %bb.0:
; X64-NEXT: movl $192, %edx
@@ -5139,18 +3093,6 @@ define i32 @length192(i8* %X, i8* %Y) nounwind {
}
define i1 @length192_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length192_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $192
-; 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
-;
; X64-SSE-LABEL: length192_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -5238,18 +3180,6 @@ define i1 @length192_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length192_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length192_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $192
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length192_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -5265,18 +3195,6 @@ define i1 @length192_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length192_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length192_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $192
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length192_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -5292,18 +3210,6 @@ define i1 @length192_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length192_eq_const(i8* %X) nounwind {
-; X86-LABEL: length192_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $192
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE-LABEL: length192_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -5395,16 +3301,6 @@ define i1 @length192_eq_const(i8* %X) nounwind {
}
define i32 @length255(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length255:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $255
-; 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: length255:
; X64: # %bb.0:
; X64-NEXT: movl $255, %edx
@@ -5414,18 +3310,6 @@ define i32 @length255(i8* %X, i8* %Y) nounwind {
}
define i1 @length255_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length255_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $255
-; 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
-;
; X64-SSE-LABEL: length255_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -5522,18 +3406,6 @@ define i1 @length255_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length255_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length255_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $255
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length255_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -5549,18 +3421,6 @@ define i1 @length255_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length255_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length255_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $255
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length255_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -5576,18 +3436,6 @@ define i1 @length255_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length255_eq_const(i8* %X) nounwind {
-; X86-LABEL: length255_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $255
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE-LABEL: length255_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -5688,16 +3536,6 @@ define i1 @length255_eq_const(i8* %X) nounwind {
}
define i32 @length256(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length256:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $256 # imm = 0x100
-; 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: length256:
; X64: # %bb.0:
; X64-NEXT: movl $256, %edx # imm = 0x100
@@ -5707,18 +3545,6 @@ define i32 @length256(i8* %X, i8* %Y) nounwind {
}
define i1 @length256_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length256_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $256 # imm = 0x100
-; 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
-;
; X64-SSE-LABEL: length256_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -5815,18 +3641,6 @@ define i1 @length256_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length256_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length256_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $256 # imm = 0x100
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length256_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -5842,18 +3656,6 @@ define i1 @length256_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length256_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length256_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $256 # imm = 0x100
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length256_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -5869,18 +3671,6 @@ define i1 @length256_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length256_eq_const(i8* %X) nounwind {
-; X86-LABEL: length256_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $256 # imm = 0x100
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE-LABEL: length256_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -5981,16 +3771,6 @@ define i1 @length256_eq_const(i8* %X) nounwind {
}
define i32 @length384(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length384:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $384 # imm = 0x180
-; 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: length384:
; X64: # %bb.0:
; X64-NEXT: movl $384, %edx # imm = 0x180
@@ -6000,18 +3780,6 @@ define i32 @length384(i8* %X, i8* %Y) nounwind {
}
define i1 @length384_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length384_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $384 # imm = 0x180
-; 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
-;
; X64-LABEL: length384_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6027,18 +3795,6 @@ define i1 @length384_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length384_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length384_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $384 # imm = 0x180
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length384_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6054,18 +3810,6 @@ define i1 @length384_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length384_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length384_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $384 # imm = 0x180
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length384_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6081,18 +3825,6 @@ define i1 @length384_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length384_eq_const(i8* %X) nounwind {
-; X86-LABEL: length384_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $384 # imm = 0x180
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length384_eq_const:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6109,16 +3841,6 @@ define i1 @length384_eq_const(i8* %X) nounwind {
}
define i32 @length511(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length511:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $511 # imm = 0x1FF
-; 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: length511:
; X64: # %bb.0:
; X64-NEXT: movl $511, %edx # imm = 0x1FF
@@ -6128,18 +3850,6 @@ define i32 @length511(i8* %X, i8* %Y) nounwind {
}
define i1 @length511_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length511_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $511 # imm = 0x1FF
-; 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
-;
; X64-LABEL: length511_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6155,18 +3865,6 @@ define i1 @length511_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length511_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length511_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $511 # imm = 0x1FF
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length511_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6182,18 +3880,6 @@ define i1 @length511_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length511_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length511_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $511 # imm = 0x1FF
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length511_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6209,18 +3895,6 @@ define i1 @length511_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length511_eq_const(i8* %X) nounwind {
-; X86-LABEL: length511_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $511 # imm = 0x1FF
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length511_eq_const:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6237,16 +3911,6 @@ define i1 @length511_eq_const(i8* %X) nounwind {
}
define i32 @length512(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length512:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $512 # imm = 0x200
-; 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: length512:
; X64: # %bb.0:
; X64-NEXT: movl $512, %edx # imm = 0x200
@@ -6256,18 +3920,6 @@ define i32 @length512(i8* %X, i8* %Y) nounwind {
}
define i1 @length512_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length512_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $512 # imm = 0x200
-; 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
-;
; X64-LABEL: length512_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6283,18 +3935,6 @@ define i1 @length512_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length512_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length512_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $512 # imm = 0x200
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length512_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6310,18 +3950,6 @@ define i1 @length512_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length512_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length512_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $512 # imm = 0x200
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length512_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6337,18 +3965,6 @@ define i1 @length512_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length512_eq_const(i8* %X) nounwind {
-; X86-LABEL: length512_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $512 # imm = 0x200
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length512_eq_const:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6366,16 +3982,6 @@ define i1 @length512_eq_const(i8* %X) nounwind {
; This checks that we do not do stupid things with huge sizes.
define i32 @huge_length(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: huge_length:
-; X86: # %bb.0:
-; X86-NEXT: pushl $2147483647 # imm = 0x7FFFFFFF
-; X86-NEXT: pushl $-1
-; 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: huge_length:
; X64: # %bb.0:
; X64-NEXT: movabsq $9223372036854775807, %rdx # imm = 0x7FFFFFFFFFFFFFFF
@@ -6385,18 +3991,6 @@ define i32 @huge_length(i8* %X, i8* %Y) nounwind {
}
define i1 @huge_length_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: huge_length_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $2147483647 # imm = 0x7FFFFFFF
-; X86-NEXT: pushl $-1
-; 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: huge_length_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -6413,10 +4007,6 @@ define i1 @huge_length_eq(i8* %X, i8* %Y) nounwind {
; This checks non-constant sizes.
define i32 @nonconst_length(i8* %X, i8* %Y, i64 %size) nounwind {
-; X86-LABEL: nonconst_length:
-; X86: # %bb.0:
-; X86-NEXT: jmp memcmp # TAILCALL
-;
; X64-LABEL: nonconst_length:
; X64: # %bb.0:
; X64-NEXT: jmp memcmp # TAILCALL
@@ -6425,18 +4015,6 @@ define i32 @nonconst_length(i8* %X, i8* %Y, i64 %size) nounwind {
}
define i1 @nonconst_length_eq(i8* %X, i8* %Y, i64 %size) nounwind {
-; X86-LABEL: nonconst_length_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; 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: nonconst_length_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
diff --git a/llvm/test/CodeGen/X86/memcmp-optsize-x32.ll b/llvm/test/CodeGen/X86/memcmp-optsize-x32.ll
new file mode 100644
index 0000000000000..2c45b8510d266
--- /dev/null
+++ b/llvm/test/CodeGen/X86/memcmp-optsize-x32.ll
@@ -0,0 +1,584 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=cmov | FileCheck %s --check-prefix=X86 --check-prefix=X86-NOSSE
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=X86 --check-prefix=X86-SSE2
+
+; This tests codegen time inlining/optimization of memcmp
+; rdar://6480398
+
+ at .str = private constant [65 x i8] c"0123456789012345678901234567890123456789012345678901234567890123\00", align 1
+
+declare dso_local i32 @memcmp(i8*, i8*, i32)
+declare dso_local i32 @bcmp(i8*, i8*, i32)
+
+define i32 @length2(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length2:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: rolw $8, %cx
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: movzwl %cx, %eax
+; X86-NEXT: movzwl %dx, %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ ret i32 %m
+}
+
+define i1 @length2_eq(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length2_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: cmpw (%eax), %cx
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_eq_const(i8* %X) nounwind optsize {
+; X86-LABEL: length2_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl (%eax), %eax
+; X86-NEXT: cmpl $12849, %eax # imm = 0x3231
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 1), i32 2) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length2_eq_nobuiltin_attr:
+; X86: # %bb.0:
+; X86-NEXT: pushl $2
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind nobuiltin
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length3(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length3:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: movzwl (%ecx), %esi
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: rolw $8, %si
+; X86-NEXT: cmpw %si, %dx
+; X86-NEXT: jne .LBB4_3
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movzbl 2(%eax), %eax
+; X86-NEXT: movzbl 2(%ecx), %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: jmp .LBB4_2
+; X86-NEXT: .LBB4_3: # %res_block
+; X86-NEXT: setae %al
+; X86-NEXT: movzbl %al, %eax
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB4_2: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 3) nounwind
+ ret i32 %m
+}
+
+define i1 @length3_eq(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length3_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %edx
+; X86-NEXT: xorw (%eax), %dx
+; X86-NEXT: movb 2(%ecx), %cl
+; X86-NEXT: xorb 2(%eax), %cl
+; X86-NEXT: movzbl %cl, %eax
+; X86-NEXT: orw %dx, %ax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 3) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length4(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length4:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: seta %al
+; X86-NEXT: sbbl $0, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ ret i32 %m
+}
+
+define i1 @length4_eq(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length4_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: cmpl (%eax), %ecx
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length4_eq_const(i8* %X) nounwind optsize {
+; X86-LABEL: length4_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: cmpl $875770417, (%eax) # imm = 0x34333231
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 1), i32 4) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length5(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length5:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: movl (%ecx), %esi
+; X86-NEXT: bswapl %edx
+; X86-NEXT: bswapl %esi
+; X86-NEXT: cmpl %esi, %edx
+; X86-NEXT: jne .LBB9_3
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movzbl 4(%eax), %eax
+; X86-NEXT: movzbl 4(%ecx), %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: jmp .LBB9_2
+; X86-NEXT: .LBB9_3: # %res_block
+; X86-NEXT: setae %al
+; X86-NEXT: movzbl %al, %eax
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB9_2: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ ret i32 %m
+}
+
+define i1 @length5_eq(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length5_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: movb 4(%ecx), %cl
+; X86-NEXT: xorb 4(%eax), %cl
+; X86-NEXT: movzbl %cl, %eax
+; X86-NEXT: orl %edx, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length8(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length8:
+; X86: # %bb.0:
+; 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 .LBB11_2
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movl 4(%esi), %ecx
+; X86-NEXT: movl 4(%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: je .LBB11_3
+; X86-NEXT: .LBB11_2: # %res_block
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: setae %al
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB11_3: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 8) nounwind
+ ret i32 %m
+}
+
+define i1 @length8_eq(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length8_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 4(%ecx), %ecx
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 4(%eax), %ecx
+; X86-NEXT: orl %edx, %ecx
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 8) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length8_eq_const(i8* %X) nounwind optsize {
+; X86-LABEL: length8_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl $858927408, %ecx # imm = 0x33323130
+; X86-NEXT: xorl (%eax), %ecx
+; X86-NEXT: movl $926299444, %edx # imm = 0x37363534
+; X86-NEXT: xorl 4(%eax), %edx
+; X86-NEXT: orl %ecx, %edx
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 8) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length12_eq(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length12_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $12
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 12) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length12(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length12:
+; X86: # %bb.0:
+; X86-NEXT: pushl $12
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 12) nounwind
+ ret i32 %m
+}
+
+; PR33329 - https://bugs.llvm.org/show_bug.cgi?id=33329
+
+define i32 @length16(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length16:
+; X86: # %bb.0:
+; X86-NEXT: pushl $16
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 16) nounwind
+ ret i32 %m
+}
+
+define i1 @length16_eq(i8* %x, i8* %y) nounwind optsize {
+; X86-NOSSE-LABEL: length16_eq:
+; X86-NOSSE: # %bb.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 $12, %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
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 16) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length16_eq_const(i8* %X) nounwind optsize {
+; X86-NOSSE-LABEL: length16_eq_const:
+; X86-NOSSE: # %bb.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 $12, %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[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 16) nounwind
+ %c = icmp eq i32 %m, 0
+ 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 $24
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 24) nounwind
+ ret i32 %m
+}
+
+define i1 @length24_eq(i8* %x, i8* %y) nounwind optsize {
+; X86-NOSSE-LABEL: length24_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $24
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE2-LABEL: length24_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 8(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 8(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 24) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length24_eq_const(i8* %X) nounwind optsize {
+; X86-NOSSE-LABEL: length24_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $24
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE2-LABEL: length24_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 8(%eax), %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 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:
+; X86-NEXT: pushl $32
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 32) nounwind
+ ret i32 %m
+}
+
+; PR33325 - https://bugs.llvm.org/show_bug.cgi?id=33325
+
+define i1 @length32_eq(i8* %x, i8* %y) nounwind optsize {
+; X86-NOSSE-LABEL: length32_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $32
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE2-LABEL: length32_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 16(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 32) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length32_eq_const(i8* %X) nounwind optsize {
+; X86-NOSSE-LABEL: length32_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $32
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE2-LABEL: length32_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 32) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length64(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: length64:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 64) nounwind
+ ret i32 %m
+}
+
+define i1 @length64_eq(i8* %x, i8* %y) nounwind optsize {
+; X86-LABEL: length64_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 64) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length64_eq_const(i8* %X) nounwind optsize {
+; X86-LABEL: length64_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 64) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @bcmp_length2(i8* %X, i8* %Y) nounwind optsize {
+; X86-LABEL: bcmp_length2:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: rolw $8, %cx
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: movzwl %cx, %eax
+; X86-NEXT: movzwl %dx, %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @bcmp(i8* %X, i8* %Y, i32 2) nounwind
+ ret i32 %m
+}
diff --git a/llvm/test/CodeGen/X86/memcmp-optsize.ll b/llvm/test/CodeGen/X86/memcmp-optsize.ll
index ab5cc93ca41c2..4c5b339859719 100644
--- a/llvm/test/CodeGen/X86/memcmp-optsize.ll
+++ b/llvm/test/CodeGen/X86/memcmp-optsize.ll
@@ -1,6 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=cmov | FileCheck %s --check-prefix=X86 --check-prefix=X86-NOSSE
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=X86 --check-prefix=X86-SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefix=X64 --check-prefix=X64-SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx | FileCheck %s --check-prefix=X64 --check-prefix=X64-AVX --check-prefix=X64-AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx2 | FileCheck %s --check-prefix=X64 --check-prefix=X64-AVX --check-prefix=X64-AVX2
@@ -14,19 +12,6 @@ declare dso_local i32 @memcmp(i8*, i8*, i64)
declare dso_local i32 @bcmp(i8*, i8*, i64)
define i32 @length2(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length2:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: rolw $8, %cx
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: movzwl %cx, %eax
-; X86-NEXT: movzwl %dx, %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length2:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -42,15 +27,6 @@ define i32 @length2(i8* %X, i8* %Y) nounwind optsize {
}
define i1 @length2_eq(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length2_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: cmpw (%eax), %cx
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_eq:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -63,14 +39,6 @@ define i1 @length2_eq(i8* %X, i8* %Y) nounwind optsize {
}
define i1 @length2_eq_const(i8* %X) nounwind optsize {
-; X86-LABEL: length2_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movzwl (%eax), %eax
-; X86-NEXT: cmpl $12849, %eax # imm = 0x3231
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_eq_const:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -83,18 +51,6 @@ define i1 @length2_eq_const(i8* %X) nounwind optsize {
}
define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length2_eq_nobuiltin_attr:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $2
-; 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: length2_eq_nobuiltin_attr:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -110,30 +66,6 @@ define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind optsize {
}
define i32 @length3(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length3:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: movzwl (%ecx), %esi
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: rolw $8, %si
-; X86-NEXT: cmpw %si, %dx
-; X86-NEXT: jne .LBB4_3
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movzbl 2(%eax), %eax
-; X86-NEXT: movzbl 2(%ecx), %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: jmp .LBB4_2
-; X86-NEXT: .LBB4_3: # %res_block
-; X86-NEXT: setae %al
-; X86-NEXT: movzbl %al, %eax
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB4_2: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length3:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -157,19 +89,6 @@ define i32 @length3(i8* %X, i8* %Y) nounwind optsize {
}
define i1 @length3_eq(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length3_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %edx
-; X86-NEXT: xorw (%eax), %dx
-; X86-NEXT: movb 2(%ecx), %cl
-; X86-NEXT: xorb 2(%eax), %cl
-; X86-NEXT: movzbl %cl, %eax
-; X86-NEXT: orw %dx, %ax
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length3_eq:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -186,20 +105,6 @@ define i1 @length3_eq(i8* %X, i8* %Y) nounwind optsize {
}
define i32 @length4(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length4:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: seta %al
-; X86-NEXT: sbbl $0, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length4:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %ecx
@@ -216,15 +121,6 @@ define i32 @length4(i8* %X, i8* %Y) nounwind optsize {
}
define i1 @length4_eq(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length4_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: cmpl (%eax), %ecx
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_eq:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -237,13 +133,6 @@ define i1 @length4_eq(i8* %X, i8* %Y) nounwind optsize {
}
define i1 @length4_eq_const(i8* %X) nounwind optsize {
-; X86-LABEL: length4_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: cmpl $875770417, (%eax) # imm = 0x34333231
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_eq_const:
; X64: # %bb.0:
; X64-NEXT: cmpl $875770417, (%rdi) # imm = 0x34333231
@@ -255,30 +144,6 @@ define i1 @length4_eq_const(i8* %X) nounwind optsize {
}
define i32 @length5(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length5:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: movl (%ecx), %esi
-; X86-NEXT: bswapl %edx
-; X86-NEXT: bswapl %esi
-; X86-NEXT: cmpl %esi, %edx
-; X86-NEXT: jne .LBB9_3
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movzbl 4(%eax), %eax
-; X86-NEXT: movzbl 4(%ecx), %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: jmp .LBB9_2
-; X86-NEXT: .LBB9_3: # %res_block
-; X86-NEXT: setae %al
-; X86-NEXT: movzbl %al, %eax
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB9_2: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length5:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -302,19 +167,6 @@ define i32 @length5(i8* %X, i8* %Y) nounwind optsize {
}
define i1 @length5_eq(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length5_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: movb 4(%ecx), %cl
-; X86-NEXT: xorb 4(%eax), %cl
-; X86-NEXT: movzbl %cl, %eax
-; X86-NEXT: orl %edx, %eax
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length5_eq:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -331,34 +183,6 @@ define i1 @length5_eq(i8* %X, i8* %Y) nounwind optsize {
}
define i32 @length8(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length8:
-; X86: # %bb.0:
-; 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 .LBB11_2
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movl 4(%esi), %ecx
-; X86-NEXT: movl 4(%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: je .LBB11_3
-; X86-NEXT: .LBB11_2: # %res_block
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: setae %al
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB11_3: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length8:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -375,18 +199,6 @@ define i32 @length8(i8* %X, i8* %Y) nounwind optsize {
}
define i1 @length8_eq(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length8_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 4(%ecx), %ecx
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 4(%eax), %ecx
-; X86-NEXT: orl %edx, %ecx
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length8_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -399,17 +211,6 @@ define i1 @length8_eq(i8* %X, i8* %Y) nounwind optsize {
}
define i1 @length8_eq_const(i8* %X) nounwind optsize {
-; X86-LABEL: length8_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl $858927408, %ecx # imm = 0x33323130
-; X86-NEXT: xorl (%eax), %ecx
-; X86-NEXT: movl $926299444, %edx # imm = 0x37363534
-; X86-NEXT: xorl 4(%eax), %edx
-; X86-NEXT: orl %ecx, %edx
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length8_eq_const:
; X64: # %bb.0:
; X64-NEXT: movabsq $3978425819141910832, %rax # imm = 0x3736353433323130
@@ -422,18 +223,6 @@ define i1 @length8_eq_const(i8* %X) nounwind optsize {
}
define i1 @length12_eq(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length12_eq:
-; 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
-;
; X64-LABEL: length12_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -449,16 +238,6 @@ define i1 @length12_eq(i8* %X, i8* %Y) nounwind optsize {
}
define i32 @length12(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length12:
-; 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:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -489,16 +268,6 @@ define i32 @length12(i8* %X, i8* %Y) nounwind optsize {
; PR33329 - https://bugs.llvm.org/show_bug.cgi?id=33329
define i32 @length16(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length16:
-; 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:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -527,30 +296,6 @@ define i32 @length16(i8* %X, i8* %Y) nounwind optsize {
}
define i1 @length16_eq(i8* %x, i8* %y) nounwind optsize {
-; 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-SSE2-LABEL: length16_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -574,28 +319,6 @@ define i1 @length16_eq(i8* %x, i8* %y) nounwind optsize {
}
define i1 @length16_eq_const(i8* %X) nounwind optsize {
-; 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[0-9]+_[0-9]+}}, %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-SSE2-LABEL: length16_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -620,16 +343,6 @@ define i1 @length16_eq_const(i8* %X) nounwind optsize {
; 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
@@ -639,34 +352,6 @@ define i32 @length24(i8* %X, i8* %Y) nounwind optsize {
}
define i1 @length24_eq(i8* %x, i8* %y) nounwind optsize {
-; X86-NOSSE-LABEL: length24_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $24
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE2-LABEL: length24_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 8(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 8(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %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-SSE2-LABEL: length24_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -698,31 +383,6 @@ define i1 @length24_eq(i8* %x, i8* %y) nounwind optsize {
}
define i1 @length24_eq_const(i8* %X) nounwind optsize {
-; X86-NOSSE-LABEL: length24_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $24
-; 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: setne %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE2-LABEL: length24_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 8(%eax), %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
; X64-SSE2-LABEL: length24_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -751,16 +411,6 @@ define i1 @length24_eq_const(i8* %X) nounwind optsize {
}
define i32 @length32(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length32:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $32
-; 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: length32:
; X64: # %bb.0:
; X64-NEXT: movl $32, %edx
@@ -772,34 +422,6 @@ define i32 @length32(i8* %X, i8* %Y) nounwind optsize {
; PR33325 - https://bugs.llvm.org/show_bug.cgi?id=33325
define i1 @length32_eq(i8* %x, i8* %y) nounwind optsize {
-; X86-NOSSE-LABEL: length32_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $32
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE2-LABEL: length32_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 16(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %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-SSE2-LABEL: length32_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -837,31 +459,6 @@ define i1 @length32_eq(i8* %x, i8* %y) nounwind optsize {
}
define i1 @length32_eq_const(i8* %X) nounwind optsize {
-; X86-NOSSE-LABEL: length32_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $32
-; 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: setne %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE2-LABEL: length32_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
; X64-SSE2-LABEL: length32_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -897,16 +494,6 @@ define i1 @length32_eq_const(i8* %X) nounwind optsize {
}
define i32 @length64(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: length64:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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: length64:
; X64: # %bb.0:
; X64-NEXT: movl $64, %edx
@@ -916,18 +503,6 @@ define i32 @length64(i8* %X, i8* %Y) nounwind optsize {
}
define i1 @length64_eq(i8* %x, i8* %y) nounwind optsize {
-; X86-LABEL: length64_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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
-;
; X64-SSE2-LABEL: length64_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: pushq %rax
@@ -967,18 +542,6 @@ define i1 @length64_eq(i8* %x, i8* %y) nounwind optsize {
}
define i1 @length64_eq_const(i8* %X) nounwind optsize {
-; X86-LABEL: length64_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE2-LABEL: length64_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: pushq %rax
@@ -1019,19 +582,6 @@ define i1 @length64_eq_const(i8* %X) nounwind optsize {
}
define i32 @bcmp_length2(i8* %X, i8* %Y) nounwind optsize {
-; X86-LABEL: bcmp_length2:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: rolw $8, %cx
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: movzwl %cx, %eax
-; X86-NEXT: movzwl %dx, %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: bcmp_length2:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -1045,4 +595,3 @@ define i32 @bcmp_length2(i8* %X, i8* %Y) nounwind optsize {
%m = tail call i32 @bcmp(i8* %X, i8* %Y, i64 2) nounwind
ret i32 %m
}
-
diff --git a/llvm/test/CodeGen/X86/memcmp-pgso-x32.ll b/llvm/test/CodeGen/X86/memcmp-pgso-x32.ll
new file mode 100644
index 0000000000000..0953e35b33979
--- /dev/null
+++ b/llvm/test/CodeGen/X86/memcmp-pgso-x32.ll
@@ -0,0 +1,601 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=cmov | FileCheck %s --check-prefix=X86 --check-prefix=X86-NOSSE
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=X86 --check-prefix=X86-SSE2
+
+; This tests codegen time inlining/optimization of memcmp
+; rdar://6480398
+
+ at .str = private constant [65 x i8] c"0123456789012345678901234567890123456789012345678901234567890123\00", align 1
+
+declare dso_local i32 @memcmp(i8*, i8*, i32)
+declare dso_local i32 @bcmp(i8*, i8*, i32)
+
+define i32 @length2(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length2:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: rolw $8, %cx
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: movzwl %cx, %eax
+; X86-NEXT: movzwl %dx, %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ ret i32 %m
+}
+
+define i1 @length2_eq(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length2_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: cmpw (%eax), %cx
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_eq_const(i8* %X) nounwind !prof !14 {
+; X86-LABEL: length2_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl (%eax), %eax
+; X86-NEXT: cmpl $12849, %eax # imm = 0x3231
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 1), i32 2) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length2_eq_nobuiltin_attr:
+; X86: # %bb.0:
+; X86-NEXT: pushl $2
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind nobuiltin
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length3(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length3:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: movzwl (%ecx), %esi
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: rolw $8, %si
+; X86-NEXT: cmpw %si, %dx
+; X86-NEXT: jne .LBB4_3
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movzbl 2(%eax), %eax
+; X86-NEXT: movzbl 2(%ecx), %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: jmp .LBB4_2
+; X86-NEXT: .LBB4_3: # %res_block
+; X86-NEXT: setae %al
+; X86-NEXT: movzbl %al, %eax
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB4_2: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 3) nounwind
+ ret i32 %m
+}
+
+define i1 @length3_eq(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length3_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %edx
+; X86-NEXT: xorw (%eax), %dx
+; X86-NEXT: movb 2(%ecx), %cl
+; X86-NEXT: xorb 2(%eax), %cl
+; X86-NEXT: movzbl %cl, %eax
+; X86-NEXT: orw %dx, %ax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 3) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length4(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length4:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: seta %al
+; X86-NEXT: sbbl $0, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ ret i32 %m
+}
+
+define i1 @length4_eq(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length4_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: cmpl (%eax), %ecx
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length4_eq_const(i8* %X) nounwind !prof !14 {
+; X86-LABEL: length4_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: cmpl $875770417, (%eax) # imm = 0x34333231
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 1), i32 4) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length5(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length5:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: movl (%ecx), %esi
+; X86-NEXT: bswapl %edx
+; X86-NEXT: bswapl %esi
+; X86-NEXT: cmpl %esi, %edx
+; X86-NEXT: jne .LBB9_3
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movzbl 4(%eax), %eax
+; X86-NEXT: movzbl 4(%ecx), %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: jmp .LBB9_2
+; X86-NEXT: .LBB9_3: # %res_block
+; X86-NEXT: setae %al
+; X86-NEXT: movzbl %al, %eax
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB9_2: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ ret i32 %m
+}
+
+define i1 @length5_eq(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length5_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: movb 4(%ecx), %cl
+; X86-NEXT: xorb 4(%eax), %cl
+; X86-NEXT: movzbl %cl, %eax
+; X86-NEXT: orl %edx, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length8(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length8:
+; X86: # %bb.0:
+; 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 .LBB11_2
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movl 4(%esi), %ecx
+; X86-NEXT: movl 4(%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: je .LBB11_3
+; X86-NEXT: .LBB11_2: # %res_block
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: setae %al
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB11_3: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 8) nounwind
+ ret i32 %m
+}
+
+define i1 @length8_eq(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length8_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 4(%ecx), %ecx
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 4(%eax), %ecx
+; X86-NEXT: orl %edx, %ecx
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 8) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length8_eq_const(i8* %X) nounwind !prof !14 {
+; X86-LABEL: length8_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl $858927408, %ecx # imm = 0x33323130
+; X86-NEXT: xorl (%eax), %ecx
+; X86-NEXT: movl $926299444, %edx # imm = 0x37363534
+; X86-NEXT: xorl 4(%eax), %edx
+; X86-NEXT: orl %ecx, %edx
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 8) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length12_eq(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length12_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $12
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 12) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length12(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length12:
+; X86: # %bb.0:
+; X86-NEXT: pushl $12
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 12) nounwind
+ ret i32 %m
+}
+
+; PR33329 - https://bugs.llvm.org/show_bug.cgi?id=33329
+
+define i32 @length16(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length16:
+; X86: # %bb.0:
+; X86-NEXT: pushl $16
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 16) nounwind
+ ret i32 %m
+}
+
+define i1 @length16_eq(i8* %x, i8* %y) nounwind !prof !14 {
+; X86-NOSSE-LABEL: length16_eq:
+; X86-NOSSE: # %bb.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 $12, %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
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 16) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length16_eq_const(i8* %X) nounwind !prof !14 {
+; X86-NOSSE-LABEL: length16_eq_const:
+; X86-NOSSE: # %bb.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 $12, %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[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 16) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+; PR33914 - https://bugs.llvm.org/show_bug.cgi?id=33914
+
+define i32 @length24(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length24:
+; X86: # %bb.0:
+; X86-NEXT: pushl $24
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 24) nounwind
+ ret i32 %m
+}
+
+define i1 @length24_eq(i8* %x, i8* %y) nounwind !prof !14 {
+; X86-NOSSE-LABEL: length24_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $24
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE2-LABEL: length24_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 8(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 8(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 24) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length24_eq_const(i8* %X) nounwind !prof !14 {
+; X86-NOSSE-LABEL: length24_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $24
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE2-LABEL: length24_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 8(%eax), %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 24) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length32(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length32:
+; X86: # %bb.0:
+; X86-NEXT: pushl $32
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 32) nounwind
+ ret i32 %m
+}
+
+; PR33325 - https://bugs.llvm.org/show_bug.cgi?id=33325
+
+define i1 @length32_eq(i8* %x, i8* %y) nounwind !prof !14 {
+; X86-NOSSE-LABEL: length32_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $32
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE2-LABEL: length32_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 16(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 32) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length32_eq_const(i8* %X) nounwind !prof !14 {
+; X86-NOSSE-LABEL: length32_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $32
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE2-LABEL: length32_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 32) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length64(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: length64:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 64) nounwind
+ ret i32 %m
+}
+
+define i1 @length64_eq(i8* %x, i8* %y) nounwind !prof !14 {
+; X86-LABEL: length64_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 64) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length64_eq_const(i8* %X) nounwind !prof !14 {
+; X86-LABEL: length64_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i32 64) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @bcmp_length2(i8* %X, i8* %Y) nounwind !prof !14 {
+; X86-LABEL: bcmp_length2:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: rolw $8, %cx
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: movzwl %cx, %eax
+; X86-NEXT: movzwl %dx, %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @bcmp(i8* %X, i8* %Y, i32 2) nounwind
+ ret i32 %m
+}
+
+!llvm.module.flags = !{!0}
+!0 = !{i32 1, !"ProfileSummary", !1}
+!1 = !{!2, !3, !4, !5, !6, !7, !8, !9}
+!2 = !{!"ProfileFormat", !"InstrProf"}
+!3 = !{!"TotalCount", i32 10000}
+!4 = !{!"MaxCount", i32 10}
+!5 = !{!"MaxInternalCount", i32 1}
+!6 = !{!"MaxFunctionCount", i32 1000}
+!7 = !{!"NumCounts", i32 3}
+!8 = !{!"NumFunctions", i32 3}
+!9 = !{!"DetailedSummary", !10}
+!10 = !{!11, !12, !13}
+!11 = !{i32 10000, i32 100, i32 1}
+!12 = !{i32 999000, i32 100, i32 1}
+!13 = !{i32 999999, i32 1, i32 2}
+!14 = !{!"function_entry_count", i32 0}
diff --git a/llvm/test/CodeGen/X86/memcmp-pgso.ll b/llvm/test/CodeGen/X86/memcmp-pgso.ll
index b39f0d70a56b9..f763d91b8f774 100644
--- a/llvm/test/CodeGen/X86/memcmp-pgso.ll
+++ b/llvm/test/CodeGen/X86/memcmp-pgso.ll
@@ -1,6 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=cmov | FileCheck %s --check-prefix=X86 --check-prefix=X86-NOSSE
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=X86 --check-prefix=X86-SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefix=X64 --check-prefix=X64-SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx | FileCheck %s --check-prefix=X64 --check-prefix=X64-AVX --check-prefix=X64-AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx2 | FileCheck %s --check-prefix=X64 --check-prefix=X64-AVX --check-prefix=X64-AVX2
@@ -14,19 +12,6 @@ declare dso_local i32 @memcmp(i8*, i8*, i64)
declare dso_local i32 @bcmp(i8*, i8*, i64)
define i32 @length2(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length2:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: rolw $8, %cx
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: movzwl %cx, %eax
-; X86-NEXT: movzwl %dx, %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length2:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -42,15 +27,6 @@ define i32 @length2(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i1 @length2_eq(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length2_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: cmpw (%eax), %cx
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_eq:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -63,14 +39,6 @@ define i1 @length2_eq(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i1 @length2_eq_const(i8* %X) nounwind !prof !14 {
-; X86-LABEL: length2_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movzwl (%eax), %eax
-; X86-NEXT: cmpl $12849, %eax # imm = 0x3231
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_eq_const:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -83,18 +51,6 @@ define i1 @length2_eq_const(i8* %X) nounwind !prof !14 {
}
define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length2_eq_nobuiltin_attr:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $2
-; 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: length2_eq_nobuiltin_attr:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -110,30 +66,6 @@ define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i32 @length3(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length3:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: movzwl (%ecx), %esi
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: rolw $8, %si
-; X86-NEXT: cmpw %si, %dx
-; X86-NEXT: jne .LBB4_3
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movzbl 2(%eax), %eax
-; X86-NEXT: movzbl 2(%ecx), %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: jmp .LBB4_2
-; X86-NEXT: .LBB4_3: # %res_block
-; X86-NEXT: setae %al
-; X86-NEXT: movzbl %al, %eax
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB4_2: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length3:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -157,19 +89,6 @@ define i32 @length3(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i1 @length3_eq(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length3_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %edx
-; X86-NEXT: xorw (%eax), %dx
-; X86-NEXT: movb 2(%ecx), %cl
-; X86-NEXT: xorb 2(%eax), %cl
-; X86-NEXT: movzbl %cl, %eax
-; X86-NEXT: orw %dx, %ax
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length3_eq:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -186,20 +105,6 @@ define i1 @length3_eq(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i32 @length4(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length4:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: seta %al
-; X86-NEXT: sbbl $0, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length4:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %ecx
@@ -216,15 +121,6 @@ define i32 @length4(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i1 @length4_eq(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length4_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: cmpl (%eax), %ecx
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_eq:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -237,13 +133,6 @@ define i1 @length4_eq(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i1 @length4_eq_const(i8* %X) nounwind !prof !14 {
-; X86-LABEL: length4_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: cmpl $875770417, (%eax) # imm = 0x34333231
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_eq_const:
; X64: # %bb.0:
; X64-NEXT: cmpl $875770417, (%rdi) # imm = 0x34333231
@@ -255,30 +144,6 @@ define i1 @length4_eq_const(i8* %X) nounwind !prof !14 {
}
define i32 @length5(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length5:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: movl (%ecx), %esi
-; X86-NEXT: bswapl %edx
-; X86-NEXT: bswapl %esi
-; X86-NEXT: cmpl %esi, %edx
-; X86-NEXT: jne .LBB9_3
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movzbl 4(%eax), %eax
-; X86-NEXT: movzbl 4(%ecx), %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: jmp .LBB9_2
-; X86-NEXT: .LBB9_3: # %res_block
-; X86-NEXT: setae %al
-; X86-NEXT: movzbl %al, %eax
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB9_2: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length5:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -302,19 +167,6 @@ define i32 @length5(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i1 @length5_eq(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length5_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: movb 4(%ecx), %cl
-; X86-NEXT: xorb 4(%eax), %cl
-; X86-NEXT: movzbl %cl, %eax
-; X86-NEXT: orl %edx, %eax
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length5_eq:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -331,34 +183,6 @@ define i1 @length5_eq(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i32 @length8(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length8:
-; X86: # %bb.0:
-; 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 .LBB11_2
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movl 4(%esi), %ecx
-; X86-NEXT: movl 4(%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: je .LBB11_3
-; X86-NEXT: .LBB11_2: # %res_block
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: setae %al
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB11_3: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length8:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -375,18 +199,6 @@ define i32 @length8(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i1 @length8_eq(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length8_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 4(%ecx), %ecx
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 4(%eax), %ecx
-; X86-NEXT: orl %edx, %ecx
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length8_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -399,17 +211,6 @@ define i1 @length8_eq(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i1 @length8_eq_const(i8* %X) nounwind !prof !14 {
-; X86-LABEL: length8_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl $858927408, %ecx # imm = 0x33323130
-; X86-NEXT: xorl (%eax), %ecx
-; X86-NEXT: movl $926299444, %edx # imm = 0x37363534
-; X86-NEXT: xorl 4(%eax), %edx
-; X86-NEXT: orl %ecx, %edx
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length8_eq_const:
; X64: # %bb.0:
; X64-NEXT: movabsq $3978425819141910832, %rax # imm = 0x3736353433323130
@@ -422,18 +223,6 @@ define i1 @length8_eq_const(i8* %X) nounwind !prof !14 {
}
define i1 @length12_eq(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length12_eq:
-; 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
-;
; X64-LABEL: length12_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -449,16 +238,6 @@ define i1 @length12_eq(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i32 @length12(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length12:
-; 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:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -489,16 +268,6 @@ define i32 @length12(i8* %X, i8* %Y) nounwind !prof !14 {
; PR33329 - https://bugs.llvm.org/show_bug.cgi?id=33329
define i32 @length16(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length16:
-; 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:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -527,30 +296,6 @@ define i32 @length16(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i1 @length16_eq(i8* %x, i8* %y) nounwind !prof !14 {
-; 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-SSE2-LABEL: length16_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -574,28 +319,6 @@ define i1 @length16_eq(i8* %x, i8* %y) nounwind !prof !14 {
}
define i1 @length16_eq_const(i8* %X) nounwind !prof !14 {
-; 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[0-9]+_[0-9]+}}, %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-SSE2-LABEL: length16_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -620,16 +343,6 @@ define i1 @length16_eq_const(i8* %X) nounwind !prof !14 {
; PR33914 - https://bugs.llvm.org/show_bug.cgi?id=33914
define i32 @length24(i8* %X, i8* %Y) nounwind !prof !14 {
-; 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
@@ -639,34 +352,6 @@ define i32 @length24(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i1 @length24_eq(i8* %x, i8* %y) nounwind !prof !14 {
-; X86-NOSSE-LABEL: length24_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $24
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE2-LABEL: length24_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 8(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 8(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %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-SSE2-LABEL: length24_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -698,31 +383,6 @@ define i1 @length24_eq(i8* %x, i8* %y) nounwind !prof !14 {
}
define i1 @length24_eq_const(i8* %X) nounwind !prof !14 {
-; X86-NOSSE-LABEL: length24_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $24
-; 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: setne %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE2-LABEL: length24_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 8(%eax), %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
; X64-SSE2-LABEL: length24_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -751,16 +411,6 @@ define i1 @length24_eq_const(i8* %X) nounwind !prof !14 {
}
define i32 @length32(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length32:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $32
-; 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: length32:
; X64: # %bb.0:
; X64-NEXT: movl $32, %edx
@@ -772,34 +422,6 @@ define i32 @length32(i8* %X, i8* %Y) nounwind !prof !14 {
; PR33325 - https://bugs.llvm.org/show_bug.cgi?id=33325
define i1 @length32_eq(i8* %x, i8* %y) nounwind !prof !14 {
-; X86-NOSSE-LABEL: length32_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $32
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE2-LABEL: length32_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 16(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %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-SSE2-LABEL: length32_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -837,31 +459,6 @@ define i1 @length32_eq(i8* %x, i8* %y) nounwind !prof !14 {
}
define i1 @length32_eq_const(i8* %X) nounwind !prof !14 {
-; X86-NOSSE-LABEL: length32_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $32
-; 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: setne %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE2-LABEL: length32_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
; X64-SSE2-LABEL: length32_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -897,16 +494,6 @@ define i1 @length32_eq_const(i8* %X) nounwind !prof !14 {
}
define i32 @length64(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: length64:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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: length64:
; X64: # %bb.0:
; X64-NEXT: movl $64, %edx
@@ -916,18 +503,6 @@ define i32 @length64(i8* %X, i8* %Y) nounwind !prof !14 {
}
define i1 @length64_eq(i8* %x, i8* %y) nounwind !prof !14 {
-; X86-LABEL: length64_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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
-;
; X64-SSE2-LABEL: length64_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: pushq %rax
@@ -967,18 +542,6 @@ define i1 @length64_eq(i8* %x, i8* %y) nounwind !prof !14 {
}
define i1 @length64_eq_const(i8* %X) nounwind !prof !14 {
-; X86-LABEL: length64_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE2-LABEL: length64_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: pushq %rax
@@ -1019,19 +582,6 @@ define i1 @length64_eq_const(i8* %X) nounwind !prof !14 {
}
define i32 @bcmp_length2(i8* %X, i8* %Y) nounwind !prof !14 {
-; X86-LABEL: bcmp_length2:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: rolw $8, %cx
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: movzwl %cx, %eax
-; X86-NEXT: movzwl %dx, %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: bcmp_length2:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
diff --git a/llvm/test/CodeGen/X86/memcmp-x32.ll b/llvm/test/CodeGen/X86/memcmp-x32.ll
new file mode 100644
index 0000000000000..ad7a1c0e0a492
--- /dev/null
+++ b/llvm/test/CodeGen/X86/memcmp-x32.ll
@@ -0,0 +1,2434 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=cmov | FileCheck %s --check-prefixes=X86,X86-NOSSE
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse | FileCheck %s --check-prefixes=X86,X86-SSE1
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=X86,X86-SSE2
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=X86,X86-SSE41
+
+; This tests codegen time inlining/optimization of memcmp
+; rdar://6480398
+
+ at .str = private constant [513 x i8] c"01234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901\00", align 1
+
+declare dso_local i32 @memcmp(i8*, i8*, i32)
+
+define i32 @length0(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length0:
+; X86: # %bb.0:
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 0) nounwind
+ ret i32 %m
+ }
+
+define i1 @length0_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length0_eq:
+; X86: # %bb.0:
+; X86-NEXT: movb $1, %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 0) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length0_lt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length0_lt:
+; X86: # %bb.0:
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 0) nounwind
+ %c = icmp slt i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length2(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: rolw $8, %cx
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: movzwl %cx, %eax
+; X86-NEXT: movzwl %dx, %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ ret i32 %m
+}
+
+define i32 @length2_const(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl (%eax), %eax
+; X86-NEXT: rolw $8, %ax
+; X86-NEXT: movzwl %ax, %eax
+; X86-NEXT: addl $-12594, %eax # imm = 0xCECE
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 1), i32 2) nounwind
+ ret i32 %m
+}
+
+define i1 @length2_gt_const(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2_gt_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl (%eax), %eax
+; X86-NEXT: rolw $8, %ax
+; X86-NEXT: movzwl %ax, %eax
+; X86-NEXT: addl $-12594, %eax # imm = 0xCECE
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 1), i32 2) nounwind
+ %c = icmp sgt i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: cmpw (%eax), %cx
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_lt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2_lt:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: rolw $8, %cx
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: movzwl %cx, %eax
+; X86-NEXT: movzwl %dx, %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ %c = icmp slt i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_gt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2_gt:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: movzwl (%eax), %eax
+; X86-NEXT: rolw $8, %cx
+; X86-NEXT: rolw $8, %ax
+; X86-NEXT: movzwl %cx, %ecx
+; X86-NEXT: movzwl %ax, %eax
+; X86-NEXT: subl %eax, %ecx
+; X86-NEXT: testl %ecx, %ecx
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind
+ %c = icmp sgt i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_eq_const(i8* %X) nounwind {
+; X86-LABEL: length2_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl (%eax), %eax
+; X86-NEXT: cmpl $12849, %eax # imm = 0x3231
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 1), i32 2) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length2_eq_nobuiltin_attr:
+; X86: # %bb.0:
+; X86-NEXT: pushl $2
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 2) nounwind nobuiltin
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length3(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length3:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl (%eax), %edx
+; X86-NEXT: movzwl (%ecx), %esi
+; X86-NEXT: rolw $8, %dx
+; X86-NEXT: rolw $8, %si
+; X86-NEXT: cmpw %si, %dx
+; X86-NEXT: jne .LBB11_3
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movzbl 2(%eax), %eax
+; X86-NEXT: movzbl 2(%ecx), %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+; X86-NEXT: .LBB11_3: # %res_block
+; X86-NEXT: setae %al
+; X86-NEXT: movzbl %al, %eax
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 3) nounwind
+ ret i32 %m
+}
+
+define i1 @length3_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length3_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %edx
+; X86-NEXT: xorw (%eax), %dx
+; X86-NEXT: movb 2(%ecx), %cl
+; X86-NEXT: xorb 2(%eax), %cl
+; X86-NEXT: movzbl %cl, %eax
+; X86-NEXT: orw %dx, %ax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 3) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length4(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length4:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: seta %al
+; X86-NEXT: sbbl $0, %eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ ret i32 %m
+}
+
+define i1 @length4_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length4_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: cmpl (%eax), %ecx
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length4_lt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length4_lt:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: seta %al
+; X86-NEXT: sbbl $0, %eax
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ %c = icmp slt i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length4_gt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length4_gt:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %ecx
+; X86-NEXT: movl (%eax), %eax
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %eax
+; X86-NEXT: xorl %edx, %edx
+; X86-NEXT: cmpl %eax, %ecx
+; X86-NEXT: seta %dl
+; X86-NEXT: sbbl $0, %edx
+; X86-NEXT: testl %edx, %edx
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 4) nounwind
+ %c = icmp sgt i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length4_eq_const(i8* %X) nounwind {
+; X86-LABEL: length4_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: cmpl $875770417, (%eax) # imm = 0x34333231
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 1), i32 4) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length5(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length5:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: movl (%ecx), %esi
+; X86-NEXT: bswapl %edx
+; X86-NEXT: bswapl %esi
+; X86-NEXT: cmpl %esi, %edx
+; X86-NEXT: jne .LBB18_3
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movzbl 4(%eax), %eax
+; X86-NEXT: movzbl 4(%ecx), %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+; X86-NEXT: .LBB18_3: # %res_block
+; X86-NEXT: setae %al
+; X86-NEXT: movzbl %al, %eax
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ ret i32 %m
+}
+
+define i1 @length5_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length5_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: movb 4(%ecx), %cl
+; X86-NEXT: xorb 4(%eax), %cl
+; X86-NEXT: movzbl %cl, %eax
+; X86-NEXT: orl %edx, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length5_lt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length5_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl (%eax), %edx
+; X86-NEXT: movl (%ecx), %esi
+; X86-NEXT: bswapl %edx
+; X86-NEXT: bswapl %esi
+; X86-NEXT: cmpl %esi, %edx
+; X86-NEXT: jne .LBB20_3
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movzbl 4(%eax), %eax
+; X86-NEXT: movzbl 4(%ecx), %ecx
+; X86-NEXT: subl %ecx, %eax
+; X86-NEXT: jmp .LBB20_2
+; X86-NEXT: .LBB20_3: # %res_block
+; X86-NEXT: setae %al
+; X86-NEXT: movzbl %al, %eax
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB20_2: # %endblock
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 5) nounwind
+ %c = icmp slt i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length7(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length7:
+; X86: # %bb.0:
+; 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 .LBB21_2
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movl 3(%esi), %ecx
+; X86-NEXT: movl 3(%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: je .LBB21_3
+; X86-NEXT: .LBB21_2: # %res_block
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: setae %al
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB21_3: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 7) nounwind
+ ret i32 %m
+}
+
+define i1 @length7_lt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length7_lt:
+; X86: # %bb.0:
+; 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 .LBB22_2
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movl 3(%esi), %ecx
+; X86-NEXT: movl 3(%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: je .LBB22_3
+; X86-NEXT: .LBB22_2: # %res_block
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: setae %al
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB22_3: # %endblock
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 7) nounwind
+ %c = icmp slt i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length7_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length7_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 3(%ecx), %ecx
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 3(%eax), %ecx
+; X86-NEXT: orl %edx, %ecx
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 7) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length8(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length8:
+; X86: # %bb.0:
+; 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 .LBB24_2
+; X86-NEXT: # %bb.1: # %loadbb1
+; X86-NEXT: movl 4(%esi), %ecx
+; X86-NEXT: movl 4(%eax), %edx
+; X86-NEXT: bswapl %ecx
+; X86-NEXT: bswapl %edx
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: je .LBB24_3
+; X86-NEXT: .LBB24_2: # %res_block
+; X86-NEXT: xorl %eax, %eax
+; X86-NEXT: cmpl %edx, %ecx
+; X86-NEXT: setae %al
+; X86-NEXT: leal -1(%eax,%eax), %eax
+; X86-NEXT: .LBB24_3: # %endblock
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 8) nounwind
+ ret i32 %m
+}
+
+define i1 @length8_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length8_eq:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl (%ecx), %edx
+; X86-NEXT: movl 4(%ecx), %ecx
+; X86-NEXT: xorl (%eax), %edx
+; X86-NEXT: xorl 4(%eax), %ecx
+; X86-NEXT: orl %edx, %ecx
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 8) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length8_eq_const(i8* %X) nounwind {
+; X86-LABEL: length8_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl $858927408, %ecx # imm = 0x33323130
+; X86-NEXT: xorl (%eax), %ecx
+; X86-NEXT: movl $926299444, %edx # imm = 0x37363534
+; X86-NEXT: xorl 4(%eax), %edx
+; X86-NEXT: orl %ecx, %edx
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 8) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length9_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length9_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $9
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 9) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length10_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length10_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $10
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 10) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length11_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length11_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $11
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 11) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length12_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length12_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $12
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 12) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length12(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length12:
+; X86: # %bb.0:
+; X86-NEXT: pushl $12
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 12) nounwind
+ ret i32 %m
+}
+
+define i1 @length13_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length13_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $13
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 13) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length14_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length14_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $14
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 14) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length15(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length15:
+; X86: # %bb.0:
+; X86-NEXT: pushl $15
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 15) nounwind
+ ret i32 %m
+}
+
+define i1 @length15_lt(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length15_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $15
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 15) nounwind
+ %c = icmp slt i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length15_const(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length15_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $15
+; X86-NEXT: pushl $.L.str+1
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 1), i32 15) nounwind
+ ret i32 %m
+}
+
+define i1 @length15_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length15_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $15
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 15) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i1 @length15_gt_const(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length15_gt_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $15
+; X86-NEXT: pushl $.L.str+1
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 1), i32 15) nounwind
+ %c = icmp sgt i32 %m, 0
+ ret i1 %c
+}
+
+; PR33329 - https://bugs.llvm.org/show_bug.cgi?id=33329
+
+define i32 @length16(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length16:
+; X86: # %bb.0:
+; X86-NEXT: pushl $16
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 16) nounwind
+ ret i32 %m
+}
+
+define i1 @length16_eq(i8* %x, i8* %y) nounwind {
+; X86-NOSSE-LABEL: length16_eq:
+; X86-NOSSE: # %bb.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 $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length16_eq:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $16
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: setne %al
+; X86-SSE1-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
+;
+; X86-SSE41-LABEL: length16_eq:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu (%eax), %xmm1
+; X86-SSE41-NEXT: pxor %xmm0, %xmm1
+; X86-SSE41-NEXT: ptest %xmm1, %xmm1
+; X86-SSE41-NEXT: setne %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 16) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length16_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length16_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $16
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 16) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length16_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length16_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $16
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 16) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length16_eq_const(i8* %X) nounwind {
+; X86-NOSSE-LABEL: length16_eq_const:
+; X86-NOSSE: # %bb.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 $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length16_eq_const:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $16
+; X86-SSE1-NEXT: pushl $.L.str
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-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[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length16_eq_const:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movdqu (%eax), %xmm0
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 16) nounwind
+ %c = icmp eq i32 %m, 0
+ 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 $24
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 24) nounwind
+ ret i32 %m
+}
+
+define i1 @length24_eq(i8* %x, i8* %y) nounwind {
+; X86-NOSSE-LABEL: length24_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $24
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length24_eq:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $24
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length24_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 8(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 8(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length24_eq:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 8(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 8(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: por %xmm2, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 24) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length24_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length24_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $24
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 24) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length24_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length24_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $24
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 24) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length24_eq_const(i8* %X) nounwind {
+; X86-NOSSE-LABEL: length24_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $24
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length24_eq_const:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $24
+; X86-SSE1-NEXT: pushl $.L.str
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: setne %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length24_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 8(%eax), %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length24_eq_const:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movdqu (%eax), %xmm0
+; X86-SSE41-NEXT: movdqu 8(%eax), %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE41-NEXT: por %xmm1, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: setne %al
+; X86-SSE41-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 24) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length31(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length31:
+; X86: # %bb.0:
+; X86-NEXT: pushl $31
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 31) nounwind
+ ret i32 %m
+}
+
+define i1 @length31_eq(i8* %x, i8* %y) nounwind {
+; X86-NOSSE-LABEL: length31_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $31
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length31_eq:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $31
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length31_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 15(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 15(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length31_eq:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 15(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 15(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: por %xmm2, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 31) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length31_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length31_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $31
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 31) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length31_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length31_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $31
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 31) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length31_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
+; X86-NOSSE-LABEL: length31_eq_prefer128:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $31
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length31_eq_prefer128:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $31
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length31_eq_prefer128:
+; 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 15(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 15(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length31_eq_prefer128:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 15(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 15(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: por %xmm2, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 31) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length31_eq_const(i8* %X) nounwind {
+; X86-NOSSE-LABEL: length31_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $31
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length31_eq_const:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $31
+; X86-SSE1-NEXT: pushl $.L.str
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: setne %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length31_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 15(%eax), %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length31_eq_const:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movdqu (%eax), %xmm0
+; X86-SSE41-NEXT: movdqu 15(%eax), %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE41-NEXT: por %xmm1, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: setne %al
+; X86-SSE41-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 31) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length32(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length32:
+; X86: # %bb.0:
+; X86-NEXT: pushl $32
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 32) nounwind
+ ret i32 %m
+}
+
+; PR33325 - https://bugs.llvm.org/show_bug.cgi?id=33325
+
+define i1 @length32_eq(i8* %x, i8* %y) nounwind {
+; X86-NOSSE-LABEL: length32_eq:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $32
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length32_eq:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $32
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length32_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 16(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length32_eq:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: por %xmm2, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 32) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length32_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length32_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $32
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 32) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length32_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length32_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $32
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 32) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length32_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
+; X86-NOSSE-LABEL: length32_eq_prefer128:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $32
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: sete %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length32_eq_prefer128:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $32
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: sete %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length32_eq_prefer128:
+; 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 16(%ecx), %xmm1
+; X86-SSE2-NEXT: movdqu (%eax), %xmm2
+; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
+; X86-SSE2-NEXT: pand %xmm2, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: sete %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length32_eq_prefer128:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
+; X86-SSE41-NEXT: movdqu (%eax), %xmm2
+; X86-SSE41-NEXT: pxor %xmm0, %xmm2
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
+; X86-SSE41-NEXT: pxor %xmm1, %xmm0
+; X86-SSE41-NEXT: por %xmm2, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: sete %al
+; X86-SSE41-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 32) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length32_eq_const(i8* %X) nounwind {
+; X86-NOSSE-LABEL: length32_eq_const:
+; X86-NOSSE: # %bb.0:
+; X86-NOSSE-NEXT: pushl $32
+; X86-NOSSE-NEXT: pushl $.L.str
+; X86-NOSSE-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NOSSE-NEXT: calll memcmp
+; X86-NOSSE-NEXT: addl $12, %esp
+; X86-NOSSE-NEXT: testl %eax, %eax
+; X86-NOSSE-NEXT: setne %al
+; X86-NOSSE-NEXT: retl
+;
+; X86-SSE1-LABEL: length32_eq_const:
+; X86-SSE1: # %bb.0:
+; X86-SSE1-NEXT: pushl $32
+; X86-SSE1-NEXT: pushl $.L.str
+; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-SSE1-NEXT: calll memcmp
+; X86-SSE1-NEXT: addl $12, %esp
+; X86-SSE1-NEXT: testl %eax, %eax
+; X86-SSE1-NEXT: setne %al
+; X86-SSE1-NEXT: retl
+;
+; X86-SSE2-LABEL: length32_eq_const:
+; X86-SSE2: # %bb.0:
+; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE2-NEXT: movdqu (%eax), %xmm0
+; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE2-NEXT: pand %xmm1, %xmm0
+; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
+; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
+; X86-SSE2-NEXT: setne %al
+; X86-SSE2-NEXT: retl
+;
+; X86-SSE41-LABEL: length32_eq_const:
+; X86-SSE41: # %bb.0:
+; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-SSE41-NEXT: movdqu (%eax), %xmm0
+; X86-SSE41-NEXT: movdqu 16(%eax), %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
+; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE41-NEXT: por %xmm1, %xmm0
+; X86-SSE41-NEXT: ptest %xmm0, %xmm0
+; X86-SSE41-NEXT: setne %al
+; X86-SSE41-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 32) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length48(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length48:
+; X86: # %bb.0:
+; X86-NEXT: pushl $48
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 48) nounwind
+ ret i32 %m
+}
+
+define i1 @length48_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length48_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $48
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 48) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length48_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length48_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $48
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 48) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length48_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length48_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $48
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 48) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length48_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
+; X86-LABEL: length48_eq_prefer128:
+; X86: # %bb.0:
+; X86-NEXT: pushl $48
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 48) nounwind
+ %cmp = icmp eq i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length48_eq_const(i8* %X) nounwind {
+; X86-LABEL: length48_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $48
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 48) nounwind
+ %c = icmp ne i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length63(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length63:
+; X86: # %bb.0:
+; X86-NEXT: pushl $63
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 63) nounwind
+ ret i32 %m
+}
+
+define i1 @length63_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length63_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $63
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 63) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length63_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length63_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $63
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 63) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length63_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length63_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $63
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 63) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length63_eq_const(i8* %X) nounwind {
+; X86-LABEL: length63_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $63
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 63) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length64(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length64:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 64) nounwind
+ ret i32 %m
+}
+
+define i1 @length64_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length64_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 64) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length64_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length64_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 64) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length64_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length64_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 64) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length64_eq_const(i8* %X) nounwind {
+; X86-LABEL: length64_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $64
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 64) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length96(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length96:
+; X86: # %bb.0:
+; X86-NEXT: pushl $96
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 96) nounwind
+ ret i32 %m
+}
+
+define i1 @length96_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length96_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $96
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 96) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length96_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length96_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $96
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 96) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length96_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length96_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $96
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 96) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length96_eq_const(i8* %X) nounwind {
+; X86-LABEL: length96_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $96
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 96) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length127(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length127:
+; X86: # %bb.0:
+; X86-NEXT: pushl $127
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 127) nounwind
+ ret i32 %m
+}
+
+define i1 @length127_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length127_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $127
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 127) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length127_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length127_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $127
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 127) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length127_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length127_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $127
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 127) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length127_eq_const(i8* %X) nounwind {
+; X86-LABEL: length127_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $127
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 127) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length128(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length128:
+; X86: # %bb.0:
+; X86-NEXT: pushl $128
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 128) nounwind
+ ret i32 %m
+}
+
+define i1 @length128_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length128_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $128
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 128) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length128_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length128_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $128
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 128) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length128_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length128_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $128
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 128) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length128_eq_const(i8* %X) nounwind {
+; X86-LABEL: length128_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $128
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 128) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length192(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length192:
+; X86: # %bb.0:
+; X86-NEXT: pushl $192
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 192) nounwind
+ ret i32 %m
+}
+
+define i1 @length192_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length192_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $192
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 192) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length192_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length192_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $192
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 192) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length192_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length192_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $192
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 192) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length192_eq_const(i8* %X) nounwind {
+; X86-LABEL: length192_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $192
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 192) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length255(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length255:
+; X86: # %bb.0:
+; X86-NEXT: pushl $255
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 255) nounwind
+ ret i32 %m
+}
+
+define i1 @length255_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length255_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $255
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 255) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length255_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length255_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $255
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 255) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length255_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length255_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $255
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 255) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length255_eq_const(i8* %X) nounwind {
+; X86-LABEL: length255_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $255
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 255) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length256(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length256:
+; X86: # %bb.0:
+; X86-NEXT: pushl $256 # imm = 0x100
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 256) nounwind
+ ret i32 %m
+}
+
+define i1 @length256_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length256_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $256 # imm = 0x100
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 256) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length256_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length256_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $256 # imm = 0x100
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 256) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length256_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length256_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $256 # imm = 0x100
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 256) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length256_eq_const(i8* %X) nounwind {
+; X86-LABEL: length256_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $256 # imm = 0x100
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 256) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length384(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length384:
+; X86: # %bb.0:
+; X86-NEXT: pushl $384 # imm = 0x180
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 384) nounwind
+ ret i32 %m
+}
+
+define i1 @length384_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length384_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $384 # imm = 0x180
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 384) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length384_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length384_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $384 # imm = 0x180
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 384) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length384_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length384_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $384 # imm = 0x180
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 384) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length384_eq_const(i8* %X) nounwind {
+; X86-LABEL: length384_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $384 # imm = 0x180
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 384) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length511(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length511:
+; X86: # %bb.0:
+; X86-NEXT: pushl $511 # imm = 0x1FF
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 511) nounwind
+ ret i32 %m
+}
+
+define i1 @length511_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length511_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $511 # imm = 0x1FF
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 511) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length511_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length511_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $511 # imm = 0x1FF
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 511) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length511_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length511_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $511 # imm = 0x1FF
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 511) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length511_eq_const(i8* %X) nounwind {
+; X86-LABEL: length511_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $511 # imm = 0x1FF
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 511) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+define i32 @length512(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: length512:
+; X86: # %bb.0:
+; X86-NEXT: pushl $512 # imm = 0x200
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 512) nounwind
+ ret i32 %m
+}
+
+define i1 @length512_eq(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length512_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $512 # imm = 0x200
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setne %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 512) nounwind
+ %cmp = icmp ne i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length512_lt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length512_lt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $512 # imm = 0x200
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: shrl $31, %eax
+; X86-NEXT: # kill: def $al killed $al killed $eax
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 512) nounwind
+ %cmp = icmp slt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length512_gt(i8* %x, i8* %y) nounwind {
+; X86-LABEL: length512_gt:
+; X86: # %bb.0:
+; X86-NEXT: pushl $512 # imm = 0x200
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: setg %al
+; X86-NEXT: retl
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 512) nounwind
+ %cmp = icmp sgt i32 %call, 0
+ ret i1 %cmp
+}
+
+define i1 @length512_eq_const(i8* %X) nounwind {
+; X86-LABEL: length512_eq_const:
+; X86: # %bb.0:
+; X86-NEXT: pushl $512 # imm = 0x200
+; X86-NEXT: pushl $.L.str
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([513 x i8], [513 x i8]* @.str, i32 0, i32 0), i32 512) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+; This checks that we do not do stupid things with huge sizes.
+define i32 @huge_length(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: huge_length:
+; X86: # %bb.0:
+; X86-NEXT: pushl $-1
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 9223372036854775807) nounwind
+ ret i32 %m
+}
+
+define i1 @huge_length_eq(i8* %X, i8* %Y) nounwind {
+; X86-LABEL: huge_length_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl $-1
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 9223372036854775807) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
+
+; This checks non-constant sizes.
+define i32 @nonconst_length(i8* %X, i8* %Y, i32 %size) nounwind {
+; X86-LABEL: nonconst_length:
+; X86: # %bb.0:
+; X86-NEXT: jmp memcmp # TAILCALL
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 %size) nounwind
+ ret i32 %m
+}
+
+define i1 @nonconst_length_eq(i8* %X, i8* %Y, i32 %size) nounwind {
+; X86-LABEL: nonconst_length_eq:
+; X86: # %bb.0:
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: pushl {{[0-9]+}}(%esp)
+; X86-NEXT: calll memcmp
+; X86-NEXT: addl $12, %esp
+; X86-NEXT: testl %eax, %eax
+; X86-NEXT: sete %al
+; X86-NEXT: retl
+ %m = tail call i32 @memcmp(i8* %X, i8* %Y, i32 %size) nounwind
+ %c = icmp eq i32 %m, 0
+ ret i1 %c
+}
diff --git a/llvm/test/CodeGen/X86/memcmp.ll b/llvm/test/CodeGen/X86/memcmp.ll
index d498d85711f80..bf11d3d18b2ed 100644
--- a/llvm/test/CodeGen/X86/memcmp.ll
+++ b/llvm/test/CodeGen/X86/memcmp.ll
@@ -1,8 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=cmov | FileCheck %s --check-prefixes=X86,X86-NOSSE
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse | FileCheck %s --check-prefixes=X86,X86-SSE1
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=X86,X86-SSE2
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=X86,X86-SSE41
; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=X64,X64-SSE,X64-SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=sse4.1 | FileCheck %s --check-prefixes=X64,X64-SSE,X64-SSE41
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx | FileCheck %s --check-prefixes=X64,X64-AVX,X64-AVX1
@@ -22,11 +18,6 @@
declare dso_local i32 @memcmp(i8*, i8*, i64)
define i32 @length0(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length0:
-; X86: # %bb.0:
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length0:
; X64: # %bb.0:
; X64-NEXT: xorl %eax, %eax
@@ -36,11 +27,6 @@ define i32 @length0(i8* %X, i8* %Y) nounwind {
}
define i1 @length0_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length0_eq:
-; X86: # %bb.0:
-; X86-NEXT: movb $1, %al
-; X86-NEXT: retl
-;
; X64-LABEL: length0_eq:
; X64: # %bb.0:
; X64-NEXT: movb $1, %al
@@ -51,11 +37,6 @@ define i1 @length0_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length0_lt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length0_lt:
-; X86: # %bb.0:
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length0_lt:
; X64: # %bb.0:
; X64-NEXT: xorl %eax, %eax
@@ -66,19 +47,6 @@ define i1 @length0_lt(i8* %X, i8* %Y) nounwind {
}
define i32 @length2(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: rolw $8, %cx
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: movzwl %cx, %eax
-; X86-NEXT: movzwl %dx, %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length2:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -94,15 +62,6 @@ define i32 @length2(i8* %X, i8* %Y) nounwind {
}
define i32 @length2_const(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movzwl (%eax), %eax
-; X86-NEXT: rolw $8, %ax
-; X86-NEXT: movzwl %ax, %eax
-; X86-NEXT: addl $-12594, %eax # imm = 0xCECE
-; X86-NEXT: retl
-;
; X64-LABEL: length2_const:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -115,17 +74,6 @@ define i32 @length2_const(i8* %X, i8* %Y) nounwind {
}
define i1 @length2_gt_const(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2_gt_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movzwl (%eax), %eax
-; X86-NEXT: rolw $8, %ax
-; X86-NEXT: movzwl %ax, %eax
-; X86-NEXT: addl $-12594, %eax # imm = 0xCECE
-; X86-NEXT: testl %eax, %eax
-; X86-NEXT: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_gt_const:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -141,15 +89,6 @@ define i1 @length2_gt_const(i8* %X, i8* %Y) nounwind {
}
define i1 @length2_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: cmpw (%eax), %cx
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_eq:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -162,21 +101,6 @@ define i1 @length2_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length2_lt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2_lt:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: rolw $8, %cx
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: movzwl %cx, %eax
-; X86-NEXT: movzwl %dx, %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length2_lt:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -195,21 +119,6 @@ define i1 @length2_lt(i8* %X, i8* %Y) nounwind {
}
define i1 @length2_gt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2_gt:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %ecx
-; X86-NEXT: movzwl (%eax), %eax
-; X86-NEXT: rolw $8, %cx
-; X86-NEXT: rolw $8, %ax
-; X86-NEXT: movzwl %cx, %ecx
-; X86-NEXT: movzwl %ax, %eax
-; X86-NEXT: subl %eax, %ecx
-; X86-NEXT: testl %ecx, %ecx
-; X86-NEXT: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_gt:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -228,14 +137,6 @@ define i1 @length2_gt(i8* %X, i8* %Y) nounwind {
}
define i1 @length2_eq_const(i8* %X) nounwind {
-; X86-LABEL: length2_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movzwl (%eax), %eax
-; X86-NEXT: cmpl $12849, %eax # imm = 0x3231
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length2_eq_const:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -248,18 +149,6 @@ define i1 @length2_eq_const(i8* %X) nounwind {
}
define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length2_eq_nobuiltin_attr:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $2
-; 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: length2_eq_nobuiltin_attr:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -275,30 +164,6 @@ define i1 @length2_eq_nobuiltin_attr(i8* %X, i8* %Y) nounwind {
}
define i32 @length3(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length3:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movzwl (%eax), %edx
-; X86-NEXT: movzwl (%ecx), %esi
-; X86-NEXT: rolw $8, %dx
-; X86-NEXT: rolw $8, %si
-; X86-NEXT: cmpw %si, %dx
-; X86-NEXT: jne .LBB11_3
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movzbl 2(%eax), %eax
-; X86-NEXT: movzbl 2(%ecx), %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-; X86-NEXT: .LBB11_3: # %res_block
-; X86-NEXT: setae %al
-; X86-NEXT: movzbl %al, %eax
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length3:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -322,19 +187,6 @@ define i32 @length3(i8* %X, i8* %Y) nounwind {
}
define i1 @length3_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length3_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl (%ecx), %edx
-; X86-NEXT: xorw (%eax), %dx
-; X86-NEXT: movb 2(%ecx), %cl
-; X86-NEXT: xorb 2(%eax), %cl
-; X86-NEXT: movzbl %cl, %eax
-; X86-NEXT: orw %dx, %ax
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length3_eq:
; X64: # %bb.0:
; X64-NEXT: movzwl (%rdi), %eax
@@ -351,20 +203,6 @@ define i1 @length3_eq(i8* %X, i8* %Y) nounwind {
}
define i32 @length4(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length4:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: seta %al
-; X86-NEXT: sbbl $0, %eax
-; X86-NEXT: retl
-;
; X64-LABEL: length4:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %ecx
@@ -381,15 +219,6 @@ define i32 @length4(i8* %X, i8* %Y) nounwind {
}
define i1 @length4_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length4_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: cmpl (%eax), %ecx
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_eq:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -402,22 +231,6 @@ define i1 @length4_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length4_lt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length4_lt:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: seta %al
-; X86-NEXT: sbbl $0, %eax
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length4_lt:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %ecx
@@ -437,22 +250,6 @@ define i1 @length4_lt(i8* %X, i8* %Y) nounwind {
}
define i1 @length4_gt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length4_gt:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %ecx
-; X86-NEXT: movl (%eax), %eax
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %eax
-; X86-NEXT: xorl %edx, %edx
-; X86-NEXT: cmpl %eax, %ecx
-; X86-NEXT: seta %dl
-; X86-NEXT: sbbl $0, %edx
-; X86-NEXT: testl %edx, %edx
-; X86-NEXT: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_gt:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -472,13 +269,6 @@ define i1 @length4_gt(i8* %X, i8* %Y) nounwind {
}
define i1 @length4_eq_const(i8* %X) nounwind {
-; X86-LABEL: length4_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: cmpl $875770417, (%eax) # imm = 0x34333231
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length4_eq_const:
; X64: # %bb.0:
; X64-NEXT: cmpl $875770417, (%rdi) # imm = 0x34333231
@@ -490,30 +280,6 @@ define i1 @length4_eq_const(i8* %X) nounwind {
}
define i32 @length5(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length5:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: movl (%ecx), %esi
-; X86-NEXT: bswapl %edx
-; X86-NEXT: bswapl %esi
-; X86-NEXT: cmpl %esi, %edx
-; X86-NEXT: jne .LBB18_3
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movzbl 4(%eax), %eax
-; X86-NEXT: movzbl 4(%ecx), %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-; X86-NEXT: .LBB18_3: # %res_block
-; X86-NEXT: setae %al
-; X86-NEXT: movzbl %al, %eax
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length5:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -537,19 +303,6 @@ define i32 @length5(i8* %X, i8* %Y) nounwind {
}
define i1 @length5_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length5_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: movb 4(%ecx), %cl
-; X86-NEXT: xorb 4(%eax), %cl
-; X86-NEXT: movzbl %cl, %eax
-; X86-NEXT: orl %edx, %eax
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length5_eq:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -566,32 +319,6 @@ define i1 @length5_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length5_lt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length5_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl %esi
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl (%eax), %edx
-; X86-NEXT: movl (%ecx), %esi
-; X86-NEXT: bswapl %edx
-; X86-NEXT: bswapl %esi
-; X86-NEXT: cmpl %esi, %edx
-; X86-NEXT: jne .LBB20_3
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movzbl 4(%eax), %eax
-; X86-NEXT: movzbl 4(%ecx), %ecx
-; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: jmp .LBB20_2
-; X86-NEXT: .LBB20_3: # %res_block
-; X86-NEXT: setae %al
-; X86-NEXT: movzbl %al, %eax
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB20_2: # %endblock
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length5_lt:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -620,34 +347,6 @@ define i1 @length5_lt(i8* %X, i8* %Y) nounwind {
}
define i32 @length7(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length7:
-; X86: # %bb.0:
-; 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 .LBB21_2
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movl 3(%esi), %ecx
-; X86-NEXT: movl 3(%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: je .LBB21_3
-; X86-NEXT: .LBB21_2: # %res_block
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: setae %al
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB21_3: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length7:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %ecx
@@ -676,36 +375,6 @@ define i32 @length7(i8* %X, i8* %Y) nounwind {
}
define i1 @length7_lt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length7_lt:
-; X86: # %bb.0:
-; 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 .LBB22_2
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movl 3(%esi), %ecx
-; X86-NEXT: movl 3(%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: je .LBB22_3
-; X86-NEXT: .LBB22_2: # %res_block
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: setae %al
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB22_3: # %endblock
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length7_lt:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %ecx
@@ -737,18 +406,6 @@ define i1 @length7_lt(i8* %X, i8* %Y) nounwind {
}
define i1 @length7_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length7_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 3(%ecx), %ecx
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 3(%eax), %ecx
-; X86-NEXT: orl %edx, %ecx
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length7_eq:
; X64: # %bb.0:
; X64-NEXT: movl (%rdi), %eax
@@ -764,34 +421,6 @@ define i1 @length7_eq(i8* %X, i8* %Y) nounwind {
}
define i32 @length8(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length8:
-; X86: # %bb.0:
-; 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 .LBB24_2
-; X86-NEXT: # %bb.1: # %loadbb1
-; X86-NEXT: movl 4(%esi), %ecx
-; X86-NEXT: movl 4(%eax), %edx
-; X86-NEXT: bswapl %ecx
-; X86-NEXT: bswapl %edx
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: je .LBB24_3
-; X86-NEXT: .LBB24_2: # %res_block
-; X86-NEXT: xorl %eax, %eax
-; X86-NEXT: cmpl %edx, %ecx
-; X86-NEXT: setae %al
-; X86-NEXT: leal -1(%eax,%eax), %eax
-; X86-NEXT: .LBB24_3: # %endblock
-; X86-NEXT: popl %esi
-; X86-NEXT: retl
-;
; X64-LABEL: length8:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -808,18 +437,6 @@ define i32 @length8(i8* %X, i8* %Y) nounwind {
}
define i1 @length8_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length8_eq:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl (%ecx), %edx
-; X86-NEXT: movl 4(%ecx), %ecx
-; X86-NEXT: xorl (%eax), %edx
-; X86-NEXT: xorl 4(%eax), %ecx
-; X86-NEXT: orl %edx, %ecx
-; X86-NEXT: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length8_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -832,17 +449,6 @@ define i1 @length8_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length8_eq_const(i8* %X) nounwind {
-; X86-LABEL: length8_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl $858927408, %ecx # imm = 0x33323130
-; X86-NEXT: xorl (%eax), %ecx
-; X86-NEXT: movl $926299444, %edx # imm = 0x37363534
-; X86-NEXT: xorl 4(%eax), %edx
-; X86-NEXT: orl %ecx, %edx
-; X86-NEXT: setne %al
-; X86-NEXT: retl
-;
; X64-LABEL: length8_eq_const:
; X64: # %bb.0:
; X64-NEXT: movabsq $3978425819141910832, %rax # imm = 0x3736353433323130
@@ -855,18 +461,6 @@ define i1 @length8_eq_const(i8* %X) nounwind {
}
define i1 @length9_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length9_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $9
-; 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: length9_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -883,18 +477,6 @@ define i1 @length9_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length10_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length10_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $10
-; 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: length10_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -911,18 +493,6 @@ define i1 @length10_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length11_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length11_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $11
-; 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: length11_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -938,18 +508,6 @@ define i1 @length11_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length12_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length12_eq:
-; 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
-;
; X64-LABEL: length12_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -965,16 +523,6 @@ define i1 @length12_eq(i8* %X, i8* %Y) nounwind {
}
define i32 @length12(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length12:
-; 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:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -1003,18 +551,6 @@ define i32 @length12(i8* %X, i8* %Y) nounwind {
}
define i1 @length13_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length13_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $13
-; 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: length13_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -1030,18 +566,6 @@ define i1 @length13_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length14_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length14_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $14
-; 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: length14_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -1057,16 +581,6 @@ define i1 @length14_eq(i8* %X, i8* %Y) nounwind {
}
define i32 @length15(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length15:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $15
-; 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: length15:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -1095,18 +609,6 @@ define i32 @length15(i8* %X, i8* %Y) nounwind {
}
define i1 @length15_lt(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length15_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $15
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length15_lt:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -1138,16 +640,6 @@ define i1 @length15_lt(i8* %X, i8* %Y) nounwind {
}
define i32 @length15_const(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length15_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $15
-; X86-NEXT: pushl $.L.str+1
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: retl
-;
; X64-LABEL: length15_const:
; X64: # %bb.0:
; X64-NEXT: movabsq $3544952156018063160, %rcx # imm = 0x3132333435363738
@@ -1174,18 +666,6 @@ define i32 @length15_const(i8* %X, i8* %Y) nounwind {
}
define i1 @length15_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length15_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $15
-; 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: length15_eq:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -1201,18 +681,6 @@ define i1 @length15_eq(i8* %X, i8* %Y) nounwind {
}
define i1 @length15_gt_const(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length15_gt_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $15
-; X86-NEXT: pushl $.L.str+1
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: testl %eax, %eax
-; X86-NEXT: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length15_gt_const:
; X64: # %bb.0:
; X64-NEXT: movabsq $3544952156018063160, %rax # imm = 0x3132333435363738
@@ -1244,16 +712,6 @@ define i1 @length15_gt_const(i8* %X, i8* %Y) nounwind {
; PR33329 - https://bugs.llvm.org/show_bug.cgi?id=33329
define i32 @length16(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length16:
-; 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:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -1282,53 +740,6 @@ define i32 @length16(i8* %X, i8* %Y) nounwind {
}
define i1 @length16_eq(i8* %x, i8* %y) nounwind {
-; 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-SSE1-LABEL: length16_eq:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $16
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: setne %al
-; X86-SSE1-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
-;
-; X86-SSE41-LABEL: length16_eq:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu (%eax), %xmm1
-; X86-SSE41-NEXT: pxor %xmm0, %xmm1
-; X86-SSE41-NEXT: ptest %xmm1, %xmm1
-; X86-SSE41-NEXT: setne %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length16_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -1371,18 +782,6 @@ define i1 @length16_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length16_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length16_lt:
-; 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: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length16_lt:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rcx
@@ -1414,18 +813,6 @@ define i1 @length16_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length16_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length16_gt:
-; 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: testl %eax, %eax
-; X86-NEXT: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length16_gt:
; X64: # %bb.0:
; X64-NEXT: movq (%rdi), %rax
@@ -1457,49 +844,6 @@ define i1 @length16_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length16_eq_const(i8* %X) nounwind {
-; 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-SSE1-LABEL: length16_eq_const:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $16
-; X86-SSE1-NEXT: pushl $.L.str
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-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[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length16_eq_const:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movdqu (%eax), %xmm0
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length16_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -1542,16 +886,6 @@ define i1 @length16_eq_const(i8* %X) nounwind {
; 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
@@ -1561,61 +895,6 @@ define i32 @length24(i8* %X, i8* %Y) nounwind {
}
define i1 @length24_eq(i8* %x, i8* %y) nounwind {
-; X86-NOSSE-LABEL: length24_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $24
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length24_eq:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $24
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length24_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 8(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 8(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length24_eq:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 8(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 8(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: por %xmm2, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length24_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -1673,18 +952,6 @@ define i1 @length24_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length24_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length24_lt:
-; 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: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length24_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -1700,18 +967,6 @@ define i1 @length24_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length24_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length24_gt:
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length24_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -1727,55 +982,6 @@ define i1 @length24_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length24_eq_const(i8* %X) nounwind {
-; X86-NOSSE-LABEL: length24_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $24
-; 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: setne %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length24_eq_const:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $24
-; X86-SSE1-NEXT: pushl $.L.str
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: setne %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length24_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 8(%eax), %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length24_eq_const:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movdqu (%eax), %xmm0
-; X86-SSE41-NEXT: movdqu 8(%eax), %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE41-NEXT: por %xmm1, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: setne %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length24_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -1828,16 +1034,6 @@ define i1 @length24_eq_const(i8* %X) nounwind {
}
define i32 @length31(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length31:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $31
-; 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: length31:
; X64: # %bb.0:
; X64-NEXT: movl $31, %edx
@@ -1847,61 +1043,6 @@ define i32 @length31(i8* %X, i8* %Y) nounwind {
}
define i1 @length31_eq(i8* %x, i8* %y) nounwind {
-; X86-NOSSE-LABEL: length31_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $31
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length31_eq:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $31
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length31_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 15(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 15(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length31_eq:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 15(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 15(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: por %xmm2, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length31_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -1958,18 +1099,6 @@ define i1 @length31_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length31_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length31_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $31
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length31_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -1985,18 +1114,6 @@ define i1 @length31_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length31_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length31_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $31
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length31_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -2012,61 +1129,6 @@ define i1 @length31_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length31_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
-; X86-NOSSE-LABEL: length31_eq_prefer128:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $31
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length31_eq_prefer128:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $31
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length31_eq_prefer128:
-; 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 15(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 15(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length31_eq_prefer128:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 15(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 15(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: por %xmm2, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length31_eq_prefer128:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -2123,55 +1185,6 @@ define i1 @length31_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"=
}
define i1 @length31_eq_const(i8* %X) nounwind {
-; X86-NOSSE-LABEL: length31_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $31
-; 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: setne %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length31_eq_const:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $31
-; X86-SSE1-NEXT: pushl $.L.str
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: setne %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length31_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 15(%eax), %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length31_eq_const:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movdqu (%eax), %xmm0
-; X86-SSE41-NEXT: movdqu 15(%eax), %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE41-NEXT: por %xmm1, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: setne %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length31_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -2224,16 +1237,6 @@ define i1 @length31_eq_const(i8* %X) nounwind {
}
define i32 @length32(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length32:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $32
-; 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: length32:
; X64: # %bb.0:
; X64-NEXT: movl $32, %edx
@@ -2245,61 +1248,6 @@ define i32 @length32(i8* %X, i8* %Y) nounwind {
; PR33325 - https://bugs.llvm.org/show_bug.cgi?id=33325
define i1 @length32_eq(i8* %x, i8* %y) nounwind {
-; X86-NOSSE-LABEL: length32_eq:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $32
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length32_eq:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $32
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length32_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 16(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length32_eq:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: por %xmm2, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length32_eq:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -2369,18 +1317,6 @@ define i1 @length32_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length32_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length32_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $32
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length32_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -2396,18 +1332,6 @@ define i1 @length32_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length32_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length32_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $32
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length32_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -2423,61 +1347,6 @@ define i1 @length32_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length32_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
-; X86-NOSSE-LABEL: length32_eq_prefer128:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $32
-; 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: sete %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length32_eq_prefer128:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $32
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: sete %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length32_eq_prefer128:
-; 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 16(%ecx), %xmm1
-; X86-SSE2-NEXT: movdqu (%eax), %xmm2
-; X86-SSE2-NEXT: pcmpeqb %xmm0, %xmm2
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE2-NEXT: pcmpeqb %xmm1, %xmm0
-; X86-SSE2-NEXT: pand %xmm2, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: sete %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length32_eq_prefer128:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE41-NEXT: movdqu (%ecx), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%ecx), %xmm1
-; X86-SSE41-NEXT: movdqu (%eax), %xmm2
-; X86-SSE41-NEXT: pxor %xmm0, %xmm2
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm0
-; X86-SSE41-NEXT: pxor %xmm1, %xmm0
-; X86-SSE41-NEXT: por %xmm2, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: sete %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length32_eq_prefer128:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -2534,55 +1403,6 @@ define i1 @length32_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"=
}
define i1 @length32_eq_const(i8* %X) nounwind {
-; X86-NOSSE-LABEL: length32_eq_const:
-; X86-NOSSE: # %bb.0:
-; X86-NOSSE-NEXT: pushl $0
-; X86-NOSSE-NEXT: pushl $32
-; 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: setne %al
-; X86-NOSSE-NEXT: retl
-;
-; X86-SSE1-LABEL: length32_eq_const:
-; X86-SSE1: # %bb.0:
-; X86-SSE1-NEXT: pushl $0
-; X86-SSE1-NEXT: pushl $32
-; X86-SSE1-NEXT: pushl $.L.str
-; X86-SSE1-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-SSE1-NEXT: calll memcmp
-; X86-SSE1-NEXT: addl $16, %esp
-; X86-SSE1-NEXT: testl %eax, %eax
-; X86-SSE1-NEXT: setne %al
-; X86-SSE1-NEXT: retl
-;
-; X86-SSE2-LABEL: length32_eq_const:
-; X86-SSE2: # %bb.0:
-; X86-SSE2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE2-NEXT: movdqu (%eax), %xmm0
-; X86-SSE2-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE2-NEXT: pcmpeqb {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE2-NEXT: pand %xmm1, %xmm0
-; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
-; X86-SSE2-NEXT: cmpl $65535, %eax # imm = 0xFFFF
-; X86-SSE2-NEXT: setne %al
-; X86-SSE2-NEXT: retl
-;
-; X86-SSE41-LABEL: length32_eq_const:
-; X86-SSE41: # %bb.0:
-; X86-SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE41-NEXT: movdqu (%eax), %xmm0
-; X86-SSE41-NEXT: movdqu 16(%eax), %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-SSE41-NEXT: pxor {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-SSE41-NEXT: por %xmm1, %xmm0
-; X86-SSE41-NEXT: ptest %xmm0, %xmm0
-; X86-SSE41-NEXT: setne %al
-; X86-SSE41-NEXT: retl
-;
; X64-SSE2-LABEL: length32_eq_const:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: movdqu (%rdi), %xmm0
@@ -2648,16 +1468,6 @@ define i1 @length32_eq_const(i8* %X) nounwind {
}
define i32 @length48(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length48:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $48
-; 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: length48:
; X64: # %bb.0:
; X64-NEXT: movl $48, %edx
@@ -2667,18 +1477,6 @@ define i32 @length48(i8* %X, i8* %Y) nounwind {
}
define i1 @length48_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length48_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $48
-; 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-SSE-LABEL: length48_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -2746,18 +1544,6 @@ define i1 @length48_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length48_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length48_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $48
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length48_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -2773,18 +1559,6 @@ define i1 @length48_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length48_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length48_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $48
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length48_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -2800,18 +1574,6 @@ define i1 @length48_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length48_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"="128" {
-; X86-LABEL: length48_eq_prefer128:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $48
-; 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: length48_eq_prefer128:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -2827,18 +1589,6 @@ define i1 @length48_eq_prefer128(i8* %x, i8* %y) nounwind "prefer-vector-width"=
}
define i1 @length48_eq_const(i8* %X) nounwind {
-; X86-LABEL: length48_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $48
-; 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-SSE-LABEL: length48_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -2904,16 +1654,6 @@ define i1 @length48_eq_const(i8* %X) nounwind {
}
define i32 @length63(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length63:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $63
-; 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: length63:
; X64: # %bb.0:
; X64-NEXT: movl $63, %edx
@@ -2923,18 +1663,6 @@ define i32 @length63(i8* %X, i8* %Y) nounwind {
}
define i1 @length63_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length63_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $63
-; 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
-;
; X64-SSE-LABEL: length63_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -2999,18 +1727,6 @@ define i1 @length63_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length63_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length63_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $63
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length63_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3026,18 +1742,6 @@ define i1 @length63_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length63_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length63_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $63
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length63_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3053,18 +1757,6 @@ define i1 @length63_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length63_eq_const(i8* %X) nounwind {
-; X86-LABEL: length63_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $63
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE-LABEL: length63_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -3130,16 +1822,6 @@ define i1 @length63_eq_const(i8* %X) nounwind {
}
define i32 @length64(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length64:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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: length64:
; X64: # %bb.0:
; X64-NEXT: movl $64, %edx
@@ -3149,18 +1831,6 @@ define i32 @length64(i8* %X, i8* %Y) nounwind {
}
define i1 @length64_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length64_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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
-;
; X64-SSE-LABEL: length64_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -3240,18 +1910,6 @@ define i1 @length64_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length64_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length64_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length64_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3267,18 +1925,6 @@ define i1 @length64_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length64_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length64_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length64_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3294,18 +1940,6 @@ define i1 @length64_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length64_eq_const(i8* %X) nounwind {
-; X86-LABEL: length64_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $64
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE-LABEL: length64_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -3386,16 +2020,6 @@ define i1 @length64_eq_const(i8* %X) nounwind {
}
define i32 @length96(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length96:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $96
-; 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: length96:
; X64: # %bb.0:
; X64-NEXT: movl $96, %edx
@@ -3405,18 +2029,6 @@ define i32 @length96(i8* %X, i8* %Y) nounwind {
}
define i1 @length96_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length96_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $96
-; 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
-;
; X64-SSE-LABEL: length96_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -3498,18 +2110,6 @@ define i1 @length96_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length96_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length96_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $96
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length96_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3525,18 +2125,6 @@ define i1 @length96_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length96_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length96_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $96
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length96_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3552,18 +2140,6 @@ define i1 @length96_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length96_eq_const(i8* %X) nounwind {
-; X86-LABEL: length96_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $96
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE-LABEL: length96_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -3646,16 +2222,6 @@ define i1 @length96_eq_const(i8* %X) nounwind {
}
define i32 @length127(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length127:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $127
-; 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: length127:
; X64: # %bb.0:
; X64-NEXT: movl $127, %edx
@@ -3665,18 +2231,6 @@ define i32 @length127(i8* %X, i8* %Y) nounwind {
}
define i1 @length127_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length127_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $127
-; 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
-;
; X64-SSE-LABEL: length127_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -3755,18 +2309,6 @@ define i1 @length127_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length127_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length127_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $127
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length127_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3782,18 +2324,6 @@ define i1 @length127_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length127_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length127_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $127
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length127_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -3809,18 +2339,6 @@ define i1 @length127_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length127_eq_const(i8* %X) nounwind {
-; X86-LABEL: length127_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $127
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE-LABEL: length127_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -3903,16 +2421,6 @@ define i1 @length127_eq_const(i8* %X) nounwind {
}
define i32 @length128(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length128:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $128
-; 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: length128:
; X64: # %bb.0:
; X64-NEXT: movl $128, %edx
@@ -3922,18 +2430,6 @@ define i32 @length128(i8* %X, i8* %Y) nounwind {
}
define i1 @length128_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length128_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $128
-; 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
-;
; X64-SSE-LABEL: length128_eq:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -4012,18 +2508,6 @@ define i1 @length128_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length128_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length128_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $128
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length128_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4039,18 +2523,6 @@ define i1 @length128_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length128_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length128_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $128
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length128_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4066,18 +2538,6 @@ define i1 @length128_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length128_eq_const(i8* %X) nounwind {
-; X86-LABEL: length128_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $128
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-SSE-LABEL: length128_eq_const:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pushq %rax
@@ -4160,16 +2620,6 @@ define i1 @length128_eq_const(i8* %X) nounwind {
}
define i32 @length192(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length192:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $192
-; 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: length192:
; X64: # %bb.0:
; X64-NEXT: movl $192, %edx
@@ -4179,18 +2629,6 @@ define i32 @length192(i8* %X, i8* %Y) nounwind {
}
define i1 @length192_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length192_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $192
-; 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
-;
; X64-LABEL: length192_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4206,18 +2644,6 @@ define i1 @length192_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length192_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length192_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $192
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length192_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4233,18 +2659,6 @@ define i1 @length192_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length192_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length192_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $192
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length192_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4260,18 +2674,6 @@ define i1 @length192_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length192_eq_const(i8* %X) nounwind {
-; X86-LABEL: length192_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $192
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length192_eq_const:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4288,16 +2690,6 @@ define i1 @length192_eq_const(i8* %X) nounwind {
}
define i32 @length255(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length255:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $255
-; 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: length255:
; X64: # %bb.0:
; X64-NEXT: movl $255, %edx
@@ -4307,18 +2699,6 @@ define i32 @length255(i8* %X, i8* %Y) nounwind {
}
define i1 @length255_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length255_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $255
-; 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
-;
; X64-LABEL: length255_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4334,18 +2714,6 @@ define i1 @length255_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length255_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length255_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $255
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length255_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4361,18 +2729,6 @@ define i1 @length255_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length255_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length255_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $255
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length255_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4388,18 +2744,6 @@ define i1 @length255_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length255_eq_const(i8* %X) nounwind {
-; X86-LABEL: length255_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $255
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length255_eq_const:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4416,16 +2760,6 @@ define i1 @length255_eq_const(i8* %X) nounwind {
}
define i32 @length256(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length256:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $256 # imm = 0x100
-; 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: length256:
; X64: # %bb.0:
; X64-NEXT: movl $256, %edx # imm = 0x100
@@ -4435,18 +2769,6 @@ define i32 @length256(i8* %X, i8* %Y) nounwind {
}
define i1 @length256_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length256_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $256 # imm = 0x100
-; 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
-;
; X64-LABEL: length256_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4462,18 +2784,6 @@ define i1 @length256_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length256_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length256_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $256 # imm = 0x100
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length256_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4489,18 +2799,6 @@ define i1 @length256_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length256_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length256_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $256 # imm = 0x100
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length256_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4516,18 +2814,6 @@ define i1 @length256_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length256_eq_const(i8* %X) nounwind {
-; X86-LABEL: length256_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $256 # imm = 0x100
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length256_eq_const:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4544,16 +2830,6 @@ define i1 @length256_eq_const(i8* %X) nounwind {
}
define i32 @length384(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length384:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $384 # imm = 0x180
-; 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: length384:
; X64: # %bb.0:
; X64-NEXT: movl $384, %edx # imm = 0x180
@@ -4563,18 +2839,6 @@ define i32 @length384(i8* %X, i8* %Y) nounwind {
}
define i1 @length384_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length384_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $384 # imm = 0x180
-; 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
-;
; X64-LABEL: length384_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4590,18 +2854,6 @@ define i1 @length384_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length384_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length384_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $384 # imm = 0x180
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length384_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4617,18 +2869,6 @@ define i1 @length384_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length384_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length384_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $384 # imm = 0x180
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length384_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4644,18 +2884,6 @@ define i1 @length384_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length384_eq_const(i8* %X) nounwind {
-; X86-LABEL: length384_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $384 # imm = 0x180
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length384_eq_const:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4672,16 +2900,6 @@ define i1 @length384_eq_const(i8* %X) nounwind {
}
define i32 @length511(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length511:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $511 # imm = 0x1FF
-; 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: length511:
; X64: # %bb.0:
; X64-NEXT: movl $511, %edx # imm = 0x1FF
@@ -4691,18 +2909,6 @@ define i32 @length511(i8* %X, i8* %Y) nounwind {
}
define i1 @length511_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length511_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $511 # imm = 0x1FF
-; 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
-;
; X64-LABEL: length511_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4718,18 +2924,6 @@ define i1 @length511_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length511_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length511_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $511 # imm = 0x1FF
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length511_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4745,18 +2939,6 @@ define i1 @length511_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length511_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length511_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $511 # imm = 0x1FF
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length511_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4772,18 +2954,6 @@ define i1 @length511_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length511_eq_const(i8* %X) nounwind {
-; X86-LABEL: length511_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $511 # imm = 0x1FF
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length511_eq_const:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4800,16 +2970,6 @@ define i1 @length511_eq_const(i8* %X) nounwind {
}
define i32 @length512(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: length512:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $512 # imm = 0x200
-; 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: length512:
; X64: # %bb.0:
; X64-NEXT: movl $512, %edx # imm = 0x200
@@ -4819,18 +2979,6 @@ define i32 @length512(i8* %X, i8* %Y) nounwind {
}
define i1 @length512_eq(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length512_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $512 # imm = 0x200
-; 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
-;
; X64-LABEL: length512_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4846,18 +2994,6 @@ define i1 @length512_eq(i8* %x, i8* %y) nounwind {
}
define i1 @length512_lt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length512_lt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $512 # imm = 0x200
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: calll memcmp
-; X86-NEXT: addl $16, %esp
-; X86-NEXT: shrl $31, %eax
-; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: retl
-;
; X64-LABEL: length512_lt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4873,18 +3009,6 @@ define i1 @length512_lt(i8* %x, i8* %y) nounwind {
}
define i1 @length512_gt(i8* %x, i8* %y) nounwind {
-; X86-LABEL: length512_gt:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $512 # imm = 0x200
-; 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: setg %al
-; X86-NEXT: retl
-;
; X64-LABEL: length512_gt:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4900,18 +3024,6 @@ define i1 @length512_gt(i8* %x, i8* %y) nounwind {
}
define i1 @length512_eq_const(i8* %X) nounwind {
-; X86-LABEL: length512_eq_const:
-; X86: # %bb.0:
-; X86-NEXT: pushl $0
-; X86-NEXT: pushl $512 # imm = 0x200
-; 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: sete %al
-; X86-NEXT: retl
-;
; X64-LABEL: length512_eq_const:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4929,16 +3041,6 @@ define i1 @length512_eq_const(i8* %X) nounwind {
; This checks that we do not do stupid things with huge sizes.
define i32 @huge_length(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: huge_length:
-; X86: # %bb.0:
-; X86-NEXT: pushl $2147483647 # imm = 0x7FFFFFFF
-; X86-NEXT: pushl $-1
-; 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: huge_length:
; X64: # %bb.0:
; X64-NEXT: movabsq $9223372036854775807, %rdx # imm = 0x7FFFFFFFFFFFFFFF
@@ -4948,18 +3050,6 @@ define i32 @huge_length(i8* %X, i8* %Y) nounwind {
}
define i1 @huge_length_eq(i8* %X, i8* %Y) nounwind {
-; X86-LABEL: huge_length_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl $2147483647 # imm = 0x7FFFFFFF
-; X86-NEXT: pushl $-1
-; 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: huge_length_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
@@ -4976,10 +3066,6 @@ define i1 @huge_length_eq(i8* %X, i8* %Y) nounwind {
; This checks non-constant sizes.
define i32 @nonconst_length(i8* %X, i8* %Y, i64 %size) nounwind {
-; X86-LABEL: nonconst_length:
-; X86: # %bb.0:
-; X86-NEXT: jmp memcmp # TAILCALL
-;
; X64-LABEL: nonconst_length:
; X64: # %bb.0:
; X64-NEXT: jmp memcmp # TAILCALL
@@ -4988,18 +3074,6 @@ define i32 @nonconst_length(i8* %X, i8* %Y, i64 %size) nounwind {
}
define i1 @nonconst_length_eq(i8* %X, i8* %Y, i64 %size) nounwind {
-; X86-LABEL: nonconst_length_eq:
-; X86: # %bb.0:
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; X86-NEXT: pushl {{[0-9]+}}(%esp)
-; 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: nonconst_length_eq:
; X64: # %bb.0:
; X64-NEXT: pushq %rax
diff --git a/llvm/test/Transforms/ExpandMemCmp/X86/memcmp-x32.ll b/llvm/test/Transforms/ExpandMemCmp/X86/memcmp-x32.ll
new file mode 100644
index 0000000000000..3c69841fbfc8e
--- /dev/null
+++ b/llvm/test/Transforms/ExpandMemCmp/X86/memcmp-x32.ll
@@ -0,0 +1,615 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -S -expandmemcmp -mtriple=i686-unknown-unknown -data-layout=e-m:o-p:32:32-f64:32:64-f80:128-n8:16:32-S128 < %s | FileCheck %s --check-prefix=X32
+
+declare i32 @memcmp(i8* nocapture, i8* nocapture, i32)
+
+define i32 @cmp2(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp2(
+; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i16*
+; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i16*
+; X32-NEXT: [[TMP3:%.*]] = load i16, i16* [[TMP1]], align 1
+; X32-NEXT: [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 1
+; X32-NEXT: [[TMP5:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP3]])
+; X32-NEXT: [[TMP6:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP4]])
+; X32-NEXT: [[TMP7:%.*]] = zext i16 [[TMP5]] to i32
+; X32-NEXT: [[TMP8:%.*]] = zext i16 [[TMP6]] to i32
+; X32-NEXT: [[TMP9:%.*]] = sub i32 [[TMP7]], [[TMP8]]
+; X32-NEXT: ret i32 [[TMP9]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 2)
+ ret i32 %call
+}
+
+define i32 @cmp2_align2(i8* nocapture readonly align 2 %x, i8* nocapture readonly align 2 %y) {
+; X32-LABEL: @cmp2_align2(
+; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i16*
+; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i16*
+; X32-NEXT: [[TMP3:%.*]] = load i16, i16* [[TMP1]], align 2
+; X32-NEXT: [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 2
+; X32-NEXT: [[TMP5:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP3]])
+; X32-NEXT: [[TMP6:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP4]])
+; X32-NEXT: [[TMP7:%.*]] = zext i16 [[TMP5]] to i32
+; X32-NEXT: [[TMP8:%.*]] = zext i16 [[TMP6]] to i32
+; X32-NEXT: [[TMP9:%.*]] = sub i32 [[TMP7]], [[TMP8]]
+; X32-NEXT: ret i32 [[TMP9]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 2)
+ ret i32 %call
+}
+
+define i32 @cmp3(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp3(
+; X32-NEXT: br label [[LOADBB:%.*]]
+; X32: res_block:
+; X32-NEXT: [[TMP1:%.*]] = icmp ult i16 [[TMP7:%.*]], [[TMP8:%.*]]
+; X32-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
+; X32-NEXT: br label [[ENDBLOCK:%.*]]
+; X32: loadbb:
+; X32-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i16*
+; X32-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i16*
+; X32-NEXT: [[TMP5:%.*]] = load i16, i16* [[TMP3]], align 1
+; X32-NEXT: [[TMP6:%.*]] = load i16, i16* [[TMP4]], align 1
+; X32-NEXT: [[TMP7]] = call i16 @llvm.bswap.i16(i16 [[TMP5]])
+; X32-NEXT: [[TMP8]] = call i16 @llvm.bswap.i16(i16 [[TMP6]])
+; X32-NEXT: [[TMP9:%.*]] = icmp eq i16 [[TMP7]], [[TMP8]]
+; X32-NEXT: br i1 [[TMP9]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
+; X32: loadbb1:
+; X32-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 2
+; X32-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 2
+; X32-NEXT: [[TMP12:%.*]] = load i8, i8* [[TMP10]], align 1
+; X32-NEXT: [[TMP13:%.*]] = load i8, i8* [[TMP11]], align 1
+; X32-NEXT: [[TMP14:%.*]] = zext i8 [[TMP12]] to i32
+; X32-NEXT: [[TMP15:%.*]] = zext i8 [[TMP13]] to i32
+; X32-NEXT: [[TMP16:%.*]] = sub i32 [[TMP14]], [[TMP15]]
+; X32-NEXT: br label [[ENDBLOCK]]
+; X32: endblock:
+; X32-NEXT: [[PHI_RES:%.*]] = phi i32 [ [[TMP16]], [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
+; X32-NEXT: ret i32 [[PHI_RES]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 3)
+ ret i32 %call
+}
+
+define i32 @cmp4(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp4(
+; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
+; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
+; X32-NEXT: [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
+; X32-NEXT: [[TMP6:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP4]])
+; X32-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+; X32-NEXT: [[TMP8:%.*]] = icmp ult i32 [[TMP5]], [[TMP6]]
+; X32-NEXT: [[TMP9:%.*]] = zext i1 [[TMP7]] to i32
+; X32-NEXT: [[TMP10:%.*]] = zext i1 [[TMP8]] to i32
+; X32-NEXT: [[TMP11:%.*]] = sub i32 [[TMP9]], [[TMP10]]
+; X32-NEXT: ret i32 [[TMP11]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 4)
+ ret i32 %call
+}
+
+define i32 @cmp5(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp5(
+; X32-NEXT: br label [[LOADBB:%.*]]
+; X32: res_block:
+; X32-NEXT: [[TMP1:%.*]] = icmp ult i32 [[TMP7:%.*]], [[TMP8:%.*]]
+; X32-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
+; X32-NEXT: br label [[ENDBLOCK:%.*]]
+; X32: loadbb:
+; X32-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X32-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X32-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 1
+; X32-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP4]], align 1
+; X32-NEXT: [[TMP7]] = call i32 @llvm.bswap.i32(i32 [[TMP5]])
+; X32-NEXT: [[TMP8]] = call i32 @llvm.bswap.i32(i32 [[TMP6]])
+; X32-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP7]], [[TMP8]]
+; X32-NEXT: br i1 [[TMP9]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
+; X32: loadbb1:
+; X32-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 4
+; X32-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 4
+; X32-NEXT: [[TMP12:%.*]] = load i8, i8* [[TMP10]], align 1
+; X32-NEXT: [[TMP13:%.*]] = load i8, i8* [[TMP11]], align 1
+; X32-NEXT: [[TMP14:%.*]] = zext i8 [[TMP12]] to i32
+; X32-NEXT: [[TMP15:%.*]] = zext i8 [[TMP13]] to i32
+; X32-NEXT: [[TMP16:%.*]] = sub i32 [[TMP14]], [[TMP15]]
+; X32-NEXT: br label [[ENDBLOCK]]
+; X32: endblock:
+; X32-NEXT: [[PHI_RES:%.*]] = phi i32 [ [[TMP16]], [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
+; X32-NEXT: ret i32 [[PHI_RES]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 5)
+ ret i32 %call
+}
+
+define i32 @cmp6(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp6(
+; X32-NEXT: br label [[LOADBB:%.*]]
+; X32: res_block:
+; X32-NEXT: [[PHI_SRC1:%.*]] = phi i32 [ [[TMP7:%.*]], [[LOADBB]] ], [ [[TMP18:%.*]], [[LOADBB1:%.*]] ]
+; X32-NEXT: [[PHI_SRC2:%.*]] = phi i32 [ [[TMP8:%.*]], [[LOADBB]] ], [ [[TMP19:%.*]], [[LOADBB1]] ]
+; X32-NEXT: [[TMP1:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
+; X32-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
+; X32-NEXT: br label [[ENDBLOCK:%.*]]
+; X32: loadbb:
+; X32-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X32-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X32-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 1
+; X32-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP4]], align 1
+; X32-NEXT: [[TMP7]] = call i32 @llvm.bswap.i32(i32 [[TMP5]])
+; X32-NEXT: [[TMP8]] = call i32 @llvm.bswap.i32(i32 [[TMP6]])
+; X32-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP7]], [[TMP8]]
+; X32-NEXT: br i1 [[TMP9]], label [[LOADBB1]], label [[RES_BLOCK:%.*]]
+; X32: loadbb1:
+; X32-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 4
+; X32-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 4
+; X32-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i16*
+; X32-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i16*
+; X32-NEXT: [[TMP14:%.*]] = load i16, i16* [[TMP12]], align 1
+; X32-NEXT: [[TMP15:%.*]] = load i16, i16* [[TMP13]], align 1
+; X32-NEXT: [[TMP16:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP14]])
+; X32-NEXT: [[TMP17:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP15]])
+; X32-NEXT: [[TMP18]] = zext i16 [[TMP16]] to i32
+; X32-NEXT: [[TMP19]] = zext i16 [[TMP17]] to i32
+; X32-NEXT: [[TMP20:%.*]] = icmp eq i32 [[TMP18]], [[TMP19]]
+; X32-NEXT: br i1 [[TMP20]], label [[ENDBLOCK]], label [[RES_BLOCK]]
+; X32: endblock:
+; X32-NEXT: [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
+; X32-NEXT: ret i32 [[PHI_RES]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 6)
+ ret i32 %call
+}
+
+define i32 @cmp7(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp7(
+; X32-NEXT: br label [[LOADBB:%.*]]
+; X32: res_block:
+; X32-NEXT: [[PHI_SRC1:%.*]] = phi i32 [ [[TMP7:%.*]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1:%.*]] ]
+; X32-NEXT: [[PHI_SRC2:%.*]] = phi i32 [ [[TMP8:%.*]], [[LOADBB]] ], [ [[TMP17:%.*]], [[LOADBB1]] ]
+; X32-NEXT: [[TMP1:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
+; X32-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
+; X32-NEXT: br label [[ENDBLOCK:%.*]]
+; X32: loadbb:
+; X32-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X32-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X32-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 1
+; X32-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP4]], align 1
+; X32-NEXT: [[TMP7]] = call i32 @llvm.bswap.i32(i32 [[TMP5]])
+; X32-NEXT: [[TMP8]] = call i32 @llvm.bswap.i32(i32 [[TMP6]])
+; X32-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP7]], [[TMP8]]
+; X32-NEXT: br i1 [[TMP9]], label [[LOADBB1]], label [[RES_BLOCK:%.*]]
+; X32: loadbb1:
+; X32-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 3
+; X32-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 3
+; X32-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i32*
+; X32-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i32*
+; X32-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP12]], align 1
+; X32-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP13]], align 1
+; X32-NEXT: [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
+; X32-NEXT: [[TMP17]] = call i32 @llvm.bswap.i32(i32 [[TMP15]])
+; X32-NEXT: [[TMP18:%.*]] = icmp eq i32 [[TMP16]], [[TMP17]]
+; X32-NEXT: br i1 [[TMP18]], label [[ENDBLOCK]], label [[RES_BLOCK]]
+; X32: endblock:
+; X32-NEXT: [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
+; X32-NEXT: ret i32 [[PHI_RES]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 7)
+ ret i32 %call
+}
+
+define i32 @cmp8(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp8(
+; X32-NEXT: br label [[LOADBB:%.*]]
+; X32: res_block:
+; X32-NEXT: [[PHI_SRC1:%.*]] = phi i32 [ [[TMP7:%.*]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1:%.*]] ]
+; X32-NEXT: [[PHI_SRC2:%.*]] = phi i32 [ [[TMP8:%.*]], [[LOADBB]] ], [ [[TMP17:%.*]], [[LOADBB1]] ]
+; X32-NEXT: [[TMP1:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
+; X32-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
+; X32-NEXT: br label [[ENDBLOCK:%.*]]
+; X32: loadbb:
+; X32-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X32-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X32-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 1
+; X32-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP4]], align 1
+; X32-NEXT: [[TMP7]] = call i32 @llvm.bswap.i32(i32 [[TMP5]])
+; X32-NEXT: [[TMP8]] = call i32 @llvm.bswap.i32(i32 [[TMP6]])
+; X32-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP7]], [[TMP8]]
+; X32-NEXT: br i1 [[TMP9]], label [[LOADBB1]], label [[RES_BLOCK:%.*]]
+; X32: loadbb1:
+; X32-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 4
+; X32-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 4
+; X32-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i32*
+; X32-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i32*
+; X32-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP12]], align 1
+; X32-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP13]], align 1
+; X32-NEXT: [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
+; X32-NEXT: [[TMP17]] = call i32 @llvm.bswap.i32(i32 [[TMP15]])
+; X32-NEXT: [[TMP18:%.*]] = icmp eq i32 [[TMP16]], [[TMP17]]
+; X32-NEXT: br i1 [[TMP18]], label [[ENDBLOCK]], label [[RES_BLOCK]]
+; X32: endblock:
+; X32-NEXT: [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
+; X32-NEXT: ret i32 [[PHI_RES]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 8)
+ ret i32 %call
+}
+
+define i32 @cmp9(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp9(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 9)
+; X32-NEXT: ret i32 [[CALL]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 9)
+ ret i32 %call
+}
+
+define i32 @cmp10(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp10(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 10)
+; X32-NEXT: ret i32 [[CALL]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 10)
+ ret i32 %call
+}
+
+define i32 @cmp11(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp11(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 11)
+; X32-NEXT: ret i32 [[CALL]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 11)
+ ret i32 %call
+}
+
+define i32 @cmp12(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp12(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 12)
+; X32-NEXT: ret i32 [[CALL]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 12)
+ ret i32 %call
+}
+
+define i32 @cmp13(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp13(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 13)
+; X32-NEXT: ret i32 [[CALL]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 13)
+ ret i32 %call
+}
+
+define i32 @cmp14(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp14(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 14)
+; X32-NEXT: ret i32 [[CALL]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 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:%.*]], i32 15)
+; X32-NEXT: ret i32 [[CALL]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 15)
+ ret i32 %call
+}
+
+define i32 @cmp16(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp16(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 16)
+; X32-NEXT: ret i32 [[CALL]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 16)
+ ret i32 %call
+}
+
+define i32 @cmp_eq2(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq2(
+; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i16*
+; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i16*
+; X32-NEXT: [[TMP3:%.*]] = load i16, i16* [[TMP1]], align 1
+; X32-NEXT: [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 1
+; X32-NEXT: [[TMP5:%.*]] = icmp ne i16 [[TMP3]], [[TMP4]]
+; X32-NEXT: [[TMP6:%.*]] = zext i1 [[TMP5]] to i32
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP6]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 2)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq3(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq3(
+; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i16*
+; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i16*
+; X32-NEXT: [[TMP3:%.*]] = load i16, i16* [[TMP1]], align 1
+; X32-NEXT: [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 1
+; X32-NEXT: [[TMP5:%.*]] = xor i16 [[TMP3]], [[TMP4]]
+; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 2
+; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 2
+; X32-NEXT: [[TMP8:%.*]] = load i8, i8* [[TMP6]], align 1
+; X32-NEXT: [[TMP9:%.*]] = load i8, i8* [[TMP7]], align 1
+; X32-NEXT: [[TMP10:%.*]] = zext i8 [[TMP8]] to i16
+; X32-NEXT: [[TMP11:%.*]] = zext i8 [[TMP9]] to i16
+; X32-NEXT: [[TMP12:%.*]] = xor i16 [[TMP10]], [[TMP11]]
+; X32-NEXT: [[TMP13:%.*]] = or i16 [[TMP5]], [[TMP12]]
+; X32-NEXT: [[TMP14:%.*]] = icmp ne i16 [[TMP13]], 0
+; X32-NEXT: [[TMP15:%.*]] = zext i1 [[TMP14]] to i32
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP15]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 3)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq4(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq4(
+; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
+; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
+; X32-NEXT: [[TMP5:%.*]] = icmp ne i32 [[TMP3]], [[TMP4]]
+; X32-NEXT: [[TMP6:%.*]] = zext i1 [[TMP5]] to i32
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP6]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 4)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq5(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq5(
+; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
+; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
+; X32-NEXT: [[TMP5:%.*]] = xor i32 [[TMP3]], [[TMP4]]
+; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 4
+; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 4
+; X32-NEXT: [[TMP8:%.*]] = load i8, i8* [[TMP6]], align 1
+; X32-NEXT: [[TMP9:%.*]] = load i8, i8* [[TMP7]], align 1
+; X32-NEXT: [[TMP10:%.*]] = zext i8 [[TMP8]] to i32
+; X32-NEXT: [[TMP11:%.*]] = zext i8 [[TMP9]] to i32
+; X32-NEXT: [[TMP12:%.*]] = xor i32 [[TMP10]], [[TMP11]]
+; X32-NEXT: [[TMP13:%.*]] = or i32 [[TMP5]], [[TMP12]]
+; X32-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
+; X32-NEXT: [[TMP15:%.*]] = zext i1 [[TMP14]] to i32
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP15]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 5)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq6(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq6(
+; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
+; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
+; X32-NEXT: [[TMP5:%.*]] = xor i32 [[TMP3]], [[TMP4]]
+; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 4
+; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 4
+; X32-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP6]] to i16*
+; X32-NEXT: [[TMP9:%.*]] = bitcast i8* [[TMP7]] to i16*
+; X32-NEXT: [[TMP10:%.*]] = load i16, i16* [[TMP8]], align 1
+; X32-NEXT: [[TMP11:%.*]] = load i16, i16* [[TMP9]], align 1
+; X32-NEXT: [[TMP12:%.*]] = zext i16 [[TMP10]] to i32
+; X32-NEXT: [[TMP13:%.*]] = zext i16 [[TMP11]] to i32
+; X32-NEXT: [[TMP14:%.*]] = xor i32 [[TMP12]], [[TMP13]]
+; X32-NEXT: [[TMP15:%.*]] = or i32 [[TMP5]], [[TMP14]]
+; X32-NEXT: [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
+; X32-NEXT: [[TMP17:%.*]] = zext i1 [[TMP16]] to i32
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP17]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 6)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq6_align4(i8* nocapture readonly align 4 %x, i8* nocapture readonly align 4 %y) {
+; X32-LABEL: @cmp_eq6_align4(
+; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 4
+; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 4
+; X32-NEXT: [[TMP5:%.*]] = xor i32 [[TMP3]], [[TMP4]]
+; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 4
+; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 4
+; X32-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP6]] to i16*
+; X32-NEXT: [[TMP9:%.*]] = bitcast i8* [[TMP7]] to i16*
+; X32-NEXT: [[TMP10:%.*]] = load i16, i16* [[TMP8]], align 4
+; X32-NEXT: [[TMP11:%.*]] = load i16, i16* [[TMP9]], align 4
+; X32-NEXT: [[TMP12:%.*]] = zext i16 [[TMP10]] to i32
+; X32-NEXT: [[TMP13:%.*]] = zext i16 [[TMP11]] to i32
+; X32-NEXT: [[TMP14:%.*]] = xor i32 [[TMP12]], [[TMP13]]
+; X32-NEXT: [[TMP15:%.*]] = or i32 [[TMP5]], [[TMP14]]
+; X32-NEXT: [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
+; X32-NEXT: [[TMP17:%.*]] = zext i1 [[TMP16]] to i32
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP17]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 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) {
+; X32-LABEL: @cmp_eq7(
+; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
+; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
+; X32-NEXT: [[TMP5:%.*]] = xor i32 [[TMP3]], [[TMP4]]
+; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 3
+; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 3
+; X32-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP6]] to i32*
+; X32-NEXT: [[TMP9:%.*]] = bitcast i8* [[TMP7]] to i32*
+; X32-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP8]], align 1
+; X32-NEXT: [[TMP11:%.*]] = load i32, i32* [[TMP9]], align 1
+; X32-NEXT: [[TMP12:%.*]] = xor i32 [[TMP10]], [[TMP11]]
+; X32-NEXT: [[TMP13:%.*]] = or i32 [[TMP5]], [[TMP12]]
+; X32-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
+; X32-NEXT: [[TMP15:%.*]] = zext i1 [[TMP14]] to i32
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP15]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 7)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq8(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq8(
+; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
+; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
+; X32-NEXT: [[TMP5:%.*]] = xor i32 [[TMP3]], [[TMP4]]
+; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 4
+; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 4
+; X32-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP6]] to i32*
+; X32-NEXT: [[TMP9:%.*]] = bitcast i8* [[TMP7]] to i32*
+; X32-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP8]], align 1
+; X32-NEXT: [[TMP11:%.*]] = load i32, i32* [[TMP9]], align 1
+; X32-NEXT: [[TMP12:%.*]] = xor i32 [[TMP10]], [[TMP11]]
+; X32-NEXT: [[TMP13:%.*]] = or i32 [[TMP5]], [[TMP12]]
+; X32-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
+; X32-NEXT: [[TMP15:%.*]] = zext i1 [[TMP14]] to i32
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP15]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 8)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq9(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq9(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 9)
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 9)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq10(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq10(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 10)
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 10)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq11(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq11(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 11)
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 11)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq12(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq12(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 12)
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 12)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq13(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq13(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 13)
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 13)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq14(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq14(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 14)
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 14)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+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:%.*]], i32 15)
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 15)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
+define i32 @cmp_eq16(i8* nocapture readonly %x, i8* nocapture readonly %y) {
+; X32-LABEL: @cmp_eq16(
+; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i32 16)
+; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
+; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X32-NEXT: ret i32 [[CONV]]
+;
+ %call = tail call i32 @memcmp(i8* %x, i8* %y, i32 16)
+ %cmp = icmp eq i32 %call, 0
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
+
diff --git a/llvm/test/Transforms/ExpandMemCmp/X86/memcmp.ll b/llvm/test/Transforms/ExpandMemCmp/X86/memcmp.ll
index df05021473585..27b3ce4f04b27 100644
--- a/llvm/test/Transforms/ExpandMemCmp/X86/memcmp.ll
+++ b/llvm/test/Transforms/ExpandMemCmp/X86/memcmp.ll
@@ -1,238 +1,205 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -S -expandmemcmp -mtriple=i686-unknown-unknown -data-layout=e-m:o-p:32:32-f64:32:64-f80:128-n8:16:32-S128 < %s | FileCheck %s --check-prefix=ALL --check-prefix=X32
-; RUN: opt -S -expandmemcmp -memcmp-num-loads-per-block=1 -mtriple=x86_64-unknown-unknown -data-layout=e-m:o-i64:64-f80:128-n8:16:32:64-S128 < %s | FileCheck %s --check-prefix=ALL --check-prefix=X64 --check-prefix=X64_1LD
-; RUN: opt -S -expandmemcmp -memcmp-num-loads-per-block=2 -mtriple=x86_64-unknown-unknown -data-layout=e-m:o-i64:64-f80:128-n8:16:32:64-S128 < %s | FileCheck %s --check-prefix=ALL --check-prefix=X64 --check-prefix=X64_2LD
+; RUN: opt -S -expandmemcmp -memcmp-num-loads-per-block=1 -mtriple=x86_64-unknown-unknown -data-layout=e-m:o-i64:64-f80:128-n8:16:32:64-S128 < %s | FileCheck %s --check-prefix=X64 --check-prefix=X64_1LD
+; RUN: opt -S -expandmemcmp -memcmp-num-loads-per-block=2 -mtriple=x86_64-unknown-unknown -data-layout=e-m:o-i64:64-f80:128-n8:16:32:64-S128 < %s | FileCheck %s --check-prefix=X64 --check-prefix=X64_2LD
declare i32 @memcmp(i8* nocapture, i8* nocapture, i64)
define i32 @cmp2(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; ALL-LABEL: @cmp2(
-; ALL-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i16*
-; ALL-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i16*
-; ALL-NEXT: [[TMP3:%.*]] = load i16, i16* [[TMP1]], align 1
-; ALL-NEXT: [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 1
-; ALL-NEXT: [[TMP5:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP3]])
-; ALL-NEXT: [[TMP6:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP4]])
-; ALL-NEXT: [[TMP7:%.*]] = zext i16 [[TMP5]] to i32
-; ALL-NEXT: [[TMP8:%.*]] = zext i16 [[TMP6]] to i32
-; ALL-NEXT: [[TMP9:%.*]] = sub i32 [[TMP7]], [[TMP8]]
-; ALL-NEXT: ret i32 [[TMP9]]
+; X64-LABEL: @cmp2(
+; X64-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i16*
+; X64-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i16*
+; X64-NEXT: [[TMP3:%.*]] = load i16, i16* [[TMP1]], align 1
+; X64-NEXT: [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 1
+; X64-NEXT: [[TMP5:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP3]])
+; X64-NEXT: [[TMP6:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP4]])
+; X64-NEXT: [[TMP7:%.*]] = zext i16 [[TMP5]] to i32
+; X64-NEXT: [[TMP8:%.*]] = zext i16 [[TMP6]] to i32
+; X64-NEXT: [[TMP9:%.*]] = sub i32 [[TMP7]], [[TMP8]]
+; X64-NEXT: ret i32 [[TMP9]]
;
%call = tail call i32 @memcmp(i8* %x, i8* %y, i64 2)
ret i32 %call
}
define i32 @cmp2_align2(i8* nocapture readonly align 2 %x, i8* nocapture readonly align 2 %y) {
-; ALL-LABEL: @cmp2_align2(
-; ALL-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i16*
-; ALL-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i16*
-; ALL-NEXT: [[TMP3:%.*]] = load i16, i16* [[TMP1]], align 2
-; ALL-NEXT: [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 2
-; ALL-NEXT: [[TMP5:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP3]])
-; ALL-NEXT: [[TMP6:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP4]])
-; ALL-NEXT: [[TMP7:%.*]] = zext i16 [[TMP5]] to i32
-; ALL-NEXT: [[TMP8:%.*]] = zext i16 [[TMP6]] to i32
-; ALL-NEXT: [[TMP9:%.*]] = sub i32 [[TMP7]], [[TMP8]]
-; ALL-NEXT: ret i32 [[TMP9]]
+; X64-LABEL: @cmp2_align2(
+; X64-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i16*
+; X64-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i16*
+; X64-NEXT: [[TMP3:%.*]] = load i16, i16* [[TMP1]], align 2
+; X64-NEXT: [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 2
+; X64-NEXT: [[TMP5:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP3]])
+; X64-NEXT: [[TMP6:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP4]])
+; X64-NEXT: [[TMP7:%.*]] = zext i16 [[TMP5]] to i32
+; X64-NEXT: [[TMP8:%.*]] = zext i16 [[TMP6]] to i32
+; X64-NEXT: [[TMP9:%.*]] = sub i32 [[TMP7]], [[TMP8]]
+; X64-NEXT: ret i32 [[TMP9]]
;
%call = tail call i32 @memcmp(i8* %x, i8* %y, i64 2)
ret i32 %call
}
define i32 @cmp3(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; ALL-LABEL: @cmp3(
-; ALL-NEXT: br label [[LOADBB:%.*]]
-; ALL: res_block:
-; ALL-NEXT: [[TMP1:%.*]] = icmp ult i16 [[TMP7:%.*]], [[TMP8:%.*]]
-; ALL-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
-; ALL-NEXT: br label [[ENDBLOCK:%.*]]
-; ALL: loadbb:
-; ALL-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i16*
-; ALL-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i16*
-; ALL-NEXT: [[TMP5:%.*]] = load i16, i16* [[TMP3]], align 1
-; ALL-NEXT: [[TMP6:%.*]] = load i16, i16* [[TMP4]], align 1
-; ALL-NEXT: [[TMP7]] = call i16 @llvm.bswap.i16(i16 [[TMP5]])
-; ALL-NEXT: [[TMP8]] = call i16 @llvm.bswap.i16(i16 [[TMP6]])
-; ALL-NEXT: [[TMP9:%.*]] = icmp eq i16 [[TMP7]], [[TMP8]]
-; ALL-NEXT: br i1 [[TMP9]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; ALL: loadbb1:
-; ALL-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 2
-; ALL-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 2
-; ALL-NEXT: [[TMP12:%.*]] = load i8, i8* [[TMP10]], align 1
-; ALL-NEXT: [[TMP13:%.*]] = load i8, i8* [[TMP11]], align 1
-; ALL-NEXT: [[TMP14:%.*]] = zext i8 [[TMP12]] to i32
-; ALL-NEXT: [[TMP15:%.*]] = zext i8 [[TMP13]] to i32
-; ALL-NEXT: [[TMP16:%.*]] = sub i32 [[TMP14]], [[TMP15]]
-; ALL-NEXT: br label [[ENDBLOCK]]
-; ALL: endblock:
-; ALL-NEXT: [[PHI_RES:%.*]] = phi i32 [ [[TMP16]], [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
-; ALL-NEXT: ret i32 [[PHI_RES]]
+; X64-LABEL: @cmp3(
+; X64-NEXT: br label [[LOADBB:%.*]]
+; X64: res_block:
+; X64-NEXT: [[TMP1:%.*]] = icmp ult i16 [[TMP7:%.*]], [[TMP8:%.*]]
+; X64-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
+; X64-NEXT: br label [[ENDBLOCK:%.*]]
+; X64: loadbb:
+; X64-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i16*
+; X64-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i16*
+; X64-NEXT: [[TMP5:%.*]] = load i16, i16* [[TMP3]], align 1
+; X64-NEXT: [[TMP6:%.*]] = load i16, i16* [[TMP4]], align 1
+; X64-NEXT: [[TMP7]] = call i16 @llvm.bswap.i16(i16 [[TMP5]])
+; X64-NEXT: [[TMP8]] = call i16 @llvm.bswap.i16(i16 [[TMP6]])
+; X64-NEXT: [[TMP9:%.*]] = icmp eq i16 [[TMP7]], [[TMP8]]
+; X64-NEXT: br i1 [[TMP9]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
+; X64: loadbb1:
+; X64-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 2
+; X64-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 2
+; X64-NEXT: [[TMP12:%.*]] = load i8, i8* [[TMP10]], align 1
+; X64-NEXT: [[TMP13:%.*]] = load i8, i8* [[TMP11]], align 1
+; X64-NEXT: [[TMP14:%.*]] = zext i8 [[TMP12]] to i32
+; X64-NEXT: [[TMP15:%.*]] = zext i8 [[TMP13]] to i32
+; X64-NEXT: [[TMP16:%.*]] = sub i32 [[TMP14]], [[TMP15]]
+; X64-NEXT: br label [[ENDBLOCK]]
+; X64: endblock:
+; X64-NEXT: [[PHI_RES:%.*]] = phi i32 [ [[TMP16]], [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
+; X64-NEXT: ret i32 [[PHI_RES]]
;
%call = tail call i32 @memcmp(i8* %x, i8* %y, i64 3)
ret i32 %call
}
define i32 @cmp4(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; ALL-LABEL: @cmp4(
-; ALL-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; ALL-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; ALL-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
-; ALL-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
-; ALL-NEXT: [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
-; ALL-NEXT: [[TMP6:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP4]])
-; ALL-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
-; ALL-NEXT: [[TMP8:%.*]] = icmp ult i32 [[TMP5]], [[TMP6]]
-; ALL-NEXT: [[TMP9:%.*]] = zext i1 [[TMP7]] to i32
-; ALL-NEXT: [[TMP10:%.*]] = zext i1 [[TMP8]] to i32
-; ALL-NEXT: [[TMP11:%.*]] = sub i32 [[TMP9]], [[TMP10]]
-; ALL-NEXT: ret i32 [[TMP11]]
+; X64-LABEL: @cmp4(
+; X64-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X64-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X64-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
+; X64-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
+; X64-NEXT: [[TMP5:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP3]])
+; X64-NEXT: [[TMP6:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP4]])
+; X64-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[TMP5]], [[TMP6]]
+; X64-NEXT: [[TMP8:%.*]] = icmp ult i32 [[TMP5]], [[TMP6]]
+; X64-NEXT: [[TMP9:%.*]] = zext i1 [[TMP7]] to i32
+; X64-NEXT: [[TMP10:%.*]] = zext i1 [[TMP8]] to i32
+; X64-NEXT: [[TMP11:%.*]] = sub i32 [[TMP9]], [[TMP10]]
+; X64-NEXT: ret i32 [[TMP11]]
;
%call = tail call i32 @memcmp(i8* %x, i8* %y, i64 4)
ret i32 %call
}
define i32 @cmp5(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; ALL-LABEL: @cmp5(
-; ALL-NEXT: br label [[LOADBB:%.*]]
-; ALL: res_block:
-; ALL-NEXT: [[TMP1:%.*]] = icmp ult i32 [[TMP7:%.*]], [[TMP8:%.*]]
-; ALL-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
-; ALL-NEXT: br label [[ENDBLOCK:%.*]]
-; ALL: loadbb:
-; ALL-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; ALL-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; ALL-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 1
-; ALL-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP4]], align 1
-; ALL-NEXT: [[TMP7]] = call i32 @llvm.bswap.i32(i32 [[TMP5]])
-; ALL-NEXT: [[TMP8]] = call i32 @llvm.bswap.i32(i32 [[TMP6]])
-; ALL-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP7]], [[TMP8]]
-; ALL-NEXT: br i1 [[TMP9]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
-; ALL: loadbb1:
-; ALL-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 4
-; ALL-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 4
-; ALL-NEXT: [[TMP12:%.*]] = load i8, i8* [[TMP10]], align 1
-; ALL-NEXT: [[TMP13:%.*]] = load i8, i8* [[TMP11]], align 1
-; ALL-NEXT: [[TMP14:%.*]] = zext i8 [[TMP12]] to i32
-; ALL-NEXT: [[TMP15:%.*]] = zext i8 [[TMP13]] to i32
-; ALL-NEXT: [[TMP16:%.*]] = sub i32 [[TMP14]], [[TMP15]]
-; ALL-NEXT: br label [[ENDBLOCK]]
-; ALL: endblock:
-; ALL-NEXT: [[PHI_RES:%.*]] = phi i32 [ [[TMP16]], [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
-; ALL-NEXT: ret i32 [[PHI_RES]]
+; X64-LABEL: @cmp5(
+; X64-NEXT: br label [[LOADBB:%.*]]
+; X64: res_block:
+; X64-NEXT: [[TMP1:%.*]] = icmp ult i32 [[TMP7:%.*]], [[TMP8:%.*]]
+; X64-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
+; X64-NEXT: br label [[ENDBLOCK:%.*]]
+; X64: loadbb:
+; X64-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X64-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X64-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 1
+; X64-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP4]], align 1
+; X64-NEXT: [[TMP7]] = call i32 @llvm.bswap.i32(i32 [[TMP5]])
+; X64-NEXT: [[TMP8]] = call i32 @llvm.bswap.i32(i32 [[TMP6]])
+; X64-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP7]], [[TMP8]]
+; X64-NEXT: br i1 [[TMP9]], label [[LOADBB1:%.*]], label [[RES_BLOCK:%.*]]
+; X64: loadbb1:
+; X64-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 4
+; X64-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 4
+; X64-NEXT: [[TMP12:%.*]] = load i8, i8* [[TMP10]], align 1
+; X64-NEXT: [[TMP13:%.*]] = load i8, i8* [[TMP11]], align 1
+; X64-NEXT: [[TMP14:%.*]] = zext i8 [[TMP12]] to i32
+; X64-NEXT: [[TMP15:%.*]] = zext i8 [[TMP13]] to i32
+; X64-NEXT: [[TMP16:%.*]] = sub i32 [[TMP14]], [[TMP15]]
+; X64-NEXT: br label [[ENDBLOCK]]
+; X64: endblock:
+; X64-NEXT: [[PHI_RES:%.*]] = phi i32 [ [[TMP16]], [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
+; X64-NEXT: ret i32 [[PHI_RES]]
;
%call = tail call i32 @memcmp(i8* %x, i8* %y, i64 5)
ret i32 %call
}
define i32 @cmp6(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; ALL-LABEL: @cmp6(
-; ALL-NEXT: br label [[LOADBB:%.*]]
-; ALL: res_block:
-; ALL-NEXT: [[PHI_SRC1:%.*]] = phi i32 [ [[TMP7:%.*]], [[LOADBB]] ], [ [[TMP18:%.*]], [[LOADBB1:%.*]] ]
-; ALL-NEXT: [[PHI_SRC2:%.*]] = phi i32 [ [[TMP8:%.*]], [[LOADBB]] ], [ [[TMP19:%.*]], [[LOADBB1]] ]
-; ALL-NEXT: [[TMP1:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
-; ALL-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
-; ALL-NEXT: br label [[ENDBLOCK:%.*]]
-; ALL: loadbb:
-; ALL-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; ALL-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; ALL-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 1
-; ALL-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP4]], align 1
-; ALL-NEXT: [[TMP7]] = call i32 @llvm.bswap.i32(i32 [[TMP5]])
-; ALL-NEXT: [[TMP8]] = call i32 @llvm.bswap.i32(i32 [[TMP6]])
-; ALL-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP7]], [[TMP8]]
-; ALL-NEXT: br i1 [[TMP9]], label [[LOADBB1]], label [[RES_BLOCK:%.*]]
-; ALL: loadbb1:
-; ALL-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 4
-; ALL-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 4
-; ALL-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i16*
-; ALL-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i16*
-; ALL-NEXT: [[TMP14:%.*]] = load i16, i16* [[TMP12]], align 1
-; ALL-NEXT: [[TMP15:%.*]] = load i16, i16* [[TMP13]], align 1
-; ALL-NEXT: [[TMP16:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP14]])
-; ALL-NEXT: [[TMP17:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP15]])
-; ALL-NEXT: [[TMP18]] = zext i16 [[TMP16]] to i32
-; ALL-NEXT: [[TMP19]] = zext i16 [[TMP17]] to i32
-; ALL-NEXT: [[TMP20:%.*]] = icmp eq i32 [[TMP18]], [[TMP19]]
-; ALL-NEXT: br i1 [[TMP20]], label [[ENDBLOCK]], label [[RES_BLOCK]]
-; ALL: endblock:
-; ALL-NEXT: [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
-; ALL-NEXT: ret i32 [[PHI_RES]]
+; X64-LABEL: @cmp6(
+; X64-NEXT: br label [[LOADBB:%.*]]
+; X64: res_block:
+; X64-NEXT: [[PHI_SRC1:%.*]] = phi i32 [ [[TMP7:%.*]], [[LOADBB]] ], [ [[TMP18:%.*]], [[LOADBB1:%.*]] ]
+; X64-NEXT: [[PHI_SRC2:%.*]] = phi i32 [ [[TMP8:%.*]], [[LOADBB]] ], [ [[TMP19:%.*]], [[LOADBB1]] ]
+; X64-NEXT: [[TMP1:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
+; X64-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
+; X64-NEXT: br label [[ENDBLOCK:%.*]]
+; X64: loadbb:
+; X64-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X64-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X64-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 1
+; X64-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP4]], align 1
+; X64-NEXT: [[TMP7]] = call i32 @llvm.bswap.i32(i32 [[TMP5]])
+; X64-NEXT: [[TMP8]] = call i32 @llvm.bswap.i32(i32 [[TMP6]])
+; X64-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP7]], [[TMP8]]
+; X64-NEXT: br i1 [[TMP9]], label [[LOADBB1]], label [[RES_BLOCK:%.*]]
+; X64: loadbb1:
+; X64-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 4
+; X64-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 4
+; X64-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i16*
+; X64-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i16*
+; X64-NEXT: [[TMP14:%.*]] = load i16, i16* [[TMP12]], align 1
+; X64-NEXT: [[TMP15:%.*]] = load i16, i16* [[TMP13]], align 1
+; X64-NEXT: [[TMP16:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP14]])
+; X64-NEXT: [[TMP17:%.*]] = call i16 @llvm.bswap.i16(i16 [[TMP15]])
+; X64-NEXT: [[TMP18]] = zext i16 [[TMP16]] to i32
+; X64-NEXT: [[TMP19]] = zext i16 [[TMP17]] to i32
+; X64-NEXT: [[TMP20:%.*]] = icmp eq i32 [[TMP18]], [[TMP19]]
+; X64-NEXT: br i1 [[TMP20]], label [[ENDBLOCK]], label [[RES_BLOCK]]
+; X64: endblock:
+; X64-NEXT: [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
+; X64-NEXT: ret i32 [[PHI_RES]]
;
%call = tail call i32 @memcmp(i8* %x, i8* %y, i64 6)
ret i32 %call
}
define i32 @cmp7(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; ALL-LABEL: @cmp7(
-; ALL-NEXT: br label [[LOADBB:%.*]]
-; ALL: res_block:
-; ALL-NEXT: [[PHI_SRC1:%.*]] = phi i32 [ [[TMP7:%.*]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1:%.*]] ]
-; ALL-NEXT: [[PHI_SRC2:%.*]] = phi i32 [ [[TMP8:%.*]], [[LOADBB]] ], [ [[TMP17:%.*]], [[LOADBB1]] ]
-; ALL-NEXT: [[TMP1:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
-; ALL-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
-; ALL-NEXT: br label [[ENDBLOCK:%.*]]
-; ALL: loadbb:
-; ALL-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; ALL-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; ALL-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]]
-; ALL-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP4]]
-; ALL-NEXT: [[TMP7]] = call i32 @llvm.bswap.i32(i32 [[TMP5]])
-; ALL-NEXT: [[TMP8]] = call i32 @llvm.bswap.i32(i32 [[TMP6]])
-; ALL-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP7]], [[TMP8]]
-; ALL-NEXT: br i1 [[TMP9]], label [[LOADBB1]], label [[RES_BLOCK:%.*]]
-; ALL: loadbb1:
-; ALL-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 3
-; ALL-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 3
-; ALL-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i32*
-; ALL-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i32*
-; ALL-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP12]]
-; ALL-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP13]]
-; ALL-NEXT: [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; ALL-NEXT: [[TMP17]] = call i32 @llvm.bswap.i32(i32 [[TMP15]])
-; ALL-NEXT: [[TMP18:%.*]] = icmp eq i32 [[TMP16]], [[TMP17]]
-; ALL-NEXT: br i1 [[TMP18]], label [[ENDBLOCK]], label [[RES_BLOCK]]
-; ALL: endblock:
-; ALL-NEXT: [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
-; ALL-NEXT: ret i32 [[PHI_RES]]
+; X64-LABEL: @cmp7(
+; X64-NEXT: br label [[LOADBB:%.*]]
+; X64: res_block:
+; X64-NEXT: [[PHI_SRC1:%.*]] = phi i32 [ [[TMP7:%.*]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1:%.*]] ]
+; X64-NEXT: [[PHI_SRC2:%.*]] = phi i32 [ [[TMP8:%.*]], [[LOADBB]] ], [ [[TMP17:%.*]], [[LOADBB1]] ]
+; X64-NEXT: [[TMP1:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
+; X64-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
+; X64-NEXT: br label [[ENDBLOCK:%.*]]
+; X64: loadbb:
+; X64-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X64-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X64-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 1
+; X64-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP4]], align 1
+; X64-NEXT: [[TMP7]] = call i32 @llvm.bswap.i32(i32 [[TMP5]])
+; X64-NEXT: [[TMP8]] = call i32 @llvm.bswap.i32(i32 [[TMP6]])
+; X64-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP7]], [[TMP8]]
+; X64-NEXT: br i1 [[TMP9]], label [[LOADBB1]], label [[RES_BLOCK:%.*]]
+; X64: loadbb1:
+; X64-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 3
+; X64-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 3
+; X64-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i32*
+; X64-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i32*
+; X64-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP12]], align 1
+; X64-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP13]], align 1
+; X64-NEXT: [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
+; X64-NEXT: [[TMP17]] = call i32 @llvm.bswap.i32(i32 [[TMP15]])
+; X64-NEXT: [[TMP18:%.*]] = icmp eq i32 [[TMP16]], [[TMP17]]
+; X64-NEXT: br i1 [[TMP18]], label [[ENDBLOCK]], label [[RES_BLOCK]]
+; X64: endblock:
+; X64-NEXT: [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
+; X64-NEXT: ret i32 [[PHI_RES]]
;
%call = tail call i32 @memcmp(i8* %x, i8* %y, i64 7)
ret i32 %call
}
define i32 @cmp8(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; X32-LABEL: @cmp8(
-; X32-NEXT: br label [[LOADBB:%.*]]
-; X32: res_block:
-; X32-NEXT: [[PHI_SRC1:%.*]] = phi i32 [ [[TMP7:%.*]], [[LOADBB]] ], [ [[TMP16:%.*]], [[LOADBB1:%.*]] ]
-; X32-NEXT: [[PHI_SRC2:%.*]] = phi i32 [ [[TMP8:%.*]], [[LOADBB]] ], [ [[TMP17:%.*]], [[LOADBB1]] ]
-; X32-NEXT: [[TMP1:%.*]] = icmp ult i32 [[PHI_SRC1]], [[PHI_SRC2]]
-; X32-NEXT: [[TMP2:%.*]] = select i1 [[TMP1]], i32 -1, i32 1
-; X32-NEXT: br label [[ENDBLOCK:%.*]]
-; X32: loadbb:
-; X32-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 1
-; X32-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP4]], align 1
-; X32-NEXT: [[TMP7]] = call i32 @llvm.bswap.i32(i32 [[TMP5]])
-; X32-NEXT: [[TMP8]] = call i32 @llvm.bswap.i32(i32 [[TMP6]])
-; X32-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP7]], [[TMP8]]
-; X32-NEXT: br i1 [[TMP9]], label [[LOADBB1]], label [[RES_BLOCK:%.*]]
-; X32: loadbb1:
-; X32-NEXT: [[TMP10:%.*]] = getelementptr i8, i8* [[X]], i64 4
-; X32-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 4
-; X32-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i32*
-; X32-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i32*
-; X32-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP12]], align 1
-; X32-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP13]], align 1
-; X32-NEXT: [[TMP16]] = call i32 @llvm.bswap.i32(i32 [[TMP14]])
-; X32-NEXT: [[TMP17]] = call i32 @llvm.bswap.i32(i32 [[TMP15]])
-; X32-NEXT: [[TMP18:%.*]] = icmp eq i32 [[TMP16]], [[TMP17]]
-; X32-NEXT: br i1 [[TMP18]], label [[ENDBLOCK]], label [[RES_BLOCK]]
-; X32: endblock:
-; X32-NEXT: [[PHI_RES:%.*]] = phi i32 [ 0, [[LOADBB1]] ], [ [[TMP2]], [[RES_BLOCK]] ]
-; X32-NEXT: ret i32 [[PHI_RES]]
-;
; X64-LABEL: @cmp8(
; X64-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i64*
; X64-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i64*
@@ -252,10 +219,6 @@ 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: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 9)
-; X32-NEXT: ret i32 [[CALL]]
-;
; X64-LABEL: @cmp9(
; X64-NEXT: br label [[LOADBB:%.*]]
; X64: res_block:
@@ -289,10 +252,6 @@ 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: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 10)
-; X32-NEXT: ret i32 [[CALL]]
-;
; X64-LABEL: @cmp10(
; X64-NEXT: br label [[LOADBB:%.*]]
; X64: res_block:
@@ -332,10 +291,6 @@ 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: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 11)
-; X32-NEXT: ret i32 [[CALL]]
-;
; X64-LABEL: @cmp11(
; X64-NEXT: br label [[LOADBB:%.*]]
; X64: res_block:
@@ -347,8 +302,8 @@ define i32 @cmp11(i8* nocapture readonly %x, i8* nocapture readonly %y) {
; X64: loadbb:
; X64-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i64*
; X64-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT: [[TMP5:%.*]] = load i64, i64* [[TMP3]]
-; X64-NEXT: [[TMP6:%.*]] = load i64, i64* [[TMP4]]
+; X64-NEXT: [[TMP5:%.*]] = load i64, i64* [[TMP3]], align 1
+; X64-NEXT: [[TMP6:%.*]] = load i64, i64* [[TMP4]], align 1
; X64-NEXT: [[TMP7]] = call i64 @llvm.bswap.i64(i64 [[TMP5]])
; X64-NEXT: [[TMP8]] = call i64 @llvm.bswap.i64(i64 [[TMP6]])
; X64-NEXT: [[TMP9:%.*]] = icmp eq i64 [[TMP7]], [[TMP8]]
@@ -358,8 +313,8 @@ define i32 @cmp11(i8* nocapture readonly %x, i8* nocapture readonly %y) {
; X64-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 3
; X64-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i64*
; X64-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i64*
-; X64-NEXT: [[TMP14:%.*]] = load i64, i64* [[TMP12]]
-; X64-NEXT: [[TMP15:%.*]] = load i64, i64* [[TMP13]]
+; X64-NEXT: [[TMP14:%.*]] = load i64, i64* [[TMP12]], align 1
+; X64-NEXT: [[TMP15:%.*]] = load i64, i64* [[TMP13]], align 1
; X64-NEXT: [[TMP16]] = call i64 @llvm.bswap.i64(i64 [[TMP14]])
; X64-NEXT: [[TMP17]] = call i64 @llvm.bswap.i64(i64 [[TMP15]])
; X64-NEXT: [[TMP18:%.*]] = icmp eq i64 [[TMP16]], [[TMP17]]
@@ -373,10 +328,6 @@ 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: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 12)
-; X32-NEXT: ret i32 [[CALL]]
-;
; X64-LABEL: @cmp12(
; X64-NEXT: br label [[LOADBB:%.*]]
; X64: res_block:
@@ -416,10 +367,6 @@ 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: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 13)
-; X32-NEXT: ret i32 [[CALL]]
-;
; X64-LABEL: @cmp13(
; X64-NEXT: br label [[LOADBB:%.*]]
; X64: res_block:
@@ -431,8 +378,8 @@ define i32 @cmp13(i8* nocapture readonly %x, i8* nocapture readonly %y) {
; X64: loadbb:
; X64-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i64*
; X64-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT: [[TMP5:%.*]] = load i64, i64* [[TMP3]]
-; X64-NEXT: [[TMP6:%.*]] = load i64, i64* [[TMP4]]
+; X64-NEXT: [[TMP5:%.*]] = load i64, i64* [[TMP3]], align 1
+; X64-NEXT: [[TMP6:%.*]] = load i64, i64* [[TMP4]], align 1
; X64-NEXT: [[TMP7]] = call i64 @llvm.bswap.i64(i64 [[TMP5]])
; X64-NEXT: [[TMP8]] = call i64 @llvm.bswap.i64(i64 [[TMP6]])
; X64-NEXT: [[TMP9:%.*]] = icmp eq i64 [[TMP7]], [[TMP8]]
@@ -442,8 +389,8 @@ define i32 @cmp13(i8* nocapture readonly %x, i8* nocapture readonly %y) {
; X64-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 5
; X64-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i64*
; X64-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i64*
-; X64-NEXT: [[TMP14:%.*]] = load i64, i64* [[TMP12]]
-; X64-NEXT: [[TMP15:%.*]] = load i64, i64* [[TMP13]]
+; X64-NEXT: [[TMP14:%.*]] = load i64, i64* [[TMP12]], align 1
+; X64-NEXT: [[TMP15:%.*]] = load i64, i64* [[TMP13]], align 1
; X64-NEXT: [[TMP16]] = call i64 @llvm.bswap.i64(i64 [[TMP14]])
; X64-NEXT: [[TMP17]] = call i64 @llvm.bswap.i64(i64 [[TMP15]])
; X64-NEXT: [[TMP18:%.*]] = icmp eq i64 [[TMP16]], [[TMP17]]
@@ -457,10 +404,6 @@ define i32 @cmp13(i8* nocapture readonly %x, i8* nocapture readonly %y) {
}
define i32 @cmp14(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; X32-LABEL: @cmp14(
-; X32-NEXT: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 14)
-; X32-NEXT: ret i32 [[CALL]]
-;
; X64-LABEL: @cmp14(
; X64-NEXT: br label [[LOADBB:%.*]]
; X64: res_block:
@@ -472,8 +415,8 @@ define i32 @cmp14(i8* nocapture readonly %x, i8* nocapture readonly %y) {
; X64: loadbb:
; X64-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i64*
; X64-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT: [[TMP5:%.*]] = load i64, i64* [[TMP3]]
-; X64-NEXT: [[TMP6:%.*]] = load i64, i64* [[TMP4]]
+; X64-NEXT: [[TMP5:%.*]] = load i64, i64* [[TMP3]], align 1
+; X64-NEXT: [[TMP6:%.*]] = load i64, i64* [[TMP4]], align 1
; X64-NEXT: [[TMP7]] = call i64 @llvm.bswap.i64(i64 [[TMP5]])
; X64-NEXT: [[TMP8]] = call i64 @llvm.bswap.i64(i64 [[TMP6]])
; X64-NEXT: [[TMP9:%.*]] = icmp eq i64 [[TMP7]], [[TMP8]]
@@ -483,8 +426,8 @@ define i32 @cmp14(i8* nocapture readonly %x, i8* nocapture readonly %y) {
; X64-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 6
; X64-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i64*
; X64-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i64*
-; X64-NEXT: [[TMP14:%.*]] = load i64, i64* [[TMP12]]
-; X64-NEXT: [[TMP15:%.*]] = load i64, i64* [[TMP13]]
+; X64-NEXT: [[TMP14:%.*]] = load i64, i64* [[TMP12]], align 1
+; X64-NEXT: [[TMP15:%.*]] = load i64, i64* [[TMP13]], align 1
; X64-NEXT: [[TMP16]] = call i64 @llvm.bswap.i64(i64 [[TMP14]])
; X64-NEXT: [[TMP17]] = call i64 @llvm.bswap.i64(i64 [[TMP15]])
; X64-NEXT: [[TMP18:%.*]] = icmp eq i64 [[TMP16]], [[TMP17]]
@@ -498,10 +441,6 @@ define i32 @cmp14(i8* nocapture readonly %x, i8* nocapture readonly %y) {
}
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: br label [[LOADBB:%.*]]
; X64: res_block:
@@ -513,8 +452,8 @@ define i32 @cmp15(i8* nocapture readonly %x, i8* nocapture readonly %y) {
; X64: loadbb:
; X64-NEXT: [[TMP3:%.*]] = bitcast i8* [[X:%.*]] to i64*
; X64-NEXT: [[TMP4:%.*]] = bitcast i8* [[Y:%.*]] to i64*
-; X64-NEXT: [[TMP5:%.*]] = load i64, i64* [[TMP3]]
-; X64-NEXT: [[TMP6:%.*]] = load i64, i64* [[TMP4]]
+; X64-NEXT: [[TMP5:%.*]] = load i64, i64* [[TMP3]], align 1
+; X64-NEXT: [[TMP6:%.*]] = load i64, i64* [[TMP4]], align 1
; X64-NEXT: [[TMP7]] = call i64 @llvm.bswap.i64(i64 [[TMP5]])
; X64-NEXT: [[TMP8]] = call i64 @llvm.bswap.i64(i64 [[TMP6]])
; X64-NEXT: [[TMP9:%.*]] = icmp eq i64 [[TMP7]], [[TMP8]]
@@ -524,8 +463,8 @@ define i32 @cmp15(i8* nocapture readonly %x, i8* nocapture readonly %y) {
; X64-NEXT: [[TMP11:%.*]] = getelementptr i8, i8* [[Y]], i64 7
; X64-NEXT: [[TMP12:%.*]] = bitcast i8* [[TMP10]] to i64*
; X64-NEXT: [[TMP13:%.*]] = bitcast i8* [[TMP11]] to i64*
-; X64-NEXT: [[TMP14:%.*]] = load i64, i64* [[TMP12]]
-; X64-NEXT: [[TMP15:%.*]] = load i64, i64* [[TMP13]]
+; X64-NEXT: [[TMP14:%.*]] = load i64, i64* [[TMP12]], align 1
+; X64-NEXT: [[TMP15:%.*]] = load i64, i64* [[TMP13]], align 1
; X64-NEXT: [[TMP16]] = call i64 @llvm.bswap.i64(i64 [[TMP14]])
; X64-NEXT: [[TMP17]] = call i64 @llvm.bswap.i64(i64 [[TMP15]])
; X64-NEXT: [[TMP18:%.*]] = icmp eq i64 [[TMP16]], [[TMP17]]
@@ -539,10 +478,6 @@ 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: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 16)
-; X32-NEXT: ret i32 [[CALL]]
-;
; X64-LABEL: @cmp16(
; X64-NEXT: br label [[LOADBB:%.*]]
; X64: res_block:
@@ -580,16 +515,16 @@ define i32 @cmp16(i8* nocapture readonly %x, i8* nocapture readonly %y) {
}
define i32 @cmp_eq2(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; ALL-LABEL: @cmp_eq2(
-; ALL-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i16*
-; ALL-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i16*
-; ALL-NEXT: [[TMP3:%.*]] = load i16, i16* [[TMP1]], align 1
-; ALL-NEXT: [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 1
-; ALL-NEXT: [[TMP5:%.*]] = icmp ne i16 [[TMP3]], [[TMP4]]
-; ALL-NEXT: [[TMP6:%.*]] = zext i1 [[TMP5]] to i32
-; ALL-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP6]], 0
-; ALL-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; ALL-NEXT: ret i32 [[CONV]]
+; X64-LABEL: @cmp_eq2(
+; X64-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i16*
+; X64-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i16*
+; X64-NEXT: [[TMP3:%.*]] = load i16, i16* [[TMP1]], align 1
+; X64-NEXT: [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 1
+; X64-NEXT: [[TMP5:%.*]] = icmp ne i16 [[TMP3]], [[TMP4]]
+; X64-NEXT: [[TMP6:%.*]] = zext i1 [[TMP5]] to i32
+; X64-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP6]], 0
+; X64-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X64-NEXT: ret i32 [[CONV]]
;
%call = tail call i32 @memcmp(i8* %x, i8* %y, i64 2)
%cmp = icmp eq i32 %call, 0
@@ -598,26 +533,6 @@ define i32 @cmp_eq2(i8* nocapture readonly %x, i8* nocapture readonly %y) {
}
define i32 @cmp_eq3(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; X32-LABEL: @cmp_eq3(
-; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i16*
-; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i16*
-; X32-NEXT: [[TMP3:%.*]] = load i16, i16* [[TMP1]], align 1
-; X32-NEXT: [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 1
-; X32-NEXT: [[TMP5:%.*]] = xor i16 [[TMP3]], [[TMP4]]
-; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 2
-; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 2
-; X32-NEXT: [[TMP8:%.*]] = load i8, i8* [[TMP6]], align 1
-; X32-NEXT: [[TMP9:%.*]] = load i8, i8* [[TMP7]], align 1
-; X32-NEXT: [[TMP10:%.*]] = zext i8 [[TMP8]] to i16
-; X32-NEXT: [[TMP11:%.*]] = zext i8 [[TMP9]] to i16
-; X32-NEXT: [[TMP12:%.*]] = xor i16 [[TMP10]], [[TMP11]]
-; X32-NEXT: [[TMP13:%.*]] = or i16 [[TMP5]], [[TMP12]]
-; X32-NEXT: [[TMP14:%.*]] = icmp ne i16 [[TMP13]], 0
-; X32-NEXT: [[TMP15:%.*]] = zext i1 [[TMP14]] to i32
-; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP15]], 0
-; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT: ret i32 [[CONV]]
-;
; X64_1LD-LABEL: @cmp_eq3(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -669,16 +584,16 @@ define i32 @cmp_eq3(i8* nocapture readonly %x, i8* nocapture readonly %y) {
}
define i32 @cmp_eq4(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; ALL-LABEL: @cmp_eq4(
-; ALL-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; ALL-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; ALL-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
-; ALL-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
-; ALL-NEXT: [[TMP5:%.*]] = icmp ne i32 [[TMP3]], [[TMP4]]
-; ALL-NEXT: [[TMP6:%.*]] = zext i1 [[TMP5]] to i32
-; ALL-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP6]], 0
-; ALL-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; ALL-NEXT: ret i32 [[CONV]]
+; X64-LABEL: @cmp_eq4(
+; X64-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
+; X64-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
+; X64-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
+; X64-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
+; X64-NEXT: [[TMP5:%.*]] = icmp ne i32 [[TMP3]], [[TMP4]]
+; X64-NEXT: [[TMP6:%.*]] = zext i1 [[TMP5]] to i32
+; X64-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP6]], 0
+; X64-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
+; X64-NEXT: ret i32 [[CONV]]
;
%call = tail call i32 @memcmp(i8* %x, i8* %y, i64 4)
%cmp = icmp eq i32 %call, 0
@@ -687,26 +602,6 @@ define i32 @cmp_eq4(i8* nocapture readonly %x, i8* nocapture readonly %y) {
}
define i32 @cmp_eq5(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; X32-LABEL: @cmp_eq5(
-; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
-; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
-; X32-NEXT: [[TMP5:%.*]] = xor i32 [[TMP3]], [[TMP4]]
-; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 4
-; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 4
-; X32-NEXT: [[TMP8:%.*]] = load i8, i8* [[TMP6]], align 1
-; X32-NEXT: [[TMP9:%.*]] = load i8, i8* [[TMP7]], align 1
-; X32-NEXT: [[TMP10:%.*]] = zext i8 [[TMP8]] to i32
-; X32-NEXT: [[TMP11:%.*]] = zext i8 [[TMP9]] to i32
-; X32-NEXT: [[TMP12:%.*]] = xor i32 [[TMP10]], [[TMP11]]
-; X32-NEXT: [[TMP13:%.*]] = or i32 [[TMP5]], [[TMP12]]
-; X32-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
-; X32-NEXT: [[TMP15:%.*]] = zext i1 [[TMP14]] to i32
-; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP15]], 0
-; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT: ret i32 [[CONV]]
-;
; X64_1LD-LABEL: @cmp_eq5(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -758,28 +653,6 @@ define i32 @cmp_eq5(i8* nocapture readonly %x, i8* nocapture readonly %y) {
}
define i32 @cmp_eq6(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; X32-LABEL: @cmp_eq6(
-; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
-; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
-; X32-NEXT: [[TMP5:%.*]] = xor i32 [[TMP3]], [[TMP4]]
-; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 4
-; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 4
-; X32-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP6]] to i16*
-; X32-NEXT: [[TMP9:%.*]] = bitcast i8* [[TMP7]] to i16*
-; X32-NEXT: [[TMP10:%.*]] = load i16, i16* [[TMP8]], align 1
-; X32-NEXT: [[TMP11:%.*]] = load i16, i16* [[TMP9]], align 1
-; X32-NEXT: [[TMP12:%.*]] = zext i16 [[TMP10]] to i32
-; X32-NEXT: [[TMP13:%.*]] = zext i16 [[TMP11]] to i32
-; X32-NEXT: [[TMP14:%.*]] = xor i32 [[TMP12]], [[TMP13]]
-; X32-NEXT: [[TMP15:%.*]] = or i32 [[TMP5]], [[TMP14]]
-; X32-NEXT: [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
-; X32-NEXT: [[TMP17:%.*]] = zext i1 [[TMP16]] to i32
-; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP17]], 0
-; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT: ret i32 [[CONV]]
-;
; X64_1LD-LABEL: @cmp_eq6(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -835,28 +708,6 @@ define i32 @cmp_eq6(i8* nocapture readonly %x, i8* nocapture readonly %y) {
}
define i32 @cmp_eq6_align4(i8* nocapture readonly align 4 %x, i8* nocapture readonly align 4 %y) {
-; X32-LABEL: @cmp_eq6_align4(
-; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 4
-; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 4
-; X32-NEXT: [[TMP5:%.*]] = xor i32 [[TMP3]], [[TMP4]]
-; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 4
-; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 4
-; X32-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP6]] to i16*
-; X32-NEXT: [[TMP9:%.*]] = bitcast i8* [[TMP7]] to i16*
-; X32-NEXT: [[TMP10:%.*]] = load i16, i16* [[TMP8]], align 4
-; X32-NEXT: [[TMP11:%.*]] = load i16, i16* [[TMP9]], align 4
-; X32-NEXT: [[TMP12:%.*]] = zext i16 [[TMP10]] to i32
-; X32-NEXT: [[TMP13:%.*]] = zext i16 [[TMP11]] to i32
-; X32-NEXT: [[TMP14:%.*]] = xor i32 [[TMP12]], [[TMP13]]
-; X32-NEXT: [[TMP15:%.*]] = or i32 [[TMP5]], [[TMP14]]
-; X32-NEXT: [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
-; X32-NEXT: [[TMP17:%.*]] = zext i1 [[TMP16]] to i32
-; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP17]], 0
-; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT: ret i32 [[CONV]]
-;
; X64_1LD-LABEL: @cmp_eq6_align4(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -912,26 +763,6 @@ define i32 @cmp_eq6_align4(i8* nocapture readonly align 4 %x, i8* nocapture read
}
define i32 @cmp_eq7(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; X32-LABEL: @cmp_eq7(
-; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
-; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
-; X32-NEXT: [[TMP5:%.*]] = xor i32 [[TMP3]], [[TMP4]]
-; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 3
-; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 3
-; X32-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP6]] to i32*
-; X32-NEXT: [[TMP9:%.*]] = bitcast i8* [[TMP7]] to i32*
-; X32-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP8]], align 1
-; X32-NEXT: [[TMP11:%.*]] = load i32, i32* [[TMP9]], align 1
-; X32-NEXT: [[TMP12:%.*]] = xor i32 [[TMP10]], [[TMP11]]
-; X32-NEXT: [[TMP13:%.*]] = or i32 [[TMP5]], [[TMP12]]
-; X32-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
-; X32-NEXT: [[TMP15:%.*]] = zext i1 [[TMP14]] to i32
-; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP15]], 0
-; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT: ret i32 [[CONV]]
-;
; X64_1LD-LABEL: @cmp_eq7(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -985,26 +816,6 @@ define i32 @cmp_eq7(i8* nocapture readonly %x, i8* nocapture readonly %y) {
}
define i32 @cmp_eq8(i8* nocapture readonly %x, i8* nocapture readonly %y) {
-; X32-LABEL: @cmp_eq8(
-; X32-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i32*
-; X32-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i32*
-; X32-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 1
-; X32-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 1
-; X32-NEXT: [[TMP5:%.*]] = xor i32 [[TMP3]], [[TMP4]]
-; X32-NEXT: [[TMP6:%.*]] = getelementptr i8, i8* [[X]], i64 4
-; X32-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[Y]], i64 4
-; X32-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP6]] to i32*
-; X32-NEXT: [[TMP9:%.*]] = bitcast i8* [[TMP7]] to i32*
-; X32-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP8]], align 1
-; X32-NEXT: [[TMP11:%.*]] = load i32, i32* [[TMP9]], align 1
-; X32-NEXT: [[TMP12:%.*]] = xor i32 [[TMP10]], [[TMP11]]
-; X32-NEXT: [[TMP13:%.*]] = or i32 [[TMP5]], [[TMP12]]
-; X32-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
-; X32-NEXT: [[TMP15:%.*]] = zext i1 [[TMP14]] to i32
-; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP15]], 0
-; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT: ret i32 [[CONV]]
-;
; X64-LABEL: @cmp_eq8(
; X64-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i64*
; X64-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i64*
@@ -1023,12 +834,6 @@ 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: [[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]]
-;
; X64_1LD-LABEL: @cmp_eq9(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -1080,12 +885,6 @@ 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: [[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]]
-;
; X64_1LD-LABEL: @cmp_eq10(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -1141,12 +940,6 @@ 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: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 11)
-; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
-; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT: ret i32 [[CONV]]
-;
; X64_1LD-LABEL: @cmp_eq11(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -1200,12 +993,6 @@ 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: [[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]]
-;
; X64_1LD-LABEL: @cmp_eq12(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -1261,12 +1048,6 @@ 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: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 13)
-; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
-; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT: ret i32 [[CONV]]
-;
; X64_1LD-LABEL: @cmp_eq13(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -1320,12 +1101,6 @@ 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: [[CALL:%.*]] = tail call i32 @memcmp(i8* [[X:%.*]], i8* [[Y:%.*]], i64 14)
-; X32-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], 0
-; X32-NEXT: [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; X32-NEXT: ret i32 [[CONV]]
-;
; X64_1LD-LABEL: @cmp_eq14(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -1379,12 +1154,6 @@ 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_1LD-LABEL: @cmp_eq15(
; X64_1LD-NEXT: br label [[LOADBB:%.*]]
; X64_1LD: res_block:
@@ -1438,12 +1207,6 @@ 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: [[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]]
-;
; X64-LABEL: @cmp_eq16(
; X64-NEXT: [[TMP1:%.*]] = bitcast i8* [[X:%.*]] to i128*
; X64-NEXT: [[TMP2:%.*]] = bitcast i8* [[Y:%.*]] to i128*
More information about the llvm-commits
mailing list