diff --git a/llvm/lib/CodeGen/PrologEpilogInserter.cpp b/llvm/lib/CodeGen/PrologEpilogInserter.cpp
--- a/llvm/lib/CodeGen/PrologEpilogInserter.cpp
+++ b/llvm/lib/CodeGen/PrologEpilogInserter.cpp
@@ -1193,7 +1193,8 @@
 
   // Mark all used registers.
   BitVector UsedRegs(TRI.getNumRegs());
-  if (OnlyUsed)
+  if (OnlyUsed) {
+    const MachineRegisterInfo &MRI = MF.getRegInfo();
     for (const MachineBasicBlock &MBB : MF)
       for (const MachineInstr &MI : MBB)
         for (const MachineOperand &MO : MI.operands()) {
@@ -1202,9 +1203,10 @@
 
           MCRegister Reg = MO.getReg();
           if (AllocatableSet[Reg] && !MO.isImplicit() &&
-              (MO.isDef() || MO.isUse()))
+              MRI.isPhysRegModified(Reg))
             UsedRegs.set(Reg);
         }
+  }
 
   // Get a list of registers that are used.
   BitVector LiveIns(TRI.getNumRegs());
diff --git a/llvm/test/CodeGen/AArch64/zero-call-used-regs.ll b/llvm/test/CodeGen/AArch64/zero-call-used-regs.ll
--- a/llvm/test/CodeGen/AArch64/zero-call-used-regs.ll
+++ b/llvm/test/CodeGen/AArch64/zero-call-used-regs.ll
@@ -22,8 +22,6 @@
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul w8, w1, w0
 ; CHECK-NEXT:    orr w0, w8, w2
-; CHECK-NEXT:    mov x1, #0
-; CHECK-NEXT:    mov x2, #0
 ; CHECK-NEXT:    ret
 
 entry:
@@ -37,8 +35,6 @@
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul w8, w1, w0
 ; CHECK-NEXT:    orr w0, w8, w2
-; CHECK-NEXT:    mov x1, #0
-; CHECK-NEXT:    mov x2, #0
 ; CHECK-NEXT:    mov x8, #0
 ; CHECK-NEXT:    ret
 
@@ -53,8 +49,6 @@
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul w8, w1, w0
 ; CHECK-NEXT:    orr w0, w8, w2
-; CHECK-NEXT:    mov x1, #0
-; CHECK-NEXT:    mov x2, #0
 ; CHECK-NEXT:    ret
 
 entry:
@@ -68,8 +62,6 @@
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mul w8, w1, w0
 ; CHECK-NEXT:    orr w0, w8, w2
-; CHECK-NEXT:    mov x1, #0
-; CHECK-NEXT:    mov x2, #0
 ; CHECK-NEXT:    mov x8, #0
 ; CHECK-NEXT:    ret
 
diff --git a/llvm/test/CodeGen/X86/zero-call-used-regs.ll b/llvm/test/CodeGen/X86/zero-call-used-regs.ll
--- a/llvm/test/CodeGen/X86/zero-call-used-regs.ll
+++ b/llvm/test/CodeGen/X86/zero-call-used-regs.ll
@@ -4,83 +4,682 @@
 
 @result = dso_local global i32 0, align 4
 
-define dso_local i32 @skip(i32 returned %x) local_unnamed_addr #0 "zero-call-used-regs"="skip" {
+declare dso_local i64 @g(ptr noundef, ptr noundef, i64 noundef) local_unnamed_addr
+
+define dso_local i32 @skip(ptr noundef %arg, i64 noundef %arg1) local_unnamed_addr #0 "zero-call-used-regs"="skip" {
 ; I386-LABEL: skip:
-; I386:       # %bb.0: # %entry
-; I386-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; I386:       # %bb.0: # %bb
+; I386-NEXT:    pushl %ebp
+; I386-NEXT:    pushl %ebx
+; I386-NEXT:    pushl %edi
+; I386-NEXT:    pushl %esi
+; I386-NEXT:    subl $12, %esp
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; I386-NEXT:    movb $0, {{[0-9]+}}(%esp)
+; I386-NEXT:    movl %edi, %eax
+; I386-NEXT:    orl %esi, %eax
+; I386-NEXT:    je .LBB0_1
+; I386-NEXT:  # %bb.2: # %bb3.preheader
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %ebp
+; I386-NEXT:    xorl %ebx, %ebx
+; I386-NEXT:    .p2align 4, 0x90
+; I386-NEXT:  .LBB0_3: # %bb3
+; I386-NEXT:    # =>This Inner Loop Header: Depth=1
+; I386-NEXT:    pushl $0
+; I386-NEXT:    pushl $1
+; I386-NEXT:    pushl %ebp
+; I386-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; I386-NEXT:    pushl %eax
+; I386-NEXT:    calll g
+; I386-NEXT:    addl $16, %esp
+; I386-NEXT:    orl %eax, %edx
+; I386-NEXT:    sete %al
+; I386-NEXT:    jne .LBB0_5
+; I386-NEXT:  # %bb.4: # %bb9
+; I386-NEXT:    # in Loop: Header=BB0_3 Depth=1
+; I386-NEXT:    incl %ebp
+; I386-NEXT:    addb {{[0-9]+}}(%esp), %bl
+; I386-NEXT:    addl $-1, %edi
+; I386-NEXT:    adcl $-1, %esi
+; I386-NEXT:    movl %edi, %ecx
+; I386-NEXT:    orl %esi, %ecx
+; I386-NEXT:    jne .LBB0_3
+; I386-NEXT:    jmp .LBB0_5
+; I386-NEXT:  .LBB0_1:
+; I386-NEXT:    movb $1, %al
+; I386-NEXT:    xorl %ebx, %ebx
+; I386-NEXT:  .LBB0_5: # %bb15
+; I386-NEXT:    testb %bl, %bl
+; I386-NEXT:    sete %cl
+; I386-NEXT:    andb %al, %cl
+; I386-NEXT:    movzbl %cl, %eax
+; I386-NEXT:    addl $12, %esp
+; I386-NEXT:    popl %esi
+; I386-NEXT:    popl %edi
+; I386-NEXT:    popl %ebx
+; I386-NEXT:    popl %ebp
 ; I386-NEXT:    retl
 ;
 ; X86-64-LABEL: skip:
-; X86-64:       # %bb.0: # %entry
-; X86-64-NEXT:    movl %edi, %eax
+; X86-64:       # %bb.0: # %bb
+; X86-64-NEXT:    pushq %rbp
+; X86-64-NEXT:    pushq %r15
+; X86-64-NEXT:    pushq %r14
+; X86-64-NEXT:    pushq %rbx
+; X86-64-NEXT:    pushq %rax
+; X86-64-NEXT:    movb $0, {{[0-9]+}}(%rsp)
+; X86-64-NEXT:    testq %rsi, %rsi
+; X86-64-NEXT:    je .LBB0_1
+; X86-64-NEXT:  # %bb.2: # %bb3.preheader
+; X86-64-NEXT:    movq %rsi, %r14
+; X86-64-NEXT:    movq %rdi, %rbx
+; X86-64-NEXT:    xorl %ebp, %ebp
+; X86-64-NEXT:    leaq {{[0-9]+}}(%rsp), %r15
+; X86-64-NEXT:    .p2align 4, 0x90
+; X86-64-NEXT:  .LBB0_3: # %bb3
+; X86-64-NEXT:    # =>This Inner Loop Header: Depth=1
+; X86-64-NEXT:    movl $1, %edx
+; X86-64-NEXT:    movq %r15, %rdi
+; X86-64-NEXT:    movq %rbx, %rsi
+; X86-64-NEXT:    callq g
+; X86-64-NEXT:    testq %rax, %rax
+; X86-64-NEXT:    sete %al
+; X86-64-NEXT:    jne .LBB0_5
+; X86-64-NEXT:  # %bb.4: # %bb9
+; X86-64-NEXT:    # in Loop: Header=BB0_3 Depth=1
+; X86-64-NEXT:    incq %rbx
+; X86-64-NEXT:    addb {{[0-9]+}}(%rsp), %bpl
+; X86-64-NEXT:    decq %r14
+; X86-64-NEXT:    jne .LBB0_3
+; X86-64-NEXT:    jmp .LBB0_5
+; X86-64-NEXT:  .LBB0_1:
+; X86-64-NEXT:    movb $1, %al
+; X86-64-NEXT:    xorl %ebp, %ebp
+; X86-64-NEXT:  .LBB0_5: # %bb15
+; X86-64-NEXT:    testb %bpl, %bpl
+; X86-64-NEXT:    sete %cl
+; X86-64-NEXT:    andb %al, %cl
+; X86-64-NEXT:    movzbl %cl, %eax
+; X86-64-NEXT:    addq $8, %rsp
+; X86-64-NEXT:    popq %rbx
+; X86-64-NEXT:    popq %r14
+; X86-64-NEXT:    popq %r15
+; X86-64-NEXT:    popq %rbp
 ; X86-64-NEXT:    retq
+bb:
+  %i = alloca i8, align 1
+  store i8 0, ptr %i, align 1
+  %i2 = icmp eq i64 %arg1, 0
+  br i1 %i2, label %bb15, label %bb3
 
-entry:
-  ret i32 %x
+bb3:                                              ; preds = %bb9, %bb
+  %i4 = phi ptr [ %i10, %bb9 ], [ %arg, %bb ]
+  %i5 = phi i64 [ %i13, %bb9 ], [ %arg1, %bb ]
+  %i6 = phi i8 [ %i12, %bb9 ], [ 0, %bb ]
+  %i7 = call i64 @g(ptr noundef nonnull %i, ptr noundef %i4, i64 noundef 1)
+  %i8 = icmp eq i64 %i7, 0
+  br i1 %i8, label %bb9, label %bb15
+
+bb9:                                              ; preds = %bb3
+  %i10 = getelementptr i8, ptr %i4, i64 1
+  %i11 = load i8, ptr %i, align 1
+  %i12 = add i8 %i11, %i6
+  %i13 = add i64 %i5, -1
+  %i14 = icmp eq i64 %i13, 0
+  br i1 %i14, label %bb15, label %bb3
+
+bb15:                                             ; preds = %bb9, %bb3, %bb
+  %i16 = phi i8 [ 0, %bb ], [ %i12, %bb9 ], [ %i6, %bb3 ]
+  %i17 = phi i1 [ true, %bb ], [ %i8, %bb9 ], [ %i8, %bb3 ]
+  %i18 = icmp eq i8 %i16, 0
+  %i19 = select i1 %i17, i1 %i18, i1 false
+  %i20 = zext i1 %i19 to i32
+  ret i32 %i20
 }
 
-define dso_local i32 @used_gpr_arg(i32 returned %x) local_unnamed_addr #0 "zero-call-used-regs"="used-gpr-arg" {
+define dso_local i32 @used_gpr_arg(ptr noundef %arg, i64 noundef %arg1) local_unnamed_addr #0 "zero-call-used-regs"="used-gpr-arg" {
 ; I386-LABEL: used_gpr_arg:
-; I386:       # %bb.0: # %entry
-; I386-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; I386:       # %bb.0: # %bb
+; I386-NEXT:    pushl %ebp
+; I386-NEXT:    pushl %ebx
+; I386-NEXT:    pushl %edi
+; I386-NEXT:    pushl %esi
+; I386-NEXT:    subl $12, %esp
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; I386-NEXT:    movb $0, {{[0-9]+}}(%esp)
+; I386-NEXT:    movl %edi, %eax
+; I386-NEXT:    orl %esi, %eax
+; I386-NEXT:    je .LBB1_1
+; I386-NEXT:  # %bb.2: # %bb3.preheader
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %ebp
+; I386-NEXT:    xorl %ebx, %ebx
+; I386-NEXT:    .p2align 4, 0x90
+; I386-NEXT:  .LBB1_3: # %bb3
+; I386-NEXT:    # =>This Inner Loop Header: Depth=1
+; I386-NEXT:    pushl $0
+; I386-NEXT:    pushl $1
+; I386-NEXT:    pushl %ebp
+; I386-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; I386-NEXT:    pushl %eax
+; I386-NEXT:    calll g
+; I386-NEXT:    addl $16, %esp
+; I386-NEXT:    orl %eax, %edx
+; I386-NEXT:    sete %al
+; I386-NEXT:    jne .LBB1_5
+; I386-NEXT:  # %bb.4: # %bb9
+; I386-NEXT:    # in Loop: Header=BB1_3 Depth=1
+; I386-NEXT:    incl %ebp
+; I386-NEXT:    addb {{[0-9]+}}(%esp), %bl
+; I386-NEXT:    addl $-1, %edi
+; I386-NEXT:    adcl $-1, %esi
+; I386-NEXT:    movl %edi, %ecx
+; I386-NEXT:    orl %esi, %ecx
+; I386-NEXT:    jne .LBB1_3
+; I386-NEXT:    jmp .LBB1_5
+; I386-NEXT:  .LBB1_1:
+; I386-NEXT:    movb $1, %al
+; I386-NEXT:    xorl %ebx, %ebx
+; I386-NEXT:  .LBB1_5: # %bb15
+; I386-NEXT:    testb %bl, %bl
+; I386-NEXT:    sete %cl
+; I386-NEXT:    andb %al, %cl
+; I386-NEXT:    movzbl %cl, %eax
+; I386-NEXT:    addl $12, %esp
+; I386-NEXT:    popl %esi
+; I386-NEXT:    popl %edi
+; I386-NEXT:    popl %ebx
+; I386-NEXT:    popl %ebp
 ; I386-NEXT:    retl
 ;
 ; X86-64-LABEL: used_gpr_arg:
-; X86-64:       # %bb.0: # %entry
-; X86-64-NEXT:    movl %edi, %eax
+; X86-64:       # %bb.0: # %bb
+; X86-64-NEXT:    pushq %rbp
+; X86-64-NEXT:    pushq %r15
+; X86-64-NEXT:    pushq %r14
+; X86-64-NEXT:    pushq %rbx
+; X86-64-NEXT:    pushq %rax
+; X86-64-NEXT:    movb $0, {{[0-9]+}}(%rsp)
+; X86-64-NEXT:    testq %rsi, %rsi
+; X86-64-NEXT:    je .LBB1_1
+; X86-64-NEXT:  # %bb.2: # %bb3.preheader
+; X86-64-NEXT:    movq %rsi, %r14
+; X86-64-NEXT:    movq %rdi, %rbx
+; X86-64-NEXT:    xorl %ebp, %ebp
+; X86-64-NEXT:    leaq {{[0-9]+}}(%rsp), %r15
+; X86-64-NEXT:    .p2align 4, 0x90
+; X86-64-NEXT:  .LBB1_3: # %bb3
+; X86-64-NEXT:    # =>This Inner Loop Header: Depth=1
+; X86-64-NEXT:    movl $1, %edx
+; X86-64-NEXT:    movq %r15, %rdi
+; X86-64-NEXT:    movq %rbx, %rsi
+; X86-64-NEXT:    callq g
+; X86-64-NEXT:    testq %rax, %rax
+; X86-64-NEXT:    sete %al
+; X86-64-NEXT:    jne .LBB1_5
+; X86-64-NEXT:  # %bb.4: # %bb9
+; X86-64-NEXT:    # in Loop: Header=BB1_3 Depth=1
+; X86-64-NEXT:    incq %rbx
+; X86-64-NEXT:    addb {{[0-9]+}}(%rsp), %bpl
+; X86-64-NEXT:    decq %r14
+; X86-64-NEXT:    jne .LBB1_3
+; X86-64-NEXT:    jmp .LBB1_5
+; X86-64-NEXT:  .LBB1_1:
+; X86-64-NEXT:    movb $1, %al
+; X86-64-NEXT:    xorl %ebp, %ebp
+; X86-64-NEXT:  .LBB1_5: # %bb15
+; X86-64-NEXT:    testb %bpl, %bpl
+; X86-64-NEXT:    sete %cl
+; X86-64-NEXT:    andb %al, %cl
+; X86-64-NEXT:    movzbl %cl, %eax
+; X86-64-NEXT:    addq $8, %rsp
+; X86-64-NEXT:    popq %rbx
+; X86-64-NEXT:    popq %r14
+; X86-64-NEXT:    popq %r15
+; X86-64-NEXT:    popq %rbp
 ; X86-64-NEXT:    xorl %edi, %edi
+; X86-64-NEXT:    xorl %esi, %esi
 ; X86-64-NEXT:    retq
+bb:
+  %i = alloca i8, align 1
+  store i8 0, ptr %i, align 1
+  %i2 = icmp eq i64 %arg1, 0
+  br i1 %i2, label %bb15, label %bb3
 
-entry:
-  ret i32 %x
+bb3:                                              ; preds = %bb9, %bb
+  %i4 = phi ptr [ %i10, %bb9 ], [ %arg, %bb ]
+  %i5 = phi i64 [ %i13, %bb9 ], [ %arg1, %bb ]
+  %i6 = phi i8 [ %i12, %bb9 ], [ 0, %bb ]
+  %i7 = call i64 @g(ptr noundef nonnull %i, ptr noundef %i4, i64 noundef 1)
+  %i8 = icmp eq i64 %i7, 0
+  br i1 %i8, label %bb9, label %bb15
+
+bb9:                                              ; preds = %bb3
+  %i10 = getelementptr i8, ptr %i4, i64 1
+  %i11 = load i8, ptr %i, align 1
+  %i12 = add i8 %i11, %i6
+  %i13 = add i64 %i5, -1
+  %i14 = icmp eq i64 %i13, 0
+  br i1 %i14, label %bb15, label %bb3
+
+bb15:                                             ; preds = %bb9, %bb3, %bb
+  %i16 = phi i8 [ 0, %bb ], [ %i12, %bb9 ], [ %i6, %bb3 ]
+  %i17 = phi i1 [ true, %bb ], [ %i8, %bb9 ], [ %i8, %bb3 ]
+  %i18 = icmp eq i8 %i16, 0
+  %i19 = select i1 %i17, i1 %i18, i1 false
+  %i20 = zext i1 %i19 to i32
+  ret i32 %i20
 }
 
-define dso_local i32 @used_gpr(i32 returned %x) local_unnamed_addr #0 "zero-call-used-regs"="used-gpr" {
+define dso_local i32 @used_gpr(ptr noundef %arg, i64 noundef %arg1) local_unnamed_addr #0 "zero-call-used-regs"="used-gpr" {
 ; I386-LABEL: used_gpr:
-; I386:       # %bb.0: # %entry
-; I386-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; I386:       # %bb.0: # %bb
+; I386-NEXT:    pushl %ebp
+; I386-NEXT:    pushl %ebx
+; I386-NEXT:    pushl %edi
+; I386-NEXT:    pushl %esi
+; I386-NEXT:    subl $12, %esp
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; I386-NEXT:    movb $0, {{[0-9]+}}(%esp)
+; I386-NEXT:    movl %edi, %eax
+; I386-NEXT:    orl %esi, %eax
+; I386-NEXT:    je .LBB2_1
+; I386-NEXT:  # %bb.2: # %bb3.preheader
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %ebp
+; I386-NEXT:    xorl %ebx, %ebx
+; I386-NEXT:    .p2align 4, 0x90
+; I386-NEXT:  .LBB2_3: # %bb3
+; I386-NEXT:    # =>This Inner Loop Header: Depth=1
+; I386-NEXT:    pushl $0
+; I386-NEXT:    pushl $1
+; I386-NEXT:    pushl %ebp
+; I386-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; I386-NEXT:    pushl %eax
+; I386-NEXT:    calll g
+; I386-NEXT:    addl $16, %esp
+; I386-NEXT:    orl %eax, %edx
+; I386-NEXT:    sete %al
+; I386-NEXT:    jne .LBB2_5
+; I386-NEXT:  # %bb.4: # %bb9
+; I386-NEXT:    # in Loop: Header=BB2_3 Depth=1
+; I386-NEXT:    incl %ebp
+; I386-NEXT:    addb {{[0-9]+}}(%esp), %bl
+; I386-NEXT:    addl $-1, %edi
+; I386-NEXT:    adcl $-1, %esi
+; I386-NEXT:    movl %edi, %ecx
+; I386-NEXT:    orl %esi, %ecx
+; I386-NEXT:    jne .LBB2_3
+; I386-NEXT:    jmp .LBB2_5
+; I386-NEXT:  .LBB2_1:
+; I386-NEXT:    movb $1, %al
+; I386-NEXT:    xorl %ebx, %ebx
+; I386-NEXT:  .LBB2_5: # %bb15
+; I386-NEXT:    testb %bl, %bl
+; I386-NEXT:    sete %cl
+; I386-NEXT:    andb %al, %cl
+; I386-NEXT:    movzbl %cl, %eax
+; I386-NEXT:    addl $12, %esp
+; I386-NEXT:    popl %esi
+; I386-NEXT:    popl %edi
+; I386-NEXT:    popl %ebx
+; I386-NEXT:    popl %ebp
+; I386-NEXT:    xorl %ecx, %ecx
+; I386-NEXT:    xorl %edx, %edx
 ; I386-NEXT:    retl
 ;
 ; X86-64-LABEL: used_gpr:
-; X86-64:       # %bb.0: # %entry
-; X86-64-NEXT:    movl %edi, %eax
+; X86-64:       # %bb.0: # %bb
+; X86-64-NEXT:    pushq %rbp
+; X86-64-NEXT:    pushq %r15
+; X86-64-NEXT:    pushq %r14
+; X86-64-NEXT:    pushq %rbx
+; X86-64-NEXT:    pushq %rax
+; X86-64-NEXT:    movb $0, {{[0-9]+}}(%rsp)
+; X86-64-NEXT:    testq %rsi, %rsi
+; X86-64-NEXT:    je .LBB2_1
+; X86-64-NEXT:  # %bb.2: # %bb3.preheader
+; X86-64-NEXT:    movq %rsi, %r14
+; X86-64-NEXT:    movq %rdi, %rbx
+; X86-64-NEXT:    xorl %ebp, %ebp
+; X86-64-NEXT:    leaq {{[0-9]+}}(%rsp), %r15
+; X86-64-NEXT:    .p2align 4, 0x90
+; X86-64-NEXT:  .LBB2_3: # %bb3
+; X86-64-NEXT:    # =>This Inner Loop Header: Depth=1
+; X86-64-NEXT:    movl $1, %edx
+; X86-64-NEXT:    movq %r15, %rdi
+; X86-64-NEXT:    movq %rbx, %rsi
+; X86-64-NEXT:    callq g
+; X86-64-NEXT:    testq %rax, %rax
+; X86-64-NEXT:    sete %al
+; X86-64-NEXT:    jne .LBB2_5
+; X86-64-NEXT:  # %bb.4: # %bb9
+; X86-64-NEXT:    # in Loop: Header=BB2_3 Depth=1
+; X86-64-NEXT:    incq %rbx
+; X86-64-NEXT:    addb {{[0-9]+}}(%rsp), %bpl
+; X86-64-NEXT:    decq %r14
+; X86-64-NEXT:    jne .LBB2_3
+; X86-64-NEXT:    jmp .LBB2_5
+; X86-64-NEXT:  .LBB2_1:
+; X86-64-NEXT:    movb $1, %al
+; X86-64-NEXT:    xorl %ebp, %ebp
+; X86-64-NEXT:  .LBB2_5: # %bb15
+; X86-64-NEXT:    testb %bpl, %bpl
+; X86-64-NEXT:    sete %cl
+; X86-64-NEXT:    andb %al, %cl
+; X86-64-NEXT:    movzbl %cl, %eax
+; X86-64-NEXT:    addq $8, %rsp
+; X86-64-NEXT:    popq %rbx
+; X86-64-NEXT:    popq %r14
+; X86-64-NEXT:    popq %r15
+; X86-64-NEXT:    popq %rbp
+; X86-64-NEXT:    xorl %ecx, %ecx
 ; X86-64-NEXT:    xorl %edi, %edi
+; X86-64-NEXT:    xorl %edx, %edx
+; X86-64-NEXT:    xorl %esi, %esi
 ; X86-64-NEXT:    retq
+bb:
+  %i = alloca i8, align 1
+  store i8 0, ptr %i, align 1
+  %i2 = icmp eq i64 %arg1, 0
+  br i1 %i2, label %bb15, label %bb3
 
-entry:
-  ret i32 %x
+bb3:                                              ; preds = %bb9, %bb
+  %i4 = phi ptr [ %i10, %bb9 ], [ %arg, %bb ]
+  %i5 = phi i64 [ %i13, %bb9 ], [ %arg1, %bb ]
+  %i6 = phi i8 [ %i12, %bb9 ], [ 0, %bb ]
+  %i7 = call i64 @g(ptr noundef nonnull %i, ptr noundef %i4, i64 noundef 1)
+  %i8 = icmp eq i64 %i7, 0
+  br i1 %i8, label %bb9, label %bb15
+
+bb9:                                              ; preds = %bb3
+  %i10 = getelementptr i8, ptr %i4, i64 1
+  %i11 = load i8, ptr %i, align 1
+  %i12 = add i8 %i11, %i6
+  %i13 = add i64 %i5, -1
+  %i14 = icmp eq i64 %i13, 0
+  br i1 %i14, label %bb15, label %bb3
+
+bb15:                                             ; preds = %bb9, %bb3, %bb
+  %i16 = phi i8 [ 0, %bb ], [ %i12, %bb9 ], [ %i6, %bb3 ]
+  %i17 = phi i1 [ true, %bb ], [ %i8, %bb9 ], [ %i8, %bb3 ]
+  %i18 = icmp eq i8 %i16, 0
+  %i19 = select i1 %i17, i1 %i18, i1 false
+  %i20 = zext i1 %i19 to i32
+  ret i32 %i20
 }
 
-define dso_local i32 @used_arg(i32 returned %x) local_unnamed_addr #0 "zero-call-used-regs"="used-arg" {
+define dso_local i32 @used_arg(ptr noundef %arg, i64 noundef %arg1) local_unnamed_addr #0 "zero-call-used-regs"="used-arg" {
 ; I386-LABEL: used_arg:
-; I386:       # %bb.0: # %entry
-; I386-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; I386:       # %bb.0: # %bb
+; I386-NEXT:    pushl %ebp
+; I386-NEXT:    pushl %ebx
+; I386-NEXT:    pushl %edi
+; I386-NEXT:    pushl %esi
+; I386-NEXT:    subl $12, %esp
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; I386-NEXT:    movb $0, {{[0-9]+}}(%esp)
+; I386-NEXT:    movl %edi, %eax
+; I386-NEXT:    orl %esi, %eax
+; I386-NEXT:    je .LBB3_1
+; I386-NEXT:  # %bb.2: # %bb3.preheader
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %ebp
+; I386-NEXT:    xorl %ebx, %ebx
+; I386-NEXT:    .p2align 4, 0x90
+; I386-NEXT:  .LBB3_3: # %bb3
+; I386-NEXT:    # =>This Inner Loop Header: Depth=1
+; I386-NEXT:    pushl $0
+; I386-NEXT:    pushl $1
+; I386-NEXT:    pushl %ebp
+; I386-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; I386-NEXT:    pushl %eax
+; I386-NEXT:    calll g
+; I386-NEXT:    addl $16, %esp
+; I386-NEXT:    orl %eax, %edx
+; I386-NEXT:    sete %al
+; I386-NEXT:    jne .LBB3_5
+; I386-NEXT:  # %bb.4: # %bb9
+; I386-NEXT:    # in Loop: Header=BB3_3 Depth=1
+; I386-NEXT:    incl %ebp
+; I386-NEXT:    addb {{[0-9]+}}(%esp), %bl
+; I386-NEXT:    addl $-1, %edi
+; I386-NEXT:    adcl $-1, %esi
+; I386-NEXT:    movl %edi, %ecx
+; I386-NEXT:    orl %esi, %ecx
+; I386-NEXT:    jne .LBB3_3
+; I386-NEXT:    jmp .LBB3_5
+; I386-NEXT:  .LBB3_1:
+; I386-NEXT:    movb $1, %al
+; I386-NEXT:    xorl %ebx, %ebx
+; I386-NEXT:  .LBB3_5: # %bb15
+; I386-NEXT:    testb %bl, %bl
+; I386-NEXT:    sete %cl
+; I386-NEXT:    andb %al, %cl
+; I386-NEXT:    movzbl %cl, %eax
+; I386-NEXT:    addl $12, %esp
+; I386-NEXT:    popl %esi
+; I386-NEXT:    popl %edi
+; I386-NEXT:    popl %ebx
+; I386-NEXT:    popl %ebp
 ; I386-NEXT:    retl
 ;
 ; X86-64-LABEL: used_arg:
-; X86-64:       # %bb.0: # %entry
-; X86-64-NEXT:    movl %edi, %eax
+; X86-64:       # %bb.0: # %bb
+; X86-64-NEXT:    pushq %rbp
+; X86-64-NEXT:    pushq %r15
+; X86-64-NEXT:    pushq %r14
+; X86-64-NEXT:    pushq %rbx
+; X86-64-NEXT:    pushq %rax
+; X86-64-NEXT:    movb $0, {{[0-9]+}}(%rsp)
+; X86-64-NEXT:    testq %rsi, %rsi
+; X86-64-NEXT:    je .LBB3_1
+; X86-64-NEXT:  # %bb.2: # %bb3.preheader
+; X86-64-NEXT:    movq %rsi, %r14
+; X86-64-NEXT:    movq %rdi, %rbx
+; X86-64-NEXT:    xorl %ebp, %ebp
+; X86-64-NEXT:    leaq {{[0-9]+}}(%rsp), %r15
+; X86-64-NEXT:    .p2align 4, 0x90
+; X86-64-NEXT:  .LBB3_3: # %bb3
+; X86-64-NEXT:    # =>This Inner Loop Header: Depth=1
+; X86-64-NEXT:    movl $1, %edx
+; X86-64-NEXT:    movq %r15, %rdi
+; X86-64-NEXT:    movq %rbx, %rsi
+; X86-64-NEXT:    callq g
+; X86-64-NEXT:    testq %rax, %rax
+; X86-64-NEXT:    sete %al
+; X86-64-NEXT:    jne .LBB3_5
+; X86-64-NEXT:  # %bb.4: # %bb9
+; X86-64-NEXT:    # in Loop: Header=BB3_3 Depth=1
+; X86-64-NEXT:    incq %rbx
+; X86-64-NEXT:    addb {{[0-9]+}}(%rsp), %bpl
+; X86-64-NEXT:    decq %r14
+; X86-64-NEXT:    jne .LBB3_3
+; X86-64-NEXT:    jmp .LBB3_5
+; X86-64-NEXT:  .LBB3_1:
+; X86-64-NEXT:    movb $1, %al
+; X86-64-NEXT:    xorl %ebp, %ebp
+; X86-64-NEXT:  .LBB3_5: # %bb15
+; X86-64-NEXT:    testb %bpl, %bpl
+; X86-64-NEXT:    sete %cl
+; X86-64-NEXT:    andb %al, %cl
+; X86-64-NEXT:    movzbl %cl, %eax
+; X86-64-NEXT:    addq $8, %rsp
+; X86-64-NEXT:    popq %rbx
+; X86-64-NEXT:    popq %r14
+; X86-64-NEXT:    popq %r15
+; X86-64-NEXT:    popq %rbp
 ; X86-64-NEXT:    xorl %edi, %edi
+; X86-64-NEXT:    xorl %esi, %esi
 ; X86-64-NEXT:    retq
+bb:
+  %i = alloca i8, align 1
+  store i8 0, ptr %i, align 1
+  %i2 = icmp eq i64 %arg1, 0
+  br i1 %i2, label %bb15, label %bb3
 
-entry:
-  ret i32 %x
+bb3:                                              ; preds = %bb9, %bb
+  %i4 = phi ptr [ %i10, %bb9 ], [ %arg, %bb ]
+  %i5 = phi i64 [ %i13, %bb9 ], [ %arg1, %bb ]
+  %i6 = phi i8 [ %i12, %bb9 ], [ 0, %bb ]
+  %i7 = call i64 @g(ptr noundef nonnull %i, ptr noundef %i4, i64 noundef 1)
+  %i8 = icmp eq i64 %i7, 0
+  br i1 %i8, label %bb9, label %bb15
+
+bb9:                                              ; preds = %bb3
+  %i10 = getelementptr i8, ptr %i4, i64 1
+  %i11 = load i8, ptr %i, align 1
+  %i12 = add i8 %i11, %i6
+  %i13 = add i64 %i5, -1
+  %i14 = icmp eq i64 %i13, 0
+  br i1 %i14, label %bb15, label %bb3
+
+bb15:                                             ; preds = %bb9, %bb3, %bb
+  %i16 = phi i8 [ 0, %bb ], [ %i12, %bb9 ], [ %i6, %bb3 ]
+  %i17 = phi i1 [ true, %bb ], [ %i8, %bb9 ], [ %i8, %bb3 ]
+  %i18 = icmp eq i8 %i16, 0
+  %i19 = select i1 %i17, i1 %i18, i1 false
+  %i20 = zext i1 %i19 to i32
+  ret i32 %i20
 }
 
-define dso_local i32 @used(i32 returned %x) local_unnamed_addr #0 "zero-call-used-regs"="used" {
+define dso_local i32 @used(ptr noundef %arg, i64 noundef %arg1) local_unnamed_addr #0 "zero-call-used-regs"="used" {
 ; I386-LABEL: used:
-; I386:       # %bb.0: # %entry
-; I386-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; I386:       # %bb.0: # %bb
+; I386-NEXT:    pushl %ebp
+; I386-NEXT:    pushl %ebx
+; I386-NEXT:    pushl %edi
+; I386-NEXT:    pushl %esi
+; I386-NEXT:    subl $12, %esp
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %edi
+; I386-NEXT:    movb $0, {{[0-9]+}}(%esp)
+; I386-NEXT:    movl %edi, %eax
+; I386-NEXT:    orl %esi, %eax
+; I386-NEXT:    je .LBB4_1
+; I386-NEXT:  # %bb.2: # %bb3.preheader
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %ebp
+; I386-NEXT:    xorl %ebx, %ebx
+; I386-NEXT:    .p2align 4, 0x90
+; I386-NEXT:  .LBB4_3: # %bb3
+; I386-NEXT:    # =>This Inner Loop Header: Depth=1
+; I386-NEXT:    pushl $0
+; I386-NEXT:    pushl $1
+; I386-NEXT:    pushl %ebp
+; I386-NEXT:    leal {{[0-9]+}}(%esp), %eax
+; I386-NEXT:    pushl %eax
+; I386-NEXT:    calll g
+; I386-NEXT:    addl $16, %esp
+; I386-NEXT:    orl %eax, %edx
+; I386-NEXT:    sete %al
+; I386-NEXT:    jne .LBB4_5
+; I386-NEXT:  # %bb.4: # %bb9
+; I386-NEXT:    # in Loop: Header=BB4_3 Depth=1
+; I386-NEXT:    incl %ebp
+; I386-NEXT:    addb {{[0-9]+}}(%esp), %bl
+; I386-NEXT:    addl $-1, %edi
+; I386-NEXT:    adcl $-1, %esi
+; I386-NEXT:    movl %edi, %ecx
+; I386-NEXT:    orl %esi, %ecx
+; I386-NEXT:    jne .LBB4_3
+; I386-NEXT:    jmp .LBB4_5
+; I386-NEXT:  .LBB4_1:
+; I386-NEXT:    movb $1, %al
+; I386-NEXT:    xorl %ebx, %ebx
+; I386-NEXT:  .LBB4_5: # %bb15
+; I386-NEXT:    testb %bl, %bl
+; I386-NEXT:    sete %cl
+; I386-NEXT:    andb %al, %cl
+; I386-NEXT:    movzbl %cl, %eax
+; I386-NEXT:    addl $12, %esp
+; I386-NEXT:    popl %esi
+; I386-NEXT:    popl %edi
+; I386-NEXT:    popl %ebx
+; I386-NEXT:    popl %ebp
+; I386-NEXT:    xorl %ecx, %ecx
+; I386-NEXT:    xorl %edx, %edx
 ; I386-NEXT:    retl
 ;
 ; X86-64-LABEL: used:
-; X86-64:       # %bb.0: # %entry
-; X86-64-NEXT:    movl %edi, %eax
+; X86-64:       # %bb.0: # %bb
+; X86-64-NEXT:    pushq %rbp
+; X86-64-NEXT:    pushq %r15
+; X86-64-NEXT:    pushq %r14
+; X86-64-NEXT:    pushq %rbx
+; X86-64-NEXT:    pushq %rax
+; X86-64-NEXT:    movb $0, {{[0-9]+}}(%rsp)
+; X86-64-NEXT:    testq %rsi, %rsi
+; X86-64-NEXT:    je .LBB4_1
+; X86-64-NEXT:  # %bb.2: # %bb3.preheader
+; X86-64-NEXT:    movq %rsi, %r14
+; X86-64-NEXT:    movq %rdi, %rbx
+; X86-64-NEXT:    xorl %ebp, %ebp
+; X86-64-NEXT:    leaq {{[0-9]+}}(%rsp), %r15
+; X86-64-NEXT:    .p2align 4, 0x90
+; X86-64-NEXT:  .LBB4_3: # %bb3
+; X86-64-NEXT:    # =>This Inner Loop Header: Depth=1
+; X86-64-NEXT:    movl $1, %edx
+; X86-64-NEXT:    movq %r15, %rdi
+; X86-64-NEXT:    movq %rbx, %rsi
+; X86-64-NEXT:    callq g
+; X86-64-NEXT:    testq %rax, %rax
+; X86-64-NEXT:    sete %al
+; X86-64-NEXT:    jne .LBB4_5
+; X86-64-NEXT:  # %bb.4: # %bb9
+; X86-64-NEXT:    # in Loop: Header=BB4_3 Depth=1
+; X86-64-NEXT:    incq %rbx
+; X86-64-NEXT:    addb {{[0-9]+}}(%rsp), %bpl
+; X86-64-NEXT:    decq %r14
+; X86-64-NEXT:    jne .LBB4_3
+; X86-64-NEXT:    jmp .LBB4_5
+; X86-64-NEXT:  .LBB4_1:
+; X86-64-NEXT:    movb $1, %al
+; X86-64-NEXT:    xorl %ebp, %ebp
+; X86-64-NEXT:  .LBB4_5: # %bb15
+; X86-64-NEXT:    testb %bpl, %bpl
+; X86-64-NEXT:    sete %cl
+; X86-64-NEXT:    andb %al, %cl
+; X86-64-NEXT:    movzbl %cl, %eax
+; X86-64-NEXT:    addq $8, %rsp
+; X86-64-NEXT:    popq %rbx
+; X86-64-NEXT:    popq %r14
+; X86-64-NEXT:    popq %r15
+; X86-64-NEXT:    popq %rbp
+; X86-64-NEXT:    xorl %ecx, %ecx
 ; X86-64-NEXT:    xorl %edi, %edi
+; X86-64-NEXT:    xorl %edx, %edx
+; X86-64-NEXT:    xorl %esi, %esi
 ; X86-64-NEXT:    retq
+bb:
+  %i = alloca i8, align 1
+  store i8 0, ptr %i, align 1
+  %i2 = icmp eq i64 %arg1, 0
+  br i1 %i2, label %bb15, label %bb3
 
-entry:
-  ret i32 %x
+bb3:                                              ; preds = %bb9, %bb
+  %i4 = phi ptr [ %i10, %bb9 ], [ %arg, %bb ]
+  %i5 = phi i64 [ %i13, %bb9 ], [ %arg1, %bb ]
+  %i6 = phi i8 [ %i12, %bb9 ], [ 0, %bb ]
+  %i7 = call i64 @g(ptr noundef nonnull %i, ptr noundef %i4, i64 noundef 1)
+  %i8 = icmp eq i64 %i7, 0
+  br i1 %i8, label %bb9, label %bb15
+
+bb9:                                              ; preds = %bb3
+  %i10 = getelementptr i8, ptr %i4, i64 1
+  %i11 = load i8, ptr %i, align 1
+  %i12 = add i8 %i11, %i6
+  %i13 = add i64 %i5, -1
+  %i14 = icmp eq i64 %i13, 0
+  br i1 %i14, label %bb15, label %bb3
+
+bb15:                                             ; preds = %bb9, %bb3, %bb
+  %i16 = phi i8 [ 0, %bb ], [ %i12, %bb9 ], [ %i6, %bb3 ]
+  %i17 = phi i1 [ true, %bb ], [ %i8, %bb9 ], [ %i8, %bb3 ]
+  %i18 = icmp eq i8 %i16, 0
+  %i19 = select i1 %i17, i1 %i18, i1 false
+  %i20 = zext i1 %i19 to i32
+  ret i32 %i20
 }
 
 define dso_local i32 @all_gpr_arg(i32 returned %x) local_unnamed_addr #0 "zero-call-used-regs"="all-gpr-arg" {
@@ -277,5 +876,5 @@
   ret i32 0
 }
 
-attributes #0 = { mustprogress nofree norecurse nosync nounwind readnone uwtable willreturn "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
-attributes #1 = { nofree norecurse nounwind uwtable "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
+attributes #0 = { mustprogress nofree norecurse nosync nounwind readnone willreturn "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
+attributes #1 = { nofree norecurse nounwind "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }