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
@@ -2,85 +2,628 @@
 ; RUN: llc < %s -mtriple=i386-unknown-linux-gnu | FileCheck %s --check-prefix=I386
 ; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu | FileCheck %s --check-prefix=X86-64
 
-@result = dso_local global i32 0, align 4
+%struct.paravirt_patch_template = type { %struct.pv_cpu_ops, %struct.pv_irq_ops, %struct.pv_mmu_ops, %struct.pv_lock_ops }
+%struct.pv_cpu_ops = type { ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr }
+%struct.pv_irq_ops = type { %struct.paravirt_callee_save, %struct.paravirt_callee_save, %struct.paravirt_callee_save, ptr, ptr }
+%struct.paravirt_callee_save = type { ptr }
+%struct.pv_mmu_ops = type { ptr, ptr, ptr, ptr, ptr, ptr, ptr, %struct.paravirt_callee_save, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, ptr, %struct.paravirt_callee_save, %struct.paravirt_callee_save, %struct.paravirt_callee_save, %struct.paravirt_callee_save, ptr, %struct.paravirt_callee_save, %struct.paravirt_callee_save, %struct.paravirt_callee_save, %struct.paravirt_callee_save, ptr, %struct.pv_lazy_ops, ptr }
+%struct.pv_lazy_ops = type { ptr, ptr, ptr }
+%struct.pv_lock_ops = type { ptr, %struct.paravirt_callee_save, ptr, ptr, %struct.paravirt_callee_save }
+%struct.anon.95 = type <{ i8, i32 }>
 
-define dso_local i32 @skip(i32 returned %x) local_unnamed_addr #0 "zero-call-used-regs"="skip" {
+@pv_ops = external dso_local global %struct.paravirt_patch_template, align 8
+
+declare dso_local void @g()
+
+declare dso_local void @h()
+
+define dso_local i32 @skip(i8 noundef zeroext %arg, ptr noundef writeonly %arg1, i64 noundef %arg2, i32 noundef %arg3) local_unnamed_addr #0 "zero-call-used-regs"="skip" {
 ; I386-LABEL: skip:
-; I386:       # %bb.0: # %entry
+; I386:       # %bb.0: # %bb
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; I386-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; I386-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
+; I386-NEXT:    movl pv_ops(,%edx,4), %edx
+; I386-NEXT:    testl %edx, %edx
+; I386-NEXT:    je .LBB0_1
+; I386-NEXT:  # %bb.2: # %bb12
+; I386-NEXT:    cmpl $h, %edx
+; I386-NEXT:    je .LBB0_3
+; I386-NEXT:  # %bb.4: # %bb14
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    jmp .LBB0_5
+; I386-NEXT:  .LBB0_1: # %bb7
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    movl $g, %edx
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:  .LBB0_5: # %bb19
+; I386-NEXT:    subl %ecx, %edx
+; I386-NEXT:    movb $-24, (%eax)
+; I386-NEXT:    addl $-5, %edx
+; I386-NEXT:    movl %edx, 1(%eax)
+; I386-NEXT:    movl $5, %eax
+; I386-NEXT:    retl
+; I386-NEXT:  .LBB0_3:
+; I386-NEXT:    xorl %eax, %eax
 ; I386-NEXT:    retl
 ;
 ; X86-64-LABEL: skip:
-; X86-64:       # %bb.0: # %entry
+; X86-64:       # %bb.0: # %bb
 ; X86-64-NEXT:    movl %edi, %eax
+; X86-64-NEXT:    movq pv_ops(,%rax,8), %rax
+; X86-64-NEXT:    testq %rax, %rax
+; X86-64-NEXT:    je .LBB0_1
+; X86-64-NEXT:  # %bb.2: # %bb12
+; X86-64-NEXT:    cmpq $h, %rax
+; X86-64-NEXT:    je .LBB0_3
+; X86-64-NEXT:  # %bb.4: # %bb14
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    jmp .LBB0_5
+; X86-64-NEXT:  .LBB0_1: # %bb7
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    movl $g, %eax
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:  .LBB0_5: # %bb19
+; X86-64-NEXT:    movb $-24, (%rsi)
+; X86-64-NEXT:    subl %edx, %eax
+; X86-64-NEXT:    addl $-5, %eax
+; X86-64-NEXT:    movl %eax, 1(%rsi)
+; X86-64-NEXT:    movl $5, %eax
+; X86-64-NEXT:    retq
+; X86-64-NEXT:  .LBB0_3:
+; X86-64-NEXT:    xorl %eax, %eax
 ; X86-64-NEXT:    retq
 
-entry:
-  ret i32 %x
+bb:
+  %i = zext i8 %arg to i64
+  %i4 = getelementptr ptr, ptr @pv_ops, i64 %i
+  %i5 = load ptr, ptr %i4, align 8
+  %i6 = icmp eq ptr %i5, null
+  br i1 %i6, label %bb7, label %bb12
+
+bb7:                                              ; preds = %bb
+  %i8 = inttoptr i64 %arg2 to ptr
+  %i9 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %arg1) #3
+  %i10 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %i8) #3
+  %i11 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @g) #3
+  br label %bb19
+
+bb12:                                             ; preds = %bb
+  %i13 = icmp eq ptr %i5, @h
+  br i1 %i13, label %bb29, label %bb14
+
+bb14:                                             ; preds = %bb12
+  %i15 = inttoptr i64 %arg2 to ptr
+  %i16 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %arg1) #3
+  %i17 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %i15) #3
+  %i18 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr nonnull %i5) #3
+  br label %bb19
+
+bb19:                                             ; preds = %bb14, %bb7
+  %i20 = phi ptr [ %i16, %bb14 ], [ %i9, %bb7 ]
+  %i21 = phi ptr [ %i18, %bb14 ], [ %i11, %bb7 ]
+  %i22 = phi ptr [ %i17, %bb14 ], [ %i10, %bb7 ]
+  store i8 -24, ptr %i20, align 1
+  %i23 = ptrtoint ptr %i21 to i64
+  %i24 = getelementptr i8, ptr %i22, i64 5
+  %i25 = ptrtoint ptr %i24 to i64
+  %i26 = sub i64 %i23, %i25
+  %i27 = trunc i64 %i26 to i32
+  %i28 = getelementptr inbounds %struct.anon.95, ptr %i20, i64 0, i32 1
+  store i32 %i27, ptr %i28, align 1
+  br label %bb29
+
+bb29:                                             ; preds = %bb19, %bb12
+  %i30 = phi i32 [ 0, %bb12 ], [ 5, %bb19 ]
+  ret i32 %i30
 }
 
-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(i8 noundef zeroext %arg, ptr noundef writeonly %arg1, i64 noundef %arg2, i32 noundef %arg3) local_unnamed_addr #0 "zero-call-used-regs"="used-gpr-arg" {
 ; I386-LABEL: used_gpr_arg:
-; I386:       # %bb.0: # %entry
+; I386:       # %bb.0: # %bb
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; I386-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; I386-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
+; I386-NEXT:    movl pv_ops(,%edx,4), %edx
+; I386-NEXT:    testl %edx, %edx
+; I386-NEXT:    je .LBB1_1
+; I386-NEXT:  # %bb.2: # %bb12
+; I386-NEXT:    cmpl $h, %edx
+; I386-NEXT:    je .LBB1_3
+; I386-NEXT:  # %bb.4: # %bb14
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    jmp .LBB1_5
+; I386-NEXT:  .LBB1_1: # %bb7
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    movl $g, %edx
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:  .LBB1_5: # %bb19
+; I386-NEXT:    subl %ecx, %edx
+; I386-NEXT:    movb $-24, (%eax)
+; I386-NEXT:    addl $-5, %edx
+; I386-NEXT:    movl %edx, 1(%eax)
+; I386-NEXT:    movl $5, %eax
+; I386-NEXT:    retl
+; I386-NEXT:  .LBB1_3:
+; I386-NEXT:    xorl %eax, %eax
 ; I386-NEXT:    retl
 ;
 ; X86-64-LABEL: used_gpr_arg:
-; X86-64:       # %bb.0: # %entry
+; X86-64:       # %bb.0: # %bb
 ; X86-64-NEXT:    movl %edi, %eax
-; X86-64-NEXT:    xorl %edi, %edi
+; X86-64-NEXT:    movq pv_ops(,%rax,8), %rax
+; X86-64-NEXT:    testq %rax, %rax
+; X86-64-NEXT:    je .LBB1_1
+; X86-64-NEXT:  # %bb.2: # %bb12
+; X86-64-NEXT:    cmpq $h, %rax
+; X86-64-NEXT:    je .LBB1_3
+; X86-64-NEXT:  # %bb.4: # %bb14
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    jmp .LBB1_5
+; X86-64-NEXT:  .LBB1_1: # %bb7
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    movl $g, %eax
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:  .LBB1_5: # %bb19
+; X86-64-NEXT:    movb $-24, (%rsi)
+; X86-64-NEXT:    subl %edx, %eax
+; X86-64-NEXT:    addl $-5, %eax
+; X86-64-NEXT:    movl %eax, 1(%rsi)
+; X86-64-NEXT:    movl $5, %eax
+; X86-64-NEXT:    jmp .LBB1_6
+; X86-64-NEXT:  .LBB1_3:
+; X86-64-NEXT:    xorl %eax, %eax
+; X86-64-NEXT:  .LBB1_6: # %bb29
+; X86-64-NEXT:    xorl %edx, %edx
+; X86-64-NEXT:    xorl %esi, %esi
 ; X86-64-NEXT:    retq
+bb:
+  %i = zext i8 %arg to i64
+  %i4 = getelementptr ptr, ptr @pv_ops, i64 %i
+  %i5 = load ptr, ptr %i4, align 8
+  %i6 = icmp eq ptr %i5, null
+  br i1 %i6, label %bb7, label %bb12
 
-entry:
-  ret i32 %x
+bb7:                                              ; preds = %bb
+  %i8 = inttoptr i64 %arg2 to ptr
+  %i9 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %arg1) #3
+  %i10 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %i8) #3
+  %i11 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @g) #3
+  br label %bb19
+
+bb12:                                             ; preds = %bb
+  %i13 = icmp eq ptr %i5, @h
+  br i1 %i13, label %bb29, label %bb14
+
+bb14:                                             ; preds = %bb12
+  %i15 = inttoptr i64 %arg2 to ptr
+  %i16 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %arg1) #3
+  %i17 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %i15) #3
+  %i18 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr nonnull %i5) #3
+  br label %bb19
+
+bb19:                                             ; preds = %bb14, %bb7
+  %i20 = phi ptr [ %i16, %bb14 ], [ %i9, %bb7 ]
+  %i21 = phi ptr [ %i18, %bb14 ], [ %i11, %bb7 ]
+  %i22 = phi ptr [ %i17, %bb14 ], [ %i10, %bb7 ]
+  store i8 -24, ptr %i20, align 1
+  %i23 = ptrtoint ptr %i21 to i64
+  %i24 = getelementptr i8, ptr %i22, i64 5
+  %i25 = ptrtoint ptr %i24 to i64
+  %i26 = sub i64 %i23, %i25
+  %i27 = trunc i64 %i26 to i32
+  %i28 = getelementptr inbounds %struct.anon.95, ptr %i20, i64 0, i32 1
+  store i32 %i27, ptr %i28, align 1
+  br label %bb29
+
+bb29:                                             ; preds = %bb19, %bb12
+  %i30 = phi i32 [ 0, %bb12 ], [ 5, %bb19 ]
+  ret i32 %i30
 }
 
-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(i8 noundef zeroext %arg, ptr noundef writeonly %arg1, i64 noundef %arg2, i32 noundef %arg3) local_unnamed_addr #0 "zero-call-used-regs"="used-gpr" {
 ; I386-LABEL: used_gpr:
-; I386:       # %bb.0: # %entry
+; I386:       # %bb.0: # %bb
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; I386-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; I386-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
+; I386-NEXT:    movl pv_ops(,%edx,4), %edx
+; I386-NEXT:    testl %edx, %edx
+; I386-NEXT:    je .LBB2_1
+; I386-NEXT:  # %bb.2: # %bb12
+; I386-NEXT:    cmpl $h, %edx
+; I386-NEXT:    je .LBB2_3
+; I386-NEXT:  # %bb.4: # %bb14
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    jmp .LBB2_5
+; I386-NEXT:  .LBB2_1: # %bb7
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    movl $g, %edx
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:  .LBB2_5: # %bb19
+; I386-NEXT:    subl %ecx, %edx
+; I386-NEXT:    movb $-24, (%eax)
+; I386-NEXT:    addl $-5, %edx
+; I386-NEXT:    movl %edx, 1(%eax)
+; I386-NEXT:    movl $5, %eax
+; I386-NEXT:    jmp .LBB2_6
+; I386-NEXT:  .LBB2_3:
+; I386-NEXT:    xorl %eax, %eax
+; I386-NEXT:  .LBB2_6: # %bb29
+; 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:       # %bb.0: # %bb
 ; X86-64-NEXT:    movl %edi, %eax
-; X86-64-NEXT:    xorl %edi, %edi
+; X86-64-NEXT:    movq pv_ops(,%rax,8), %rax
+; X86-64-NEXT:    testq %rax, %rax
+; X86-64-NEXT:    je .LBB2_1
+; X86-64-NEXT:  # %bb.2: # %bb12
+; X86-64-NEXT:    cmpq $h, %rax
+; X86-64-NEXT:    je .LBB2_3
+; X86-64-NEXT:  # %bb.4: # %bb14
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    jmp .LBB2_5
+; X86-64-NEXT:  .LBB2_1: # %bb7
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    movl $g, %eax
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:  .LBB2_5: # %bb19
+; X86-64-NEXT:    movb $-24, (%rsi)
+; X86-64-NEXT:    subl %edx, %eax
+; X86-64-NEXT:    addl $-5, %eax
+; X86-64-NEXT:    movl %eax, 1(%rsi)
+; X86-64-NEXT:    movl $5, %eax
+; X86-64-NEXT:    jmp .LBB2_6
+; X86-64-NEXT:  .LBB2_3:
+; X86-64-NEXT:    xorl %eax, %eax
+; X86-64-NEXT:  .LBB2_6: # %bb29
+; X86-64-NEXT:    xorl %edx, %edx
+; X86-64-NEXT:    xorl %esi, %esi
 ; X86-64-NEXT:    retq
+bb:
+  %i = zext i8 %arg to i64
+  %i4 = getelementptr ptr, ptr @pv_ops, i64 %i
+  %i5 = load ptr, ptr %i4, align 8
+  %i6 = icmp eq ptr %i5, null
+  br i1 %i6, label %bb7, label %bb12
 
-entry:
-  ret i32 %x
+bb7:                                              ; preds = %bb
+  %i8 = inttoptr i64 %arg2 to ptr
+  %i9 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %arg1) #3
+  %i10 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %i8) #3
+  %i11 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @g) #3
+  br label %bb19
+
+bb12:                                             ; preds = %bb
+  %i13 = icmp eq ptr %i5, @h
+  br i1 %i13, label %bb29, label %bb14
+
+bb14:                                             ; preds = %bb12
+  %i15 = inttoptr i64 %arg2 to ptr
+  %i16 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %arg1) #3
+  %i17 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %i15) #3
+  %i18 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr nonnull %i5) #3
+  br label %bb19
+
+bb19:                                             ; preds = %bb14, %bb7
+  %i20 = phi ptr [ %i16, %bb14 ], [ %i9, %bb7 ]
+  %i21 = phi ptr [ %i18, %bb14 ], [ %i11, %bb7 ]
+  %i22 = phi ptr [ %i17, %bb14 ], [ %i10, %bb7 ]
+  store i8 -24, ptr %i20, align 1
+  %i23 = ptrtoint ptr %i21 to i64
+  %i24 = getelementptr i8, ptr %i22, i64 5
+  %i25 = ptrtoint ptr %i24 to i64
+  %i26 = sub i64 %i23, %i25
+  %i27 = trunc i64 %i26 to i32
+  %i28 = getelementptr inbounds %struct.anon.95, ptr %i20, i64 0, i32 1
+  store i32 %i27, ptr %i28, align 1
+  br label %bb29
+
+bb29:                                             ; preds = %bb19, %bb12
+  %i30 = phi i32 [ 0, %bb12 ], [ 5, %bb19 ]
+  ret i32 %i30
 }
 
-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(i8 noundef zeroext %arg, ptr noundef writeonly %arg1, i64 noundef %arg2, i32 noundef %arg3) local_unnamed_addr #0 "zero-call-used-regs"="used-arg" {
 ; I386-LABEL: used_arg:
-; I386:       # %bb.0: # %entry
+; I386:       # %bb.0: # %bb
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; I386-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; I386-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
+; I386-NEXT:    movl pv_ops(,%edx,4), %edx
+; I386-NEXT:    testl %edx, %edx
+; I386-NEXT:    je .LBB3_1
+; I386-NEXT:  # %bb.2: # %bb12
+; I386-NEXT:    cmpl $h, %edx
+; I386-NEXT:    je .LBB3_3
+; I386-NEXT:  # %bb.4: # %bb14
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    jmp .LBB3_5
+; I386-NEXT:  .LBB3_1: # %bb7
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    movl $g, %edx
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:  .LBB3_5: # %bb19
+; I386-NEXT:    subl %ecx, %edx
+; I386-NEXT:    movb $-24, (%eax)
+; I386-NEXT:    addl $-5, %edx
+; I386-NEXT:    movl %edx, 1(%eax)
+; I386-NEXT:    movl $5, %eax
+; I386-NEXT:    retl
+; I386-NEXT:  .LBB3_3:
+; I386-NEXT:    xorl %eax, %eax
 ; I386-NEXT:    retl
 ;
 ; X86-64-LABEL: used_arg:
-; X86-64:       # %bb.0: # %entry
+; X86-64:       # %bb.0: # %bb
 ; X86-64-NEXT:    movl %edi, %eax
-; X86-64-NEXT:    xorl %edi, %edi
+; X86-64-NEXT:    movq pv_ops(,%rax,8), %rax
+; X86-64-NEXT:    testq %rax, %rax
+; X86-64-NEXT:    je .LBB3_1
+; X86-64-NEXT:  # %bb.2: # %bb12
+; X86-64-NEXT:    cmpq $h, %rax
+; X86-64-NEXT:    je .LBB3_3
+; X86-64-NEXT:  # %bb.4: # %bb14
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    jmp .LBB3_5
+; X86-64-NEXT:  .LBB3_1: # %bb7
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    movl $g, %eax
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:  .LBB3_5: # %bb19
+; X86-64-NEXT:    movb $-24, (%rsi)
+; X86-64-NEXT:    subl %edx, %eax
+; X86-64-NEXT:    addl $-5, %eax
+; X86-64-NEXT:    movl %eax, 1(%rsi)
+; X86-64-NEXT:    movl $5, %eax
+; X86-64-NEXT:    jmp .LBB3_6
+; X86-64-NEXT:  .LBB3_3:
+; X86-64-NEXT:    xorl %eax, %eax
+; X86-64-NEXT:  .LBB3_6: # %bb29
+; X86-64-NEXT:    xorl %edx, %edx
+; X86-64-NEXT:    xorl %esi, %esi
 ; X86-64-NEXT:    retq
+bb:
+  %i = zext i8 %arg to i64
+  %i4 = getelementptr ptr, ptr @pv_ops, i64 %i
+  %i5 = load ptr, ptr %i4, align 8
+  %i6 = icmp eq ptr %i5, null
+  br i1 %i6, label %bb7, label %bb12
 
-entry:
-  ret i32 %x
+bb7:                                              ; preds = %bb
+  %i8 = inttoptr i64 %arg2 to ptr
+  %i9 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %arg1) #3
+  %i10 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %i8) #3
+  %i11 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @g) #3
+  br label %bb19
+
+bb12:                                             ; preds = %bb
+  %i13 = icmp eq ptr %i5, @h
+  br i1 %i13, label %bb29, label %bb14
+
+bb14:                                             ; preds = %bb12
+  %i15 = inttoptr i64 %arg2 to ptr
+  %i16 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %arg1) #3
+  %i17 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %i15) #3
+  %i18 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr nonnull %i5) #3
+  br label %bb19
+
+bb19:                                             ; preds = %bb14, %bb7
+  %i20 = phi ptr [ %i16, %bb14 ], [ %i9, %bb7 ]
+  %i21 = phi ptr [ %i18, %bb14 ], [ %i11, %bb7 ]
+  %i22 = phi ptr [ %i17, %bb14 ], [ %i10, %bb7 ]
+  store i8 -24, ptr %i20, align 1
+  %i23 = ptrtoint ptr %i21 to i64
+  %i24 = getelementptr i8, ptr %i22, i64 5
+  %i25 = ptrtoint ptr %i24 to i64
+  %i26 = sub i64 %i23, %i25
+  %i27 = trunc i64 %i26 to i32
+  %i28 = getelementptr inbounds %struct.anon.95, ptr %i20, i64 0, i32 1
+  store i32 %i27, ptr %i28, align 1
+  br label %bb29
+
+bb29:                                             ; preds = %bb19, %bb12
+  %i30 = phi i32 [ 0, %bb12 ], [ 5, %bb19 ]
+  ret i32 %i30
 }
 
-define dso_local i32 @used(i32 returned %x) local_unnamed_addr #0 "zero-call-used-regs"="used" {
+define dso_local i32 @used(i8 noundef zeroext %arg, ptr noundef writeonly %arg1, i64 noundef %arg2, i32 noundef %arg3) local_unnamed_addr #0 "zero-call-used-regs"="used" {
 ; I386-LABEL: used:
-; I386:       # %bb.0: # %entry
+; I386:       # %bb.0: # %bb
+; I386-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; I386-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; I386-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
+; I386-NEXT:    movl pv_ops(,%edx,4), %edx
+; I386-NEXT:    testl %edx, %edx
+; I386-NEXT:    je .LBB4_1
+; I386-NEXT:  # %bb.2: # %bb12
+; I386-NEXT:    cmpl $h, %edx
+; I386-NEXT:    je .LBB4_3
+; I386-NEXT:  # %bb.4: # %bb14
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    jmp .LBB4_5
+; I386-NEXT:  .LBB4_1: # %bb7
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:    movl $g, %edx
+; I386-NEXT:    #APP
+; I386-NEXT:    #NO_APP
+; I386-NEXT:  .LBB4_5: # %bb19
+; I386-NEXT:    subl %ecx, %edx
+; I386-NEXT:    movb $-24, (%eax)
+; I386-NEXT:    addl $-5, %edx
+; I386-NEXT:    movl %edx, 1(%eax)
+; I386-NEXT:    movl $5, %eax
+; I386-NEXT:    jmp .LBB4_6
+; I386-NEXT:  .LBB4_3:
+; I386-NEXT:    xorl %eax, %eax
+; I386-NEXT:  .LBB4_6: # %bb29
+; I386-NEXT:    xorl %ecx, %ecx
+; I386-NEXT:    xorl %edx, %edx
 ; I386-NEXT:    retl
 ;
 ; X86-64-LABEL: used:
-; X86-64:       # %bb.0: # %entry
+; X86-64:       # %bb.0: # %bb
 ; X86-64-NEXT:    movl %edi, %eax
-; X86-64-NEXT:    xorl %edi, %edi
+; X86-64-NEXT:    movq pv_ops(,%rax,8), %rax
+; X86-64-NEXT:    testq %rax, %rax
+; X86-64-NEXT:    je .LBB4_1
+; X86-64-NEXT:  # %bb.2: # %bb12
+; X86-64-NEXT:    cmpq $h, %rax
+; X86-64-NEXT:    je .LBB4_3
+; X86-64-NEXT:  # %bb.4: # %bb14
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    jmp .LBB4_5
+; X86-64-NEXT:  .LBB4_1: # %bb7
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:    movl $g, %eax
+; X86-64-NEXT:    #APP
+; X86-64-NEXT:    #NO_APP
+; X86-64-NEXT:  .LBB4_5: # %bb19
+; X86-64-NEXT:    movb $-24, (%rsi)
+; X86-64-NEXT:    subl %edx, %eax
+; X86-64-NEXT:    addl $-5, %eax
+; X86-64-NEXT:    movl %eax, 1(%rsi)
+; X86-64-NEXT:    movl $5, %eax
+; X86-64-NEXT:    jmp .LBB4_6
+; X86-64-NEXT:  .LBB4_3:
+; X86-64-NEXT:    xorl %eax, %eax
+; X86-64-NEXT:  .LBB4_6: # %bb29
+; X86-64-NEXT:    xorl %edx, %edx
+; X86-64-NEXT:    xorl %esi, %esi
 ; X86-64-NEXT:    retq
+bb:
+  %i = zext i8 %arg to i64
+  %i4 = getelementptr ptr, ptr @pv_ops, i64 %i
+  %i5 = load ptr, ptr %i4, align 8
+  %i6 = icmp eq ptr %i5, null
+  br i1 %i6, label %bb7, label %bb12
 
-entry:
-  ret i32 %x
+bb7:                                              ; preds = %bb
+  %i8 = inttoptr i64 %arg2 to ptr
+  %i9 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %arg1) #3
+  %i10 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %i8) #3
+  %i11 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr nonnull @g) #3
+  br label %bb19
+
+bb12:                                             ; preds = %bb
+  %i13 = icmp eq ptr %i5, @h
+  br i1 %i13, label %bb29, label %bb14
+
+bb14:                                             ; preds = %bb12
+  %i15 = inttoptr i64 %arg2 to ptr
+  %i16 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %arg1) #3
+  %i17 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr %i15) #3
+  %i18 = tail call ptr asm "", "=r,0,~{dirflag},~{fpsr},~{flags}"(ptr nonnull %i5) #3
+  br label %bb19
+
+bb19:                                             ; preds = %bb14, %bb7
+  %i20 = phi ptr [ %i16, %bb14 ], [ %i9, %bb7 ]
+  %i21 = phi ptr [ %i18, %bb14 ], [ %i11, %bb7 ]
+  %i22 = phi ptr [ %i17, %bb14 ], [ %i10, %bb7 ]
+  store i8 -24, ptr %i20, align 1
+  %i23 = ptrtoint ptr %i21 to i64
+  %i24 = getelementptr i8, ptr %i22, i64 5
+  %i25 = ptrtoint ptr %i24 to i64
+  %i26 = sub i64 %i23, %i25
+  %i27 = trunc i64 %i26 to i32
+  %i28 = getelementptr inbounds %struct.anon.95, ptr %i20, i64 0, i32 1
+  store i32 %i27, ptr %i28, align 1
+  br label %bb29
+
+bb29:                                             ; preds = %bb19, %bb12
+  %i30 = phi i32 [ 0, %bb12 ], [ 5, %bb19 ]
+  ret i32 %i30
 }
 
 define dso_local i32 @all_gpr_arg(i32 returned %x) local_unnamed_addr #0 "zero-call-used-regs"="all-gpr-arg" {
@@ -277,5 +820,6 @@
   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" }
+attributes #2 = { fn_ret_thunk_extern noredzone nounwind null_pointer_is_valid sspstrong "fentry-call"="true" "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,+retpoline-external-thunk,+retpoline-indirect-branches,+retpoline-indirect-calls,-3dnow,-3dnowa,-aes,-avx,-avx2,-avx512bf16,-avx512bitalg,-avx512bw,-avx512cd,-avx512dq,-avx512er,-avx512f,-avx512fp16,-avx512ifma,-avx512pf,-avx512vbmi,-avx512vbmi2,-avx512vl,-avx512vnni,-avx512vp2intersect,-avx512vpopcntdq,-avxvnni,-f16c,-fma,-fma4,-gfni,-kl,-mmx,-pclmul,-sha,-sse,-sse2,-sse3,-sse4.1,-sse4.2,-sse4a,-ssse3,-vaes,-vpclmulqdq,-widekl,-x87,-xop" "tune-cpu"="generic" "warn-stack-size"="2048" }