; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=-bmi2 | FileCheck %s
;
; For UMUL_LOHI lowering without BMI2, MUL/IMUL use RAX as an implicit source.
; If one operand already lives in RAX (e.g. call result), avoid shuffling it out
; and back before the multiply.

declare i32 @foo32()
declare i64 @foo64()

define i32 @mul32_no_implicit_copy(i32 %a0) nounwind {
; CHECK-LABEL: mul32_no_implicit_copy:
; CHECK:       # %bb.0:
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    movl %edi, %ebx
; CHECK-NEXT:    callq foo32@PLT
; CHECK-NEXT:    mull %ebx
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    retq
  %a1 = call i32 @foo32()
  %a2 = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 %a0, i32 %a1)
  %a3 = extractvalue { i32, i1 } %a2, 0
  ret i32 %a3
}

define i64 @mul64_no_implicit_copy(i64 %a0) nounwind {
; CHECK-LABEL: mul64_no_implicit_copy:
; CHECK:       # %bb.0:
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    movq %rdi, %rbx
; CHECK-NEXT:    callq foo64@PLT
; CHECK-NEXT:    mulq %rbx
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    retq
  %a1 = call i64 @foo64()
  %a2 = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 %a0, i64 %a1)
  %a3 = extractvalue { i64, i1 } %a2, 0
  ret i64 %a3
}

define i64 @mul64_add_hi_order_a(ptr %x, i64 %y) {
; CHECK-LABEL: mul64_add_hi_order_a:
; CHECK:       # %bb.0:
; CHECK-NEXT:    movq %rsi, %rax
; CHECK-NEXT:    imulq 8(%rdi), %rsi
; CHECK-NEXT:    mulq (%rdi)
; CHECK-NEXT:    leaq (%rdx,%rsi), %rax
; CHECK-NEXT:    retq
  %p1 = getelementptr inbounds i64, ptr %x, i64 1
  %qv = load i64, ptr %p1, align 8
  %q = mul i64 %qv, %y
  %pv = load i64, ptr %x, align 8
  %pv.zext = zext i64 %pv to i128
  %y.zext = zext i64 %y to i128
  %prod = mul nuw i128 %pv.zext, %y.zext
  %prod.hi = lshr i128 %prod, 64
  %p = trunc i128 %prod.hi to i64
  %sum = add i64 %q, %p
  ret i64 %sum
}

define i64 @mul64_add_hi_order_b(ptr %x, i64 %y) {
; CHECK-LABEL: mul64_add_hi_order_b:
; CHECK:       # %bb.0:
; CHECK-NEXT:    movq %rsi, %rax
; CHECK-NEXT:    mulq (%rdi)
; CHECK-NEXT:    imulq 8(%rdi), %rsi
; CHECK-NEXT:    leaq (%rsi,%rdx), %rax
; CHECK-NEXT:    retq
  %pv = load i64, ptr %x, align 8
  %pv.zext = zext i64 %pv to i128
  %y.zext = zext i64 %y to i128
  %prod = mul nuw i128 %pv.zext, %y.zext
  %prod.hi = lshr i128 %prod, 64
  %p = trunc i128 %prod.hi to i64
  %p1 = getelementptr inbounds i64, ptr %x, i64 1
  %qv = load i64, ptr %p1, align 8
  %q = mul i64 %qv, %y
  %sum = add i64 %q, %p
  ret i64 %sum
}