; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
; RUN: llc < %s -mtriple=x86_64-unknown-linux | FileCheck %s

; This test verifies that the experimental.vector.extract.last.active intrinsic
; doesn't cause an infinite loop during legalization when the step vector type
; needs widening (e.g., v4i8 -> v16i8 on X86).

define i32 @extract_last_active_v4i32(<4 x i32> %a, <4 x i1> %c) {
; CHECK-LABEL: extract_last_active_v4i32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[2,3,2,3]
; CHECK-NEXT:    por %xmm1, %xmm2
; CHECK-NEXT:    pslld $31, %xmm1
; CHECK-NEXT:    psrad $31, %xmm1
; CHECK-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648,2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm1, %xmm3
; CHECK-NEXT:    por %xmm0, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,3,2,3]
; CHECK-NEXT:    por %xmm4, %xmm0
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm3
; CHECK-NEXT:    pand %xmm3, %xmm1
; CHECK-NEXT:    pandn %xmm4, %xmm3
; CHECK-NEXT:    por %xmm1, %xmm3
; CHECK-NEXT:    movd %xmm3, %eax
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
; CHECK-NEXT:    movd %xmm0, %ecx
; CHECK-NEXT:    cmpl %ecx, %eax
; CHECK-NEXT:    cmoval %eax, %ecx
; CHECK-NEXT:    andl $3, %ecx
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
; CHECK-NEXT:    por %xmm2, %xmm0
; CHECK-NEXT:    movd %xmm0, %edx
; CHECK-NEXT:    andb $1, %dl
; CHECK-NEXT:    xorl %eax, %eax
; CHECK-NEXT:    cmpb $1, %dl
; CHECK-NEXT:    sbbl %eax, %eax
; CHECK-NEXT:    orl -24(%rsp,%rcx,4), %eax
; CHECK-NEXT:    retq
  %res = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> %a, <4 x i1> %c, i32 -1)
  ret i32 %res
}

define i32 @extract_last_active_v4i32_no_default(<4 x i32> %a, <4 x i1> %c) {
; CHECK-LABEL: extract_last_active_v4i32_no_default:
; CHECK:       # %bb.0:
; CHECK-NEXT:    pslld $31, %xmm1
; CHECK-NEXT:    psrad $31, %xmm1
; CHECK-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648,2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[2,3,2,3]
; CHECK-NEXT:    por %xmm3, %xmm0
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm1
; CHECK-NEXT:    pandn %xmm3, %xmm2
; CHECK-NEXT:    por %xmm1, %xmm2
; CHECK-NEXT:    movd %xmm2, %eax
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
; CHECK-NEXT:    movd %xmm0, %ecx
; CHECK-NEXT:    cmpl %ecx, %eax
; CHECK-NEXT:    cmoval %eax, %ecx
; CHECK-NEXT:    andl $3, %ecx
; CHECK-NEXT:    movl -24(%rsp,%rcx,4), %eax
; CHECK-NEXT:    retq
  %res = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> %a, <4 x i1> %c, i32 poison)
  ret i32 %res
}

; Test v2i32 - smaller vector.
define i32 @extract_last_active_v2i32(<2 x i32> %a, <2 x i1> %c) {
; CHECK-LABEL: extract_last_active_v2i32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[2,3,2,3]
; CHECK-NEXT:    por %xmm1, %xmm2
; CHECK-NEXT:    psllq $63, %xmm1
; CHECK-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    movq %xmm2, %rcx
; CHECK-NEXT:    andb $1, %cl
; CHECK-NEXT:    xorl %eax, %eax
; CHECK-NEXT:    cmpb $1, %cl
; CHECK-NEXT:    sbbl %eax, %eax
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[3,3,3,3]
; CHECK-NEXT:    psrld $31, %xmm0
; CHECK-NEXT:    movq %xmm0, %rcx
; CHECK-NEXT:    movl %ecx, %ecx
; CHECK-NEXT:    orl -24(%rsp,%rcx,4), %eax
; CHECK-NEXT:    retq
  %res = call i32 @llvm.experimental.vector.extract.last.active.v2i32(<2 x i32> %a, <2 x i1> %c, i32 -1)
  ret i32 %res
}

; Test v3i32 - non-power-of-2 element count that requires mask widening
; (v3i1 -> v4i1) via WidenVecOp_VECTOR_FIND_LAST_ACTIVE.
define i32 @extract_last_active_v3i32(<3 x i32> %a, <3 x i1> %c) {
; CHECK-LABEL: extract_last_active_v3i32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    movd %esi, %xmm1
; CHECK-NEXT:    movd %edi, %xmm2
; CHECK-NEXT:    movdqa %xmm2, %xmm4
; CHECK-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; CHECK-NEXT:    movd %edx, %xmm3
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm3[0]
; CHECK-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    pslld $31, %xmm4
; CHECK-NEXT:    psrad $31, %xmm4
; CHECK-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4
; CHECK-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648,2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm4, %xmm5
; CHECK-NEXT:    por %xmm0, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[2,3,2,3]
; CHECK-NEXT:    por %xmm6, %xmm0
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm5
; CHECK-NEXT:    pand %xmm5, %xmm4
; CHECK-NEXT:    pandn %xmm6, %xmm5
; CHECK-NEXT:    por %xmm4, %xmm5
; CHECK-NEXT:    movd %xmm5, %eax
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
; CHECK-NEXT:    movd %xmm0, %ecx
; CHECK-NEXT:    cmpl %ecx, %eax
; CHECK-NEXT:    cmoval %eax, %ecx
; CHECK-NEXT:    andl $3, %ecx
; CHECK-NEXT:    por %xmm3, %xmm2
; CHECK-NEXT:    por %xmm1, %xmm2
; CHECK-NEXT:    movd %xmm2, %edx
; CHECK-NEXT:    andb $1, %dl
; CHECK-NEXT:    xorl %eax, %eax
; CHECK-NEXT:    cmpb $1, %dl
; CHECK-NEXT:    sbbl %eax, %eax
; CHECK-NEXT:    orl -24(%rsp,%rcx,4), %eax
; CHECK-NEXT:    retq
  %res = call i32 @llvm.experimental.vector.extract.last.active.v3i32(<3 x i32> %a, <3 x i1> %c, i32 -1)
  ret i32 %res
}

; Test v8i32 - larger vector where step vector type doesn't need widening.
define i32 @extract_last_active_v8i32(<8 x i32> %a, <8 x i1> %c) {
; CHECK-LABEL: extract_last_active_v8i32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[2,3,2,3]
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    psllw $15, %xmm2
; CHECK-NEXT:    psraw $15, %xmm2
; CHECK-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
; CHECK-NEXT:    por %xmm3, %xmm0
; CHECK-NEXT:    movdqa %xmm0, %xmm1
; CHECK-NEXT:    psrld $16, %xmm1
; CHECK-NEXT:    por %xmm0, %xmm1
; CHECK-NEXT:    movd %xmm1, %ecx
; CHECK-NEXT:    andb $1, %cl
; CHECK-NEXT:    xorl %eax, %eax
; CHECK-NEXT:    cmpb $1, %cl
; CHECK-NEXT:    sbbl %eax, %eax
; CHECK-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
; CHECK-NEXT:    psubusw %xmm2, %xmm0
; CHECK-NEXT:    paddw %xmm2, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; CHECK-NEXT:    psubusw %xmm0, %xmm1
; CHECK-NEXT:    paddw %xmm0, %xmm1
; CHECK-NEXT:    pextrw $1, %xmm1, %ecx
; CHECK-NEXT:    movd %xmm1, %edx
; CHECK-NEXT:    cmpw %cx, %dx
; CHECK-NEXT:    cmoval %edx, %ecx
; CHECK-NEXT:    andl $7, %ecx
; CHECK-NEXT:    orl -40(%rsp,%rcx,4), %eax
; CHECK-NEXT:    retq
  %res = call i32 @llvm.experimental.vector.extract.last.active.v8i32(<8 x i32> %a, <8 x i1> %c, i32 -1)
  ret i32 %res
}

; Test v16i32 - even larger vector.
define i32 @extract_last_active_v16i32(<16 x i32> %a, <16 x i1> %c) {
; CHECK-LABEL: extract_last_active_v16i32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[2,3,2,3]
; CHECK-NEXT:    por %xmm4, %xmm5
; CHECK-NEXT:    psllw $7, %xmm4
; CHECK-NEXT:    pxor %xmm6, %xmm6
; CHECK-NEXT:    pcmpgtb %xmm4, %xmm6
; CHECK-NEXT:    movaps %xmm3, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    movaps %xmm2, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm6
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
; CHECK-NEXT:    pmaxub %xmm6, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; CHECK-NEXT:    pmaxub %xmm0, %xmm1
; CHECK-NEXT:    movdqa %xmm1, %xmm0
; CHECK-NEXT:    psrld $16, %xmm0
; CHECK-NEXT:    pmaxub %xmm1, %xmm0
; CHECK-NEXT:    movdqa %xmm0, %xmm1
; CHECK-NEXT:    psrlw $8, %xmm1
; CHECK-NEXT:    pmaxub %xmm0, %xmm1
; CHECK-NEXT:    movd %xmm1, %ecx
; CHECK-NEXT:    andl $15, %ecx
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
; CHECK-NEXT:    por %xmm5, %xmm0
; CHECK-NEXT:    movdqa %xmm0, %xmm1
; CHECK-NEXT:    psrld $16, %xmm1
; CHECK-NEXT:    por %xmm0, %xmm1
; CHECK-NEXT:    movdqa %xmm1, %xmm0
; CHECK-NEXT:    psrlw $8, %xmm0
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    movd %xmm0, %edx
; CHECK-NEXT:    andb $1, %dl
; CHECK-NEXT:    xorl %eax, %eax
; CHECK-NEXT:    cmpb $1, %dl
; CHECK-NEXT:    sbbl %eax, %eax
; CHECK-NEXT:    orl -72(%rsp,%rcx,4), %eax
; CHECK-NEXT:    retq
  %res = call i32 @llvm.experimental.vector.extract.last.active.v16i32(<16 x i32> %a, <16 x i1> %c, i32 -1)
  ret i32 %res
}

; Test for older x86 CPUs (pre-AVX) where the mask type legalization creates
; larger element types, requiring the step vector type adjustment.
define i32 @extract_last_active_v4i32_penryn(<4 x i32> %a, <4 x i1> %c) "target-cpu"="penryn" {
; CHECK-LABEL: extract_last_active_v4i32_penryn:
; CHECK:       # %bb.0:
; CHECK-NEXT:    pslld $31, %xmm1
; CHECK-NEXT:    psrad $31, %xmm1
; CHECK-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; CHECK-NEXT:    pmaxud %xmm0, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
; CHECK-NEXT:    pmaxud %xmm0, %xmm1
; CHECK-NEXT:    movd %xmm1, %eax
; CHECK-NEXT:    andl $3, %eax
; CHECK-NEXT:    movl -24(%rsp,%rax,4), %eax
; CHECK-NEXT:    retq
  %res = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> %a, <4 x i1> %c, i32 poison)
  ret i32 %res
}

define i8 @extract_last_active_split(<32 x i8> %data, <32 x i8> %mask, i8 %passthru) {
; CHECK-LABEL: extract_last_active_split:
; CHECK:       # %bb.0:
; CHECK-NEXT:    movdqa %xmm2, %xmm4
; CHECK-NEXT:    por %xmm3, %xmm4
; CHECK-NEXT:    pxor %xmm5, %xmm5
; CHECK-NEXT:    pcmpeqb %xmm5, %xmm3
; CHECK-NEXT:    pcmpeqb %xmm5, %xmm2
; CHECK-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
; CHECK-NEXT:    pandn %xmm0, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,3,2,3]
; CHECK-NEXT:    pmaxub %xmm2, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
; CHECK-NEXT:    pmaxub %xmm1, %xmm2
; CHECK-NEXT:    movdqa %xmm2, %xmm1
; CHECK-NEXT:    psrld $16, %xmm1
; CHECK-NEXT:    pmaxub %xmm2, %xmm1
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    psrlw $8, %xmm2
; CHECK-NEXT:    pmaxub %xmm1, %xmm2
; CHECK-NEXT:    movd %xmm2, %eax
; CHECK-NEXT:    pmovmskb %xmm3, %ecx
; CHECK-NEXT:    pandn %xmm0, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,3,2,3]
; CHECK-NEXT:    pmaxub %xmm3, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; CHECK-NEXT:    pmaxub %xmm0, %xmm1
; CHECK-NEXT:    movdqa %xmm1, %xmm0
; CHECK-NEXT:    psrld $16, %xmm0
; CHECK-NEXT:    pmaxub %xmm1, %xmm0
; CHECK-NEXT:    movdqa %xmm0, %xmm1
; CHECK-NEXT:    psrlw $8, %xmm1
; CHECK-NEXT:    pmaxub %xmm0, %xmm1
; CHECK-NEXT:    movd %xmm1, %edx
; CHECK-NEXT:    addl $16, %edx
; CHECK-NEXT:    cmpl $65535, %ecx # imm = 0xFFFF
; CHECK-NEXT:    cmoveq %rax, %rdx
; CHECK-NEXT:    andl $31, %edx
; CHECK-NEXT:    movzbl -40(%rsp,%rdx), %eax
; CHECK-NEXT:    pcmpeqb %xmm5, %xmm4
; CHECK-NEXT:    pmovmskb %xmm4, %ecx
; CHECK-NEXT:    cmpl $65535, %ecx # imm = 0xFFFF
; CHECK-NEXT:    cmovel %edi, %eax
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    retq
  %notzero = icmp ne <32 x i8> %mask, zeroinitializer
  %res = call i8 @llvm.experimental.vector.extract.last.active.v16i8(<32 x i8> %data, <32 x i1> %notzero, i8 %passthru)
  ret i8 %res
}