; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc --mtriple=loongarch64 --mattr=+d --target-abi=lp64d < %s \
; RUN:   | FileCheck --check-prefix=LA64-FPELIM %s
; RUN: llc --mtriple=loongarch64 --mattr=+d --target-abi=lp64d < %s \
; RUN:     --frame-pointer=all < %s \
; RUN:   | FileCheck --check-prefix=LA64-WITHFP %s

declare void @llvm.va_start(ptr)
declare void @llvm.va_end(ptr)

declare void @notdead(ptr)

define i64 @va1(ptr %fmt, ...) {
; LA64-FPELIM-LABEL: va1:
; LA64-FPELIM:       # %bb.0:
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, -80
; LA64-FPELIM-NEXT:    .cfi_def_cfa_offset 80
; LA64-FPELIM-NEXT:    move $a0, $a1
; LA64-FPELIM-NEXT:    st.d $a7, $sp, 72
; LA64-FPELIM-NEXT:    st.d $a6, $sp, 64
; LA64-FPELIM-NEXT:    st.d $a5, $sp, 56
; LA64-FPELIM-NEXT:    st.d $a4, $sp, 48
; LA64-FPELIM-NEXT:    st.d $a3, $sp, 40
; LA64-FPELIM-NEXT:    st.d $a2, $sp, 32
; LA64-FPELIM-NEXT:    addi.d $a1, $sp, 32
; LA64-FPELIM-NEXT:    st.d $a1, $sp, 8
; LA64-FPELIM-NEXT:    st.d $a0, $sp, 24
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, 80
; LA64-FPELIM-NEXT:    ret
;
; LA64-WITHFP-LABEL: va1:
; LA64-WITHFP:       # %bb.0:
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, -96
; LA64-WITHFP-NEXT:    .cfi_def_cfa_offset 96
; LA64-WITHFP-NEXT:    st.d $ra, $sp, 24 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    st.d $fp, $sp, 16 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    .cfi_offset 1, -72
; LA64-WITHFP-NEXT:    .cfi_offset 22, -80
; LA64-WITHFP-NEXT:    addi.d $fp, $sp, 32
; LA64-WITHFP-NEXT:    .cfi_def_cfa 22, 64
; LA64-WITHFP-NEXT:    move $a0, $a1
; LA64-WITHFP-NEXT:    st.d $a7, $fp, 56
; LA64-WITHFP-NEXT:    st.d $a6, $fp, 48
; LA64-WITHFP-NEXT:    st.d $a5, $fp, 40
; LA64-WITHFP-NEXT:    st.d $a4, $fp, 32
; LA64-WITHFP-NEXT:    st.d $a3, $fp, 24
; LA64-WITHFP-NEXT:    st.d $a2, $fp, 16
; LA64-WITHFP-NEXT:    addi.d $a1, $fp, 16
; LA64-WITHFP-NEXT:    st.d $a1, $fp, -24
; LA64-WITHFP-NEXT:    st.d $a0, $fp, 8
; LA64-WITHFP-NEXT:    ld.d $fp, $sp, 16 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    ld.d $ra, $sp, 24 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, 96
; LA64-WITHFP-NEXT:    ret
  %va = alloca ptr, align 8
  call void @llvm.va_start(ptr %va)
  %argp.cur = load ptr, ptr %va, align 8
  %argp.next = getelementptr inbounds i64, ptr %argp.cur, i32 1
  store ptr %argp.next, ptr %va, align 8
  %1 = load i64, ptr %argp.cur, align 8
  call void @llvm.va_end(ptr %va)
  ret i64 %1
}

define i64 @va1_va_arg(ptr %fmt, ...) nounwind {
; LA64-FPELIM-LABEL: va1_va_arg:
; LA64-FPELIM:       # %bb.0:
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, -80
; LA64-FPELIM-NEXT:    move $a0, $a1
; LA64-FPELIM-NEXT:    st.d $a7, $sp, 72
; LA64-FPELIM-NEXT:    st.d $a6, $sp, 64
; LA64-FPELIM-NEXT:    st.d $a5, $sp, 56
; LA64-FPELIM-NEXT:    st.d $a4, $sp, 48
; LA64-FPELIM-NEXT:    st.d $a3, $sp, 40
; LA64-FPELIM-NEXT:    st.d $a2, $sp, 32
; LA64-FPELIM-NEXT:    addi.d $a1, $sp, 32
; LA64-FPELIM-NEXT:    st.d $a1, $sp, 8
; LA64-FPELIM-NEXT:    st.d $a0, $sp, 24
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, 80
; LA64-FPELIM-NEXT:    ret
;
; LA64-WITHFP-LABEL: va1_va_arg:
; LA64-WITHFP:       # %bb.0:
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, -96
; LA64-WITHFP-NEXT:    st.d $ra, $sp, 24 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    st.d $fp, $sp, 16 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    addi.d $fp, $sp, 32
; LA64-WITHFP-NEXT:    move $a0, $a1
; LA64-WITHFP-NEXT:    st.d $a7, $fp, 56
; LA64-WITHFP-NEXT:    st.d $a6, $fp, 48
; LA64-WITHFP-NEXT:    st.d $a5, $fp, 40
; LA64-WITHFP-NEXT:    st.d $a4, $fp, 32
; LA64-WITHFP-NEXT:    st.d $a3, $fp, 24
; LA64-WITHFP-NEXT:    st.d $a2, $fp, 16
; LA64-WITHFP-NEXT:    addi.d $a1, $fp, 16
; LA64-WITHFP-NEXT:    st.d $a1, $fp, -24
; LA64-WITHFP-NEXT:    st.d $a0, $fp, 8
; LA64-WITHFP-NEXT:    ld.d $fp, $sp, 16 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    ld.d $ra, $sp, 24 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, 96
; LA64-WITHFP-NEXT:    ret
  %va = alloca ptr, align 8
  call void @llvm.va_start(ptr %va)
  %1 = va_arg ptr %va, i64
  call void @llvm.va_end(ptr %va)
  ret i64 %1
}

;; Ensure the adjustment when restoring the stack pointer using the frame
;; pointer is correct

define i64 @va1_va_arg_alloca(ptr %fmt, ...) nounwind {
; LA64-FPELIM-LABEL: va1_va_arg_alloca:
; LA64-FPELIM:       # %bb.0:
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, -96
; LA64-FPELIM-NEXT:    st.d $ra, $sp, 24 # 8-byte Folded Spill
; LA64-FPELIM-NEXT:    st.d $fp, $sp, 16 # 8-byte Folded Spill
; LA64-FPELIM-NEXT:    st.d $s0, $sp, 8 # 8-byte Folded Spill
; LA64-FPELIM-NEXT:    addi.d $fp, $sp, 32
; LA64-FPELIM-NEXT:    move $s0, $a1
; LA64-FPELIM-NEXT:    st.d $a7, $fp, 56
; LA64-FPELIM-NEXT:    st.d $a6, $fp, 48
; LA64-FPELIM-NEXT:    st.d $a5, $fp, 40
; LA64-FPELIM-NEXT:    st.d $a4, $fp, 32
; LA64-FPELIM-NEXT:    st.d $a3, $fp, 24
; LA64-FPELIM-NEXT:    st.d $a2, $fp, 16
; LA64-FPELIM-NEXT:    addi.d $a0, $fp, 16
; LA64-FPELIM-NEXT:    st.d $a0, $fp, -32
; LA64-FPELIM-NEXT:    st.d $a1, $fp, 8
; LA64-FPELIM-NEXT:    addi.d $a0, $a1, 15
; LA64-FPELIM-NEXT:    bstrins.d $a0, $zero, 3, 0
; LA64-FPELIM-NEXT:    sub.d $a0, $sp, $a0
; LA64-FPELIM-NEXT:    move $sp, $a0
; LA64-FPELIM-NEXT:    bl %plt(notdead)
; LA64-FPELIM-NEXT:    move $a0, $s0
; LA64-FPELIM-NEXT:    addi.d $sp, $fp, -32
; LA64-FPELIM-NEXT:    ld.d $s0, $sp, 8 # 8-byte Folded Reload
; LA64-FPELIM-NEXT:    ld.d $fp, $sp, 16 # 8-byte Folded Reload
; LA64-FPELIM-NEXT:    ld.d $ra, $sp, 24 # 8-byte Folded Reload
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, 96
; LA64-FPELIM-NEXT:    ret
;
; LA64-WITHFP-LABEL: va1_va_arg_alloca:
; LA64-WITHFP:       # %bb.0:
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, -96
; LA64-WITHFP-NEXT:    st.d $ra, $sp, 24 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    st.d $fp, $sp, 16 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    st.d $s0, $sp, 8 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    addi.d $fp, $sp, 32
; LA64-WITHFP-NEXT:    move $s0, $a1
; LA64-WITHFP-NEXT:    st.d $a7, $fp, 56
; LA64-WITHFP-NEXT:    st.d $a6, $fp, 48
; LA64-WITHFP-NEXT:    st.d $a5, $fp, 40
; LA64-WITHFP-NEXT:    st.d $a4, $fp, 32
; LA64-WITHFP-NEXT:    st.d $a3, $fp, 24
; LA64-WITHFP-NEXT:    st.d $a2, $fp, 16
; LA64-WITHFP-NEXT:    addi.d $a0, $fp, 16
; LA64-WITHFP-NEXT:    st.d $a0, $fp, -32
; LA64-WITHFP-NEXT:    st.d $a1, $fp, 8
; LA64-WITHFP-NEXT:    addi.d $a0, $a1, 15
; LA64-WITHFP-NEXT:    bstrins.d $a0, $zero, 3, 0
; LA64-WITHFP-NEXT:    sub.d $a0, $sp, $a0
; LA64-WITHFP-NEXT:    move $sp, $a0
; LA64-WITHFP-NEXT:    bl %plt(notdead)
; LA64-WITHFP-NEXT:    move $a0, $s0
; LA64-WITHFP-NEXT:    addi.d $sp, $fp, -32
; LA64-WITHFP-NEXT:    ld.d $s0, $sp, 8 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    ld.d $fp, $sp, 16 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    ld.d $ra, $sp, 24 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, 96
; LA64-WITHFP-NEXT:    ret
  %va = alloca ptr, align 8
  call void @llvm.va_start(ptr %va)
  %1 = va_arg ptr %va, i64
  %2 = alloca i8, i64 %1
  call void @notdead(ptr %2)
  call void @llvm.va_end(ptr %va)
  ret i64 %1
}

define void @va1_caller() nounwind {
; LA64-FPELIM-LABEL: va1_caller:
; LA64-FPELIM:       # %bb.0:
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, -16
; LA64-FPELIM-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
; LA64-FPELIM-NEXT:    lu52i.d $a1, $zero, 1023
; LA64-FPELIM-NEXT:    ori $a2, $zero, 2
; LA64-FPELIM-NEXT:    bl %plt(va1)
; LA64-FPELIM-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, 16
; LA64-FPELIM-NEXT:    ret
;
; LA64-WITHFP-LABEL: va1_caller:
; LA64-WITHFP:       # %bb.0:
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, -16
; LA64-WITHFP-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    st.d $fp, $sp, 0 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    addi.d $fp, $sp, 16
; LA64-WITHFP-NEXT:    lu52i.d $a1, $zero, 1023
; LA64-WITHFP-NEXT:    ori $a2, $zero, 2
; LA64-WITHFP-NEXT:    bl %plt(va1)
; LA64-WITHFP-NEXT:    ld.d $fp, $sp, 0 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, 16
; LA64-WITHFP-NEXT:    ret
  %1 = call i64 (ptr, ...) @va1(ptr undef, double 1.0, i64 2)
  ret void
}

;; Ensure a named 2*GRLen argument is passed in a1 and a2, while the
;; vararg long double is passed in a4 and a5 (rather than a3 and a4)

declare i64 @va_aligned_register(i64 %a, i128 %b, ...)

define void @va_aligned_register_caller() nounwind {
; LA64-FPELIM-LABEL: va_aligned_register_caller:
; LA64-FPELIM:       # %bb.0:
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, -16
; LA64-FPELIM-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
; LA64-FPELIM-NEXT:    lu12i.w $a0, 335544
; LA64-FPELIM-NEXT:    ori $a0, $a0, 1311
; LA64-FPELIM-NEXT:    lu32i.d $a0, 335544
; LA64-FPELIM-NEXT:    lu52i.d $a4, $a0, -328
; LA64-FPELIM-NEXT:    lu12i.w $a0, -503317
; LA64-FPELIM-NEXT:    ori $a0, $a0, 2129
; LA64-FPELIM-NEXT:    lu32i.d $a0, 37355
; LA64-FPELIM-NEXT:    lu52i.d $a5, $a0, 1024
; LA64-FPELIM-NEXT:    ori $a0, $zero, 2
; LA64-FPELIM-NEXT:    ori $a1, $zero, 1111
; LA64-FPELIM-NEXT:    move $a2, $zero
; LA64-FPELIM-NEXT:    bl %plt(va_aligned_register)
; LA64-FPELIM-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, 16
; LA64-FPELIM-NEXT:    ret
;
; LA64-WITHFP-LABEL: va_aligned_register_caller:
; LA64-WITHFP:       # %bb.0:
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, -16
; LA64-WITHFP-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    st.d $fp, $sp, 0 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    addi.d $fp, $sp, 16
; LA64-WITHFP-NEXT:    lu12i.w $a0, 335544
; LA64-WITHFP-NEXT:    ori $a0, $a0, 1311
; LA64-WITHFP-NEXT:    lu32i.d $a0, 335544
; LA64-WITHFP-NEXT:    lu52i.d $a4, $a0, -328
; LA64-WITHFP-NEXT:    lu12i.w $a0, -503317
; LA64-WITHFP-NEXT:    ori $a0, $a0, 2129
; LA64-WITHFP-NEXT:    lu32i.d $a0, 37355
; LA64-WITHFP-NEXT:    lu52i.d $a5, $a0, 1024
; LA64-WITHFP-NEXT:    ori $a0, $zero, 2
; LA64-WITHFP-NEXT:    ori $a1, $zero, 1111
; LA64-WITHFP-NEXT:    move $a2, $zero
; LA64-WITHFP-NEXT:    bl %plt(va_aligned_register)
; LA64-WITHFP-NEXT:    ld.d $fp, $sp, 0 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, 16
; LA64-WITHFP-NEXT:    ret
 %1 = call i64 (i64, i128, ...) @va_aligned_register(i64 2, i128 1111,
   fp128 0xLEB851EB851EB851F400091EB851EB851)
 ret void
}

;; Check 2*GRLen values are aligned appropriately when passed on the stack
;; in a vararg call

declare i32 @va_aligned_stack_callee(i32, ...)

define void @va_aligned_stack_caller() nounwind {
; LA64-FPELIM-LABEL: va_aligned_stack_caller:
; LA64-FPELIM:       # %bb.0:
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, -112
; LA64-FPELIM-NEXT:    st.d $ra, $sp, 104 # 8-byte Folded Spill
; LA64-FPELIM-NEXT:    ori $a0, $zero, 17
; LA64-FPELIM-NEXT:    st.d $a0, $sp, 48
; LA64-FPELIM-NEXT:    ori $a0, $zero, 16
; LA64-FPELIM-NEXT:    st.d $a0, $sp, 40
; LA64-FPELIM-NEXT:    ori $a0, $zero, 15
; LA64-FPELIM-NEXT:    st.d $a0, $sp, 32
; LA64-FPELIM-NEXT:    ori $a0, $zero, 14
; LA64-FPELIM-NEXT:    st.d $a0, $sp, 0
; LA64-FPELIM-NEXT:    lu12i.w $a0, -503317
; LA64-FPELIM-NEXT:    ori $a0, $a0, 2129
; LA64-FPELIM-NEXT:    lu32i.d $a0, 37355
; LA64-FPELIM-NEXT:    lu52i.d $a0, $a0, 1024
; LA64-FPELIM-NEXT:    st.d $a0, $sp, 24
; LA64-FPELIM-NEXT:    lu12i.w $a0, 335544
; LA64-FPELIM-NEXT:    ori $a0, $a0, 1311
; LA64-FPELIM-NEXT:    lu32i.d $a0, 335544
; LA64-FPELIM-NEXT:    lu52i.d $a0, $a0, -328
; LA64-FPELIM-NEXT:    st.d $a0, $sp, 16
; LA64-FPELIM-NEXT:    ori $a0, $zero, 1000
; LA64-FPELIM-NEXT:    st.d $a0, $sp, 64
; LA64-FPELIM-NEXT:    st.d $zero, $sp, 88
; LA64-FPELIM-NEXT:    st.d $zero, $sp, 80
; LA64-FPELIM-NEXT:    st.d $zero, $sp, 72
; LA64-FPELIM-NEXT:    ori $a1, $zero, 11
; LA64-FPELIM-NEXT:    addi.d $a2, $sp, 64
; LA64-FPELIM-NEXT:    ori $a3, $zero, 12
; LA64-FPELIM-NEXT:    ori $a4, $zero, 13
; LA64-FPELIM-NEXT:    ori $a0, $zero, 1
; LA64-FPELIM-NEXT:    move $a6, $zero
; LA64-FPELIM-NEXT:    move $a7, $a0
; LA64-FPELIM-NEXT:    bl %plt(va_aligned_stack_callee)
; LA64-FPELIM-NEXT:    ld.d $ra, $sp, 104 # 8-byte Folded Reload
; LA64-FPELIM-NEXT:    addi.d $sp, $sp, 112
; LA64-FPELIM-NEXT:    ret
;
; LA64-WITHFP-LABEL: va_aligned_stack_caller:
; LA64-WITHFP:       # %bb.0:
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, -112
; LA64-WITHFP-NEXT:    st.d $ra, $sp, 104 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    st.d $fp, $sp, 96 # 8-byte Folded Spill
; LA64-WITHFP-NEXT:    addi.d $fp, $sp, 112
; LA64-WITHFP-NEXT:    ori $a0, $zero, 17
; LA64-WITHFP-NEXT:    st.d $a0, $sp, 48
; LA64-WITHFP-NEXT:    ori $a0, $zero, 16
; LA64-WITHFP-NEXT:    st.d $a0, $sp, 40
; LA64-WITHFP-NEXT:    ori $a0, $zero, 15
; LA64-WITHFP-NEXT:    st.d $a0, $sp, 32
; LA64-WITHFP-NEXT:    ori $a0, $zero, 14
; LA64-WITHFP-NEXT:    st.d $a0, $sp, 0
; LA64-WITHFP-NEXT:    lu12i.w $a0, -503317
; LA64-WITHFP-NEXT:    ori $a0, $a0, 2129
; LA64-WITHFP-NEXT:    lu32i.d $a0, 37355
; LA64-WITHFP-NEXT:    lu52i.d $a0, $a0, 1024
; LA64-WITHFP-NEXT:    st.d $a0, $sp, 24
; LA64-WITHFP-NEXT:    lu12i.w $a0, 335544
; LA64-WITHFP-NEXT:    ori $a0, $a0, 1311
; LA64-WITHFP-NEXT:    lu32i.d $a0, 335544
; LA64-WITHFP-NEXT:    lu52i.d $a0, $a0, -328
; LA64-WITHFP-NEXT:    st.d $a0, $sp, 16
; LA64-WITHFP-NEXT:    ori $a0, $zero, 1000
; LA64-WITHFP-NEXT:    st.d $a0, $fp, -48
; LA64-WITHFP-NEXT:    st.d $zero, $fp, -24
; LA64-WITHFP-NEXT:    st.d $zero, $fp, -32
; LA64-WITHFP-NEXT:    st.d $zero, $fp, -40
; LA64-WITHFP-NEXT:    ori $a1, $zero, 11
; LA64-WITHFP-NEXT:    addi.d $a2, $fp, -48
; LA64-WITHFP-NEXT:    ori $a3, $zero, 12
; LA64-WITHFP-NEXT:    ori $a4, $zero, 13
; LA64-WITHFP-NEXT:    ori $a0, $zero, 1
; LA64-WITHFP-NEXT:    move $a6, $zero
; LA64-WITHFP-NEXT:    move $a7, $a0
; LA64-WITHFP-NEXT:    bl %plt(va_aligned_stack_callee)
; LA64-WITHFP-NEXT:    ld.d $fp, $sp, 96 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    ld.d $ra, $sp, 104 # 8-byte Folded Reload
; LA64-WITHFP-NEXT:    addi.d $sp, $sp, 112
; LA64-WITHFP-NEXT:    ret
  %1 = call i32 (i32, ...) @va_aligned_stack_callee(i32 1, i32 11,
    i256 1000, i32 12, i32 13, i128 18446744073709551616, i32 14,
    fp128 0xLEB851EB851EB851F400091EB851EB851, i64 15,
    [2 x i64] [i64 16, i64 17])
  ret void
}