[RISCV][GISEL] Add vararg.ll LLVM IR -> ASM test

This test is added to be the counterpart of the SelectionDAG
llvm/test/CodeGen/RISCV/vararg.ll test. Minor changes were made compared
to the other version, all which are commented in the test file added in
this commit.

1 files changed, 1008 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/vararg.ll b/llvm/test/CodeGen/RISCV/GlobalISel/vararg.ll
new file mode 100644
index 0000000..be5c1cd
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/GlobalISel/vararg.ll
@@ -0,0 +1,1008 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -global-isel  -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefixes=RV32,ILP32 %s
+; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -global-isel  -mattr=+d -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefixes=RV32,RV32D-ILP32 %s
+; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -global-isel  -mattr=+d -target-abi ilp32f \
+; RUN:     -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefixes=RV32,RV32D-ILP32F %s
+; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -global-isel  -mattr=+d -target-abi ilp32d \
+; RUN:     -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefixes=RV32,RV32D-ILP32D %s
+; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -global-isel  -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefixes=RV64,LP64 %s
+; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -global-isel  -mattr=+d -target-abi lp64f \
+; RUN:     -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefixes=RV64,LP64F %s
+; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -global-isel  -mattr=+d -target-abi lp64d \
+; RUN:     -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefixes=RV64,LP64D %s
+
+; The same vararg calling convention is used for ilp32/ilp32f/ilp32d and for
+; lp64/lp64f/lp64d. Different CHECK lines are required due to slight
+; codegen differences due to the way the f64 load operations are lowered and
+; because the PseudoCALL specifies the calling  convention.
+; The nounwind attribute is omitted for some of the tests, to check that CFI
+; directives are correctly generated.
+
+declare void @llvm.va_start(ptr)
+declare void @llvm.va_end(ptr)
+
+declare void @notdead(ptr)
+
+; Although frontends are recommended to not generate va_arg due to the lack of
+; support for aggregate types, we test simple cases here to ensure they are
+; lowered correctly
+
+define i32 @va1(ptr %fmt, ...) {
+; RV32-LABEL: va1:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -48
+; RV32-NEXT:    .cfi_def_cfa_offset 48
+; RV32-NEXT:    sw a1, 20(sp)
+; RV32-NEXT:    sw a2, 24(sp)
+; RV32-NEXT:    sw a3, 28(sp)
+; RV32-NEXT:    sw a4, 32(sp)
+; RV32-NEXT:    addi a0, sp, 20
+; RV32-NEXT:    sw a0, 12(sp)
+; RV32-NEXT:    lw a0, 12(sp)
+; RV32-NEXT:    sw a5, 36(sp)
+; RV32-NEXT:    sw a6, 40(sp)
+; RV32-NEXT:    sw a7, 44(sp)
+; RV32-NEXT:    addi a1, a0, 4
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    lw a0, 0(a0)
+; RV32-NEXT:    addi sp, sp, 48
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: va1:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -80
+; RV64-NEXT:    .cfi_def_cfa_offset 80
+; RV64-NEXT:    sd a1, 24(sp)
+; RV64-NEXT:    sd a2, 32(sp)
+; RV64-NEXT:    sd a3, 40(sp)
+; RV64-NEXT:    sd a4, 48(sp)
+; RV64-NEXT:    sd a5, 56(sp)
+; RV64-NEXT:    addi a0, sp, 24
+; RV64-NEXT:    sd a0, 8(sp)
+; RV64-NEXT:    lw a0, 12(sp)
+; RV64-NEXT:    lwu a1, 8(sp)
+; RV64-NEXT:    sd a6, 64(sp)
+; RV64-NEXT:    sd a7, 72(sp)
+; RV64-NEXT:    slli a0, a0, 32
+; RV64-NEXT:    or a0, a0, a1
+; RV64-NEXT:    addi a1, a0, 4
+; RV64-NEXT:    srli a2, a1, 32
+; RV64-NEXT:    sw a1, 8(sp)
+; RV64-NEXT:    sw a2, 12(sp)
+; RV64-NEXT:    lw a0, 0(a0)
+; RV64-NEXT:    addi sp, sp, 80
+; RV64-NEXT:    ret
+  %va = alloca ptr
+  call void @llvm.va_start(ptr %va)
+  %argp.cur = load ptr, ptr %va, align 4
+  %argp.next = getelementptr inbounds i8, ptr %argp.cur, i32 4
+  store ptr %argp.next, ptr %va, align 4
+  %1 = load i32, ptr %argp.cur, align 4
+  call void @llvm.va_end(ptr %va)
+  ret i32 %1
+}
+
+define i32 @va1_va_arg(ptr %fmt, ...) nounwind {
+; RV32-LABEL: va1_va_arg:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -48
+; RV32-NEXT:    sw a1, 20(sp)
+; RV32-NEXT:    sw a2, 24(sp)
+; RV32-NEXT:    sw a3, 28(sp)
+; RV32-NEXT:    sw a4, 32(sp)
+; RV32-NEXT:    sw a5, 36(sp)
+; RV32-NEXT:    sw a6, 40(sp)
+; RV32-NEXT:    sw a7, 44(sp)
+; RV32-NEXT:    addi a0, sp, 20
+; RV32-NEXT:    sw a0, 12(sp)
+; RV32-NEXT:    lw a0, 12(sp)
+; RV32-NEXT:    addi a0, a0, 3
+; RV32-NEXT:    andi a0, a0, -4
+; RV32-NEXT:    addi a1, a0, 4
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    lw a0, 0(a0)
+; RV32-NEXT:    addi sp, sp, 48
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: va1_va_arg:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -80
+; RV64-NEXT:    sd a1, 24(sp)
+; RV64-NEXT:    sd a2, 32(sp)
+; RV64-NEXT:    sd a3, 40(sp)
+; RV64-NEXT:    sd a4, 48(sp)
+; RV64-NEXT:    sd a5, 56(sp)
+; RV64-NEXT:    sd a6, 64(sp)
+; RV64-NEXT:    sd a7, 72(sp)
+; RV64-NEXT:    addi a0, sp, 24
+; RV64-NEXT:    sd a0, 8(sp)
+; RV64-NEXT:    ld a0, 8(sp)
+; RV64-NEXT:    addi a0, a0, 3
+; RV64-NEXT:    andi a0, a0, -4
+; RV64-NEXT:    addi a1, a0, 4
+; RV64-NEXT:    sd a1, 8(sp)
+; RV64-NEXT:    lw a0, 0(a0)
+; RV64-NEXT:    addi sp, sp, 80
+; RV64-NEXT:    ret
+  %va = alloca ptr
+  call void @llvm.va_start(ptr %va)
+  %1 = va_arg ptr %va, i32
+  call void @llvm.va_end(ptr %va)
+  ret i32 %1
+}
+
+; Ensure the adjustment when restoring the stack pointer using the frame
+; pointer is correct
+define i32 @va1_va_arg_alloca(ptr %fmt, ...) nounwind {
+; RV32-LABEL: va1_va_arg_alloca:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -48
+; RV32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32-NEXT:    sw s1, 4(sp) # 4-byte Folded Spill
+; RV32-NEXT:    addi s0, sp, 16
+; RV32-NEXT:    sw a1, 4(s0)
+; RV32-NEXT:    sw a2, 8(s0)
+; RV32-NEXT:    sw a3, 12(s0)
+; RV32-NEXT:    sw a4, 16(s0)
+; RV32-NEXT:    sw a5, 20(s0)
+; RV32-NEXT:    sw a6, 24(s0)
+; RV32-NEXT:    sw a7, 28(s0)
+; RV32-NEXT:    addi a0, s0, 4
+; RV32-NEXT:    sw a0, -16(s0)
+; RV32-NEXT:    lw a0, -16(s0)
+; RV32-NEXT:    addi a0, a0, 3
+; RV32-NEXT:    andi a0, a0, -4
+; RV32-NEXT:    addi a1, a0, 4
+; RV32-NEXT:    sw a1, -16(s0)
+; RV32-NEXT:    lw s1, 0(a0)
+; RV32-NEXT:    addi a0, s1, 15
+; RV32-NEXT:    andi a0, a0, -16
+; RV32-NEXT:    sub a0, sp, a0
+; RV32-NEXT:    mv sp, a0
+; RV32-NEXT:    call notdead@plt
+; RV32-NEXT:    mv a0, s1
+; RV32-NEXT:    addi sp, s0, -16
+; RV32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
+; RV32-NEXT:    lw s1, 4(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 48
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: va1_va_arg_alloca:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -96
+; RV64-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
+; RV64-NEXT:    sd s0, 16(sp) # 8-byte Folded Spill
+; RV64-NEXT:    sd s1, 8(sp) # 8-byte Folded Spill
+; RV64-NEXT:    addi s0, sp, 32
+; RV64-NEXT:    sd a1, 8(s0)
+; RV64-NEXT:    sd a2, 16(s0)
+; RV64-NEXT:    sd a3, 24(s0)
+; RV64-NEXT:    sd a4, 32(s0)
+; RV64-NEXT:    sd a5, 40(s0)
+; RV64-NEXT:    sd a6, 48(s0)
+; RV64-NEXT:    sd a7, 56(s0)
+; RV64-NEXT:    addi a0, s0, 8
+; RV64-NEXT:    sd a0, -32(s0)
+; RV64-NEXT:    ld a0, -32(s0)
+; RV64-NEXT:    addi a0, a0, 3
+; RV64-NEXT:    andi a0, a0, -4
+; RV64-NEXT:    addi a1, a0, 4
+; RV64-NEXT:    sd a1, -32(s0)
+; RV64-NEXT:    lw s1, 0(a0)
+; RV64-NEXT:    slli a0, s1, 32
+; RV64-NEXT:    srli a0, a0, 32
+; RV64-NEXT:    addi a0, a0, 15
+; RV64-NEXT:    andi a0, a0, -16
+; RV64-NEXT:    sub a0, sp, a0
+; RV64-NEXT:    mv sp, a0
+; RV64-NEXT:    call notdead@plt
+; RV64-NEXT:    mv a0, s1
+; RV64-NEXT:    addi sp, s0, -32
+; RV64-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
+; RV64-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
+; RV64-NEXT:    ld s1, 8(sp) # 8-byte Folded Reload
+; RV64-NEXT:    addi sp, sp, 96
+; RV64-NEXT:    ret
+  %va = alloca ptr
+  call void @llvm.va_start(ptr %va)
+  %1 = va_arg ptr %va, i32
+  %2 = alloca i8, i32 %1
+  call void @notdead(ptr %2)
+  call void @llvm.va_end(ptr %va)
+  ret i32 %1
+}
+
+define void @va1_caller() nounwind {
+; RV32-LABEL: va1_caller:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT:    lui a3, 261888
+; RV32-NEXT:    li a4, 2
+; RV32-NEXT:    li a2, 0
+; RV32-NEXT:    call va1@plt
+; RV32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; LP64-LABEL: va1_caller:
+; LP64:       # %bb.0:
+; LP64-NEXT:    addi sp, sp, -16
+; LP64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; LP64-NEXT:    lui a0, %hi(.LCPI3_0)
+; LP64-NEXT:    ld a1, %lo(.LCPI3_0)(a0)
+; LP64-NEXT:    li a2, 2
+; LP64-NEXT:    call va1@plt
+; LP64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; LP64-NEXT:    addi sp, sp, 16
+; LP64-NEXT:    ret
+;
+; LP64F-LABEL: va1_caller:
+; LP64F:       # %bb.0:
+; LP64F-NEXT:    addi sp, sp, -16
+; LP64F-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; LP64F-NEXT:    li a0, 1023
+; LP64F-NEXT:    slli a0, a0, 52
+; LP64F-NEXT:    fmv.d.x fa5, a0
+; LP64F-NEXT:    li a2, 2
+; LP64F-NEXT:    fmv.x.d a1, fa5
+; LP64F-NEXT:    call va1@plt
+; LP64F-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; LP64F-NEXT:    addi sp, sp, 16
+; LP64F-NEXT:    ret
+;
+; LP64D-LABEL: va1_caller:
+; LP64D:       # %bb.0:
+; LP64D-NEXT:    addi sp, sp, -16
+; LP64D-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; LP64D-NEXT:    li a0, 1023
+; LP64D-NEXT:    slli a0, a0, 52
+; LP64D-NEXT:    fmv.d.x fa5, a0
+; LP64D-NEXT:    li a2, 2
+; LP64D-NEXT:    fmv.x.d a1, fa5
+; LP64D-NEXT:    call va1@plt
+; LP64D-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; LP64D-NEXT:    addi sp, sp, 16
+; LP64D-NEXT:    ret
+  %1 = call i32 (ptr, ...) @va1(ptr undef, double 1.0, i32 2)
+  ret void
+}
+
+; Ensure that 2x xlen size+alignment varargs are accessed via an "aligned"
+; register pair (where the first register is even-numbered).
+
+define i64 @va2(ptr %fmt, ...) nounwind {
+; ILP32-LABEL: va2:
+; ILP32:       # %bb.0:
+; ILP32-NEXT:    addi sp, sp, -48
+; ILP32-NEXT:    sw a1, 20(sp)
+; ILP32-NEXT:    sw a2, 24(sp)
+; ILP32-NEXT:    sw a3, 28(sp)
+; ILP32-NEXT:    sw a4, 32(sp)
+; ILP32-NEXT:    addi a0, sp, 20
+; ILP32-NEXT:    sw a0, 12(sp)
+; ILP32-NEXT:    lw a0, 12(sp)
+; ILP32-NEXT:    sw a5, 36(sp)
+; ILP32-NEXT:    sw a6, 40(sp)
+; ILP32-NEXT:    sw a7, 44(sp)
+; ILP32-NEXT:    addi a1, a0, 7
+; ILP32-NEXT:    andi a1, a1, -8
+; ILP32-NEXT:    addi a0, a0, 15
+; ILP32-NEXT:    sw a0, 12(sp)
+; ILP32-NEXT:    lw a0, 0(a1)
+; ILP32-NEXT:    lw a1, 4(a1)
+; ILP32-NEXT:    addi sp, sp, 48
+; ILP32-NEXT:    ret
+;
+; RV32D-ILP32-LABEL: va2:
+; RV32D-ILP32:       # %bb.0:
+; RV32D-ILP32-NEXT:    addi sp, sp, -48
+; RV32D-ILP32-NEXT:    sw a1, 20(sp)
+; RV32D-ILP32-NEXT:    sw a2, 24(sp)
+; RV32D-ILP32-NEXT:    sw a3, 28(sp)
+; RV32D-ILP32-NEXT:    sw a4, 32(sp)
+; RV32D-ILP32-NEXT:    addi a0, sp, 20
+; RV32D-ILP32-NEXT:    sw a0, 12(sp)
+; RV32D-ILP32-NEXT:    lw a0, 12(sp)
+; RV32D-ILP32-NEXT:    sw a5, 36(sp)
+; RV32D-ILP32-NEXT:    sw a6, 40(sp)
+; RV32D-ILP32-NEXT:    sw a7, 44(sp)
+; RV32D-ILP32-NEXT:    addi a1, a0, 7
+; RV32D-ILP32-NEXT:    andi a1, a1, -8
+; RV32D-ILP32-NEXT:    fld fa5, 0(a1)
+; RV32D-ILP32-NEXT:    addi a0, a0, 15
+; RV32D-ILP32-NEXT:    sw a0, 12(sp)
+; RV32D-ILP32-NEXT:    fsd fa5, 0(sp)
+; RV32D-ILP32-NEXT:    lw a0, 0(sp)
+; RV32D-ILP32-NEXT:    lw a1, 4(sp)
+; RV32D-ILP32-NEXT:    addi sp, sp, 48
+; RV32D-ILP32-NEXT:    ret
+;
+; RV32D-ILP32F-LABEL: va2:
+; RV32D-ILP32F:       # %bb.0:
+; RV32D-ILP32F-NEXT:    addi sp, sp, -48
+; RV32D-ILP32F-NEXT:    sw a1, 20(sp)
+; RV32D-ILP32F-NEXT:    sw a2, 24(sp)
+; RV32D-ILP32F-NEXT:    sw a3, 28(sp)
+; RV32D-ILP32F-NEXT:    sw a4, 32(sp)
+; RV32D-ILP32F-NEXT:    addi a0, sp, 20
+; RV32D-ILP32F-NEXT:    sw a0, 12(sp)
+; RV32D-ILP32F-NEXT:    lw a0, 12(sp)
+; RV32D-ILP32F-NEXT:    sw a5, 36(sp)
+; RV32D-ILP32F-NEXT:    sw a6, 40(sp)
+; RV32D-ILP32F-NEXT:    sw a7, 44(sp)
+; RV32D-ILP32F-NEXT:    addi a1, a0, 7
+; RV32D-ILP32F-NEXT:    andi a1, a1, -8
+; RV32D-ILP32F-NEXT:    fld fa5, 0(a1)
+; RV32D-ILP32F-NEXT:    addi a0, a0, 15
+; RV32D-ILP32F-NEXT:    sw a0, 12(sp)
+; RV32D-ILP32F-NEXT:    fsd fa5, 0(sp)
+; RV32D-ILP32F-NEXT:    lw a0, 0(sp)
+; RV32D-ILP32F-NEXT:    lw a1, 4(sp)
+; RV32D-ILP32F-NEXT:    addi sp, sp, 48
+; RV32D-ILP32F-NEXT:    ret
+;
+; RV32D-ILP32D-LABEL: va2:
+; RV32D-ILP32D:       # %bb.0:
+; RV32D-ILP32D-NEXT:    addi sp, sp, -48
+; RV32D-ILP32D-NEXT:    sw a1, 20(sp)
+; RV32D-ILP32D-NEXT:    sw a2, 24(sp)
+; RV32D-ILP32D-NEXT:    sw a3, 28(sp)
+; RV32D-ILP32D-NEXT:    sw a4, 32(sp)
+; RV32D-ILP32D-NEXT:    addi a0, sp, 20
+; RV32D-ILP32D-NEXT:    sw a0, 12(sp)
+; RV32D-ILP32D-NEXT:    lw a0, 12(sp)
+; RV32D-ILP32D-NEXT:    sw a5, 36(sp)
+; RV32D-ILP32D-NEXT:    sw a6, 40(sp)
+; RV32D-ILP32D-NEXT:    sw a7, 44(sp)
+; RV32D-ILP32D-NEXT:    addi a1, a0, 7
+; RV32D-ILP32D-NEXT:    andi a1, a1, -8
+; RV32D-ILP32D-NEXT:    fld fa5, 0(a1)
+; RV32D-ILP32D-NEXT:    addi a0, a0, 15
+; RV32D-ILP32D-NEXT:    sw a0, 12(sp)
+; RV32D-ILP32D-NEXT:    fsd fa5, 0(sp)
+; RV32D-ILP32D-NEXT:    lw a0, 0(sp)
+; RV32D-ILP32D-NEXT:    lw a1, 4(sp)
+; RV32D-ILP32D-NEXT:    addi sp, sp, 48
+; RV32D-ILP32D-NEXT:    ret
+;
+; RV64-LABEL: va2:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -80
+; RV64-NEXT:    sd a1, 24(sp)
+; RV64-NEXT:    sd a2, 32(sp)
+; RV64-NEXT:    sd a3, 40(sp)
+; RV64-NEXT:    sd a4, 48(sp)
+; RV64-NEXT:    addi a0, sp, 24
+; RV64-NEXT:    sd a0, 8(sp)
+; RV64-NEXT:    ld a0, 8(sp)
+; RV64-NEXT:    sd a5, 56(sp)
+; RV64-NEXT:    sd a6, 64(sp)
+; RV64-NEXT:    sd a7, 72(sp)
+; RV64-NEXT:    addi a1, a0, 7
+; RV64-NEXT:    andi a1, a1, -8
+; RV64-NEXT:    addi a0, a0, 15
+; RV64-NEXT:    sd a0, 8(sp)
+; RV64-NEXT:    ld a0, 0(a1)
+; RV64-NEXT:    addi sp, sp, 80
+; RV64-NEXT:    ret
+  %va = alloca ptr
+  call void @llvm.va_start(ptr %va)
+  %argp.cur = load ptr, ptr %va
+  %ptrint = ptrtoint ptr %argp.cur to iXLen
+  %1 = add iXLen %ptrint, 7
+  %2 = and iXLen %1, -8
+  %argp.cur.aligned = inttoptr iXLen %1 to ptr
+  %argp.next = getelementptr inbounds i8, ptr %argp.cur.aligned, i32 8
+  store ptr %argp.next, ptr %va
+  %3 = inttoptr iXLen %2 to ptr
+  %4 = load double, ptr %3, align 8
+  %5 = bitcast double %4 to i64
+  call void @llvm.va_end(ptr %va)
+  ret i64 %5
+}
+
+; This test is slightly different than the SelectionDAG counterpart because
+; narrowScalar and widenScalar for G_VAARG on types outside of [s32, sXLen]
+; are not implemented yet.
+define i64 @va2_va_arg(ptr %fmt, ...) nounwind {
+; RV32-LABEL: va2_va_arg:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -48
+; RV32-NEXT:    sw a1, 20(sp)
+; RV32-NEXT:    sw a2, 24(sp)
+; RV32-NEXT:    sw a3, 28(sp)
+; RV32-NEXT:    sw a4, 32(sp)
+; RV32-NEXT:    sw a5, 36(sp)
+; RV32-NEXT:    sw a6, 40(sp)
+; RV32-NEXT:    sw a7, 44(sp)
+; RV32-NEXT:    addi a0, sp, 20
+; RV32-NEXT:    sw a0, 12(sp)
+; RV32-NEXT:    lw a0, 12(sp)
+; RV32-NEXT:    addi a0, a0, 3
+; RV32-NEXT:    andi a0, a0, -4
+; RV32-NEXT:    addi a1, a0, 4
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    lw a0, 0(a0)
+; RV32-NEXT:    li a1, 0
+; RV32-NEXT:    addi sp, sp, 48
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: va2_va_arg:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -80
+; RV64-NEXT:    sd a1, 24(sp)
+; RV64-NEXT:    sd a2, 32(sp)
+; RV64-NEXT:    sd a3, 40(sp)
+; RV64-NEXT:    sd a4, 48(sp)
+; RV64-NEXT:    sd a5, 56(sp)
+; RV64-NEXT:    sd a6, 64(sp)
+; RV64-NEXT:    sd a7, 72(sp)
+; RV64-NEXT:    addi a0, sp, 24
+; RV64-NEXT:    sd a0, 8(sp)
+; RV64-NEXT:    ld a0, 8(sp)
+; RV64-NEXT:    addi a0, a0, 3
+; RV64-NEXT:    andi a0, a0, -4
+; RV64-NEXT:    addi a1, a0, 4
+; RV64-NEXT:    sd a1, 8(sp)
+; RV64-NEXT:    lw a0, 0(a0)
+; RV64-NEXT:    slli a0, a0, 32
+; RV64-NEXT:    srli a0, a0, 32
+; RV64-NEXT:    addi sp, sp, 80
+; RV64-NEXT:    ret
+  %va = alloca ptr
+  call void @llvm.va_start(ptr %va)
+  %1 = va_arg ptr %va, i32
+  call void @llvm.va_end(ptr %va)
+  %2 = zext i32 %1 to i64
+  ret i64 %2
+}
+
+define void @va2_caller() nounwind {
+; RV32-LABEL: va2_caller:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT:    li a1, 1
+; RV32-NEXT:    call va2@plt
+; RV32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: va2_caller:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -16
+; RV64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64-NEXT:    li a1, 1
+; RV64-NEXT:    call va2@plt
+; RV64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64-NEXT:    addi sp, sp, 16
+; RV64-NEXT:    ret
+ %1 = call i64 (ptr, ...) @va2(ptr undef, i32 1)
+ ret void
+}
+
+; On RV32, Ensure a named 2*xlen argument is passed in a1 and a2, while the
+; vararg double is passed in a4 and a5 (rather than a3 and a4)
+
+define i64 @va3(i32 %a, i64 %b, ...) nounwind {
+; ILP32-LABEL: va3:
+; ILP32:       # %bb.0:
+; ILP32-NEXT:    addi sp, sp, -32
+; ILP32-NEXT:    sw a3, 12(sp)
+; ILP32-NEXT:    sw a4, 16(sp)
+; ILP32-NEXT:    addi a0, sp, 12
+; ILP32-NEXT:    sw a0, 4(sp)
+; ILP32-NEXT:    lw a0, 4(sp)
+; ILP32-NEXT:    sw a5, 20(sp)
+; ILP32-NEXT:    sw a6, 24(sp)
+; ILP32-NEXT:    sw a7, 28(sp)
+; ILP32-NEXT:    addi a3, a0, 7
+; ILP32-NEXT:    andi a3, a3, -8
+; ILP32-NEXT:    addi a0, a0, 15
+; ILP32-NEXT:    sw a0, 4(sp)
+; ILP32-NEXT:    lw a4, 0(a3)
+; ILP32-NEXT:    lw a3, 4(a3)
+; ILP32-NEXT:    add a0, a1, a4
+; ILP32-NEXT:    sltu a1, a0, a4
+; ILP32-NEXT:    add a2, a2, a3
+; ILP32-NEXT:    add a1, a2, a1
+; ILP32-NEXT:    addi sp, sp, 32
+; ILP32-NEXT:    ret
+;
+; RV32D-ILP32-LABEL: va3:
+; RV32D-ILP32:       # %bb.0:
+; RV32D-ILP32-NEXT:    addi sp, sp, -48
+; RV32D-ILP32-NEXT:    sw a3, 28(sp)
+; RV32D-ILP32-NEXT:    sw a4, 32(sp)
+; RV32D-ILP32-NEXT:    addi a0, sp, 28
+; RV32D-ILP32-NEXT:    sw a0, 20(sp)
+; RV32D-ILP32-NEXT:    lw a0, 20(sp)
+; RV32D-ILP32-NEXT:    sw a5, 36(sp)
+; RV32D-ILP32-NEXT:    sw a6, 40(sp)
+; RV32D-ILP32-NEXT:    sw a7, 44(sp)
+; RV32D-ILP32-NEXT:    addi a3, a0, 7
+; RV32D-ILP32-NEXT:    andi a3, a3, -8
+; RV32D-ILP32-NEXT:    fld fa5, 0(a3)
+; RV32D-ILP32-NEXT:    addi a0, a0, 15
+; RV32D-ILP32-NEXT:    sw a0, 20(sp)
+; RV32D-ILP32-NEXT:    fsd fa5, 8(sp)
+; RV32D-ILP32-NEXT:    lw a3, 8(sp)
+; RV32D-ILP32-NEXT:    lw a4, 12(sp)
+; RV32D-ILP32-NEXT:    add a0, a1, a3
+; RV32D-ILP32-NEXT:    sltu a1, a0, a3
+; RV32D-ILP32-NEXT:    add a2, a2, a4
+; RV32D-ILP32-NEXT:    add a1, a2, a1
+; RV32D-ILP32-NEXT:    addi sp, sp, 48
+; RV32D-ILP32-NEXT:    ret
+;
+; RV32D-ILP32F-LABEL: va3:
+; RV32D-ILP32F:       # %bb.0:
+; RV32D-ILP32F-NEXT:    addi sp, sp, -48
+; RV32D-ILP32F-NEXT:    sw a3, 28(sp)
+; RV32D-ILP32F-NEXT:    sw a4, 32(sp)
+; RV32D-ILP32F-NEXT:    addi a0, sp, 28
+; RV32D-ILP32F-NEXT:    sw a0, 20(sp)
+; RV32D-ILP32F-NEXT:    lw a0, 20(sp)
+; RV32D-ILP32F-NEXT:    sw a5, 36(sp)
+; RV32D-ILP32F-NEXT:    sw a6, 40(sp)
+; RV32D-ILP32F-NEXT:    sw a7, 44(sp)
+; RV32D-ILP32F-NEXT:    addi a3, a0, 7
+; RV32D-ILP32F-NEXT:    andi a3, a3, -8
+; RV32D-ILP32F-NEXT:    fld fa5, 0(a3)
+; RV32D-ILP32F-NEXT:    addi a0, a0, 15
+; RV32D-ILP32F-NEXT:    sw a0, 20(sp)
+; RV32D-ILP32F-NEXT:    fsd fa5, 8(sp)
+; RV32D-ILP32F-NEXT:    lw a3, 8(sp)
+; RV32D-ILP32F-NEXT:    lw a4, 12(sp)
+; RV32D-ILP32F-NEXT:    add a0, a1, a3
+; RV32D-ILP32F-NEXT:    sltu a1, a0, a3
+; RV32D-ILP32F-NEXT:    add a2, a2, a4
+; RV32D-ILP32F-NEXT:    add a1, a2, a1
+; RV32D-ILP32F-NEXT:    addi sp, sp, 48
+; RV32D-ILP32F-NEXT:    ret
+;
+; RV32D-ILP32D-LABEL: va3:
+; RV32D-ILP32D:       # %bb.0:
+; RV32D-ILP32D-NEXT:    addi sp, sp, -48
+; RV32D-ILP32D-NEXT:    sw a3, 28(sp)
+; RV32D-ILP32D-NEXT:    sw a4, 32(sp)
+; RV32D-ILP32D-NEXT:    addi a0, sp, 28
+; RV32D-ILP32D-NEXT:    sw a0, 20(sp)
+; RV32D-ILP32D-NEXT:    lw a0, 20(sp)
+; RV32D-ILP32D-NEXT:    sw a5, 36(sp)
+; RV32D-ILP32D-NEXT:    sw a6, 40(sp)
+; RV32D-ILP32D-NEXT:    sw a7, 44(sp)
+; RV32D-ILP32D-NEXT:    addi a3, a0, 7
+; RV32D-ILP32D-NEXT:    andi a3, a3, -8
+; RV32D-ILP32D-NEXT:    fld fa5, 0(a3)
+; RV32D-ILP32D-NEXT:    addi a0, a0, 15
+; RV32D-ILP32D-NEXT:    sw a0, 20(sp)
+; RV32D-ILP32D-NEXT:    fsd fa5, 8(sp)
+; RV32D-ILP32D-NEXT:    lw a3, 8(sp)
+; RV32D-ILP32D-NEXT:    lw a4, 12(sp)
+; RV32D-ILP32D-NEXT:    add a0, a1, a3
+; RV32D-ILP32D-NEXT:    sltu a1, a0, a3
+; RV32D-ILP32D-NEXT:    add a2, a2, a4
+; RV32D-ILP32D-NEXT:    add a1, a2, a1
+; RV32D-ILP32D-NEXT:    addi sp, sp, 48
+; RV32D-ILP32D-NEXT:    ret
+;
+; RV64-LABEL: va3:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -64
+; RV64-NEXT:    sd a2, 16(sp)
+; RV64-NEXT:    sd a3, 24(sp)
+; RV64-NEXT:    sd a4, 32(sp)
+; RV64-NEXT:    addi a0, sp, 16
+; RV64-NEXT:    sd a0, 8(sp)
+; RV64-NEXT:    ld a0, 8(sp)
+; RV64-NEXT:    sd a5, 40(sp)
+; RV64-NEXT:    sd a6, 48(sp)
+; RV64-NEXT:    sd a7, 56(sp)
+; RV64-NEXT:    addi a2, a0, 7
+; RV64-NEXT:    andi a2, a2, -8
+; RV64-NEXT:    addi a0, a0, 15
+; RV64-NEXT:    sd a0, 8(sp)
+; RV64-NEXT:    ld a0, 0(a2)
+; RV64-NEXT:    add a0, a1, a0
+; RV64-NEXT:    addi sp, sp, 64
+; RV64-NEXT:    ret
+  %va = alloca ptr
+  call void @llvm.va_start(ptr %va)
+  %argp.cur = load ptr, ptr %va
+  %ptrint = ptrtoint ptr %argp.cur to iXLen
+  %1 = add iXLen %ptrint, 7
+  %2 = and iXLen %1, -8
+  %argp.cur.aligned = inttoptr iXLen %1 to ptr
+  %argp.next = getelementptr inbounds i8, ptr %argp.cur.aligned, i32 8
+  store ptr %argp.next, ptr %va
+  %3 = inttoptr iXLen %2 to ptr
+  %4 = load double, ptr %3, align 8
+  call void @llvm.va_end(ptr %va)
+  %5 = bitcast double %4 to i64
+  %6 = add i64 %b, %5
+  ret i64 %6
+}
+
+; This test is slightly different than the SelectionDAG counterpart because
+; narrowScalar and widenScalar for G_VAARG on types outside of [s32, sXLen]
+; are not implemented yet.
+define i64 @va3_va_arg(i32 %a, i64 %b, ...) nounwind {
+; RV32-LABEL: va3_va_arg:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -32
+; RV32-NEXT:    sw a3, 12(sp)
+; RV32-NEXT:    sw a4, 16(sp)
+; RV32-NEXT:    sw a5, 20(sp)
+; RV32-NEXT:    sw a6, 24(sp)
+; RV32-NEXT:    sw a7, 28(sp)
+; RV32-NEXT:    addi a0, sp, 12
+; RV32-NEXT:    sw a0, 4(sp)
+; RV32-NEXT:    lw a0, 4(sp)
+; RV32-NEXT:    addi a0, a0, 3
+; RV32-NEXT:    andi a0, a0, -4
+; RV32-NEXT:    addi a3, a0, 4
+; RV32-NEXT:    sw a3, 4(sp)
+; RV32-NEXT:    lw a3, 0(a0)
+; RV32-NEXT:    add a0, a1, a3
+; RV32-NEXT:    sltu a1, a0, a3
+; RV32-NEXT:    add a1, a2, a1
+; RV32-NEXT:    addi sp, sp, 32
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: va3_va_arg:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -64
+; RV64-NEXT:    sd a2, 16(sp)
+; RV64-NEXT:    sd a3, 24(sp)
+; RV64-NEXT:    sd a4, 32(sp)
+; RV64-NEXT:    sd a5, 40(sp)
+; RV64-NEXT:    sd a6, 48(sp)
+; RV64-NEXT:    sd a7, 56(sp)
+; RV64-NEXT:    addi a0, sp, 16
+; RV64-NEXT:    sd a0, 8(sp)
+; RV64-NEXT:    ld a0, 8(sp)
+; RV64-NEXT:    addi a0, a0, 3
+; RV64-NEXT:    andi a0, a0, -4
+; RV64-NEXT:    addi a2, a0, 4
+; RV64-NEXT:    sd a2, 8(sp)
+; RV64-NEXT:    lw a0, 0(a0)
+; RV64-NEXT:    slli a0, a0, 32
+; RV64-NEXT:    srli a0, a0, 32
+; RV64-NEXT:    add a0, a1, a0
+; RV64-NEXT:    addi sp, sp, 64
+; RV64-NEXT:    ret
+  %va = alloca ptr
+  call void @llvm.va_start(ptr %va)
+  %1 = va_arg ptr %va, i32
+  call void @llvm.va_end(ptr %va)
+  %2 = zext i32 %1 to i64
+  %3 = add i64 %b, %2
+  ret i64 %3
+}
+
+define void @va3_caller() nounwind {
+; RV32-LABEL: va3_caller:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT:    lui a0, 5
+; RV32-NEXT:    addi a3, a0, -480
+; RV32-NEXT:    li a0, 2
+; RV32-NEXT:    li a1, 1111
+; RV32-NEXT:    li a2, 0
+; RV32-NEXT:    call va3@plt
+; RV32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: va3_caller:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -16
+; RV64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64-NEXT:    lui a0, 5
+; RV64-NEXT:    addiw a2, a0, -480
+; RV64-NEXT:    li a0, 2
+; RV64-NEXT:    li a1, 1111
+; RV64-NEXT:    call va3@plt
+; RV64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64-NEXT:    addi sp, sp, 16
+; RV64-NEXT:    ret
+ %1 = call i64 (i32, i64, ...) @va3(i32 2, i64 1111, i32 20000)
+ ret void
+}
+
+declare void @llvm.va_copy(ptr, ptr)
+
+define i32 @va4_va_copy(i32 %argno, ...) nounwind {
+; RV32-LABEL: va4_va_copy:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -48
+; RV32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT:    sw s0, 8(sp) # 4-byte Folded Spill
+; RV32-NEXT:    sw a1, 20(sp)
+; RV32-NEXT:    sw a2, 24(sp)
+; RV32-NEXT:    sw a3, 28(sp)
+; RV32-NEXT:    sw a4, 32(sp)
+; RV32-NEXT:    sw a5, 36(sp)
+; RV32-NEXT:    sw a6, 40(sp)
+; RV32-NEXT:    sw a7, 44(sp)
+; RV32-NEXT:    addi a0, sp, 20
+; RV32-NEXT:    sw a0, 4(sp)
+; RV32-NEXT:    lw a0, 4(sp)
+; RV32-NEXT:    addi a0, a0, 3
+; RV32-NEXT:    andi a0, a0, -4
+; RV32-NEXT:    addi a1, a0, 4
+; RV32-NEXT:    sw a1, 4(sp)
+; RV32-NEXT:    lw a1, 4(sp)
+; RV32-NEXT:    mv a2, sp
+; RV32-NEXT:    lw s0, 0(a0)
+; RV32-NEXT:    sw a2, 0(a1)
+; RV32-NEXT:    lw a0, 0(sp)
+; RV32-NEXT:    call notdead@plt
+; RV32-NEXT:    lw a0, 4(sp)
+; RV32-NEXT:    addi a0, a0, 3
+; RV32-NEXT:    andi a0, a0, -4
+; RV32-NEXT:    addi a1, a0, 4
+; RV32-NEXT:    sw a1, 4(sp)
+; RV32-NEXT:    lw a1, 4(sp)
+; RV32-NEXT:    lw a0, 0(a0)
+; RV32-NEXT:    addi a1, a1, 3
+; RV32-NEXT:    andi a1, a1, -4
+; RV32-NEXT:    addi a2, a1, 4
+; RV32-NEXT:    sw a2, 4(sp)
+; RV32-NEXT:    lw a2, 4(sp)
+; RV32-NEXT:    lw a1, 0(a1)
+; RV32-NEXT:    addi a2, a2, 3
+; RV32-NEXT:    andi a2, a2, -4
+; RV32-NEXT:    addi a3, a2, 4
+; RV32-NEXT:    sw a3, 4(sp)
+; RV32-NEXT:    lw a2, 0(a2)
+; RV32-NEXT:    add a0, a0, s0
+; RV32-NEXT:    add a1, a1, a2
+; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-NEXT:    lw s0, 8(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 48
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: va4_va_copy:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -96
+; RV64-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
+; RV64-NEXT:    sd s0, 16(sp) # 8-byte Folded Spill
+; RV64-NEXT:    sd a1, 40(sp)
+; RV64-NEXT:    sd a2, 48(sp)
+; RV64-NEXT:    sd a3, 56(sp)
+; RV64-NEXT:    sd a4, 64(sp)
+; RV64-NEXT:    sd a5, 72(sp)
+; RV64-NEXT:    sd a6, 80(sp)
+; RV64-NEXT:    sd a7, 88(sp)
+; RV64-NEXT:    addi a0, sp, 40
+; RV64-NEXT:    sd a0, 8(sp)
+; RV64-NEXT:    ld a0, 8(sp)
+; RV64-NEXT:    addi a0, a0, 3
+; RV64-NEXT:    andi a0, a0, -4
+; RV64-NEXT:    addi a1, a0, 4
+; RV64-NEXT:    sd a1, 8(sp)
+; RV64-NEXT:    ld a1, 8(sp)
+; RV64-NEXT:    mv a2, sp
+; RV64-NEXT:    lw s0, 0(a0)
+; RV64-NEXT:    sd a2, 0(a1)
+; RV64-NEXT:    lw a0, 4(sp)
+; RV64-NEXT:    lwu a1, 0(sp)
+; RV64-NEXT:    slli a0, a0, 32
+; RV64-NEXT:    or a0, a0, a1
+; RV64-NEXT:    call notdead@plt
+; RV64-NEXT:    ld a0, 8(sp)
+; RV64-NEXT:    addi a0, a0, 3
+; RV64-NEXT:    andi a0, a0, -4
+; RV64-NEXT:    addi a1, a0, 4
+; RV64-NEXT:    sd a1, 8(sp)
+; RV64-NEXT:    ld a1, 8(sp)
+; RV64-NEXT:    lw a0, 0(a0)
+; RV64-NEXT:    addi a1, a1, 3
+; RV64-NEXT:    andi a1, a1, -4
+; RV64-NEXT:    addi a2, a1, 4
+; RV64-NEXT:    sd a2, 8(sp)
+; RV64-NEXT:    ld a2, 8(sp)
+; RV64-NEXT:    lw a1, 0(a1)
+; RV64-NEXT:    addi a2, a2, 3
+; RV64-NEXT:    andi a2, a2, -4
+; RV64-NEXT:    addi a3, a2, 4
+; RV64-NEXT:    sd a3, 8(sp)
+; RV64-NEXT:    lw a2, 0(a2)
+; RV64-NEXT:    add a0, a0, s0
+; RV64-NEXT:    add a1, a1, a2
+; RV64-NEXT:    addw a0, a0, a1
+; RV64-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
+; RV64-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
+; RV64-NEXT:    addi sp, sp, 96
+; RV64-NEXT:    ret
+  %vargs = alloca ptr
+  %wargs = alloca ptr
+  call void @llvm.va_start(ptr %vargs)
+  %1 = va_arg ptr %vargs, i32
+  call void @llvm.va_copy(ptr %wargs, ptr %vargs)
+  %2 = load ptr, ptr %wargs, align 4
+  call void @notdead(ptr %2)
+  %3 = va_arg ptr %vargs, i32
+  %4 = va_arg ptr %vargs, i32
+  %5 = va_arg ptr %vargs, i32
+  call void @llvm.va_end(ptr %vargs)
+  call void @llvm.va_end(ptr %wargs)
+  %add1 = add i32 %3, %1
+  %add2 = add i32 %add1, %4
+  %add3 = add i32 %add2, %5
+  ret i32 %add3
+}
+
+; The va5_aligned_stack_callee and caller function are ommitted from this file
+; since they were not included in the IR lowering test when vararg calls were
+; initially added.
+
+; A function with no fixed arguments is not valid C, but can be
+; specified in LLVM IR. We must ensure the vararg save area is
+; still set up correctly.
+
+define i32 @va6_no_fixed_args(...) nounwind {
+; RV32-LABEL: va6_no_fixed_args:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -48
+; RV32-NEXT:    sw a0, 16(sp)
+; RV32-NEXT:    sw a1, 20(sp)
+; RV32-NEXT:    sw a2, 24(sp)
+; RV32-NEXT:    sw a3, 28(sp)
+; RV32-NEXT:    sw a4, 32(sp)
+; RV32-NEXT:    sw a5, 36(sp)
+; RV32-NEXT:    sw a6, 40(sp)
+; RV32-NEXT:    sw a7, 44(sp)
+; RV32-NEXT:    addi a0, sp, 16
+; RV32-NEXT:    sw a0, 12(sp)
+; RV32-NEXT:    lw a0, 12(sp)
+; RV32-NEXT:    addi a0, a0, 3
+; RV32-NEXT:    andi a0, a0, -4
+; RV32-NEXT:    addi a1, a0, 4
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    lw a0, 0(a0)
+; RV32-NEXT:    addi sp, sp, 48
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: va6_no_fixed_args:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -80
+; RV64-NEXT:    sd a0, 16(sp)
+; RV64-NEXT:    sd a1, 24(sp)
+; RV64-NEXT:    sd a2, 32(sp)
+; RV64-NEXT:    sd a3, 40(sp)
+; RV64-NEXT:    sd a4, 48(sp)
+; RV64-NEXT:    sd a5, 56(sp)
+; RV64-NEXT:    sd a6, 64(sp)
+; RV64-NEXT:    sd a7, 72(sp)
+; RV64-NEXT:    addi a0, sp, 16
+; RV64-NEXT:    sd a0, 8(sp)
+; RV64-NEXT:    ld a0, 8(sp)
+; RV64-NEXT:    addi a0, a0, 3
+; RV64-NEXT:    andi a0, a0, -4
+; RV64-NEXT:    addi a1, a0, 4
+; RV64-NEXT:    sd a1, 8(sp)
+; RV64-NEXT:    lw a0, 0(a0)
+; RV64-NEXT:    addi sp, sp, 80
+; RV64-NEXT:    ret
+  %va = alloca ptr
+  call void @llvm.va_start(ptr %va)
+  %1 = va_arg ptr %va, i32
+  call void @llvm.va_end(ptr %va)
+  ret i32 %1
+}
+
+; TODO: improve constant materialization of stack addresses
+
+define i32 @va_large_stack(ptr %fmt, ...) {
+; RV32-LABEL: va_large_stack:
+; RV32:       # %bb.0:
+; RV32-NEXT:    lui a0, 24414
+; RV32-NEXT:    addi a0, a0, 304
+; RV32-NEXT:    sub sp, sp, a0
+; RV32-NEXT:    .cfi_def_cfa_offset 100000048
+; RV32-NEXT:    lui a0, 24414
+; RV32-NEXT:    add a0, sp, a0
+; RV32-NEXT:    sw a1, 276(a0)
+; RV32-NEXT:    lui a0, 24414
+; RV32-NEXT:    add a0, sp, a0
+; RV32-NEXT:    sw a2, 280(a0)
+; RV32-NEXT:    lui a0, 24414
+; RV32-NEXT:    add a0, sp, a0
+; RV32-NEXT:    sw a3, 284(a0)
+; RV32-NEXT:    lui a0, 24414
+; RV32-NEXT:    add a0, sp, a0
+; RV32-NEXT:    sw a4, 288(a0)
+; RV32-NEXT:    lui a0, 24414
+; RV32-NEXT:    addi a0, a0, 276
+; RV32-NEXT:    add a0, sp, a0
+; RV32-NEXT:    sw a0, 12(sp)
+; RV32-NEXT:    lw a0, 12(sp)
+; RV32-NEXT:    lui a1, 24414
+; RV32-NEXT:    add a1, sp, a1
+; RV32-NEXT:    sw a5, 292(a1)
+; RV32-NEXT:    lui a1, 24414
+; RV32-NEXT:    add a1, sp, a1
+; RV32-NEXT:    sw a6, 296(a1)
+; RV32-NEXT:    lui a1, 24414
+; RV32-NEXT:    add a1, sp, a1
+; RV32-NEXT:    sw a7, 300(a1)
+; RV32-NEXT:    addi a1, a0, 4
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    lw a0, 0(a0)
+; RV32-NEXT:    lui a1, 24414
+; RV32-NEXT:    addi a1, a1, 304
+; RV32-NEXT:    add sp, sp, a1
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: va_large_stack:
+; RV64:       # %bb.0:
+; RV64-NEXT:    lui a0, 24414
+; RV64-NEXT:    addiw a0, a0, 336
+; RV64-NEXT:    sub sp, sp, a0
+; RV64-NEXT:    .cfi_def_cfa_offset 100000080
+; RV64-NEXT:    lui a0, 24414
+; RV64-NEXT:    add a0, sp, a0
+; RV64-NEXT:    sd a1, 280(a0)
+; RV64-NEXT:    lui a0, 24414
+; RV64-NEXT:    add a0, sp, a0
+; RV64-NEXT:    sd a2, 288(a0)
+; RV64-NEXT:    lui a0, 24414
+; RV64-NEXT:    add a0, sp, a0
+; RV64-NEXT:    sd a3, 296(a0)
+; RV64-NEXT:    lui a0, 24414
+; RV64-NEXT:    add a0, sp, a0
+; RV64-NEXT:    sd a4, 304(a0)
+; RV64-NEXT:    lui a0, 24414
+; RV64-NEXT:    add a0, sp, a0
+; RV64-NEXT:    sd a5, 312(a0)
+; RV64-NEXT:    lui a0, 24414
+; RV64-NEXT:    addiw a0, a0, 280
+; RV64-NEXT:    add a0, sp, a0
+; RV64-NEXT:    sd a0, 8(sp)
+; RV64-NEXT:    lw a0, 12(sp)
+; RV64-NEXT:    lwu a1, 8(sp)
+; RV64-NEXT:    lui a2, 24414
+; RV64-NEXT:    add a2, sp, a2
+; RV64-NEXT:    sd a6, 320(a2)
+; RV64-NEXT:    lui a2, 24414
+; RV64-NEXT:    add a2, sp, a2
+; RV64-NEXT:    sd a7, 328(a2)
+; RV64-NEXT:    slli a0, a0, 32
+; RV64-NEXT:    or a0, a0, a1
+; RV64-NEXT:    addi a1, a0, 4
+; RV64-NEXT:    srli a2, a1, 32
+; RV64-NEXT:    sw a1, 8(sp)
+; RV64-NEXT:    sw a2, 12(sp)
+; RV64-NEXT:    lw a0, 0(a0)
+; RV64-NEXT:    lui a1, 24414
+; RV64-NEXT:    addiw a1, a1, 336
+; RV64-NEXT:    add sp, sp, a1
+; RV64-NEXT:    ret
+  %large = alloca [ 100000000 x i8 ]
+  %va = alloca ptr
+  call void @llvm.va_start(ptr %va)
+  %argp.cur = load ptr, ptr %va, align 4
+  %argp.next = getelementptr inbounds i8, ptr %argp.cur, i32 4
+  store ptr %argp.next, ptr %va, align 4
+  %1 = load i32, ptr %argp.cur, align 4
+  call void @llvm.va_end(ptr %va)
+  ret i32 %1
+}
+
+
+