[AArch64] [GlobalISel] Fix clobbered callee saved registers with win64 varargs

This fixes a regression since 1c10d5b
/ https://reviews.llvm.org/D130903 by applying the same fix from
SelectionDAG from 8cb3667 /
https://reviews.llvm.org/D35720.

This could possibly have been detected if the existing testcases
in win64_vararg.ll had been tested with GlobalISel too, but all
the IR snippets there fail to be translated with GlobalISel.

This adds a separate testcase based on real world LLVM IR (instead of
hand-reduced IR), which GlobalISel does translate happily - tested
with both SelectionDAG and GlobalISel.

Before this change, the stack object locations (visible in MIR
with "llc -print-after-all") didn't match with what the prologue
emitted by AArch64FrameLowering actually looked like, which caused
clobbered callee saved registers when function local stack objects
aliased the actual location of the callee saved registers.

This fixes llvm#64740.

Differential Revision: https://reviews.llvm.org/D158272

(cherry picked from commit 955d761)

Author: mstorsjo

llvm/lib/Target/AArch64/GISel/AArch64CallLowering.cpp

@@ -564,6 +564,11 @@ void AArch64CallLowering::saveVarArgRegisters(
     if (IsWin64CC) {
       GPRIdx = MFI.CreateFixedObject(GPRSaveSize,
                                      -static_cast<int>(GPRSaveSize), false);
+      if (GPRSaveSize & 15)
+        // The extra size here, if triggered, will always be 8.
+        MFI.CreateFixedObject(16 - (GPRSaveSize & 15),
+                              -static_cast<int>(alignTo(GPRSaveSize, 16)),
+                              false);
     } else
       GPRIdx = MFI.CreateStackObject(GPRSaveSize, Align(8), false);
 

llvm/test/CodeGen/AArch64/win64_vararg2.ll

@@ -0,0 +1,82 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=aarch64-pc-win32 | FileCheck %s
+; RUN: llc < %s -global-isel -mtriple=aarch64-pc-win32 | FileCheck %s --check-prefix=GISEL
+
+; Function Attrs: mustprogress noinline nounwind optnone uwtable
+define i1 @va_func(i32 %a, i8 %b, i8 %c, ...) {
+; CHECK-LABEL: va_func:
+; CHECK:       .seh_proc va_func
+; CHECK-NEXT:  // %bb.0:
+; CHECK-NEXT:    sub sp, sp, #80
+; CHECK-NEXT:    .seh_stackalloc 80
+; CHECK-NEXT:    str x19, [sp, #16] // 8-byte Folded Spill
+; CHECK-NEXT:    .seh_save_reg x19, 16
+; CHECK-NEXT:    str x30, [sp, #24] // 8-byte Folded Spill
+; CHECK-NEXT:    .seh_save_reg x30, 24
+; CHECK-NEXT:    .seh_endprologue
+; CHECK-NEXT:    mov w19, w0
+; CHECK-NEXT:    stp x3, x4, [sp, #40]
+; CHECK-NEXT:    stp x5, x6, [sp, #56]
+; CHECK-NEXT:    str x7, [sp, #72]
+; CHECK-NEXT:    str w0, [sp, #12]
+; CHECK-NEXT:    strb w1, [sp, #11]
+; CHECK-NEXT:    strb w2, [sp, #10]
+; CHECK-NEXT:    bl other
+; CHECK-NEXT:    cmp w19, w0
+; CHECK-NEXT:    cset w0, ls
+; CHECK-NEXT:    .seh_startepilogue
+; CHECK-NEXT:    ldr x30, [sp, #24] // 8-byte Folded Reload
+; CHECK-NEXT:    .seh_save_reg x30, 24
+; CHECK-NEXT:    ldr x19, [sp, #16] // 8-byte Folded Reload
+; CHECK-NEXT:    .seh_save_reg x19, 16
+; CHECK-NEXT:    add sp, sp, #80
+; CHECK-NEXT:    .seh_stackalloc 80
+; CHECK-NEXT:    .seh_endepilogue
+; CHECK-NEXT:    ret
+; CHECK-NEXT:    .seh_endfunclet
+; CHECK-NEXT:    .seh_endproc
+;
+; GISEL-LABEL: va_func:
+; GISEL:       .seh_proc va_func
+; GISEL-NEXT:  // %bb.0:
+; GISEL-NEXT:    sub sp, sp, #80
+; GISEL-NEXT:    .seh_stackalloc 80
+; GISEL-NEXT:    str x19, [sp, #16] // 8-byte Folded Spill
+; GISEL-NEXT:    .seh_save_reg x19, 16
+; GISEL-NEXT:    str x30, [sp, #24] // 8-byte Folded Spill
+; GISEL-NEXT:    .seh_save_reg x30, 24
+; GISEL-NEXT:    .seh_endprologue
+; GISEL-NEXT:    stp x3, x4, [sp, #40]
+; GISEL-NEXT:    mov w19, w0
+; GISEL-NEXT:    stp x5, x6, [sp, #56]
+; GISEL-NEXT:    str w0, [sp, #12]
+; GISEL-NEXT:    str x7, [sp, #72]
+; GISEL-NEXT:    strb w1, [sp, #11]
+; GISEL-NEXT:    strb w2, [sp, #10]
+; GISEL-NEXT:    bl other
+; GISEL-NEXT:    cmp w19, w0
+; GISEL-NEXT:    cset w0, ls
+; GISEL-NEXT:    .seh_startepilogue
+; GISEL-NEXT:    ldr x30, [sp, #24] // 8-byte Folded Reload
+; GISEL-NEXT:    .seh_save_reg x30, 24
+; GISEL-NEXT:    ldr x19, [sp, #16] // 8-byte Folded Reload
+; GISEL-NEXT:    .seh_save_reg x19, 16
+; GISEL-NEXT:    add sp, sp, #80
+; GISEL-NEXT:    .seh_stackalloc 80
+; GISEL-NEXT:    .seh_endepilogue
+; GISEL-NEXT:    ret
+; GISEL-NEXT:    .seh_endfunclet
+; GISEL-NEXT:    .seh_endproc
+  %a_alloc = alloca i32, align 4
+  %b_alloc = alloca i8, align 1
+  %c_alloc = alloca i8, align 1
+  store i32 %a, ptr %a_alloc, align 4
+  store i8 %b, ptr %b_alloc, align 1
+  store i8 %c, ptr %c_alloc, align 1
+  %a_load = load i32, ptr %a_alloc, align 4
+  %ret = call noundef i32 @other()
+  %cmp = icmp ule i32 %a_load, %ret
+  ret i1 %cmp
+}
+
+declare i32 @other()