Merging r197503, r197505, r197520:

------------------------------------------------------------------------
r197520 | dexonsmith | 2013-12-17 12:28:21 -0800 (Tue, 17 Dec 2013) | 7
lines

Assert that the last operand is actually EFLAGS

This is another follow-up to r197503, after a post-commit review by
Andy.

<rdar://problem/15627766>

------------------------------------------------------------------------
r197505 | dexonsmith | 2013-12-17 08:20:37 -0800 (Tue, 17 Dec 2013) | 6
lines

Setting the CPU in the new vaargs test

Trying to fix buildbots after r197503 (test passes locally).

<rdar://problem/15627766>

------------------------------------------------------------------------
r197503 | dexonsmith | 2013-12-17 07:54:45 -0800 (Tue, 17 Dec 2013) | 17
lines

Revert "Revert "Mark vastart_save_xmm_regs as changing EFLAGS""

This reverts commit r197481, recommiting r197469 with an extra fix.

The vastart_save_xmm_regs pseudo-instruction expands to a test and a
branch, so it modifies EFLAGS.  Mark it so, or else the scheduler might
place it in the middle of another test+branch.

This fixes a bug exposed by r192750, which changed the initial scheduler
to source-order as part of enabling the MI Scheduler for X86.

This re-commit changes the VASTART_SAVE_XMM_REGS custom inserter not to
try to save %flags, and adds a test that catches the bad behavior of
r197469.

<rdar://problem/15627766>

------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_34@202060 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 90 insertions, 0 deletions

diff --git a/test/CodeGen/X86/vaargs.ll b/test/CodeGen/X86/vaargs.ll
new file mode 100644
index 0000000000..ddeb7a336d
--- /dev/null
+++ b/test/CodeGen/X86/vaargs.ll
@@ -0,0 +1,67 @@
+; RUN: llc -mcpu=corei7-avx %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=NO-FLAGS
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.9.0"
+
+%struct.__va_list_tag = type { i32, i32, i8*, i8* }
+
+; Check that vastart gets the right thing.
+define i32 @sum(i32 %count, ...) nounwind optsize ssp uwtable {
+; CHECK:      testb   %al, %al
+; CHECK-NEXT: je
+; CHECK-NEXT: ## BB#{{[0-9]+}}:
+; CHECK-NEXT: vmovaps %xmm0, 48(%rsp)
+; CHECK-NEXT: vmovaps %xmm1, 64(%rsp)
+; CHECK-NEXT: vmovaps %xmm2, 80(%rsp)
+; CHECK-NEXT: vmovaps %xmm3, 96(%rsp)
+; CHECK-NEXT: vmovaps %xmm4, 112(%rsp)
+; CHECK-NEXT: vmovaps %xmm5, 128(%rsp)
+; CHECK-NEXT: vmovaps %xmm6, 144(%rsp)
+; CHECK-NEXT: vmovaps %xmm7, 160(%rsp)
+
+; Check that [EFLAGS] hasn't been pulled in.
+; NO-FLAGS-NOT: %flags
+
+  %ap = alloca [1 x %struct.__va_list_tag], align 16
+  %1 = bitcast [1 x %struct.__va_list_tag]* %ap to i8*
+  call void @llvm.va_start(i8* %1)
+  %2 = icmp sgt i32 %count, 0
+  br i1 %2, label %.lr.ph, label %._crit_edge
+
+.lr.ph:                                           ; preds = %0
+  %3 = getelementptr inbounds [1 x %struct.__va_list_tag]* %ap, i64 0, i64 0, i32 0
+  %4 = getelementptr inbounds [1 x %struct.__va_list_tag]* %ap, i64 0, i64 0, i32 2
+  %.pre = load i32* %3, align 16
+  br label %5
+
+; <label>:5                                       ; preds = %.lr.ph, %13
+  %6 = phi i32 [ %.pre, %.lr.ph ], [ %14, %13 ]
+  %.01 = phi i32 [ %count, %.lr.ph ], [ %15, %13 ]
+  %7 = icmp ult i32 %6, 41
+  br i1 %7, label %8, label %10
+
+; <label>:8                                       ; preds = %5
+  %9 = add i32 %6, 8
+  store i32 %9, i32* %3, align 16
+  br label %13
+
+; <label>:10                                      ; preds = %5
+  %11 = load i8** %4, align 8
+  %12 = getelementptr i8* %11, i64 8
+  store i8* %12, i8** %4, align 8
+  br label %13
+
+; <label>:13                                      ; preds = %10, %8
+  %14 = phi i32 [ %6, %10 ], [ %9, %8 ]
+  %15 = add nsw i32 %.01, 1
+  %16 = icmp sgt i32 %15, 0
+  br i1 %16, label %5, label %._crit_edge
+
+._crit_edge:                                      ; preds = %13, %0
+  %.0.lcssa = phi i32 [ %count, %0 ], [ %15, %13 ]
+  call void @llvm.va_end(i8* %1)
+  ret i32 %.0.lcssa
+}
+
+declare void @llvm.va_start(i8*) nounwind
+
+declare void @llvm.va_end(i8*) nounwind
diff --git a/test/CodeGen/X86/vastart-defs-eflags.ll b/test/CodeGen/X86/vastart-defs-eflags.ll
new file mode 100644
index 0000000000..6017753fc8
--- /dev/null
+++ b/test/CodeGen/X86/vastart-defs-eflags.ll
@@ -0,0 +1,23 @@
+; RUN: llc %s -o - | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.10.0"
+
+; Check that vastart handling doesn't get between testb and je for the branch.
+define i32 @check_flag(i32 %flags, ...) nounwind {
+entry:
+; CHECK: {{^}} testb $2, %bh
+; CHECK-NOT: test
+; CHECK: {{^}} je
+  %and = and i32 %flags, 512
+  %tobool = icmp eq i32 %and, 0
+  br i1 %tobool, label %if.end, label %if.then
+
+if.then:                                          ; preds = %entry
+  br label %if.end
+
+if.end:                                           ; preds = %entry, %if.then
+  %hasflag = phi i32 [ 1, %if.then ], [ 0, %entry ]
+  ret i32 %hasflag
+}
+