[WebAssembly] Move register stackification and coloring to a late phase.
Move the register stackification and coloring passes to run very late, after
PEI, tail duplication, and most other passes. This means that all code emitted
and expanded by those passes is now exposed to these passes. This also
eliminates the need for prologue/epilogue code to be manually stackified,
which significantly simplifies the code.
This does require running LiveIntervals a second time. It's useful to think
of these late passes not as late optimization passes, but as a domain-specific
compression algorithm based on knowledge of liveness information. It's used to
compress the code after all conventional optimizations are complete, which is
why it uses LiveIntervals at a phase when actual optimization passes don't
typically need it.
Differential Revision: http://reviews.llvm.org/D20075
llvm-svn: 269012
diff --git a/llvm/test/CodeGen/WebAssembly/mem-intrinsics.ll b/llvm/test/CodeGen/WebAssembly/mem-intrinsics.ll
index f71903c..3a211a5 100644
--- a/llvm/test/CodeGen/WebAssembly/mem-intrinsics.ll
+++ b/llvm/test/CodeGen/WebAssembly/mem-intrinsics.ll
@@ -61,8 +61,8 @@
; CHECK-LABEL: frame_index:
-; CHECK: i32.call $discard=, memset@FUNCTION, $pop12, $pop1, $pop0{{$}}
-; CHECK: i32.call $discard=, memset@FUNCTION, $0, $pop3, $pop2{{$}}
+; CHECK: i32.call $discard=, memset@FUNCTION, $pop{{[0-9]+}}, $pop{{[0-9]+}}, $pop{{[0-9]+}}{{$}}
+; CHECK: i32.call $push{{[0-9]+}}=, memset@FUNCTION, ${{[0-9]+}}, $pop{{[0-9]+}}, $pop{{[0-9]+}}{{$}}
; CHECK: return{{$}}
define void @frame_index() {
entry:
@@ -76,11 +76,13 @@
}
; If the result value of memset doesn't get stackified, it should be marked
-; $discard.
+; $discard. Note that we use a call to prevent tail dup so that we can test
+; this specific functionality.
; CHECK-LABEL: discard_result:
; CHECK: i32.call $discard=, memset@FUNCTION, $0, $1, $2
declare i8* @def()
+declare void @block_tail_dup()
define i8* @discard_result(i8* %arg, i8 %arg1, i32 %arg2, i32 %arg3, i32 %arg4) {
bb:
%tmp = icmp eq i32 %arg3, 0
@@ -103,5 +105,36 @@
bb11:
%tmp12 = phi i8* [ %arg, %bb7 ], [ %arg, %bb8 ], [ %tmp10, %bb9 ]
+ call void @block_tail_dup()
+ ret i8* %tmp12
+}
+
+; This is the same as discard_result, except we let tail dup happen, so the
+; result of the memset *is* stackified.
+
+; CHECK-LABEL: tail_dup_to_reuse_result:
+; CHECK: i32.call $push{{[0-9]+}}=, memset@FUNCTION, $0, $1, $2
+define i8* @tail_dup_to_reuse_result(i8* %arg, i8 %arg1, i32 %arg2, i32 %arg3, i32 %arg4) {
+bb:
+ %tmp = icmp eq i32 %arg3, 0
+ br i1 %tmp, label %bb5, label %bb9
+
+bb5:
+ %tmp6 = icmp eq i32 %arg4, 0
+ br i1 %tmp6, label %bb7, label %bb8
+
+bb7:
+ call void @llvm.memset.p0i8.i32(i8* %arg, i8 %arg1, i32 %arg2, i32 1, i1 false)
+ br label %bb11
+
+bb8:
+ br label %bb11
+
+bb9:
+ %tmp10 = call i8* @def()
+ br label %bb11
+
+bb11:
+ %tmp12 = phi i8* [ %arg, %bb7 ], [ %arg, %bb8 ], [ %tmp10, %bb9 ]
ret i8* %tmp12
}