[WebAssembly] Implement prolog/epilog insertion and FrameIndex elimination

Summary:
Use the SP32 physical register as the base for FrameIndex
lowering. Update it and the __stack_pointer global var in the prolog and
epilog. Extend the mapping of virtual registers to wasm locals to
include the physical registers.

Rather than modify the target-independent PrologEpilogInserter (which
asserts that there are no virtual registers left) include a
slightly-modified copy for Wasm that does not have this assertion and
only clears the virtual registers if scavenging was needed (which of
course it isn't for wasm).

Differential Revision: http://reviews.llvm.org/D15344

llvm-svn: 255392

1 files changed, 70 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/WebAssembly/userstack.ll b/llvm/test/CodeGen/WebAssembly/userstack.ll
new file mode 100644
index 00000000000..27ff1e192e6
--- /dev/null
+++ b/llvm/test/CodeGen/WebAssembly/userstack.ll
@@ -0,0 +1,70 @@
+; RUN: llc < %s -asm-verbose=false | FileCheck %s
+; RUN: llc < %s -asm-verbose=false -fast-isel | FileCheck %s
+
+
+target datalayout = "e-p:32:32-i64:64-n32:64-S128"
+target triple = "wasm32-unknown-unknown"
+
+; CHECK-LABEL: alloca32:
+define void @alloca32() {
+ ; CHECK: i32.const [[L1:.+]]=, __stack_pointer
+ ; CHECK-NEXT: i32.load [[L1]]=, 0([[L1]])
+ ; CHECK-NEXT: i32.const [[L2:.+]]=, 16
+ ; CHECK-NEXT: i32.sub [[SP:.+]]=, [[L1]], [[L2]]
+ %retval = alloca i32
+ ; CHECK: i32.const $push[[L3:.+]]=, 0
+ ; CHECK: i32.store {{.*}}=, 12([[SP]]), $pop[[L3]]
+ store i32 0, i32* %retval
+ ; CHECK: i32.const [[L4:.+]]=, 16
+ ; CHECK-NEXT: i32.add [[SP]]=, [[SP]], [[L4]]
+ ; CHECK-NEXT: i32.const [[L5:.+]]=, __stack_pointer
+ ; CHECK-NEXT: i32.store [[SP]]=, 0([[L5]]), [[SP]]
+ ret void
+}
+
+; CHECK-LABEL: alloca3264:
+define void @alloca3264() {
+ ; CHECK: i32.const [[L1:.+]]=, __stack_pointer
+ ; CHECK-NEXT: i32.load [[L1]]=, 0([[L1]])
+ ; CHECK-NEXT: i32.const [[L2:.+]]=, 16
+ ; CHECK-NEXT: i32.sub [[SP:.+]]=, [[L1]], [[L2]]
+ %r1 = alloca i32
+ %r2 = alloca double
+ ; CHECK: i32.const $push[[L3:.+]]=, 0
+ ; CHECK: i32.store {{.*}}=, 12([[SP]]), $pop[[L3]]
+ store i32 0, i32* %r1
+ ; CHECK: i64.const $push[[L4:.+]]=, 0
+ ; CHECK: i64.store {{.*}}=, 0([[SP]]), $pop[[L4]]
+ store double 0.0, double* %r2
+ ; CHECK: i32.const [[L4:.+]]=, 16
+ ; CHECK-NEXT: i32.add [[SP]]=, [[SP]], [[L4]]
+ ; CHECK-NEXT: i32.const [[L5:.+]]=, __stack_pointer
+ ; CHECK-NEXT: i32.store [[SP]]=, 0([[L5]]), [[SP]]
+ ret void
+}
+
+define void @allocarray() {
+ ; CHECK: i32.const [[L1:.+]]=, __stack_pointer
+ ; CHECK-NEXT: i32.load [[L1]]=, 0([[L1]])
+ ; CHECK-NEXT: i32.const [[L2:.+]]=, 32
+ ; CHECK-NEXT: i32.sub [[SP:.+]]=, [[L1]], [[L2]]
+ %r = alloca [5 x i32]
+ ; CHECK: i32.const $push[[L3:.+]]=, 1
+ ; CHECK: i32.store {{.*}}=, 12([[SP]]), $pop[[L3]]
+ %p = getelementptr [5 x i32], [5 x i32]* %r, i32 0, i32 0
+ store i32 1, i32* %p
+ ; CHECK: i32.const $push[[L4:.+]]=, 4
+ ; CHECK: i32.const [[L5:.+]]=, 12
+ ; CHECK: i32.add [[L5]]=, [[SP]], [[L5]]
+ ; CHECK: i32.add $push[[L6:.+]]=, [[L5]], $pop[[L4]]
+ ; CHECK: i32.store {{.*}}=, 0($pop[[L6]]), ${{.+}}
+ %p2 = getelementptr [5 x i32], [5 x i32]* %r, i32 0, i32 1
+ store i32 1, i32* %p2
+ ; CHECK: i32.const [[L7:.+]]=, 32
+ ; CHECK-NEXT: i32.add [[SP]]=, [[SP]], [[L7]]
+ ; CHECK-NEXT: i32.const [[L8:.+]]=, __stack_pointer
+ ; CHECK-NEXT: i32.store [[SP]]=, 0([[L7]]), [[SP]]
+ ret void
+}
+
+; TODO: test aligned alloc