[CIR][FlattenCFG] Let results of CIR ScopeOp forwarded during Flatten…

…CFG (#1147) This PR implements NYI in CIRScopeOpFlattening. It seems to me the best way is to let results of ScopeOp forwarded as block arguments of the last block split from the cir.scope block.
llvm · Nov 23, 2024 · a04cf10 · a04cf10
1 parent b0fdf3d
commit a04cf10
Show file tree

Hide file tree

Showing 2 changed files with 36 additions and 6 deletions.
diff --git a/clang/lib/CIR/Dialect/Transforms/FlattenCFG.cpp b/clang/lib/CIR/Dialect/Transforms/FlattenCFG.cpp
@@ -135,13 +135,10 @@ class CIRScopeOpFlattening : public mlir::OpRewritePattern<cir::ScopeOp> {
     // Split the current block before the ScopeOp to create the inlining
     // point.
     auto *currentBlock = rewriter.getInsertionBlock();
-    auto *remainingOpsBlock =
+    mlir::Block *continueBlock =
         rewriter.splitBlock(currentBlock, rewriter.getInsertionPoint());
-    mlir::Block *continueBlock;
-    if (scopeOp.getNumResults() == 0)
-      continueBlock = remainingOpsBlock;
-    else
-      llvm_unreachable("NYI");
+    if (scopeOp.getNumResults() > 0)
+      continueBlock->addArguments(scopeOp.getResultTypes(), loc);
 
     // Inline body region.
     auto *beforeBody = &scopeOp.getRegion().front();

diff --git a/clang/test/CIR/CodeGen/fullexpr.cpp b/clang/test/CIR/CodeGen/fullexpr.cpp
@@ -1,5 +1,10 @@
 // RUN: %clang_cc1 -std=c++17 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir %s -o %t.cir
 // RUN: FileCheck --input-file=%t.cir %s
+// RUN: %clang_cc1 -std=c++17 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir-flat %s -o %t.cir.flat
+// RUN: FileCheck --check-prefix=FLAT  --input-file=%t.cir.flat %s
+// RUN: %clang_cc1 -std=c++17 -triple x86_64-unknown-linux-gnu -fclangir -emit-llvm -o - %s \
+// RUN: | opt -S -passes=instcombine,mem2reg,simplifycfg -o %t.ll
+// RUN: FileCheck --check-prefix=LLVM --input-file=%t.ll %s
 
 int go(int const& val);
 
@@ -18,3 +23,31 @@ int go1() {
 // CHECK-NEXT:   cir.yield %[[#RValTmp]] : !s32i
 // CHECK-NEXT: }
 // CHECK-NEXT: cir.store %[[#RVal]], %[[#XAddr]] : !s32i, !cir.ptr<!s32i>
+
+// FLAT: cir.func @_Z3go1v() -> !s32i
+// FLAT: %[[#TmpAddr:]] = cir.alloca !s32i, !cir.ptr<!s32i>, ["ref.tmp0", init] {alignment = 4 : i64}
+// FLAT: %[[#XAddr:]] = cir.alloca !s32i, !cir.ptr<!s32i>, ["x", init] {alignment = 4 : i64}
+// FLAT: cir.br ^[[before_body:.*]]{{ loc.*}}
+// FLAT-NEXT: ^[[before_body]]:  // pred: ^bb0
+// FLAT-NEXT:   %[[#One:]] = cir.const #cir.int<1> : !s32i
+// FLAT-NEXT:   cir.store %[[#One]], %[[#TmpAddr]] : !s32i, !cir.ptr<!s32i>
+// FLAT-NEXT:   %[[#RValTmp:]] = cir.call @_Z2goRKi(%[[#TmpAddr]]) : (!cir.ptr<!s32i>) -> !s32i
+// FLAT-NEXT:   cir.br ^[[continue_block:.*]](%[[#RValTmp]] : !s32i) {{loc.*}}
+// FLAT-NEXT: ^[[continue_block]](%[[#BlkArgRval:]]: !s32i {{loc.*}}):  // pred: ^[[before_body]]
+// FLAT-NEXT:   cir.store %[[#BlkArgRval]], %[[#XAddr]] : !s32i, !cir.ptr<!s32i>
+
+// LLVM-LABEL: @_Z3go1v()
+// LLVM-NEXT: %[[#TmpAddr:]] = alloca i32, i64 1, align 4
+// LLVM: br label %[[before_body:[0-9]+]]
+// LLVM: [[before_body]]:
+// LLVM-NEXT: store i32 1, ptr %[[#TmpAddr]], align 4
+// LLVM-NEXT: %[[#RValTmp:]] = call i32 @_Z2goRKi(ptr %[[#TmpAddr]])
+// LLVM-NEXT: br label %[[continue_block:[0-9]+]] 
+
+// LLVM: [[continue_block]]:
+// LLVM-NEXT: [[PHI:%.*]] = phi i32 [ %[[#RValTmp]], %[[before_body]] ]
+// LLVM: store i32 [[PHI]], ptr [[TMP0:%.*]], align 4
+// LLVM: [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
+// LLVM: store i32 [[TMP1]], ptr [[TMP2:%.*]], align 4
+// LLVM: [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
+// LLVM: ret i32 [[TMP3]]