[PATCH] D109707: [HIP] [AlwaysInliner] Disable AlwaysInliner to eliminate undefined symbols

[Anshil Gandhi via Phabricator via cfe-commits]

gandhi21299 updated this revision to Diff 375079.
gandhi21299 added a comment.

- Declare an unhandled call lowering in SelectionDAG when a callee is encountered which cannot be casted into a Function
- I am still investigating the effects on GlobalISel side of things, there seems to be a problem when lowering a call to `@func` in `@kernel` as well.
- inline-calls.ll is expected to fail with this patch, we could turn it into a negative test depending on how the work goes.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D109707/new/

https://reviews.llvm.org/D109707

Files:
  clang/lib/Driver/ToolChains/Clang.cpp
  clang/test/CodeGenCUDA/amdgpu-alias-undef-symbols.cu
  llvm/lib/Target/AMDGPU/AMDGPUAlwaysInlinePass.cpp
  llvm/lib/Target/AMDGPU/SIISelLowering.cpp
  llvm/test/CodeGen/AMDGPU/inline-calls.ll


Index: llvm/test/CodeGen/AMDGPU/inline-calls.ll
===================================================================
--- llvm/test/CodeGen/AMDGPU/inline-calls.ll
+++ llvm/test/CodeGen/AMDGPU/inline-calls.ll
@@ -1,6 +1,4 @@
 ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck  %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck  %s
-; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck %s
 
 ; ALL-NOT: {{^}}func:
 define internal i32 @func(i32 %a) {
@@ -18,8 +16,8 @@
   ret void
 }
 
-; CHECK-NOT: func_alias
-; ALL-NOT: func_alias
+; CHECK: func_alias
+; ALL: func_alias
 @func_alias = alias i32 (i32), i32 (i32)* @func
 
 ; ALL: {{^}}kernel3:
Index: llvm/lib/Target/AMDGPU/SIISelLowering.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -3007,6 +3007,13 @@
   bool IsSibCall = false;
   bool IsThisReturn = false;
   MachineFunction &MF = DAG.getMachineFunction();
+  GlobalAddressSDNode *GSD = dyn_cast<GlobalAddressSDNode>(Callee);
+
+  if (GSD) {
+    const GlobalValue *GV = GSD->getGlobal();
+    if (!isa<Function>(GV))
+      return lowerUnhandledCall(CLI, InVals, "callee is not a function ");
+  }
 
   if (Callee.isUndef() || isNullConstant(Callee)) {
     if (!CLI.IsTailCall) {
@@ -3264,7 +3271,7 @@
   Ops.push_back(Callee);
   // Add a redundant copy of the callee global which will not be legalized, as
   // we need direct access to the callee later.
-  if (GlobalAddressSDNode *GSD = dyn_cast<GlobalAddressSDNode>(Callee)) {
+  if (GSD) {
     const GlobalValue *GV = GSD->getGlobal();
     Ops.push_back(DAG.getTargetGlobalAddress(GV, DL, MVT::i64));
   } else {
Index: llvm/lib/Target/AMDGPU/AMDGPUAlwaysInlinePass.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPUAlwaysInlinePass.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPUAlwaysInlinePass.cpp
@@ -93,6 +93,8 @@
 
   for (GlobalAlias &A : M.aliases()) {
     if (Function* F = dyn_cast<Function>(A.getAliasee())) {
+      if (A.getLinkage() != GlobalValue::InternalLinkage)
+        continue;
       A.replaceAllUsesWith(F);
       AliasesToRemove.push_back(&A);
     }
Index: clang/test/CodeGenCUDA/amdgpu-alias-undef-symbols.cu
===================================================================
--- /dev/null
+++ clang/test/CodeGenCUDA/amdgpu-alias-undef-symbols.cu
@@ -0,0 +1,15 @@
+// RUN: %clang --offload-arch=gfx906 --cuda-device-only -x hip -emit-llvm -S -o - %s \
+// RUN:   -fgpu-rdc -O3 -mllvm -amdgpu-early-inline-all=true -mllvm -amdgpu-function-calls=false | \
+// RUN:   FileCheck %s
+
+#include "Inputs/cuda.h"
+
+// CHECK: %struct.B = type { i8 }
+struct B {
+
+  // CHECK: @_ZN1BC1Ei = hidden unnamed_addr alias void (%struct.B*, i32), void (%struct.B*, i32)* @_ZN1BC2Ei
+  __device__ B(int x);
+};
+
+__device__ B::B(int x) {
+}
Index: clang/lib/Driver/ToolChains/Clang.cpp
===================================================================
--- clang/lib/Driver/ToolChains/Clang.cpp
+++ clang/lib/Driver/ToolChains/Clang.cpp
@@ -5084,9 +5084,9 @@
   }
 
   // Enable -mconstructor-aliases except on darwin, where we have to work around
-  // a linker bug (see <rdar://problem/7651567>), and CUDA/AMDGPU device code,
-  // where aliases aren't supported.
-  if (!RawTriple.isOSDarwin() && !RawTriple.isNVPTX() && !RawTriple.isAMDGPU())
+  // a linker bug (see <rdar://problem/7651567>), and CUDA device code, where
+  // aliases aren't supported.
+  if (!RawTriple.isOSDarwin() && !RawTriple.isNVPTX())
     CmdArgs.push_back("-mconstructor-aliases");
 
   // Darwin's kernel doesn't support guard variables; just die if we


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D109707.375079.patch
Type: text/x-patch
Size: 3762 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20210926/3a35330e/attachment-0001.bin>