[AMDGPU] Fix +DumpCode to print an entry label for the first function

[foad]

Summary

The +DumpCode attribute is a horrible hack in AMDGPU to embed the
disassembly of the generated code into the elf file. It is used by LLPC
to implement an extension that allows the application to read back the
disassembly of the code.

It tries to print an entry label at the start of every function, but
that didn't work for the first function in the module because
DumpCodeInstEmitter wasn't initialised until EmitFunctionBodyStart
which is too late.

Change-Id: I790d73ddf4f51fd02ab32529380c7cb7c607c4ee

[arsenm]

I'm not a huge fan of new patches for DumpCode. Can we finally just delete it? It shouldn't be difficult for users to switch to using the proper assembler/disassembler

[tpr]

I don't think anyone is a fan of dumpcode. But we're still in the position that the proper disassembler does not support gfx6 or gfx7, and we need to get this particular problem fixed in the short term.

[arsenm]

In D63712#1555460, @tpr wrote:

I don't think anyone is a fan of dumpcode. But we're still in the position that the proper disassembler does not support gfx6 or gfx7, and we need to get this particular problem fixed in the short term.

You should go the other way, emit text asm and then assemble

[tpr]

Possibly, but our majority use case is to not need the disassembly, and our thinking is that we don't want to change the tool flow between the majority case and the minority case. So can we get this fix in please?

[arsenm]

LGTM, but I really don't like how this keeps getting put off

llvm/trunk/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp

if (!STM.hasCodeObjectV3() && STM.isAmdHsaOrMesa(F) &&
        F.getCallingConv() == CallingConv::SPIR_KERNEL)) {
     amd_kernel_code_t KernelCode;
     getAmdKernelCode(KernelCode, CurrentProgramInfo, *MF);
     getTargetStreamer()->EmitAMDKernelCodeT(KernelCode);
   }
 
   if (STM.isAmdHsaOS())
     HSAMetadataStream->emitKernel(*MF, CurrentProgramInfo);
-
-  DumpCodeInstEmitter = nullptr;
-  if (STM.dumpCode()) {
-    // For -dumpcode, get the assembler out of the streamer, even if it does
-    // not really want to let us have it. This only works with -filetype=obj.
-    bool SaveFlag = OutStreamer->getUseAssemblerInfoForParsing();
-    OutStreamer->setUseAssemblerInfoForParsing(true);
-    MCAssembler *Assembler = OutStreamer->getAssemblerPtr();
-    OutStreamer->setUseAssemblerInfoForParsing(SaveFlag);
-    if (Assembler)
-      DumpCodeInstEmitter = Assembler->getEmitterPtr();
-  }
 }
 
 void AMDGPUAsmPrinter::EmitFunctionBodyEnd() {
   const SIMachineFunctionInfo &MFI = *MF->getInfo<SIMachineFunctionInfo>();
   if (!MFI.isEntryFunction())
     return;
 
   if (!IsaInfo::hasCodeObjectV3(getGlobalSTI()) ||
bool AMDGPUAsmPrinter::runOnMachineFunction(MachineFunction &MF) {
   }
 
   if (STM.isAmdPalOS())
     EmitPALMetadata(MF, CurrentProgramInfo);
   else if (!STM.isAmdHsaOS()) {
     EmitProgramInfoSI(MF, CurrentProgramInfo);
   }
 
+  DumpCodeInstEmitter = nullptr;
+  if (STM.dumpCode()) {
+    // For -dumpcode, get the assembler out of the streamer, even if it does
+    // not really want to let us have it. This only works with -filetype=obj.
+    bool SaveFlag = OutStreamer->getUseAssemblerInfoForParsing();
+    OutStreamer->setUseAssemblerInfoForParsing(true);
+    MCAssembler *Assembler = OutStreamer->getAssemblerPtr();
+    OutStreamer->setUseAssemblerInfoForParsing(SaveFlag);
+    if (Assembler)
+      DumpCodeInstEmitter = Assembler->getEmitterPtr();
+  }
+
   DisasmLines.clear();
   HexLines.clear();
   DisasmLineMaxLen = 0;
 
   EmitFunctionBody();
 
   if (isVerbose()) {
     MCSectionELF *CommentSection =

llvm/trunk/test/CodeGen/AMDGPU/debug.ll

 ; RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs -mattr=dumpcode -filetype=obj | FileCheck --check-prefix=SI --check-prefix=FUNC %s
 ; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs -mattr=dumpcode -filetype=obj | FileCheck --check-prefix=SI --check-prefix=FUNC %s
 
 ; Test for a crash in the custom assembly dump code.
 
+; SI: test:
+; SI: BB0_0:
 ; SI: s_endpgm
 define amdgpu_kernel void @test(i32 addrspace(1)* %out) {
   store i32 0, i32 addrspace(1)* %out
   ret void
 }