[AMDGPU] Construct memory clauses before RA

Memory clauses are formed into bundles in presence of xnack.
Their source operands are marked as early-clobber.

This allows to allocate distinct source and destination registers
within a clause and prevent breaking the clause with s_nop in the
hazard recognizer.

Clauses are undone before post-RA scheduler to allow some rescheduling,
which will not break the clause since artificial edges are created in
the dag to keep memory operations together. Yet this allows a better
ILP in some cases.

Differential Revision: https://reviews.llvm.org/D47511

llvm-svn: 333691

1 files changed, 3 insertions, 0 deletions

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
index 74a1de03dda..d361f51bca1 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -171,6 +171,7 @@ extern "C" void LLVMInitializeAMDGPUTarget() {
   initializeSIDebuggerInsertNopsPass(*PR);
   initializeSIOptimizeExecMaskingPass(*PR);
   initializeSIFixWWMLivenessPass(*PR);
+  initializeSIFormMemoryClausesPass(*PR);
   initializeAMDGPUUnifyDivergentExitNodesPass(*PR);
   initializeAMDGPUAAWrapperPassPass(*PR);
   initializeAMDGPUUseNativeCallsPass(*PR);
@@ -847,6 +848,8 @@ void GCNPassConfig::addFastRegAlloc(FunctionPass *RegAllocPass) {
 void GCNPassConfig::addOptimizedRegAlloc(FunctionPass *RegAllocPass) {
   insertPass(&MachineSchedulerID, &SIOptimizeExecMaskingPreRAID);
 
+  insertPass(&SIOptimizeExecMaskingPreRAID, &SIFormMemoryClausesID);
+
   // This must be run immediately after phi elimination and before
   // TwoAddressInstructions, otherwise the processing of the tied operand of
   // SI_ELSE will introduce a copy of the tied operand source after the else.