Enable FeatureFlatForGlobal on Volcanic Islands

This switches to the workaround that HSA defaults to
for the mesa path.

This should be applied to the 4.0 branch.

Patch by Vedran Miletić <vedran@miletic.net>

llvm-svn: 292982

3 files changed, 17 insertions, 2 deletions

diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.td b/llvm/lib/Target/AMDGPU/AMDGPU.td
index 0c5f9e72b52..371042d8b4c 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPU.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.td
@@ -305,10 +305,23 @@ def FeatureEnableSIScheduler : SubtargetFeature<"si-scheduler",
   "Enable SI Machine Scheduler"
 >;
 
+def FeatureNoAddr64 : SubtargetFeature<"mubuf-no-addr64",
+  "NoAddr64",
+  "true",
+  "MUBUF instructions have addr64 bit"
+>;
+
+// Unless +-flat-for-global is specified, turn on FlatForGlobal for
+// all OS-es on VI and newer hardware to avoid assertion failures due
+// to missing ADDR64 variants of MUBUF instructions.
+// FIXME: moveToVALU should be able to handle converting addr64 MUBUF
+// instructions.
+
 def FeatureFlatForGlobal : SubtargetFeature<"flat-for-global",
   "FlatForGlobal",
   "true",
-  "Force to generate flat instruction for global"
+  "Force to generate flat instruction for global",
+  [FeatureNoAddr64]
 >;
 
 // Dummy feature used to disable assembler instructions.
@@ -361,7 +374,7 @@ def FeatureVolcanicIslands : SubtargetFeatureGeneration<"VOLCANIC_ISLANDS",
    FeatureGCN3Encoding, FeatureCIInsts, Feature16BitInsts,
    FeatureSMemRealTime, FeatureVGPRIndexMode, FeatureMovrel,
    FeatureScalarStores, FeatureInv2PiInlineImm, FeatureSDWA,
-   FeatureDPP
+   FeatureDPP, FeatureNoAddr64
   ]
 >;
 
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
index 4848b3b86bd..5064d9e8e7a 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
@@ -82,6 +82,7 @@ AMDGPUSubtarget::AMDGPUSubtarget(const Triple &TT, StringRef GPU, StringRef FS,
     FP64FP16Denormals(false),
     FPExceptions(false),
     FlatForGlobal(false),
+    NoAddr64(false),
     UnalignedScratchAccess(false),
     UnalignedBufferAccess(false),
 
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
index c8414f97808..bc43a6851e2 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -85,6 +85,7 @@ protected:
   bool FP64FP16Denormals;
   bool FPExceptions;
   bool FlatForGlobal;
+  bool NoAddr64;
   bool UnalignedScratchAccess;
   bool UnalignedBufferAccess;
   bool EnableXNACK;