[PATCH] D20335: [AMDGPU] Emit debugger prologue and emit the rest of the debugger fields in the kernel code header

[Tom Stellard via llvm-commits]

tstellarAMD added inline comments.

================
Comment at: lib/Target/AMDGPU/SIInstructions.td:1904-1920
@@ -1903,2 +1903,19 @@
 
+let mayStore = 1, UseNamedOperandTable = 1, Uses = [EXEC] in {
+  // Used for storing work group ID.
+  def SI_DEBUGGER_STORE_WGID: InstSI <
+    (outs),
+    (ins SReg_32:$src, i32imm:$frame_idx, SReg_128:$scratch_rsrc,
+         SReg_32:$scratch_offset),
+    "", []
+  >;
+  // Used for storing work item ID.
+  def SI_DEBUGGER_STORE_WIID: InstSI <
+    (outs),
+    (ins VGPR_32:$src, i32imm:$frame_idx, SReg_128:$scratch_rsrc,
+         SReg_32:$scratch_offset),
+    "", []
+  >;
+} // End let mayStore = 1, UseNamedOperandTable = 1, Uses = [EXEC]
+
 // For use in patterns
----------------
Why can't we use the existing spill pseudos for this?

================
Comment at: lib/Target/AMDGPU/SIMachineFunctionInfo.cpp:93-103
@@ -90,13 +92,13 @@
 
-  if (F->hasFnAttribute("amdgpu-work-group-id-y"))
+  if (F->hasFnAttribute("amdgpu-work-group-id-y") || ST.debuggerEmitPrologue())
     WorkGroupIDY = true;
 
-  if (F->hasFnAttribute("amdgpu-work-group-id-z"))
+  if (F->hasFnAttribute("amdgpu-work-group-id-z") || ST.debuggerEmitPrologue())
     WorkGroupIDZ = true;
 
-  if (F->hasFnAttribute("amdgpu-work-item-id-y"))
+  if (F->hasFnAttribute("amdgpu-work-item-id-y") || ST.debuggerEmitPrologue())
     WorkItemIDY = true;
 
-  if (F->hasFnAttribute("amdgpu-work-item-id-z"))
+  if (F->hasFnAttribute("amdgpu-work-item-id-z") || ST.debuggerEmitPrologue())
     WorkItemIDZ = true;
 
----------------
I think it would be better to have AMDGPUAnnotateKernelFeatures check ST.debuggerEmitPrologue() and add the attribute in that pass.

================
Comment at: lib/Target/AMDGPU/SIMachineFunctionInfo.cpp:111
@@ -108,3 +110,3 @@
   bool MaySpill = ST.isVGPRSpillingEnabled(*F);
-  bool HasStackObjects = FrameInfo->hasStackObjects();
+  bool HasStackObjects = FrameInfo->hasStackObjects() || ST.debuggerEmitPrologue();
 
----------------
We are creating stack objects during ISel for the debugger prologue, so doesn't that make ST.debuggerEmitPrologue() redundant?


http://reviews.llvm.org/D20335