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

[Konstantin Zhuravlyov via llvm-commits]

kzhuravl added inline comments.

================
Comment at: lib/Target/AMDGPU/AMDGPUAsmPrinter.h:80-86
@@ -77,1 +79,9 @@
 
+    // Fixed SGPR number used to hold wave scratch offset for entire kernel
+    // execution, or uint16_t(-1) if the register is not used or not known.
+    uint16_t DebuggerWavefrontPrivateSegmentOffsetSGPR;
+    // Fixed SGPR number of the first 4 SGPRs used to hold scratch V# for entire
+    // kernel execution, or uint16_t(-1) if the register is not used or not
+    // known.
+    uint16_t DebuggerPrivateSegmentBufferSGPR;
+
----------------
Because we have to be compliant with our existing spec, which says to put SGPR number or (uint16_t)-1.

================
Comment at: lib/Target/AMDGPU/SIISelLowering.cpp:603
@@ +602,3 @@
+      // Create fixed stack object for work group ID.
+      ObjectIdx = MF.getFrameInfo()->CreateFixedObject(4, i * 4, true);
+      Info->setDebuggerWorkGroupIDStackObjectIndex(i, ObjectIdx);
----------------
Is there any advantage to having a single object?

The reason for having multiple objects is so we can use storeRegToStackSlot, which creates spills.


http://reviews.llvm.org/D20335