[llvm] r190574 - R600: Don't use trans slot for instructions that read LDS source registers

[Tom Stellard]

Author: tstellar
Date: Wed Sep 11 21:55:06 2013
New Revision: 190574

URL: http://llvm.org/viewvc/llvm-project?rev=190574&view=rev
Log:
R600: Don't use trans slot for instructions that read LDS source registers

This fixes some regressions in the piglit local memory store tests
introduced by recent commits which made the scheduler aware of the trans
slot.

It's not possible to test this using lit, because there is no way to
determine from the assembly dumps whether or not an instruction is in
the trans slot.

Even if this were possible, the test would be highly sensitive to
changes in the scheduler and might generate confusing false negatives.

Reviewed-by: Vincent Lejeune<vljn at ovi.com>

Modified:
    llvm/trunk/lib/Target/R600/R600InstrInfo.cpp
    llvm/trunk/lib/Target/R600/R600InstrInfo.h
    llvm/trunk/lib/Target/R600/R600MachineScheduler.cpp
    llvm/trunk/lib/Target/R600/R600Packetizer.cpp
    llvm/trunk/lib/Target/R600/R600RegisterInfo.td

Modified: llvm/trunk/lib/Target/R600/R600InstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/R600/R600InstrInfo.cpp?rev=190574&r1=190573&r2=190574&view=diff
==============================================================================
--- llvm/trunk/lib/Target/R600/R600InstrInfo.cpp (original)
+++ llvm/trunk/lib/Target/R600/R600InstrInfo.cpp Wed Sep 11 21:55:06 2013
@@ -204,6 +204,22 @@ bool R600InstrInfo::mustBeLastInClause(u
   }
 }
 
+bool R600InstrInfo::readsLDSSrcReg(const MachineInstr *MI) const {
+  if (!isALUInstr(MI->getOpcode())) {
+    return false;
+  }
+  for (MachineInstr::const_mop_iterator I = MI->operands_begin(),
+                                        E = MI->operands_end(); I != E; ++I) {
+    if (!I->isReg() || !I->isUse() ||
+        TargetRegisterInfo::isVirtualRegister(I->getReg()))
+      continue;
+
+    if (AMDGPU::R600_LDS_SRC_REGRegClass.contains(I->getReg()))
+      return true;
+  }
+  return false;
+}
+
 int R600InstrInfo::getSrcIdx(unsigned Opcode, unsigned SrcNum) const {
   static const unsigned OpTable[] = {
     AMDGPU::OpName::src0,

Modified: llvm/trunk/lib/Target/R600/R600InstrInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/R600/R600InstrInfo.h?rev=190574&r1=190573&r2=190574&view=diff
==============================================================================
--- llvm/trunk/lib/Target/R600/R600InstrInfo.h (original)
+++ llvm/trunk/lib/Target/R600/R600InstrInfo.h Wed Sep 11 21:55:06 2013
@@ -78,6 +78,7 @@ namespace llvm {
   bool usesTextureCache(const MachineInstr *MI) const;
 
   bool mustBeLastInClause(unsigned Opcode) const;
+  bool readsLDSSrcReg(const MachineInstr *MI) const;
 
   /// \returns The operand index for the given source number.  Legal values
   /// for SrcNum are 0, 1, and 2.

Modified: llvm/trunk/lib/Target/R600/R600MachineScheduler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/R600/R600MachineScheduler.cpp?rev=190574&r1=190573&r2=190574&view=diff
==============================================================================
--- llvm/trunk/lib/Target/R600/R600MachineScheduler.cpp (original)
+++ llvm/trunk/lib/Target/R600/R600MachineScheduler.cpp Wed Sep 11 21:55:06 2013
@@ -314,6 +314,10 @@ R600SchedStrategy::AluKind R600SchedStra
     if (regBelongsToClass(DestReg, &AMDGPU::R600_Reg128RegClass))
       return AluT_XYZW;
 
+    // LDS src registers cannot be used in the Trans slot.
+    if (TII->readsLDSSrcReg(MI))
+      return AluT_XYZW;
+
     return AluAny;
 
 }

Modified: llvm/trunk/lib/Target/R600/R600Packetizer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/R600/R600Packetizer.cpp?rev=190574&r1=190573&r2=190574&view=diff
==============================================================================
--- llvm/trunk/lib/Target/R600/R600Packetizer.cpp (original)
+++ llvm/trunk/lib/Target/R600/R600Packetizer.cpp Wed Sep 11 21:55:06 2013
@@ -272,6 +272,10 @@ public:
       return false;
     }
 
+    // We cannot read LDS source registrs from the Trans slot.
+    if (isTransSlot && TII->readsLDSSrcReg(MI))
+      return false;
+
     CurrentPacketMIs.pop_back();
     return true;
   }

Modified: llvm/trunk/lib/Target/R600/R600RegisterInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/R600/R600RegisterInfo.td?rev=190574&r1=190573&r2=190574&view=diff
==============================================================================
--- llvm/trunk/lib/Target/R600/R600RegisterInfo.td (original)
+++ llvm/trunk/lib/Target/R600/R600RegisterInfo.td Wed Sep 11 21:55:06 2013
@@ -95,6 +95,12 @@ foreach Index = 448-480 in {
 
 // Special Registers
 
+def OQA : R600Reg<"OQA", 219>;
+def OQB : R600Reg<"OQB", 220>;
+def OQAP : R600Reg<"OQAP", 221>;
+def OQBP : R600Reg<"OQAP", 222>;
+def LDS_DIRECT_A : R600Reg<"LDS_DIRECT_A", 223>;
+def LDS_DIRECT_B : R600Reg<"LDS_DIRECT_B", 224>;
 def ZERO : R600Reg<"0.0", 248>;
 def ONE : R600Reg<"1.0", 249>;
 def NEG_ONE : R600Reg<"-1.0", 249>;
@@ -115,7 +121,6 @@ def PRED_SEL_OFF: R600Reg<"Pred_sel_off"
 def PRED_SEL_ZERO : R600Reg<"Pred_sel_zero", 2>;
 def PRED_SEL_ONE : R600Reg<"Pred_sel_one", 3>;
 def AR_X : R600Reg<"AR.x", 0>;
-def OQAP : R600Reg<"OQAP", 221>;
 
 def R600_ArrayBase : RegisterClass <"AMDGPU", [f32, i32], 32,
                           (add (sequence "ArrayBase%u", 448, 480))>;
@@ -130,6 +135,9 @@ let isAllocatable = 0 in {
 // XXX: Only use the X channel, until we support wider stack widths
 def R600_Addr : RegisterClass <"AMDGPU", [i32], 127, (add (sequence "Addr%u_X", 0, 127))>;
 
+def R600_LDS_SRC_REG : RegisterClass<"AMDGPU", [i32], 32,
+  (add OQA, OQB, OQAP, OQBP, LDS_DIRECT_A, LDS_DIRECT_B)>;
+
 } // End isAllocatable = 0
 
 def R600_KC0_X : RegisterClass <"AMDGPU", [f32, i32], 32,