R600/SI: Add compute support for CI v2

v2:
  - Fix LDS size calculation

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193621 91177308-0d34-0410-b5e6-96231b3b80d8

7 files changed, 35 insertions, 16 deletions

diff --git a/lib/Target/R600/AMDGPU.td b/lib/Target/R600/AMDGPU.td
index a722f55..f63617f 100644
--- a/lib/Target/R600/AMDGPU.td
+++ b/lib/Target/R600/AMDGPU.td
@@ -87,6 +87,8 @@ def FeatureNorthernIslands : SubtargetFeatureGeneration<"NORTHERN_ISLANDS",
 def FeatureSouthernIslands : SubtargetFeatureGeneration<"SOUTHERN_ISLANDS",
         [Feature64BitPtr, FeatureFP64]>;
 
+def FeatureSeaIslands : SubtargetFeatureGeneration<"SEA_ISLANDS",
+        [Feature64BitPtr, FeatureFP64]>;
 //===----------------------------------------------------------------------===//
 
 def AMDGPUInstrInfo : InstrInfo {
diff --git a/lib/Target/R600/AMDGPUAsmPrinter.cpp b/lib/Target/R600/AMDGPUAsmPrinter.cpp
index 5522a6b..3057790 100644
--- a/lib/Target/R600/AMDGPUAsmPrinter.cpp
+++ b/lib/Target/R600/AMDGPUAsmPrinter.cpp
@@ -167,6 +167,7 @@ void AMDGPUAsmPrinter::EmitProgramInfoR600(MachineFunction &MF) {
 }
 
 void AMDGPUAsmPrinter::EmitProgramInfoSI(MachineFunction &MF) {
+  const AMDGPUSubtarget &STM = TM.getSubtarget<AMDGPUSubtarget>();
   unsigned MaxSGPR = 0;
   unsigned MaxVGPR = 0;
   bool VCCUsed = false;
@@ -267,13 +268,24 @@ void AMDGPUAsmPrinter::EmitProgramInfoSI(MachineFunction &MF) {
   OutStreamer.EmitIntValue(RsrcReg, 4);
   OutStreamer.EmitIntValue(S_00B028_VGPRS(MaxVGPR / 4) | S_00B028_SGPRS(MaxSGPR / 8), 4);
 
+  unsigned LDSAlignShift;
+  if (STM.getGeneration() < AMDGPUSubtarget::SEA_ISLANDS) {
+    // LDS is allocated in 64 dword blocks
+    LDSAlignShift = 8;
+  } else {
+    // LDS is allocated in 128 dword blocks
+    LDSAlignShift = 9;
+  }
+  unsigned LDSBlocks =
+          RoundUpToAlignment(MFI->LDSSize, 1 << LDSAlignShift) >> LDSAlignShift;
+
   if (MFI->ShaderType == ShaderType::COMPUTE) {
     OutStreamer.EmitIntValue(R_00B84C_COMPUTE_PGM_RSRC2, 4);
-    OutStreamer.EmitIntValue(S_00B84C_LDS_SIZE(RoundUpToAlignment(MFI->LDSSize, 256) >> 8), 4);
+    OutStreamer.EmitIntValue(S_00B84C_LDS_SIZE(LDSBlocks), 4);
   }
   if (MFI->ShaderType == ShaderType::PIXEL) {
     OutStreamer.EmitIntValue(R_00B02C_SPI_SHADER_PGM_RSRC2_PS, 4);
-    OutStreamer.EmitIntValue(S_00B02C_EXTRA_LDS_SIZE(RoundUpToAlignment(MFI->LDSSize, 256) >> 8), 4);
+    OutStreamer.EmitIntValue(S_00B02C_EXTRA_LDS_SIZE(LDSBlocks), 4);
     OutStreamer.EmitIntValue(R_0286CC_SPI_PS_INPUT_ENA, 4);
     OutStreamer.EmitIntValue(MFI->PSInputAddr, 4);
   }
diff --git a/lib/Target/R600/AMDGPUCallingConv.td b/lib/Target/R600/AMDGPUCallingConv.td
index 3535e35..fbcdc6a 100644
--- a/lib/Target/R600/AMDGPUCallingConv.td
+++ b/lib/Target/R600/AMDGPUCallingConv.td
@@ -48,7 +48,7 @@ def CC_AMDGPU_Kernel : CallingConv<[
 ]>;
 
 def CC_AMDGPU : CallingConv<[
-  CCIf<"State.getTarget().getSubtarget<AMDGPUSubtarget>().getGeneration() == "
+  CCIf<"State.getTarget().getSubtarget<AMDGPUSubtarget>().getGeneration() >= "
        "AMDGPUSubtarget::SOUTHERN_ISLANDS && "
        "State.getMachineFunction().getInfo<SIMachineFunctionInfo>()->"#
        "ShaderType == ShaderType::COMPUTE", CCDelegateTo<CC_AMDGPU_Kernel>>,
@@ -57,5 +57,5 @@ def CC_AMDGPU : CallingConv<[
        "State.getMachineFunction().getInfo<R600MachineFunctionInfo>()->"
        "ShaderType == ShaderType::COMPUTE", CCDelegateTo<CC_AMDGPU_Kernel>>,
   CCIf<"State.getTarget().getSubtarget<AMDGPUSubtarget>()"#
-       ".getGeneration() == AMDGPUSubtarget::SOUTHERN_ISLANDS", CCDelegateTo<CC_SI>>
+       ".getGeneration() >= AMDGPUSubtarget::SOUTHERN_ISLANDS", CCDelegateTo<CC_SI>>
 ]>;
diff --git a/lib/Target/R600/AMDGPUSubtarget.h b/lib/Target/R600/AMDGPUSubtarget.h
index c5345cc..c08cd6a 100644
--- a/lib/Target/R600/AMDGPUSubtarget.h
+++ b/lib/Target/R600/AMDGPUSubtarget.h
@@ -33,7 +33,8 @@ public:
     R700,
     EVERGREEN,
     NORTHERN_ISLANDS,
-    SOUTHERN_ISLANDS
+    SOUTHERN_ISLANDS,
+    SEA_ISLANDS
   };
 
 private:
diff --git a/lib/Target/R600/Processors.td b/lib/Target/R600/Processors.td
index 4631c04..31079d3 100644
--- a/lib/Target/R600/Processors.td
+++ b/lib/Target/R600/Processors.td
@@ -48,6 +48,6 @@ def : Proc<"pitcairn",   SI_Itin, [FeatureSouthernIslands]>;
 def : Proc<"verde",      SI_Itin, [FeatureSouthernIslands]>;
 def : Proc<"oland",      SI_Itin, [FeatureSouthernIslands]>;
 def : Proc<"hainan",     SI_Itin, [FeatureSouthernIslands]>;
-def : Proc<"bonaire",    SI_Itin, [FeatureSouthernIslands]>;
-def : Proc<"kabini",     SI_Itin, [FeatureSouthernIslands]>;
-def : Proc<"kaveri",     SI_Itin, [FeatureSouthernIslands]>;
+def : Proc<"bonaire",    SI_Itin, [FeatureSeaIslands]>;
+def : Proc<"kabini",     SI_Itin, [FeatureSeaIslands]>;
+def : Proc<"kaveri",     SI_Itin, [FeatureSeaIslands]>;
diff --git a/lib/Target/R600/SIInstructions.td b/lib/Target/R600/SIInstructions.td
index 92a53dd..048c157 100644
--- a/lib/Target/R600/SIInstructions.td
+++ b/lib/Target/R600/SIInstructions.td
@@ -23,7 +23,7 @@ def InterpSlot : Operand<i32> {
 }
 
 def isSI : Predicate<"Subtarget.getGeneration() "
-                      "== AMDGPUSubtarget::SOUTHERN_ISLANDS">;
+                      ">= AMDGPUSubtarget::SOUTHERN_ISLANDS">;
 
 def WAIT_FLAG : InstFlag<"printWaitFlag">;
 
diff --git a/test/CodeGen/R600/local-memory.ll b/test/CodeGen/R600/local-memory.ll
index b0b00bb..2060c7a 100644
--- a/test/CodeGen/R600/local-memory.ll
+++ b/test/CodeGen/R600/local-memory.ll
@@ -1,16 +1,20 @@
 ; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck --check-prefix=EG-CHECK %s
 ; RUN: llc < %s -march=r600 -mcpu=verde -verify-machineinstrs | FileCheck --check-prefix=SI-CHECK %s
+; RUN: llc < %s -march=r600 -mcpu=bonaire -verify-machineinstrs | FileCheck --check-prefix=CI-CHECK %s
 
-@local_memory.local_mem = internal addrspace(3) unnamed_addr global [16 x i32] zeroinitializer, align 4
+@local_memory.local_mem = internal addrspace(3) unnamed_addr global [128 x i32] zeroinitializer, align 4
 
-; EG-CHECK: @local_memory
-; SI-CHECK: @local_memory
+; EG-CHECK-LABEL: @local_memory
+; SI-CHECK-LABEL: @local_memory
+; CI-CHECK-LABEL: @local_memory
 
 ; Check that the LDS size emitted correctly
 ; EG-CHECK: .long 166120
-; EG-CHECK-NEXT: .long 16
+; EG-CHECK-NEXT: .long 128
 ; SI-CHECK: .long 47180
-; SI-CHECK-NEXT: .long 32768
+; SI-CHECK-NEXT: .long 65536
+; CI-CHECK: .long 47180
+; CI-CHECK-NEXT: .long 32768
 
 ; EG-CHECK: LDS_WRITE
 ; SI-CHECK_NOT: S_WQM_B64
@@ -27,13 +31,13 @@
 define void @local_memory(i32 addrspace(1)* %out) {
 entry:
   %y.i = call i32 @llvm.r600.read.tidig.x() #0
-  %arrayidx = getelementptr inbounds [16 x i32] addrspace(3)* @local_memory.local_mem, i32 0, i32 %y.i
+  %arrayidx = getelementptr inbounds [128 x i32] addrspace(3)* @local_memory.local_mem, i32 0, i32 %y.i
   store i32 %y.i, i32 addrspace(3)* %arrayidx, align 4
   %add = add nsw i32 %y.i, 1
   %cmp = icmp eq i32 %add, 16
   %.add = select i1 %cmp, i32 0, i32 %add
   call void @llvm.AMDGPU.barrier.local()
-  %arrayidx1 = getelementptr inbounds [16 x i32] addrspace(3)* @local_memory.local_mem, i32 0, i32 %.add
+  %arrayidx1 = getelementptr inbounds [128 x i32] addrspace(3)* @local_memory.local_mem, i32 0, i32 %.add
   %0 = load i32 addrspace(3)* %arrayidx1, align 4
   %arrayidx2 = getelementptr inbounds i32 addrspace(1)* %out, i32 %y.i
   store i32 %0, i32 addrspace(1)* %arrayidx2, align 4