[AMDGPU] Fix CS scratch setup on pre-GCN3 ASICs

[critson]

Summary

Prior to GCN3 s_load_dword offsets are in dwords rather than bytes.
Thus the scratch buffer descriptor offset must be adjusted for pre-GCN3 ASICs.

[critson]

Ping

[nhaehnle]

LGTM

lib/Target/AMDGPU/SIFrameLowering.cpp

if (ST.isAmdPalOS()) {
     MachinePointerInfo PtrInfo(UndefValue::get(PtrTy));
     const MCInstrDesc &LoadDwordX4 = TII->get(AMDGPU::S_LOAD_DWORDX4_IMM);
     auto MMO = MF.getMachineMemOperand(PtrInfo,
                                        MachineMemOperand::MOLoad |
                                        MachineMemOperand::MOInvariant |
                                        MachineMemOperand::MODereferenceable,
                                        0, 0);
     unsigned Offset = Fn.getCallingConv() == CallingConv::AMDGPU_CS ? 16 : 0;
+    const GCNSubtarget &Subtarget = MF.getSubtarget<GCNSubtarget>();
+    unsigned EncodedOffset = AMDGPU::getSMRDEncodedOffset(Subtarget, Offset);
     BuildMI(MBB, I, DL, LoadDwordX4, ScratchRsrcReg)
       .addReg(Rsrc01)
-      .addImm(Offset) // offset
+      .addImm(EncodedOffset) // offset
       .addImm(0) // glc
       .addReg(ScratchRsrcReg, RegState::ImplicitDefine)
       .addMemOperand(MMO);
     return;
   }
   if (ST.isMesaGfxShader(Fn)
       || (PreloadedPrivateBufferReg == AMDGPU::NoRegister)) {
     assert(!ST.isAmdHsaOrMesa(Fn));

test/CodeGen/AMDGPU/amdpal.ll

-; RUN: llc < %s -mtriple=amdgcn--amdpal -mcpu=tahiti | FileCheck --check-prefix=PAL --enable-var-scope %s
+; RUN: llc < %s -mtriple=amdgcn--amdpal -mcpu=tahiti | FileCheck --check-prefixes=PAL,CI --enable-var-scope %s
+; RUN: llc < %s -mtriple=amdgcn--amdpal -mcpu=tonga | FileCheck --check-prefixes=PAL,VI --enable-var-scope %s
 
 ; PAL-NOT: .AMDGPU.config
 ; PAL-LABEL: {{^}}simple:
 define amdgpu_kernel void @simple(i32 addrspace(1)* %out) {
 entry:
   store i32 0, i32 addrspace(1)* %out
   ret void
 }
entry:
   %x = load i32, i32 addrspace(5)* %e
   store i32 %x, i32 addrspace(5)* %out
   ret void
 }
 
 ; Check code sequence for amdpal use of scratch for alloca in a compute shader.
 ; The scratch descriptor is loaded from offset 0x10 of the GIT, rather than offset
 ; 0 in a graphics shader.
+; Prior to GCN3 s_load_dword offsets are dwords, so the offset will be 0x4.
 
 ; PAL-LABEL: {{^}}scratch2_cs:
 ; PAL: s_movk_i32 s{{[0-9]+}}, 0x1234
 ; PAL: s_mov_b32 s[[GITPTR:[0-9]+]], s0
-; PAL: s_load_dwordx4 s{{\[}}[[SCRATCHDESC:[0-9]+]]:{{[0-9]+]}}, s{{\[}}[[GITPTR]]:{{[0-9]+\]}}, 0x10
+; CI: s_load_dwordx4 s{{\[}}[[SCRATCHDESC:[0-9]+]]:{{[0-9]+]}}, s{{\[}}[[GITPTR]]:{{[0-9]+\]}}, 0x4
+; VI: s_load_dwordx4 s{{\[}}[[SCRATCHDESC:[0-9]+]]:{{[0-9]+]}}, s{{\[}}[[GITPTR]]:{{[0-9]+\]}}, 0x10
 ; PAL: buffer_store{{.*}}, s{{\[}}[[SCRATCHDESC]]:
 
 define amdgpu_cs void @scratch2_cs(i32 inreg, i32 inreg, i32 inreg, <3 x i32> inreg, i32 inreg, <3 x i32> %coord, <2 x i32> %in, i32 %extra, i32 %idx) #0 {
 entry:
   %v = alloca [3 x i32], addrspace(5)
   %v0 = getelementptr [3 x i32], [3 x i32] addrspace(5)* %v, i32 0, i32 0
   %v1 = getelementptr [3 x i32], [3 x i32] addrspace(5)* %v, i32 0, i32 1
   store i32 %extra, i32 addrspace(5)* %v0