[AMDGPU] - Add s_bitreplicate intrinsic (#69209)

Add intrinsic for s_bitreplicate. Lower to S_BITREPLICATE_B64_B32
machine instruction in both GISel and Selection DAG.

Support VGPR arguments by inserting a `v_readfirstlane`.

Author: OutOfCache

llvm/include/llvm/IR/IntrinsicsAMDGPU.td

@@ -1927,6 +1927,11 @@ def int_amdgcn_inverse_ballot :
   Intrinsic<[llvm_i1_ty], [llvm_anyint_ty],
             [IntrNoMem, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
 
+// Lowers to S_BITREPLICATE_B64_B32.
+// The argument must be uniform; otherwise, the result is undefined.
+def int_amdgcn_s_bitreplicate :
+  DefaultAttrsIntrinsic<[llvm_i64_ty], [llvm_i32_ty], [IntrNoMem, IntrConvergent]>;
+
 class AMDGPUWaveReduce<LLVMType data_ty = llvm_anyint_ty> : Intrinsic<
     [data_ty],
     [

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

@@ -2994,6 +2994,7 @@ void AMDGPURegisterBankInfo::applyMappingImpl(
       applyMappingBFE(B, OpdMapper, false);
       return;
     case Intrinsic::amdgcn_inverse_ballot:
+    case Intrinsic::amdgcn_s_bitreplicate:
       applyDefaultMapping(OpdMapper);
       constrainOpWithReadfirstlane(B, MI, 2); // Mask
       return;
@@ -4546,6 +4547,11 @@ AMDGPURegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
       OpdsMapping[2] = AMDGPU::getValueMapping(regBankID, OpSize);
       break;
     }
+    case Intrinsic::amdgcn_s_bitreplicate:
+      Register MaskReg = MI.getOperand(2).getReg();
+      unsigned MaskBank = getRegBankID(MaskReg, MRI, AMDGPU::SGPRRegBankID);
+      OpdsMapping[0] = AMDGPU::getValueMapping(AMDGPU::SGPRRegBankID, 64);
+      OpdsMapping[2] = AMDGPU::getValueMapping(MaskBank, 32);
     }
     break;
   }

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

@@ -6473,6 +6473,14 @@ SIInstrInfo::legalizeOperands(MachineInstr &MI,
     return CreatedBB;
   }
 
+  // Legalize S_BITREPLICATE
+  if (MI.getOpcode() == AMDGPU::S_BITREPLICATE_B64_B32) {
+    MachineOperand &Src = MI.getOperand(1);
+    if (Src.isReg() && RI.hasVectorRegisters(MRI.getRegClass(Src.getReg())))
+      Src.setReg(readlaneVGPRToSGPR(Src.getReg(), MI, MRI));
+    return CreatedBB;
+  }
+
   // Legalize MIMG and MUBUF/MTBUF for shaders.
   //
   // Shaders only generate MUBUF/MTBUF instructions via intrinsics or via

llvm/lib/Target/AMDGPU/SOPInstructions.td

@@ -362,7 +362,8 @@ let SubtargetPredicate = isGFX9Plus in {
   } // End hasSideEffects = 1, Defs = [EXEC, SCC], Uses = [EXEC]
 
   let isReMaterializable = 1 in
-  def S_BITREPLICATE_B64_B32 : SOP1_64_32<"s_bitreplicate_b64_b32">;
+  def S_BITREPLICATE_B64_B32 : SOP1_64_32<"s_bitreplicate_b64_b32",
+  [(set i64:$sdst, (int_amdgcn_s_bitreplicate i32:$src0))]>;
 } // End SubtargetPredicate = isGFX9Plus
 
 let SubtargetPredicate = isGFX10Plus in {

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.bitreplicate.ll

@@ -0,0 +1,45 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 -global-isel=1 -verify-machineinstrs < %s | FileCheck  -check-prefixes=GFX11 %s
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 -global-isel=0 -verify-machineinstrs < %s | FileCheck  -check-prefixes=GFX11 %s
+
+declare i64 @llvm.amdgcn.s.bitreplicate(i32)
+
+define i64 @test_s_bitreplicate_constant() {
+; GFX11-LABEL: test_s_bitreplicate_constant:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    s_bitreplicate_b64_b32 s[0:1], 0x85fe3a92
+; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+entry:
+  %br = call i64 @llvm.amdgcn.s.bitreplicate(i32 u0x85FE3A92)
+  ret i64 %br
+}
+
+define amdgpu_cs void @test_s_bitreplicate_sgpr(i32 inreg %mask, ptr addrspace(1) %out) {
+; GFX11-LABEL: test_s_bitreplicate_sgpr:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_bitreplicate_b64_b32 s[0:1], s0
+; GFX11-NEXT:    v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
+; GFX11-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+entry:
+  %br = call i64 @llvm.amdgcn.s.bitreplicate(i32 %mask)
+  store i64 %br, ptr addrspace(1) %out
+  ret void
+}
+
+define i64 @test_s_bitreplicate_vgpr(i32 %mask) {
+; GFX11-LABEL: test_s_bitreplicate_vgpr:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX11-NEXT:    s_bitreplicate_b64_b32 s[0:1], s0
+; GFX11-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+entry:
+  %br = call i64 @llvm.amdgcn.s.bitreplicate(i32 %mask)
+  ret i64 %br
+}