================
@@ -66,9 +81,232 @@ FunctionPass *llvm::createAMDGPURegBankSelectPass() {
return new AMDGPURegBankSelect();
}
+class RegBankSelectHelper {
+ MachineIRBuilder &B;
+ MachineRegisterInfo &MRI;
+ AMDGPU::IntrinsicLaneMaskAnalyzer &ILMA;
+ const MachineUniformityInfo &MUI;
+ const SIRegisterInfo &TRI;
+ const RegisterBank *SgprRB;
+ const RegisterBank *VgprRB;
+ const RegisterBank *VccRB;
+
+public:
+ RegBankSelectHelper(MachineIRBuilder &B,
+ AMDGPU::IntrinsicLaneMaskAnalyzer &ILMA,
+ const MachineUniformityInfo &MUI,
+ const SIRegisterInfo &TRI, const RegisterBankInfo &RBI)
+ : B(B), MRI(*B.getMRI()), ILMA(ILMA), MUI(MUI), TRI(TRI),
+ SgprRB(&RBI.getRegBank(AMDGPU::SGPRRegBankID)),
+ VgprRB(&RBI.getRegBank(AMDGPU::VGPRRegBankID)),
+ VccRB(&RBI.getRegBank(AMDGPU::VCCRegBankID)) {}
+
+ bool shouldRegBankSelect(MachineInstr &MI) {
+ return MI.isPreISelOpcode() || MI.isCopy();
+ }
+
+ // Temporal divergence copy: COPY to vgpr with implicit use of $exec inside
of
+ // the cycle
+ // Note: uniformity analysis does not consider that registers with vgpr def
+ // are divergent (you can have uniform value in vgpr).
+ // - TODO: implicit use of $exec could be implemented as indicator that
+ // instruction is divergent
+ bool isTemporalDivergenceCopy(Register Reg) {
+ MachineInstr *MI = MRI.getVRegDef(Reg);
+ if (!MI->isCopy())
+ return false;
+
+ for (auto Op : MI->implicit_operands()) {
----------------
petar-avramovic wrote:
There is nothing useful in real operands, it is a simple sgpr to vgpr COPY
which is quite common. What makes that COPY special is implicit exec, for
example:
`%21:vgpr_32(s32) = COPY %8:sgpr(s32), implicit $exec_lo`
implicit exec is there to stop other passes from moving the COPY outside of the
loop.
https://github.com/llvm/llvm-project/pull/112863
_______________________________________________
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
