Lines Matching defs:Kernel

151 static const omp::GV &getGridValue(const Triple &T, Function *Kernel) {
154         Kernel->getFnAttribute("target-features").getValueAsString();
6016   Function *Kernel = Builder.GetInsertBlock()->getParent();
6021     writeTeamsForKernel(T, *Kernel, MinTeamsVal, MaxTeamsVal);
6026         int32_t(getGridValue(T, Kernel).GV_Default_WG_Size), MinThreadsVal);
6029     writeThreadBoundsForKernel(T, *Kernel, MinThreadsVal, MaxThreadsVal);
6039   StringRef KernelName = Kernel->getName();
6096   Value *KernelLaunchEnvironment = Kernel->getArg(0);
6145   Function *Kernel = Builder.GetInsertBlock()->getParent();
6147   StringRef KernelName = Kernel->getName();
6164 static MDNode *getNVPTXMDNode(Function &Kernel, StringRef Name) {
6165   Module &M = *Kernel.getParent();
6171     if (!KernelOp || KernelOp->getValue() != &Kernel)
6181 static void updateNVPTXMetadata(Function &Kernel, StringRef Name, int32_t Value,
6184   MDNode *ExistingOp = getNVPTXMDNode(Kernel, Name);
6193     LLVMContext &Ctx = Kernel.getContext();
6194     Metadata *MDVals[] = {ConstantAsMetadata::get(&Kernel),
6199     Module &M = *Kernel.getParent();
6206 OpenMPIRBuilder::readThreadBoundsForKernel(const Triple &T, Function &Kernel) {
6208       Kernel.getFnAttributeAsParsedInteger("omp_target_thread_limit");
6211     const auto &Attr = Kernel.getFnAttribute("amdgpu-flat-work-group-size");
6224   if (MDNode *ExistingOp = getNVPTXMDNode(Kernel, "maxntidx")) {
6233                                                  Function &Kernel, int32_t LB,
6235   Kernel.addFnAttr("omp_target_thread_limit", std::to_string(UB));
6238     Kernel.addFnAttr("amdgpu-flat-work-group-size",
6243   updateNVPTXMetadata(Kernel, "maxntidx", UB, true);
6247 OpenMPIRBuilder::readTeamBoundsForKernel(const Triple &, Function &Kernel) {
6249   return {0, Kernel.getFnAttributeAsParsedInteger("omp_target_num_teams")};
6252 void OpenMPIRBuilder::writeTeamsForKernel(const Triple &T, Function &Kernel,
6256       updateNVPTXMetadata(Kernel, "maxclusterrank", UB, true);
6258     Kernel.addFnAttr("amdgpu-max-num-workgroups", llvm::utostr(LB) + ",1,1");
6260   Kernel.addFnAttr("omp_target_num_teams", std::to_string(LB));