[ORC] Add std::tuple support to SimplePackedSerialization.
[llvm-project.git] / llvm / lib / Target / NVPTX / NVPTXSubtarget.cpp
blob05c20369abf45a458c645ffba4447a769b1da4c7
1 //===- NVPTXSubtarget.cpp - NVPTX Subtarget Information -------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file implements the NVPTX specific subclass of TargetSubtarget.
11 //===----------------------------------------------------------------------===//
13 #include "NVPTXSubtarget.h"
14 #include "NVPTXTargetMachine.h"
16 using namespace llvm;
18 #define DEBUG_TYPE "nvptx-subtarget"
20 #define GET_SUBTARGETINFO_ENUM
21 #define GET_SUBTARGETINFO_TARGET_DESC
22 #define GET_SUBTARGETINFO_CTOR
23 #include "NVPTXGenSubtargetInfo.inc"
25 static cl::opt<bool>
26 NoF16Math("nvptx-no-f16-math", cl::ZeroOrMore, cl::Hidden,
27 cl::desc("NVPTX Specific: Disable generation of f16 math ops."),
28 cl::init(false));
30 // Pin the vtable to this file.
31 void NVPTXSubtarget::anchor() {}
33 NVPTXSubtarget &NVPTXSubtarget::initializeSubtargetDependencies(StringRef CPU,
34 StringRef FS) {
35 // Provide the default CPU if we don't have one.
36 TargetName = std::string(CPU.empty() ? "sm_20" : CPU);
38 ParseSubtargetFeatures(TargetName, /*TuneCPU*/ TargetName, FS);
40 // Set default to PTX 3.2 (CUDA 5.5)
41 if (PTXVersion == 0) {
42 PTXVersion = 32;
45 return *this;
48 NVPTXSubtarget::NVPTXSubtarget(const Triple &TT, const std::string &CPU,
49 const std::string &FS,
50 const NVPTXTargetMachine &TM)
51 : NVPTXGenSubtargetInfo(TT, CPU, /*TuneCPU*/ CPU, FS), PTXVersion(0),
52 SmVersion(20), TM(TM), InstrInfo(),
53 TLInfo(TM, initializeSubtargetDependencies(CPU, FS)), FrameLowering() {}
55 bool NVPTXSubtarget::hasImageHandles() const {
56 // Enable handles for Kepler+, where CUDA supports indirect surfaces and
57 // textures
58 if (TM.getDrvInterface() == NVPTX::CUDA)
59 return (SmVersion >= 30);
61 // Disabled, otherwise
62 return false;
65 bool NVPTXSubtarget::allowFP16Math() const {
66 return hasFP16Math() && NoF16Math == false;