summaryrefslogtreecommitdiff
path: root/lib/Target/R600/MCTargetDesc/R600MCCodeEmitter.cpp
blob: 34048444359d4f7c1fa57a8fb3c5c0e8c3a7a48e (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
//===- R600MCCodeEmitter.cpp - Code Emitter for R600->Cayman GPU families -===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
/// \file
///
/// \brief The R600 code emitter produces machine code that can be executed
/// directly on the GPU device.
//
//===----------------------------------------------------------------------===//

#include "R600Defines.h"
#include "MCTargetDesc/AMDGPUMCCodeEmitter.h"
#include "MCTargetDesc/AMDGPUMCTargetDesc.h"
#include "llvm/MC/MCCodeEmitter.h"
#include "llvm/MC/MCContext.h"
#include "llvm/MC/MCInst.h"
#include "llvm/MC/MCInstrInfo.h"
#include "llvm/MC/MCRegisterInfo.h"
#include "llvm/MC/MCSubtargetInfo.h"
#include "llvm/Support/raw_ostream.h"
#include <stdio.h>

using namespace llvm;

namespace {

class R600MCCodeEmitter : public AMDGPUMCCodeEmitter {
  R600MCCodeEmitter(const R600MCCodeEmitter &) LLVM_DELETED_FUNCTION;
  void operator=(const R600MCCodeEmitter &) LLVM_DELETED_FUNCTION;
  const MCInstrInfo &MCII;
  const MCRegisterInfo &MRI;

public:

  R600MCCodeEmitter(const MCInstrInfo &mcii, const MCRegisterInfo &mri)
    : MCII(mcii), MRI(mri) { }

  /// \brief Encode the instruction and write it to the OS.
  virtual void EncodeInstruction(const MCInst &MI, raw_ostream &OS,
                         SmallVectorImpl<MCFixup> &Fixups) const;

  /// \returns the encoding for an MCOperand.
  virtual uint64_t getMachineOpValue(const MCInst &MI, const MCOperand &MO,
                                     SmallVectorImpl<MCFixup> &Fixups) const;
private:

  void EmitByte(unsigned int byte, raw_ostream &OS) const;

  void Emit(uint32_t value, raw_ostream &OS) const;
  void Emit(uint64_t value, raw_ostream &OS) const;

  unsigned getHWRegChan(unsigned reg) const;
  unsigned getHWReg(unsigned regNo) const;

};

} // End anonymous namespace

enum RegElement {
  ELEMENT_X = 0,
  ELEMENT_Y,
  ELEMENT_Z,
  ELEMENT_W
};

enum FCInstr {
  FC_IF_PREDICATE = 0,
  FC_ELSE,
  FC_ENDIF,
  FC_BGNLOOP,
  FC_ENDLOOP,
  FC_BREAK_PREDICATE,
  FC_CONTINUE
};

enum TextureTypes {
  TEXTURE_1D = 1,
  TEXTURE_2D,
  TEXTURE_3D,
  TEXTURE_CUBE,
  TEXTURE_RECT,
  TEXTURE_SHADOW1D,
  TEXTURE_SHADOW2D,
  TEXTURE_SHADOWRECT,
  TEXTURE_1D_ARRAY,
  TEXTURE_2D_ARRAY,
  TEXTURE_SHADOW1D_ARRAY,
  TEXTURE_SHADOW2D_ARRAY
};

MCCodeEmitter *llvm::createR600MCCodeEmitter(const MCInstrInfo &MCII,
                                           const MCRegisterInfo &MRI) {
  return new R600MCCodeEmitter(MCII, MRI);
}

void R600MCCodeEmitter::EncodeInstruction(const MCInst &MI, raw_ostream &OS,
                                       SmallVectorImpl<MCFixup> &Fixups) const {
  const MCInstrDesc &Desc = MCII.get(MI.getOpcode());
  if (MI.getOpcode() == AMDGPU::RETURN ||
    MI.getOpcode() == AMDGPU::FETCH_CLAUSE ||
    MI.getOpcode() == AMDGPU::ALU_CLAUSE ||
    MI.getOpcode() == AMDGPU::BUNDLE ||
    MI.getOpcode() == AMDGPU::KILL) {
    return;
  } else if (IS_VTX(Desc)) {
    uint64_t InstWord01 = getBinaryCodeForInstr(MI, Fixups);
    uint32_t InstWord2 = MI.getOperand(2).getImm(); // Offset
    InstWord2 |= 1 << 19;

    Emit(InstWord01, OS);
    Emit(InstWord2, OS);
    Emit((u_int32_t) 0, OS);
  } else if (IS_TEX(Desc)) {
    unsigned Opcode = MI.getOpcode();
    bool HasOffsets = (Opcode == AMDGPU::TEX_LD);
    unsigned OpOffset = HasOffsets ? 3 : 0;
    int64_t Sampler = MI.getOperand(OpOffset + 3).getImm();
    int64_t TextureType = MI.getOperand(OpOffset + 4).getImm();

    uint32_t SrcSelect[4] = {0, 1, 2, 3};
    uint32_t Offsets[3] = {0, 0, 0};
    uint64_t CoordType[4] = {1, 1, 1, 1};

    if (HasOffsets)
      for (unsigned i = 0; i < 3; i++) {
        int SignedOffset = MI.getOperand(i + 2).getImm();
        Offsets[i] = (SignedOffset & 0x1F);
      }

    if (TextureType == TEXTURE_RECT ||
        TextureType == TEXTURE_SHADOWRECT) {
      CoordType[ELEMENT_X] = 0;
      CoordType[ELEMENT_Y] = 0;
    }

    if (TextureType == TEXTURE_1D_ARRAY ||
        TextureType == TEXTURE_SHADOW1D_ARRAY) {
      if (Opcode == AMDGPU::TEX_SAMPLE_C_L ||
          Opcode == AMDGPU::TEX_SAMPLE_C_LB) {
        CoordType[ELEMENT_Y] = 0;
      } else {
        CoordType[ELEMENT_Z] = 0;
        SrcSelect[ELEMENT_Z] = ELEMENT_Y;
      }
    } else if (TextureType == TEXTURE_2D_ARRAY ||
        TextureType == TEXTURE_SHADOW2D_ARRAY) {
      CoordType[ELEMENT_Z] = 0;
    }


    if ((TextureType == TEXTURE_SHADOW1D ||
        TextureType == TEXTURE_SHADOW2D ||
        TextureType == TEXTURE_SHADOWRECT ||
        TextureType == TEXTURE_SHADOW1D_ARRAY) &&
        Opcode != AMDGPU::TEX_SAMPLE_C_L &&
        Opcode != AMDGPU::TEX_SAMPLE_C_LB) {
      SrcSelect[ELEMENT_W] = ELEMENT_Z;
    }

    uint64_t Word01 = getBinaryCodeForInstr(MI, Fixups) |
        CoordType[ELEMENT_X] << 60 | CoordType[ELEMENT_Y] << 61 |
        CoordType[ELEMENT_Z] << 62 | CoordType[ELEMENT_W] << 63;
    uint32_t Word2 = Sampler << 15 | SrcSelect[ELEMENT_X] << 20 |
        SrcSelect[ELEMENT_Y] << 23 | SrcSelect[ELEMENT_Z] << 26 |
        SrcSelect[ELEMENT_W] << 29 | Offsets[0] << 0 | Offsets[1] << 5 |
        Offsets[2] << 10;

    Emit(Word01, OS);
    Emit(Word2, OS);
    Emit((u_int32_t) 0, OS);
  } else {
    uint64_t Inst = getBinaryCodeForInstr(MI, Fixups);
    Emit(Inst, OS);
  }
}

void R600MCCodeEmitter::EmitByte(unsigned int Byte, raw_ostream &OS) const {
  OS.write((uint8_t) Byte & 0xff);
}

void R600MCCodeEmitter::Emit(uint32_t Value, raw_ostream &OS) const {
  for (unsigned i = 0; i < 4; i++) {
    OS.write((uint8_t) ((Value >> (8 * i)) & 0xff));
  }
}

void R600MCCodeEmitter::Emit(uint64_t Value, raw_ostream &OS) const {
  for (unsigned i = 0; i < 8; i++) {
    EmitByte((Value >> (8 * i)) & 0xff, OS);
  }
}

unsigned R600MCCodeEmitter::getHWRegChan(unsigned reg) const {
  return MRI.getEncodingValue(reg) >> HW_CHAN_SHIFT;
}

unsigned R600MCCodeEmitter::getHWReg(unsigned RegNo) const {
  return MRI.getEncodingValue(RegNo) & HW_REG_MASK;
}

uint64_t R600MCCodeEmitter::getMachineOpValue(const MCInst &MI,
                                              const MCOperand &MO,
                                        SmallVectorImpl<MCFixup> &Fixup) const {
  if (MO.isReg()) {
    if (HAS_NATIVE_OPERANDS(MCII.get(MI.getOpcode()).TSFlags)) {
      return MRI.getEncodingValue(MO.getReg());
    } else {
      return getHWReg(MO.getReg());
    }
  } else if (MO.isImm()) {
    return MO.getImm();
  } else {
    assert(0);
    return 0;
  }
}

#include "AMDGPUGenMCCodeEmitter.inc"