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
|
//===-- SIAssignInterpRegs.cpp - Assign interpolation registers -----------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This pass maps the pseudo interpolation registers to the correct physical
// registers. Prior to executing a fragment shader, the GPU loads interpolation
// parameters into physical registers. The specific physical register that each
// interpolation parameter ends up in depends on the type of the interpolation
// parameter as well as how many interpolation parameters are used by the
// shader.
//
//===----------------------------------------------------------------------===//
#include "AMDGPU.h"
#include "AMDGPUUtil.h"
#include "AMDIL.h"
#include "SIMachineFunctionInfo.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
using namespace llvm;
namespace {
class SIAssignInterpRegsPass : public MachineFunctionPass {
private:
static char ID;
TargetMachine &TM;
public:
SIAssignInterpRegsPass(TargetMachine &tm) :
MachineFunctionPass(ID), TM(tm) { }
virtual bool runOnMachineFunction(MachineFunction &MF);
const char *getPassName() const { return "SI Assign intrpolation registers"; }
};
} // End anonymous namespace
char SIAssignInterpRegsPass::ID = 0;
#define INTERP_VALUES 16
struct interp_info {
bool enabled;
unsigned regs[3];
unsigned reg_count;
};
FunctionPass *llvm::createSIAssignInterpRegsPass(TargetMachine &tm) {
return new SIAssignInterpRegsPass(tm);
}
bool SIAssignInterpRegsPass::runOnMachineFunction(MachineFunction &MF)
{
struct interp_info InterpUse[INTERP_VALUES] = {
{false, {AMDIL::PERSP_SAMPLE_I, AMDIL::PERSP_SAMPLE_J}, 2},
{false, {AMDIL::PERSP_CENTER_I, AMDIL::PERSP_CENTER_J}, 2},
{false, {AMDIL::PERSP_CENTROID_I, AMDIL::PERSP_CENTROID_J}, 2},
{false, {AMDIL::PERSP_I_W, AMDIL::PERSP_J_W, AMDIL::PERSP_1_W}, 3},
{false, {AMDIL::LINEAR_SAMPLE_I, AMDIL::LINEAR_SAMPLE_J}, 2},
{false, {AMDIL::LINEAR_CENTER_I, AMDIL::LINEAR_CENTER_J}, 2},
{false, {AMDIL::LINEAR_CENTROID_I, AMDIL::LINEAR_CENTROID_J}, 2},
{false, {AMDIL::LINE_STIPPLE_TEX_COORD}, 1},
{false, {AMDIL::POS_X_FLOAT}, 1},
{false, {AMDIL::POS_Y_FLOAT}, 1},
{false, {AMDIL::POS_Z_FLOAT}, 1},
{false, {AMDIL::POS_W_FLOAT}, 1},
{false, {AMDIL::FRONT_FACE}, 1},
{false, {AMDIL::ANCILLARY}, 1},
{false, {AMDIL::SAMPLE_COVERAGE}, 1},
{false, {AMDIL::POS_FIXED_PT}, 1}
};
SIMachineFunctionInfo * MFI = MF.getInfo<SIMachineFunctionInfo>();
MachineRegisterInfo &MRI = MF.getRegInfo();
/* First pass, mark the interpolation values that are used. */
for (unsigned interp_idx = 0; interp_idx < INTERP_VALUES; interp_idx++) {
for (unsigned reg_idx = 0; reg_idx < InterpUse[interp_idx].reg_count;
reg_idx++) {
InterpUse[interp_idx].enabled =
!MRI.use_empty(InterpUse[interp_idx].regs[reg_idx]);
}
}
unsigned used_vgprs = 0;
/* Second pass, replace with VGPRs. */
for (unsigned interp_idx = 0; interp_idx < INTERP_VALUES; interp_idx++) {
if (!InterpUse[interp_idx].enabled) {
continue;
}
MFI->spi_ps_input_addr |= (1 << interp_idx);
for (unsigned reg_idx = 0; reg_idx < InterpUse[interp_idx].reg_count;
reg_idx++, used_vgprs++) {
unsigned new_reg = AMDIL::VReg_32RegisterClass->getRegister(used_vgprs);
unsigned virt_reg = MRI.createVirtualRegister(AMDIL::VReg_32RegisterClass);
MRI.replaceRegWith(InterpUse[interp_idx].regs[reg_idx], virt_reg);
AMDGPU::utilAddLiveIn(&MF, MRI, TM.getInstrInfo(), new_reg, virt_reg);
}
}
return false;
}
|
