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
|
// Copyright 2021 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "shader_recompiler/backend/glasm/emit_context.h"
#include "shader_recompiler/backend/glasm/emit_glasm_instructions.h"
#include "shader_recompiler/frontend/ir/value.h"
#include "shader_recompiler/profile.h"
namespace Shader::Backend::GLASM {
void EmitLaneId(EmitContext& ctx, IR::Inst& inst) {
ctx.Add("MOV.S {}.x,{}.threadid;", inst, ctx.stage_name);
}
void EmitVoteAll(EmitContext& ctx, IR::Inst& inst, ScalarS32 pred) {
ctx.Add("TGALL.S {}.x,{};", inst, pred);
}
void EmitVoteAny(EmitContext& ctx, IR::Inst& inst, ScalarS32 pred) {
ctx.Add("TGANY.S {}.x,{};", inst, pred);
}
void EmitVoteEqual(EmitContext& ctx, IR::Inst& inst, ScalarS32 pred) {
ctx.Add("TGEQ.S {}.x,{};", inst, pred);
}
void EmitSubgroupBallot(EmitContext& ctx, IR::Inst& inst, ScalarS32 pred) {
ctx.Add("TGBALLOT {}.x,{};", inst, pred);
}
void EmitSubgroupEqMask(EmitContext& ctx, IR::Inst& inst) {
ctx.Add("MOV.U {},{}.threadeqmask;", inst, ctx.stage_name);
}
void EmitSubgroupLtMask(EmitContext& ctx, IR::Inst& inst) {
ctx.Add("MOV.U {},{}.threadltmask;", inst, ctx.stage_name);
}
void EmitSubgroupLeMask(EmitContext& ctx, IR::Inst& inst) {
ctx.Add("MOV.U {},{}.threadlemask;", inst, ctx.stage_name);
}
void EmitSubgroupGtMask(EmitContext& ctx, IR::Inst& inst) {
ctx.Add("MOV.U {},{}.threadgtmask;", inst, ctx.stage_name);
}
void EmitSubgroupGeMask(EmitContext& ctx, IR::Inst& inst) {
ctx.Add("MOV.U {},{}.threadgemask;", inst, ctx.stage_name);
}
static void Shuffle(EmitContext& ctx, IR::Inst& inst, ScalarU32 value, ScalarU32 index,
const IR::Value& clamp, const IR::Value& segmentation_mask,
std::string_view op) {
IR::Inst* const in_bounds{inst.GetAssociatedPseudoOperation(IR::Opcode::GetInBoundsFromOp)};
if (in_bounds) {
in_bounds->Invalidate();
}
std::string mask;
if (clamp.IsImmediate() && segmentation_mask.IsImmediate()) {
mask = fmt::to_string(clamp.U32() | (segmentation_mask.U32() << 8));
} else {
mask = "RC";
ctx.Add("BFI.U RC.x,{{5,8,0,0}},{},{};",
ScalarU32{ctx.reg_alloc.Consume(segmentation_mask)},
ScalarU32{ctx.reg_alloc.Consume(clamp)});
}
const Register value_ret{ctx.reg_alloc.Define(inst)};
if (in_bounds) {
const Register bounds_ret{ctx.reg_alloc.Define(*in_bounds)};
ctx.Add("SHF{}.U {},{},{},{};"
"MOV.U {}.x,{}.y;",
op, bounds_ret, value, index, mask, value_ret, bounds_ret);
} else {
ctx.Add("SHF{}.U {},{},{},{};"
"MOV.U {}.x,{}.y;",
op, value_ret, value, index, mask, value_ret, value_ret);
}
}
void EmitShuffleIndex(EmitContext& ctx, IR::Inst& inst, ScalarU32 value, ScalarU32 index,
const IR::Value& clamp, const IR::Value& segmentation_mask) {
Shuffle(ctx, inst, value, index, clamp, segmentation_mask, "IDX");
}
void EmitShuffleUp(EmitContext& ctx, IR::Inst& inst, ScalarU32 value, ScalarU32 index,
const IR::Value& clamp, const IR::Value& segmentation_mask) {
Shuffle(ctx, inst, value, index, clamp, segmentation_mask, "UP");
}
void EmitShuffleDown(EmitContext& ctx, IR::Inst& inst, ScalarU32 value, ScalarU32 index,
const IR::Value& clamp, const IR::Value& segmentation_mask) {
Shuffle(ctx, inst, value, index, clamp, segmentation_mask, "DOWN");
}
void EmitShuffleButterfly(EmitContext& ctx, IR::Inst& inst, ScalarU32 value, ScalarU32 index,
const IR::Value& clamp, const IR::Value& segmentation_mask) {
Shuffle(ctx, inst, value, index, clamp, segmentation_mask, "XOR");
}
void EmitFSwizzleAdd(EmitContext& ctx, IR::Inst& inst, ScalarF32 op_a, ScalarF32 op_b,
ScalarU32 swizzle) {
const auto ret{ctx.reg_alloc.Define(inst)};
ctx.Add("AND.U RC.z,{}.threadid,3;"
"SHL.U RC.z,RC.z,1;"
"SHR.U RC.z,{},RC.z;"
"AND.U RC.z,RC.z,3;"
"MUL.F RC.x,{},FSWZA[RC.z];"
"MUL.F RC.y,{},FSWZB[RC.z];"
"ADD.F {}.x,RC.x,RC.y;",
ctx.stage_name, swizzle, op_a, op_b, ret);
}
void EmitDPdxFine(EmitContext& ctx, IR::Inst& inst, ScalarF32 p) {
if (ctx.profile.support_derivative_control) {
ctx.Add("DDX.FINE {}.x,{};", inst, p);
} else {
LOG_WARNING(Shader_GLASM, "Fine derivatives not supported by device");
ctx.Add("DDX {}.x,{};", inst, p);
}
}
void EmitDPdyFine(EmitContext& ctx, IR::Inst& inst, ScalarF32 p) {
if (ctx.profile.support_derivative_control) {
ctx.Add("DDY.FINE {}.x,{};", inst, p);
} else {
LOG_WARNING(Shader_GLASM, "Fine derivatives not supported by device");
ctx.Add("DDY {}.x,{};", inst, p);
}
}
void EmitDPdxCoarse(EmitContext& ctx, IR::Inst& inst, ScalarF32 p) {
if (ctx.profile.support_derivative_control) {
ctx.Add("DDX.COARSE {}.x,{};", inst, p);
} else {
LOG_WARNING(Shader_GLASM, "Coarse derivatives not supported by device");
ctx.Add("DDX {}.x,{};", inst, p);
}
}
void EmitDPdyCoarse(EmitContext& ctx, IR::Inst& inst, ScalarF32 p) {
if (ctx.profile.support_derivative_control) {
ctx.Add("DDY.COARSE {}.x,{};", inst, p);
} else {
LOG_WARNING(Shader_GLASM, "Coarse derivatives not supported by device");
ctx.Add("DDY {}.x,{};", inst, p);
}
}
} // namespace Shader::Backend::GLASM
|
