Initial support of Geometry shaders (#1244)

* video_core: initial GS support

* fix for components mapping; missing prim type
This commit is contained in:
psucien 2024-10-06 00:26:50 +02:00 committed by GitHub
parent 5bb45dc7ba
commit 927bb0c175
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
40 changed files with 944 additions and 268 deletions

View file

@ -19,6 +19,7 @@
#include "common/types.h"
#include "common/unique_function.h"
#include "shader_recompiler/params.h"
#include "types.h"
#include "video_core/amdgpu/pixel_format.h"
#include "video_core/amdgpu/resource.h"
@ -842,26 +843,6 @@ struct Liverpool {
}
};
enum class PrimitiveType : u32 {
None = 0,
PointList = 1,
LineList = 2,
LineStrip = 3,
TriangleList = 4,
TriangleFan = 5,
TriangleStrip = 6,
PatchPrimitive = 9,
AdjLineList = 10,
AdjLineStrip = 11,
AdjTriangleList = 12,
AdjTriangleStrip = 13,
RectList = 17,
LineLoop = 18,
QuadList = 19,
QuadStrip = 20,
Polygon = 21,
};
enum ContextRegs : u32 {
DbZInfo = 0xA010,
CbColor0Base = 0xA318,
@ -936,7 +917,12 @@ struct Liverpool {
};
union ShaderStageEnable {
u32 raw;
enum VgtStages : u32 {
Vs = 0u, // always enabled
EsGs = 0xB0u,
};
VgtStages raw;
BitField<0, 2, u32> ls_en;
BitField<2, 1, u32> hs_en;
BitField<3, 2, u32> es_en;
@ -962,6 +948,81 @@ struct Liverpool {
}
};
union GsInstances {
u32 raw;
struct {
u32 enable : 2;
u32 count : 6;
};
bool IsEnabled() const {
return enable && count > 0;
}
};
union GsOutPrimitiveType {
u32 raw;
struct {
GsOutputPrimitiveType outprim_type : 6;
GsOutputPrimitiveType outprim_type1 : 6;
GsOutputPrimitiveType outprim_type2 : 6;
GsOutputPrimitiveType outprim_type3 : 6;
u32 reserved : 3;
u32 unique_type_per_stream : 1;
};
GsOutputPrimitiveType GetPrimitiveType(u32 stream) const {
if (unique_type_per_stream == 0) {
return outprim_type;
}
switch (stream) {
case 0:
return outprim_type;
case 1:
return outprim_type1;
case 2:
return outprim_type2;
case 3:
return outprim_type3;
default:
UNREACHABLE();
}
}
};
union GsMode {
u32 raw;
BitField<0, 3, u32> mode;
BitField<3, 2, u32> cut_mode;
BitField<22, 2, u32> onchip;
};
union StreamOutConfig {
u32 raw;
struct {
u32 streamout_0_en : 1;
u32 streamout_1_en : 1;
u32 streamout_2_en : 1;
u32 streamout_3_en : 1;
u32 rast_stream : 3;
u32 : 1;
u32 rast_stream_mask : 4;
u32 : 19;
u32 use_rast_stream_mask : 1;
};
};
union StreamOutBufferConfig {
u32 raw;
struct {
u32 stream_0_buf_en : 4;
u32 stream_1_buf_en : 4;
u32 stream_2_buf_en : 4;
u32 stream_3_buf_en : 4;
};
};
union Eqaa {
u32 raw;
BitField<0, 1, u32> max_anchor_samples;
@ -1053,9 +1114,13 @@ struct Liverpool {
PolygonControl polygon_control;
ViewportControl viewport_control;
VsOutputControl vs_output_control;
INSERT_PADDING_WORDS(0xA292 - 0xA207 - 1);
INSERT_PADDING_WORDS(0xA290 - 0xA207 - 1);
GsMode vgt_gs_mode;
INSERT_PADDING_WORDS(1);
ModeControl mode_control;
INSERT_PADDING_WORDS(0xA29D - 0xA292 - 1);
INSERT_PADDING_WORDS(8);
GsOutPrimitiveType vgt_gs_out_prim_type;
INSERT_PADDING_WORDS(1);
u32 index_size;
u32 max_index_size;
IndexBufferType index_buffer_type;
@ -1066,11 +1131,21 @@ struct Liverpool {
INSERT_PADDING_WORDS(0xA2A8 - 0xA2A5 - 1);
u32 vgt_instance_step_rate_0;
u32 vgt_instance_step_rate_1;
INSERT_PADDING_WORDS(0xA2D5 - 0xA2A9 - 1);
INSERT_PADDING_WORDS(0xA2AB - 0xA2A9 - 1);
u32 vgt_esgs_ring_itemsize;
u32 vgt_gsvs_ring_itemsize;
INSERT_PADDING_WORDS(0xA2CE - 0xA2AC - 1);
BitField<0, 11, u32> vgt_gs_max_vert_out;
INSERT_PADDING_WORDS(0xA2D5 - 0xA2CE - 1);
ShaderStageEnable stage_enable;
INSERT_PADDING_WORDS(9);
INSERT_PADDING_WORDS(1);
u32 vgt_gs_vert_itemsize[4];
INSERT_PADDING_WORDS(4);
PolygonOffset poly_offset;
INSERT_PADDING_WORDS(0xA2F8 - 0xA2DF - 5);
GsInstances vgt_gs_instance_cnt;
StreamOutConfig vgt_strmout_config;
StreamOutBufferConfig vgt_strmout_buffer_config;
INSERT_PADDING_WORDS(0xA2F8 - 0xA2E6 - 1);
AaConfig aa_config;
INSERT_PADDING_WORDS(0xA318 - 0xA2F8 - 1);
ColorBuffer color_buffers[NumColorBuffers];
@ -1291,15 +1366,24 @@ static_assert(GFX6_3D_REG_INDEX(color_control) == 0xA202);
static_assert(GFX6_3D_REG_INDEX(clipper_control) == 0xA204);
static_assert(GFX6_3D_REG_INDEX(viewport_control) == 0xA206);
static_assert(GFX6_3D_REG_INDEX(vs_output_control) == 0xA207);
static_assert(GFX6_3D_REG_INDEX(vgt_gs_mode) == 0xA290);
static_assert(GFX6_3D_REG_INDEX(mode_control) == 0xA292);
static_assert(GFX6_3D_REG_INDEX(vgt_gs_out_prim_type) == 0xA29B);
static_assert(GFX6_3D_REG_INDEX(index_size) == 0xA29D);
static_assert(GFX6_3D_REG_INDEX(index_buffer_type) == 0xA29F);
static_assert(GFX6_3D_REG_INDEX(enable_primitive_id) == 0xA2A1);
static_assert(GFX6_3D_REG_INDEX(enable_primitive_restart) == 0xA2A5);
static_assert(GFX6_3D_REG_INDEX(vgt_instance_step_rate_0) == 0xA2A8);
static_assert(GFX6_3D_REG_INDEX(vgt_instance_step_rate_1) == 0xA2A9);
static_assert(GFX6_3D_REG_INDEX(vgt_esgs_ring_itemsize) == 0xA2AB);
static_assert(GFX6_3D_REG_INDEX(vgt_gsvs_ring_itemsize) == 0xA2AC);
static_assert(GFX6_3D_REG_INDEX(vgt_gs_max_vert_out) == 0xA2CE);
static_assert(GFX6_3D_REG_INDEX(stage_enable) == 0xA2D5);
static_assert(GFX6_3D_REG_INDEX(vgt_gs_vert_itemsize[0]) == 0xA2D7);
static_assert(GFX6_3D_REG_INDEX(poly_offset) == 0xA2DF);
static_assert(GFX6_3D_REG_INDEX(vgt_gs_instance_cnt) == 0xA2E4);
static_assert(GFX6_3D_REG_INDEX(vgt_strmout_config) == 0xA2E5);
static_assert(GFX6_3D_REG_INDEX(vgt_strmout_buffer_config) == 0xA2E6);
static_assert(GFX6_3D_REG_INDEX(aa_config) == 0xA2F8);
static_assert(GFX6_3D_REG_INDEX(color_buffers[0].base_address) == 0xA318);
static_assert(GFX6_3D_REG_INDEX(color_buffers[0].pitch) == 0xA319);

View file

@ -6,78 +6,10 @@
#include <string_view>
#include <fmt/format.h>
#include "common/types.h"
#include "video_core/amdgpu/types.h"
namespace AmdGpu {
// Table 8.13 Data and Image Formats [Sea Islands Series Instruction Set Architecture]
enum class DataFormat : u32 {
FormatInvalid = 0,
Format8 = 1,
Format16 = 2,
Format8_8 = 3,
Format32 = 4,
Format16_16 = 5,
Format10_11_11 = 6,
Format11_11_10 = 7,
Format10_10_10_2 = 8,
Format2_10_10_10 = 9,
Format8_8_8_8 = 10,
Format32_32 = 11,
Format16_16_16_16 = 12,
Format32_32_32 = 13,
Format32_32_32_32 = 14,
Format5_6_5 = 16,
Format1_5_5_5 = 17,
Format5_5_5_1 = 18,
Format4_4_4_4 = 19,
Format8_24 = 20,
Format24_8 = 21,
FormatX24_8_32 = 22,
FormatGB_GR = 32,
FormatBG_RG = 33,
Format5_9_9_9 = 34,
FormatBc1 = 35,
FormatBc2 = 36,
FormatBc3 = 37,
FormatBc4 = 38,
FormatBc5 = 39,
FormatBc6 = 40,
FormatBc7 = 41,
FormatFmask8_1 = 47,
FormatFmask8_2 = 48,
FormatFmask8_4 = 49,
FormatFmask16_1 = 50,
FormatFmask16_2 = 51,
FormatFmask32_2 = 52,
FormatFmask32_4 = 53,
FormatFmask32_8 = 54,
FormatFmask64_4 = 55,
FormatFmask64_8 = 56,
Format4_4 = 57,
Format6_5_5 = 58,
Format1 = 59,
Format1_Reversed = 60,
Format32_As_8 = 61,
Format32_As_8_8 = 62,
Format32_As_32_32_32_32 = 63,
};
enum class NumberFormat : u32 {
Unorm = 0,
Snorm = 1,
Uscaled = 2,
Sscaled = 3,
Uint = 4,
Sint = 5,
SnormNz = 6,
Float = 7,
Srgb = 9,
Ubnorm = 10,
UbnromNz = 11,
Ubint = 12,
Ubscaled = 13,
};
[[nodiscard]] constexpr bool IsInteger(NumberFormat nfmt) {
return nfmt == AmdGpu::NumberFormat::Sint || nfmt == AmdGpu::NumberFormat::Uint;
}

View file

@ -0,0 +1,106 @@
// SPDX-FileCopyrightText: Copyright 2024 shadPS4 Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include "common/types.h"
namespace AmdGpu {
// See `VGT_PRIMITIVE_TYPE` description in [Radeon Sea Islands 3D/Compute Register Reference Guide]
enum class PrimitiveType : u32 {
None = 0,
PointList = 1,
LineList = 2,
LineStrip = 3,
TriangleList = 4,
TriangleFan = 5,
TriangleStrip = 6,
PatchPrimitive = 9,
AdjLineList = 10,
AdjLineStrip = 11,
AdjTriangleList = 12,
AdjTriangleStrip = 13,
RectList = 17,
LineLoop = 18,
QuadList = 19,
QuadStrip = 20,
Polygon = 21,
};
enum class GsOutputPrimitiveType : u32 {
PointList = 0,
LineStrip = 1,
TriangleStrip = 2,
};
// Table 8.13 Data and Image Formats [Sea Islands Series Instruction Set Architecture]
enum class DataFormat : u32 {
FormatInvalid = 0,
Format8 = 1,
Format16 = 2,
Format8_8 = 3,
Format32 = 4,
Format16_16 = 5,
Format10_11_11 = 6,
Format11_11_10 = 7,
Format10_10_10_2 = 8,
Format2_10_10_10 = 9,
Format8_8_8_8 = 10,
Format32_32 = 11,
Format16_16_16_16 = 12,
Format32_32_32 = 13,
Format32_32_32_32 = 14,
Format5_6_5 = 16,
Format1_5_5_5 = 17,
Format5_5_5_1 = 18,
Format4_4_4_4 = 19,
Format8_24 = 20,
Format24_8 = 21,
FormatX24_8_32 = 22,
FormatGB_GR = 32,
FormatBG_RG = 33,
Format5_9_9_9 = 34,
FormatBc1 = 35,
FormatBc2 = 36,
FormatBc3 = 37,
FormatBc4 = 38,
FormatBc5 = 39,
FormatBc6 = 40,
FormatBc7 = 41,
FormatFmask8_1 = 47,
FormatFmask8_2 = 48,
FormatFmask8_4 = 49,
FormatFmask16_1 = 50,
FormatFmask16_2 = 51,
FormatFmask32_2 = 52,
FormatFmask32_4 = 53,
FormatFmask32_8 = 54,
FormatFmask64_4 = 55,
FormatFmask64_8 = 56,
Format4_4 = 57,
Format6_5_5 = 58,
Format1 = 59,
Format1_Reversed = 60,
Format32_As_8 = 61,
Format32_As_8_8 = 62,
Format32_As_32_32_32_32 = 63,
};
enum class NumberFormat : u32 {
Unorm = 0,
Snorm = 1,
Uscaled = 2,
Sscaled = 3,
Uint = 4,
Sint = 5,
SnormNz = 6,
Float = 7,
Srgb = 9,
Ubnorm = 10,
UbnromNz = 11,
Ubint = 12,
Ubscaled = 13,
};
} // namespace AmdGpu