summaryrefslogtreecommitdiffstats
path: root/src/video_core
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core')
-rw-r--r--src/video_core/CMakeLists.txt3
-rw-r--r--src/video_core/engines/maxwell_3d.h19
-rw-r--r--src/video_core/engines/shader_bytecode.h10
-rw-r--r--src/video_core/engines/shader_header.h11
-rw-r--r--src/video_core/gpu.cpp2
-rw-r--r--src/video_core/macro_interpreter.cpp29
-rw-r--r--src/video_core/macro_interpreter.h4
-rw-r--r--src/video_core/memory_manager.cpp7
-rw-r--r--src/video_core/memory_manager.h3
-rw-r--r--src/video_core/morton.cpp353
-rw-r--r--src/video_core/morton.h21
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp49
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.h31
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.cpp240
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.h1
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.cpp1
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.h3
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp1138
-rw-r--r--src/video_core/renderer_opengl/gl_shader_gen.h1
-rw-r--r--src/video_core/renderer_opengl/gl_state.cpp37
-rw-r--r--src/video_core/renderer_opengl/gl_state.h9
-rw-r--r--src/video_core/renderer_opengl/maxwell_to_gl.h5
-rw-r--r--src/video_core/renderer_opengl/renderer_opengl.cpp10
-rw-r--r--src/video_core/utils.h164
24 files changed, 1194 insertions, 957 deletions
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index a780215c1..3f906a517 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -21,6 +21,8 @@ add_library(video_core STATIC
macro_interpreter.h
memory_manager.cpp
memory_manager.h
+ morton.cpp
+ morton.h
rasterizer_cache.cpp
rasterizer_cache.h
rasterizer_interface.h
@@ -62,7 +64,6 @@ add_library(video_core STATIC
textures/decoders.cpp
textures/decoders.h
textures/texture.h
- utils.h
video_core.cpp
video_core.h
)
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index 9e480dc39..4f137e693 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -389,6 +389,13 @@ public:
ReverseSubtract = 3,
Min = 4,
Max = 5,
+
+ // These values are used by Nouveau and some games.
+ AddGL = 0x8006,
+ SubtractGL = 0x8007,
+ ReverseSubtractGL = 0x8008,
+ MinGL = 0x800a,
+ MaxGL = 0x800b
};
enum class Factor : u32 {
@@ -624,7 +631,16 @@ public:
}
} zeta;
- INSERT_PADDING_WORDS(0x5B);
+ INSERT_PADDING_WORDS(0x41);
+
+ union {
+ BitField<0, 4, u32> stencil;
+ BitField<4, 4, u32> unknown;
+ BitField<8, 4, u32> scissor;
+ BitField<12, 4, u32> viewport;
+ } clear_flags;
+
+ INSERT_PADDING_WORDS(0x19);
std::array<VertexAttribute, NumVertexAttributes> vertex_attrib_format;
@@ -1127,6 +1143,7 @@ ASSERT_REG_POSITION(stencil_back_func_mask, 0x3D7);
ASSERT_REG_POSITION(color_mask_common, 0x3E4);
ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB);
ASSERT_REG_POSITION(zeta, 0x3F8);
+ASSERT_REG_POSITION(clear_flags, 0x43E);
ASSERT_REG_POSITION(vertex_attrib_format, 0x458);
ASSERT_REG_POSITION(rt_control, 0x487);
ASSERT_REG_POSITION(zeta_width, 0x48a);
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 5b84bcb24..52d03aee8 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -82,6 +82,8 @@ union Attribute {
Position = 7,
Attribute_0 = 8,
Attribute_31 = 39,
+ ClipDistances0123 = 44,
+ ClipDistances4567 = 45,
PointCoord = 46,
// This attribute contains a tuple of (~, ~, InstanceId, VertexId) when inside a vertex
// shader, and a tuple of (TessCoord.x, TessCoord.y, TessCoord.z, ~) when inside a Tess Eval
@@ -153,6 +155,7 @@ enum class PredCondition : u64 {
NotEqual = 5,
GreaterEqual = 6,
LessThanWithNan = 9,
+ LessEqualWithNan = 11,
GreaterThanWithNan = 12,
NotEqualWithNan = 13,
GreaterEqualWithNan = 14,
@@ -261,7 +264,7 @@ enum class FlowCondition : u64 {
Fcsm_Tr = 0x1C, // TODO(bunnei): What is this used for?
};
-enum class ControlCode : u64 {
+enum class ConditionCode : u64 {
F = 0,
LT = 1,
EQ = 2,
@@ -574,7 +577,6 @@ union Instruction {
BitField<39, 2, u64> tab5cb8_2;
BitField<41, 3, u64> tab5c68_1;
BitField<44, 2, u64> tab5c68_0;
- BitField<47, 1, u64> cc;
BitField<48, 1, u64> negate_b;
} fmul;
@@ -836,7 +838,7 @@ union Instruction {
union {
BitField<0, 3, u64> pred0;
BitField<3, 3, u64> pred3;
- BitField<8, 5, ControlCode> cc; // flag in cc
+ BitField<8, 5, ConditionCode> cc; // flag in cc
BitField<39, 3, u64> pred39;
BitField<42, 1, u64> neg_pred39;
BitField<45, 4, PredOperation> op; // op with pred39
@@ -1246,7 +1248,7 @@ union Instruction {
BitField<60, 1, u64> is_b_gpr;
BitField<59, 1, u64> is_c_gpr;
BitField<20, 24, s64> smem_imm;
- BitField<0, 5, ControlCode> flow_control_code;
+ BitField<0, 5, ConditionCode> flow_condition_code;
Attribute attribute;
Sampler sampler;
diff --git a/src/video_core/engines/shader_header.h b/src/video_core/engines/shader_header.h
index a0e015c4b..99c34649f 100644
--- a/src/video_core/engines/shader_header.h
+++ b/src/video_core/engines/shader_header.h
@@ -62,7 +62,16 @@ struct Header {
INSERT_PADDING_BYTES(1); // ImapSystemValuesB
INSERT_PADDING_BYTES(16); // ImapGenericVector[32]
INSERT_PADDING_BYTES(2); // ImapColor
- INSERT_PADDING_BYTES(2); // ImapSystemValuesC
+ union {
+ BitField<0, 8, u16> clip_distances;
+ BitField<8, 1, u16> point_sprite_s;
+ BitField<9, 1, u16> point_sprite_t;
+ BitField<10, 1, u16> fog_coordinate;
+ BitField<12, 1, u16> tessellation_eval_point_u;
+ BitField<13, 1, u16> tessellation_eval_point_v;
+ BitField<14, 1, u16> instance_id;
+ BitField<15, 1, u16> vertex_id;
+ };
INSERT_PADDING_BYTES(5); // ImapFixedFncTexture[10]
INSERT_PADDING_BYTES(1); // ImapReserved
INSERT_PADDING_BYTES(3); // OmapSystemValuesA
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index 83c7e5b0b..51b3904f6 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -17,6 +17,8 @@ u32 FramebufferConfig::BytesPerPixel(PixelFormat format) {
switch (format) {
case PixelFormat::ABGR8:
return 4;
+ default:
+ return 4;
}
UNREACHABLE();
diff --git a/src/video_core/macro_interpreter.cpp b/src/video_core/macro_interpreter.cpp
index 335a8d407..2b0dea5cd 100644
--- a/src/video_core/macro_interpreter.cpp
+++ b/src/video_core/macro_interpreter.cpp
@@ -35,6 +35,7 @@ void MacroInterpreter::Reset() {
// The next parameter index starts at 1, because $r1 already has the value of the first
// parameter.
next_parameter_index = 1;
+ carry_flag = false;
}
bool MacroInterpreter::Step(u32 offset, bool is_delay_slot) {
@@ -135,14 +136,28 @@ MacroInterpreter::Opcode MacroInterpreter::GetOpcode(u32 offset) const {
return {macro_memory[offset + pc / sizeof(u32)]};
}
-u32 MacroInterpreter::GetALUResult(ALUOperation operation, u32 src_a, u32 src_b) const {
+u32 MacroInterpreter::GetALUResult(ALUOperation operation, u32 src_a, u32 src_b) {
switch (operation) {
- case ALUOperation::Add:
- return src_a + src_b;
- // TODO(Subv): Implement AddWithCarry
- case ALUOperation::Subtract:
- return src_a - src_b;
- // TODO(Subv): Implement SubtractWithBorrow
+ case ALUOperation::Add: {
+ const u64 result{static_cast<u64>(src_a) + src_b};
+ carry_flag = result > 0xffffffff;
+ return static_cast<u32>(result);
+ }
+ case ALUOperation::AddWithCarry: {
+ const u64 result{static_cast<u64>(src_a) + src_b + (carry_flag ? 1ULL : 0ULL)};
+ carry_flag = result > 0xffffffff;
+ return static_cast<u32>(result);
+ }
+ case ALUOperation::Subtract: {
+ const u64 result{static_cast<u64>(src_a) - src_b};
+ carry_flag = result < 0x100000000;
+ return static_cast<u32>(result);
+ }
+ case ALUOperation::SubtractWithBorrow: {
+ const u64 result{static_cast<u64>(src_a) - src_b - (carry_flag ? 0ULL : 1ULL)};
+ carry_flag = result < 0x100000000;
+ return static_cast<u32>(result);
+ }
case ALUOperation::Xor:
return src_a ^ src_b;
case ALUOperation::Or:
diff --git a/src/video_core/macro_interpreter.h b/src/video_core/macro_interpreter.h
index 62d1ce289..cde360288 100644
--- a/src/video_core/macro_interpreter.h
+++ b/src/video_core/macro_interpreter.h
@@ -117,7 +117,7 @@ private:
bool Step(u32 offset, bool is_delay_slot);
/// Calculates the result of an ALU operation. src_a OP src_b;
- u32 GetALUResult(ALUOperation operation, u32 src_a, u32 src_b) const;
+ u32 GetALUResult(ALUOperation operation, u32 src_a, u32 src_b);
/// Performs the result operation on the input result and stores it in the specified register
/// (if necessary).
@@ -165,5 +165,7 @@ private:
std::vector<u32> parameters;
/// Index of the next parameter that will be fetched by the 'parm' instruction.
u32 next_parameter_index = 0;
+
+ bool carry_flag{};
};
} // namespace Tegra
diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp
index 77a20bb84..47247f097 100644
--- a/src/video_core/memory_manager.cpp
+++ b/src/video_core/memory_manager.cpp
@@ -9,6 +9,13 @@
namespace Tegra {
+MemoryManager::MemoryManager() {
+ // Mark the first page as reserved, so that 0 is not a valid GPUVAddr. Otherwise, games might
+ // try to use 0 as a valid address, which is also used to mean nullptr. This fixes a bug with
+ // Undertale using 0 for a render target.
+ PageSlot(0) = static_cast<u64>(PageStatus::Reserved);
+}
+
GPUVAddr MemoryManager::AllocateSpace(u64 size, u64 align) {
const std::optional<GPUVAddr> gpu_addr{FindFreeBlock(0, size, align, PageStatus::Unmapped)};
diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h
index 4eb338aa2..fb03497ca 100644
--- a/src/video_core/memory_manager.h
+++ b/src/video_core/memory_manager.h
@@ -18,7 +18,7 @@ using GPUVAddr = u64;
class MemoryManager final {
public:
- MemoryManager() = default;
+ MemoryManager();
GPUVAddr AllocateSpace(u64 size, u64 align);
GPUVAddr AllocateSpace(GPUVAddr gpu_addr, u64 size, u64 align);
@@ -37,6 +37,7 @@ private:
enum class PageStatus : u64 {
Unmapped = 0xFFFFFFFFFFFFFFFFULL,
Allocated = 0xFFFFFFFFFFFFFFFEULL,
+ Reserved = 0xFFFFFFFFFFFFFFFDULL,
};
std::optional<GPUVAddr> FindFreeBlock(GPUVAddr region_start, u64 size, u64 align,
diff --git a/src/video_core/morton.cpp b/src/video_core/morton.cpp
new file mode 100644
index 000000000..f14abba7d
--- /dev/null
+++ b/src/video_core/morton.cpp
@@ -0,0 +1,353 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+#include <cstring>
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "core/memory.h"
+#include "video_core/morton.h"
+#include "video_core/surface.h"
+#include "video_core/textures/decoders.h"
+
+namespace VideoCore {
+
+using Surface::GetBytesPerPixel;
+using Surface::PixelFormat;
+
+using MortonCopyFn = void (*)(u32, u32, u32, u32, u32, u8*, std::size_t, VAddr);
+using ConversionArray = std::array<MortonCopyFn, Surface::MaxPixelFormat>;
+
+template <bool morton_to_linear, PixelFormat format>
+static void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth, u32 depth,
+ u8* buffer, std::size_t buffer_size, VAddr addr) {
+ constexpr u32 bytes_per_pixel = GetBytesPerPixel(format);
+
+ // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual
+ // pixel values.
+ const u32 tile_size_x{GetDefaultBlockWidth(format)};
+ const u32 tile_size_y{GetDefaultBlockHeight(format)};
+
+ if constexpr (morton_to_linear) {
+ Tegra::Texture::UnswizzleTexture(buffer, addr, tile_size_x, tile_size_y, bytes_per_pixel,
+ stride, height, depth, block_height, block_depth);
+ } else {
+ Tegra::Texture::CopySwizzledData((stride + tile_size_x - 1) / tile_size_x,
+ (height + tile_size_y - 1) / tile_size_y, depth,
+ bytes_per_pixel, bytes_per_pixel, Memory::GetPointer(addr),
+ buffer, false, block_height, block_depth);
+ }
+}
+
+static constexpr ConversionArray morton_to_linear_fns = {
+ // clang-format off
+ MortonCopy<true, PixelFormat::ABGR8U>,
+ MortonCopy<true, PixelFormat::ABGR8S>,
+ MortonCopy<true, PixelFormat::ABGR8UI>,
+ MortonCopy<true, PixelFormat::B5G6R5U>,
+ MortonCopy<true, PixelFormat::A2B10G10R10U>,
+ MortonCopy<true, PixelFormat::A1B5G5R5U>,
+ MortonCopy<true, PixelFormat::R8U>,
+ MortonCopy<true, PixelFormat::R8UI>,
+ MortonCopy<true, PixelFormat::RGBA16F>,
+ MortonCopy<true, PixelFormat::RGBA16U>,
+ MortonCopy<true, PixelFormat::RGBA16UI>,
+ MortonCopy<true, PixelFormat::R11FG11FB10F>,
+ MortonCopy<true, PixelFormat::RGBA32UI>,
+ MortonCopy<true, PixelFormat::DXT1>,
+ MortonCopy<true, PixelFormat::DXT23>,
+ MortonCopy<true, PixelFormat::DXT45>,
+ MortonCopy<true, PixelFormat::DXN1>,
+ MortonCopy<true, PixelFormat::DXN2UNORM>,
+ MortonCopy<true, PixelFormat::DXN2SNORM>,
+ MortonCopy<true, PixelFormat::BC7U>,
+ MortonCopy<true, PixelFormat::BC6H_UF16>,
+ MortonCopy<true, PixelFormat::BC6H_SF16>,
+ MortonCopy<true, PixelFormat::ASTC_2D_4X4>,
+ MortonCopy<true, PixelFormat::G8R8U>,
+ MortonCopy<true, PixelFormat::G8R8S>,
+ MortonCopy<true, PixelFormat::BGRA8>,
+ MortonCopy<true, PixelFormat::RGBA32F>,
+ MortonCopy<true, PixelFormat::RG32F>,
+ MortonCopy<true, PixelFormat::R32F>,
+ MortonCopy<true, PixelFormat::R16F>,
+ MortonCopy<true, PixelFormat::R16U>,
+ MortonCopy<true, PixelFormat::R16S>,
+ MortonCopy<true, PixelFormat::R16UI>,
+ MortonCopy<true, PixelFormat::R16I>,
+ MortonCopy<true, PixelFormat::RG16>,
+ MortonCopy<true, PixelFormat::RG16F>,
+ MortonCopy<true, PixelFormat::RG16UI>,
+ MortonCopy<true, PixelFormat::RG16I>,
+ MortonCopy<true, PixelFormat::RG16S>,
+ MortonCopy<true, PixelFormat::RGB32F>,
+ MortonCopy<true, PixelFormat::RGBA8_SRGB>,
+ MortonCopy<true, PixelFormat::RG8U>,
+ MortonCopy<true, PixelFormat::RG8S>,
+ MortonCopy<true, PixelFormat::RG32UI>,
+ MortonCopy<true, PixelFormat::R32UI>,
+ MortonCopy<true, PixelFormat::ASTC_2D_8X8>,
+ MortonCopy<true, PixelFormat::ASTC_2D_8X5>,
+ MortonCopy<true, PixelFormat::ASTC_2D_5X4>,
+ MortonCopy<true, PixelFormat::BGRA8_SRGB>,
+ MortonCopy<true, PixelFormat::DXT1_SRGB>,
+ MortonCopy<true, PixelFormat::DXT23_SRGB>,
+ MortonCopy<true, PixelFormat::DXT45_SRGB>,
+ MortonCopy<true, PixelFormat::BC7U_SRGB>,
+ MortonCopy<true, PixelFormat::ASTC_2D_4X4_SRGB>,
+ MortonCopy<true, PixelFormat::ASTC_2D_8X8_SRGB>,
+ MortonCopy<true, PixelFormat::ASTC_2D_8X5_SRGB>,
+ MortonCopy<true, PixelFormat::ASTC_2D_5X4_SRGB>,
+ MortonCopy<true, PixelFormat::ASTC_2D_5X5>,
+ MortonCopy<true, PixelFormat::ASTC_2D_5X5_SRGB>,
+ MortonCopy<true, PixelFormat::ASTC_2D_10X8>,
+ MortonCopy<true, PixelFormat::ASTC_2D_10X8_SRGB>,
+ MortonCopy<true, PixelFormat::Z32F>,
+ MortonCopy<true, PixelFormat::Z16>,
+ MortonCopy<true, PixelFormat::Z24S8>,
+ MortonCopy<true, PixelFormat::S8Z24>,
+ MortonCopy<true, PixelFormat::Z32FS8>,
+ // clang-format on
+};
+
+static constexpr ConversionArray linear_to_morton_fns = {
+ // clang-format off
+ MortonCopy<false, PixelFormat::ABGR8U>,
+ MortonCopy<false, PixelFormat::ABGR8S>,
+ MortonCopy<false, PixelFormat::ABGR8UI>,
+ MortonCopy<false, PixelFormat::B5G6R5U>,
+ MortonCopy<false, PixelFormat::A2B10G10R10U>,
+ MortonCopy<false, PixelFormat::A1B5G5R5U>,
+ MortonCopy<false, PixelFormat::R8U>,
+ MortonCopy<false, PixelFormat::R8UI>,
+ MortonCopy<false, PixelFormat::RGBA16F>,
+ MortonCopy<false, PixelFormat::RGBA16U>,
+ MortonCopy<false, PixelFormat::RGBA16UI>,
+ MortonCopy<false, PixelFormat::R11FG11FB10F>,
+ MortonCopy<false, PixelFormat::RGBA32UI>,
+ MortonCopy<false, PixelFormat::DXT1>,
+ MortonCopy<false, PixelFormat::DXT23>,
+ MortonCopy<false, PixelFormat::DXT45>,
+ MortonCopy<false, PixelFormat::DXN1>,
+ MortonCopy<false, PixelFormat::DXN2UNORM>,
+ MortonCopy<false, PixelFormat::DXN2SNORM>,
+ MortonCopy<false, PixelFormat::BC7U>,
+ MortonCopy<false, PixelFormat::BC6H_UF16>,
+ MortonCopy<false, PixelFormat::BC6H_SF16>,
+ // TODO(Subv): Swizzling ASTC formats are not supported
+ nullptr,
+ MortonCopy<false, PixelFormat::G8R8U>,
+ MortonCopy<false, PixelFormat::G8R8S>,
+ MortonCopy<false, PixelFormat::BGRA8>,
+ MortonCopy<false, PixelFormat::RGBA32F>,
+ MortonCopy<false, PixelFormat::RG32F>,
+ MortonCopy<false, PixelFormat::R32F>,
+ MortonCopy<false, PixelFormat::R16F>,
+ MortonCopy<false, PixelFormat::R16U>,
+ MortonCopy<false, PixelFormat::R16S>,
+ MortonCopy<false, PixelFormat::R16UI>,
+ MortonCopy<false, PixelFormat::R16I>,
+ MortonCopy<false, PixelFormat::RG16>,
+ MortonCopy<false, PixelFormat::RG16F>,
+ MortonCopy<false, PixelFormat::RG16UI>,
+ MortonCopy<false, PixelFormat::RG16I>,
+ MortonCopy<false, PixelFormat::RG16S>,
+ MortonCopy<false, PixelFormat::RGB32F>,
+ MortonCopy<false, PixelFormat::RGBA8_SRGB>,
+ MortonCopy<false, PixelFormat::RG8U>,
+ MortonCopy<false, PixelFormat::RG8S>,
+ MortonCopy<false, PixelFormat::RG32UI>,
+ MortonCopy<false, PixelFormat::R32UI>,
+ nullptr,
+ nullptr,
+ nullptr,
+ MortonCopy<false, PixelFormat::BGRA8_SRGB>,
+ MortonCopy<false, PixelFormat::DXT1_SRGB>,
+ MortonCopy<false, PixelFormat::DXT23_SRGB>,
+ MortonCopy<false, PixelFormat::DXT45_SRGB>,
+ MortonCopy<false, PixelFormat::BC7U_SRGB>,
+ nullptr,
+ nullptr,
+ nullptr,
+ nullptr,
+ nullptr,
+ nullptr,
+ nullptr,
+ nullptr,
+ MortonCopy<false, PixelFormat::Z32F>,
+ MortonCopy<false, PixelFormat::Z16>,
+ MortonCopy<false, PixelFormat::Z24S8>,
+ MortonCopy<false, PixelFormat::S8Z24>,
+ MortonCopy<false, PixelFormat::Z32FS8>,
+ // clang-format on
+};
+
+constexpr MortonCopyFn GetSwizzleFunction(MortonSwizzleMode mode, Surface::PixelFormat format) {
+ switch (mode) {
+ case MortonSwizzleMode::MortonToLinear:
+ return morton_to_linear_fns[static_cast<std::size_t>(format)];
+ case MortonSwizzleMode::LinearToMorton:
+ return linear_to_morton_fns[static_cast<std::size_t>(format)];
+ }
+}
+
+/// 8x8 Z-Order coordinate from 2D coordinates
+static u32 MortonInterleave(u32 x, u32 y) {
+ static const u32 xlut[] = {0x00, 0x01, 0x04, 0x05, 0x10, 0x11, 0x14, 0x15};
+ static const u32 ylut[] = {0x00, 0x02, 0x08, 0x0a, 0x20, 0x22, 0x28, 0x2a};
+ return xlut[x % 8] + ylut[y % 8];
+}
+
+/// Calculates the offset of the position of the pixel in Morton order
+static u32 GetMortonOffset(u32 x, u32 y, u32 bytes_per_pixel) {
+ // Images are split into 8x8 tiles. Each tile is composed of four 4x4 subtiles each
+ // of which is composed of four 2x2 subtiles each of which is composed of four texels.
+ // Each structure is embedded into the next-bigger one in a diagonal pattern, e.g.
+ // texels are laid out in a 2x2 subtile like this:
+ // 2 3
+ // 0 1
+ //
+ // The full 8x8 tile has the texels arranged like this:
+ //
+ // 42 43 46 47 58 59 62 63
+ // 40 41 44 45 56 57 60 61
+ // 34 35 38 39 50 51 54 55
+ // 32 33 36 37 48 49 52 53
+ // 10 11 14 15 26 27 30 31
+ // 08 09 12 13 24 25 28 29
+ // 02 03 06 07 18 19 22 23
+ // 00 01 04 05 16 17 20 21
+ //
+ // This pattern is what's called Z-order curve, or Morton order.
+
+ const unsigned int block_height = 8;
+ const unsigned int coarse_x = x & ~7;
+
+ u32 i = MortonInterleave(x, y);
+
+ const unsigned int offset = coarse_x * block_height;
+
+ return (i + offset) * bytes_per_pixel;
+}
+
+static u32 MortonInterleave128(u32 x, u32 y) {
+ // 128x128 Z-Order coordinate from 2D coordinates
+ static constexpr u32 xlut[] = {
+ 0x0000, 0x0001, 0x0002, 0x0003, 0x0008, 0x0009, 0x000a, 0x000b, 0x0040, 0x0041, 0x0042,
+ 0x0043, 0x0048, 0x0049, 0x004a, 0x004b, 0x0800, 0x0801, 0x0802, 0x0803, 0x0808, 0x0809,
+ 0x080a, 0x080b, 0x0840, 0x0841, 0x0842, 0x0843, 0x0848, 0x0849, 0x084a, 0x084b, 0x1000,
+ 0x1001, 0x1002, 0x1003, 0x1008, 0x1009, 0x100a, 0x100b, 0x1040, 0x1041, 0x1042, 0x1043,
+ 0x1048, 0x1049, 0x104a, 0x104b, 0x1800, 0x1801, 0x1802, 0x1803, 0x1808, 0x1809, 0x180a,
+ 0x180b, 0x1840, 0x1841, 0x1842, 0x1843, 0x1848, 0x1849, 0x184a, 0x184b, 0x2000, 0x2001,
+ 0x2002, 0x2003, 0x2008, 0x2009, 0x200a, 0x200b, 0x2040, 0x2041, 0x2042, 0x2043, 0x2048,
+ 0x2049, 0x204a, 0x204b, 0x2800, 0x2801, 0x2802, 0x2803, 0x2808, 0x2809, 0x280a, 0x280b,
+ 0x2840, 0x2841, 0x2842, 0x2843, 0x2848, 0x2849, 0x284a, 0x284b, 0x3000, 0x3001, 0x3002,
+ 0x3003, 0x3008, 0x3009, 0x300a, 0x300b, 0x3040, 0x3041, 0x3042, 0x3043, 0x3048, 0x3049,
+ 0x304a, 0x304b, 0x3800, 0x3801, 0x3802, 0x3803, 0x3808, 0x3809, 0x380a, 0x380b, 0x3840,
+ 0x3841, 0x3842, 0x3843, 0x3848, 0x3849, 0x384a, 0x384b, 0x0000, 0x0001, 0x0002, 0x0003,
+ 0x0008, 0x0009, 0x000a, 0x000b, 0x0040, 0x0041, 0x0042, 0x0043, 0x0048, 0x0049, 0x004a,
+ 0x004b, 0x0800, 0x0801, 0x0802, 0x0803, 0x0808, 0x0809, 0x080a, 0x080b, 0x0840, 0x0841,
+ 0x0842, 0x0843, 0x0848, 0x0849, 0x084a, 0x084b, 0x1000, 0x1001, 0x1002, 0x1003, 0x1008,
+ 0x1009, 0x100a, 0x100b, 0x1040, 0x1041, 0x1042, 0x1043, 0x1048, 0x1049, 0x104a, 0x104b,
+ 0x1800, 0x1801, 0x1802, 0x1803, 0x1808, 0x1809, 0x180a, 0x180b, 0x1840, 0x1841, 0x1842,
+ 0x1843, 0x1848, 0x1849, 0x184a, 0x184b, 0x2000, 0x2001, 0x2002, 0x2003, 0x2008, 0x2009,
+ 0x200a, 0x200b, 0x2040, 0x2041, 0x2042, 0x2043, 0x2048, 0x2049, 0x204a, 0x204b, 0x2800,
+ 0x2801, 0x2802, 0x2803, 0x2808, 0x2809, 0x280a, 0x280b, 0x2840, 0x2841, 0x2842, 0x2843,
+ 0x2848, 0x2849, 0x284a, 0x284b, 0x3000, 0x3001, 0x3002, 0x3003, 0x3008, 0x3009, 0x300a,
+ 0x300b, 0x3040, 0x3041, 0x3042, 0x3043, 0x3048, 0x3049, 0x304a, 0x304b, 0x3800, 0x3801,
+ 0x3802, 0x3803, 0x3808, 0x3809, 0x380a, 0x380b, 0x3840, 0x3841, 0x3842, 0x3843, 0x3848,
+ 0x3849, 0x384a, 0x384b, 0x0000, 0x0001, 0x0002, 0x0003, 0x0008, 0x0009, 0x000a, 0x000b,
+ 0x0040, 0x0041, 0x0042, 0x0043, 0x0048, 0x0049, 0x004a, 0x004b, 0x0800, 0x0801, 0x0802,
+ 0x0803, 0x0808, 0x0809, 0x080a, 0x080b, 0x0840, 0x0841, 0x0842, 0x0843, 0x0848, 0x0849,
+ 0x084a, 0x084b, 0x1000, 0x1001, 0x1002, 0x1003, 0x1008, 0x1009, 0x100a, 0x100b, 0x1040,
+ 0x1041, 0x1042, 0x1043, 0x1048, 0x1049, 0x104a, 0x104b, 0x1800, 0x1801, 0x1802, 0x1803,
+ 0x1808, 0x1809, 0x180a, 0x180b, 0x1840, 0x1841, 0x1842, 0x1843, 0x1848, 0x1849, 0x184a,
+ 0x184b, 0x2000, 0x2001, 0x2002, 0x2003, 0x2008, 0x2009, 0x200a, 0x200b, 0x2040, 0x2041,
+ 0x2042, 0x2043, 0x2048, 0x2049, 0x204a, 0x204b, 0x2800, 0x2801, 0x2802, 0x2803, 0x2808,
+ 0x2809, 0x280a, 0x280b, 0x2840, 0x2841, 0x2842, 0x2843, 0x2848, 0x2849, 0x284a, 0x284b,
+ 0x3000, 0x3001, 0x3002, 0x3003, 0x3008, 0x3009, 0x300a, 0x300b, 0x3040, 0x3041, 0x3042,
+ 0x3043, 0x3048, 0x3049, 0x304a, 0x304b, 0x3800, 0x3801, 0x3802, 0x3803, 0x3808, 0x3809,
+ 0x380a, 0x380b, 0x3840, 0x3841, 0x3842, 0x3843, 0x3848, 0x3849, 0x384a, 0x384b,
+ };
+ static constexpr u32 ylut[] = {
+ 0x0000, 0x0004, 0x0010, 0x0014, 0x0020, 0x0024, 0x0030, 0x0034, 0x0080, 0x0084, 0x0090,
+ 0x0094, 0x00a0, 0x00a4, 0x00b0, 0x00b4, 0x0100, 0x0104, 0x0110, 0x0114, 0x0120, 0x0124,
+ 0x0130, 0x0134, 0x0180, 0x0184, 0x0190, 0x0194, 0x01a0, 0x01a4, 0x01b0, 0x01b4, 0x0200,
+ 0x0204, 0x0210, 0x0214, 0x0220, 0x0224, 0x0230, 0x0234, 0x0280, 0x0284, 0x0290, 0x0294,
+ 0x02a0, 0x02a4, 0x02b0, 0x02b4, 0x0300, 0x0304, 0x0310, 0x0314, 0x0320, 0x0324, 0x0330,
+ 0x0334, 0x0380, 0x0384, 0x0390, 0x0394, 0x03a0, 0x03a4, 0x03b0, 0x03b4, 0x0400, 0x0404,
+ 0x0410, 0x0414, 0x0420, 0x0424, 0x0430, 0x0434, 0x0480, 0x0484, 0x0490, 0x0494, 0x04a0,
+ 0x04a4, 0x04b0, 0x04b4, 0x0500, 0x0504, 0x0510, 0x0514, 0x0520, 0x0524, 0x0530, 0x0534,
+ 0x0580, 0x0584, 0x0590, 0x0594, 0x05a0, 0x05a4, 0x05b0, 0x05b4, 0x0600, 0x0604, 0x0610,
+ 0x0614, 0x0620, 0x0624, 0x0630, 0x0634, 0x0680, 0x0684, 0x0690, 0x0694, 0x06a0, 0x06a4,
+ 0x06b0, 0x06b4, 0x0700, 0x0704, 0x0710, 0x0714, 0x0720, 0x0724, 0x0730, 0x0734, 0x0780,
+ 0x0784, 0x0790, 0x0794, 0x07a0, 0x07a4, 0x07b0, 0x07b4, 0x0000, 0x0004, 0x0010, 0x0014,
+ 0x0020, 0x0024, 0x0030, 0x0034, 0x0080, 0x0084, 0x0090, 0x0094, 0x00a0, 0x00a4, 0x00b0,
+ 0x00b4, 0x0100, 0x0104, 0x0110, 0x0114, 0x0120, 0x0124, 0x0130, 0x0134, 0x0180, 0x0184,
+ 0x0190, 0x0194, 0x01a0, 0x01a4, 0x01b0, 0x01b4, 0x0200, 0x0204, 0x0210, 0x0214, 0x0220,
+ 0x0224, 0x0230, 0x0234, 0x0280, 0x0284, 0x0290, 0x0294, 0x02a0, 0x02a4, 0x02b0, 0x02b4,
+ 0x0300, 0x0304, 0x0310, 0x0314, 0x0320, 0x0324, 0x0330, 0x0334, 0x0380, 0x0384, 0x0390,
+ 0x0394, 0x03a0, 0x03a4, 0x03b0, 0x03b4, 0x0400, 0x0404, 0x0410, 0x0414, 0x0420, 0x0424,
+ 0x0430, 0x0434, 0x0480, 0x0484, 0x0490, 0x0494, 0x04a0, 0x04a4, 0x04b0, 0x04b4, 0x0500,
+ 0x0504, 0x0510, 0x0514, 0x0520, 0x0524, 0x0530, 0x0534, 0x0580, 0x0584, 0x0590, 0x0594,
+ 0x05a0, 0x05a4, 0x05b0, 0x05b4, 0x0600, 0x0604, 0x0610, 0x0614, 0x0620, 0x0624, 0x0630,
+ 0x0634, 0x0680, 0x0684, 0x0690, 0x0694, 0x06a0, 0x06a4, 0x06b0, 0x06b4, 0x0700, 0x0704,
+ 0x0710, 0x0714, 0x0720, 0x0724, 0x0730, 0x0734, 0x0780, 0x0784, 0x0790, 0x0794, 0x07a0,
+ 0x07a4, 0x07b0, 0x07b4, 0x0000, 0x0004, 0x0010, 0x0014, 0x0020, 0x0024, 0x0030, 0x0034,
+ 0x0080, 0x0084, 0x0090, 0x0094, 0x00a0, 0x00a4, 0x00b0, 0x00b4, 0x0100, 0x0104, 0x0110,
+ 0x0114, 0x0120, 0x0124, 0x0130, 0x0134, 0x0180, 0x0184, 0x0190, 0x0194, 0x01a0, 0x01a4,
+ 0x01b0, 0x01b4, 0x0200, 0x0204, 0x0210, 0x0214, 0x0220, 0x0224, 0x0230, 0x0234, 0x0280,
+ 0x0284, 0x0290, 0x0294, 0x02a0, 0x02a4, 0x02b0, 0x02b4, 0x0300, 0x0304, 0x0310, 0x0314,
+ 0x0320, 0x0324, 0x0330, 0x0334, 0x0380, 0x0384, 0x0390, 0x0394, 0x03a0, 0x03a4, 0x03b0,
+ 0x03b4, 0x0400, 0x0404, 0x0410, 0x0414, 0x0420, 0x0424, 0x0430, 0x0434, 0x0480, 0x0484,
+ 0x0490, 0x0494, 0x04a0, 0x04a4, 0x04b0, 0x04b4, 0x0500, 0x0504, 0x0510, 0x0514, 0x0520,
+ 0x0524, 0x0530, 0x0534, 0x0580, 0x0584, 0x0590, 0x0594, 0x05a0, 0x05a4, 0x05b0, 0x05b4,
+ 0x0600, 0x0604, 0x0610, 0x0614, 0x0620, 0x0624, 0x0630, 0x0634, 0x0680, 0x0684, 0x0690,
+ 0x0694, 0x06a0, 0x06a4, 0x06b0, 0x06b4, 0x0700, 0x0704, 0x0710, 0x0714, 0x0720, 0x0724,
+ 0x0730, 0x0734, 0x0780, 0x0784, 0x0790, 0x0794, 0x07a0, 0x07a4, 0x07b0, 0x07b4,
+ };
+ return xlut[x % 128] + ylut[y % 128];
+}
+
+static u32 GetMortonOffset128(u32 x, u32 y, u32 bytes_per_pixel) {
+ // Calculates the offset of the position of the pixel in Morton order
+ // Framebuffer images are split into 128x128 tiles.
+
+ constexpr u32 block_height = 128;
+ const u32 coarse_x = x & ~127;
+
+ const u32 i = MortonInterleave128(x, y);
+
+ const u32 offset = coarse_x * block_height;
+
+ return (i + offset) * bytes_per_pixel;
+}
+
+void MortonSwizzle(MortonSwizzleMode mode, Surface::PixelFormat format, u32 stride,
+ u32 block_height, u32 height, u32 block_depth, u32 depth, u8* buffer,
+ std::size_t buffer_size, VAddr addr) {
+
+ GetSwizzleFunction(mode, format)(stride, block_height, height, block_depth, depth, buffer,
+ buffer_size, addr);
+}
+
+void MortonCopyPixels128(u32 width, u32 height, u32 bytes_per_pixel, u32 linear_bytes_per_pixel,
+ u8* morton_data, u8* linear_data, bool morton_to_linear) {
+ u8* data_ptrs[2];
+ for (u32 y = 0; y < height; ++y) {
+ for (u32 x = 0; x < width; ++x) {
+ const u32 coarse_y = y & ~127;
+ const u32 morton_offset =
+ GetMortonOffset128(x, y, bytes_per_pixel) + coarse_y * width * bytes_per_pixel;
+ const u32 linear_pixel_index = (x + y * width) * linear_bytes_per_pixel;
+
+ data_ptrs[morton_to_linear ? 1 : 0] = morton_data + morton_offset;
+ data_ptrs[morton_to_linear ? 0 : 1] = &linear_data[linear_pixel_index];
+
+ std::memcpy(data_ptrs[0], data_ptrs[1], bytes_per_pixel);
+ }
+ }
+}
+
+} // namespace VideoCore \ No newline at end of file
diff --git a/src/video_core/morton.h b/src/video_core/morton.h
new file mode 100644
index 000000000..b9b9eca86
--- /dev/null
+++ b/src/video_core/morton.h
@@ -0,0 +1,21 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+#include "video_core/surface.h"
+
+namespace VideoCore {
+
+enum class MortonSwizzleMode { MortonToLinear, LinearToMorton };
+
+void MortonSwizzle(MortonSwizzleMode mode, VideoCore::Surface::PixelFormat format, u32 stride,
+ u32 block_height, u32 height, u32 block_depth, u32 depth, u8* buffer,
+ std::size_t buffer_size, VAddr addr);
+
+void MortonCopyPixels128(u32 width, u32 height, u32 bytes_per_pixel, u32 linear_bytes_per_pixel,
+ u8* morton_data, u8* linear_data, bool morton_to_linear);
+
+} // namespace VideoCore \ No newline at end of file
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index ae6aaee4c..630a58e49 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -98,14 +98,9 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, ScreenInfo
has_ARB_direct_state_access = true;
} else if (extension == "GL_ARB_multi_bind") {
has_ARB_multi_bind = true;
- } else if (extension == "GL_ARB_separate_shader_objects") {
- has_ARB_separate_shader_objects = true;
- } else if (extension == "GL_ARB_vertex_attrib_binding") {
- has_ARB_vertex_attrib_binding = true;
}
}
- ASSERT_MSG(has_ARB_separate_shader_objects, "has_ARB_separate_shader_objects is unsupported");
OpenGLState::ApplyDefaultState();
// Create render framebuffer
@@ -542,6 +537,30 @@ void RasterizerOpenGL::Clear() {
ASSERT_MSG(regs.zeta_enable != 0, "Tried to clear stencil but buffer is not enabled!");
use_stencil = true;
clear_state.stencil.test_enabled = true;
+ if (regs.clear_flags.stencil) {
+ // Stencil affects the clear so fill it with the used masks
+ clear_state.stencil.front.test_func = GL_ALWAYS;
+ clear_state.stencil.front.test_mask = regs.stencil_front_func_mask;
+ clear_state.stencil.front.action_stencil_fail = GL_KEEP;
+ clear_state.stencil.front.action_depth_fail = GL_KEEP;
+ clear_state.stencil.front.action_depth_pass = GL_KEEP;
+ clear_state.stencil.front.write_mask = regs.stencil_front_mask;
+ if (regs.stencil_two_side_enable) {
+ clear_state.stencil.back.test_func = GL_ALWAYS;
+ clear_state.stencil.back.test_mask = regs.stencil_back_func_mask;
+ clear_state.stencil.back.action_stencil_fail = GL_KEEP;
+ clear_state.stencil.back.action_depth_fail = GL_KEEP;
+ clear_state.stencil.back.action_depth_pass = GL_KEEP;
+ clear_state.stencil.back.write_mask = regs.stencil_back_mask;
+ } else {
+ clear_state.stencil.back.test_func = GL_ALWAYS;
+ clear_state.stencil.back.test_mask = 0xFFFFFFFF;
+ clear_state.stencil.back.write_mask = 0xFFFFFFFF;
+ clear_state.stencil.back.action_stencil_fail = GL_KEEP;
+ clear_state.stencil.back.action_depth_fail = GL_KEEP;
+ clear_state.stencil.back.action_depth_pass = GL_KEEP;
+ }
+ }
}
if (!use_color && !use_depth && !use_stencil) {
@@ -553,6 +572,14 @@ void RasterizerOpenGL::Clear() {
ConfigureFramebuffers(clear_state, use_color, use_depth || use_stencil, false,
regs.clear_buffers.RT.Value());
+ if (regs.clear_flags.scissor) {
+ SyncScissorTest(clear_state);
+ }
+
+ if (regs.clear_flags.viewport) {
+ clear_state.EmulateViewportWithScissor();
+ }
+
clear_state.Apply();
if (use_color) {
@@ -588,7 +615,7 @@ void RasterizerOpenGL::DrawArrays() {
SyncLogicOpState();
SyncCullMode();
SyncPrimitiveRestart();
- SyncScissorTest();
+ SyncScissorTest(state);
// Alpha Testing is synced on shaders.
SyncTransformFeedback();
SyncPointState();
@@ -815,7 +842,7 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntr
}
const u32 bias = config.mip_lod_bias.Value();
// Sign extend the 13-bit value.
- const u32 mask = 1U << (13 - 1);
+ constexpr u32 mask = 1U << (13 - 1);
const float bias_lod = static_cast<s32>((bias ^ mask) - mask) / 256.f;
if (lod_bias != bias_lod) {
lod_bias = bias_lod;
@@ -947,8 +974,8 @@ void RasterizerOpenGL::SyncViewport(OpenGLState& current_state) {
auto& viewport = current_state.viewports[i];
viewport.x = viewport_rect.left;
viewport.y = viewport_rect.bottom;
- viewport.width = static_cast<GLfloat>(viewport_rect.GetWidth());
- viewport.height = static_cast<GLfloat>(viewport_rect.GetHeight());
+ viewport.width = viewport_rect.GetWidth();
+ viewport.height = viewport_rect.GetHeight();
viewport.depth_range_far = regs.viewports[i].depth_range_far;
viewport.depth_range_near = regs.viewports[i].depth_range_near;
}
@@ -1120,11 +1147,11 @@ void RasterizerOpenGL::SyncLogicOpState() {
state.logic_op.operation = MaxwellToGL::LogicOp(regs.logic_op.operation);
}
-void RasterizerOpenGL::SyncScissorTest() {
+void RasterizerOpenGL::SyncScissorTest(OpenGLState& current_state) {
const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumViewports; i++) {
const auto& src = regs.scissor_test[i];
- auto& dst = state.viewports[i].scissor;
+ auto& dst = current_state.viewports[i].scissor;
dst.enabled = (src.enable != 0);
if (dst.enabled == 0) {
return;
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index 6e78ab4cd..f4354289c 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -91,19 +91,20 @@ private:
void SyncWithConfig(const Tegra::Texture::TSCEntry& info);
private:
- Tegra::Texture::TextureFilter mag_filter;
- Tegra::Texture::TextureFilter min_filter;
- Tegra::Texture::TextureMipmapFilter mip_filter;
- Tegra::Texture::WrapMode wrap_u;
- Tegra::Texture::WrapMode wrap_v;
- Tegra::Texture::WrapMode wrap_p;
- bool uses_depth_compare;
- Tegra::Texture::DepthCompareFunc depth_compare_func;
- GLvec4 border_color;
- float min_lod;
- float max_lod;
- float lod_bias;
- float max_anisotropic;
+ Tegra::Texture::TextureFilter mag_filter = Tegra::Texture::TextureFilter::Nearest;
+ Tegra::Texture::TextureFilter min_filter = Tegra::Texture::TextureFilter::Nearest;
+ Tegra::Texture::TextureMipmapFilter mip_filter = Tegra::Texture::TextureMipmapFilter::None;
+ Tegra::Texture::WrapMode wrap_u = Tegra::Texture::WrapMode::ClampToEdge;
+ Tegra::Texture::WrapMode wrap_v = Tegra::Texture::WrapMode::ClampToEdge;
+ Tegra::Texture::WrapMode wrap_p = Tegra::Texture::WrapMode::ClampToEdge;
+ bool uses_depth_compare = false;
+ Tegra::Texture::DepthCompareFunc depth_compare_func =
+ Tegra::Texture::DepthCompareFunc::Always;
+ GLvec4 border_color = {};
+ float min_lod = 0.0f;
+ float max_lod = 16.0f;
+ float lod_bias = 0.0f;
+ float max_anisotropic = 1.0f;
};
/**
@@ -171,7 +172,7 @@ private:
void SyncMultiSampleState();
/// Syncs the scissor test state to match the guest state
- void SyncScissorTest();
+ void SyncScissorTest(OpenGLState& current_state);
/// Syncs the transform feedback state to match the guest state
void SyncTransformFeedback();
@@ -187,8 +188,6 @@ private:
bool has_ARB_direct_state_access = false;
bool has_ARB_multi_bind = false;
- bool has_ARB_separate_shader_objects = false;
- bool has_ARB_vertex_attrib_binding = false;
OpenGLState state;
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 9ca82c06c..d458f77e4 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -15,6 +15,7 @@
#include "core/memory.h"
#include "core/settings.h"
#include "video_core/engines/maxwell_3d.h"
+#include "video_core/morton.h"
#include "video_core/renderer_opengl/gl_rasterizer.h"
#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
#include "video_core/renderer_opengl/gl_state.h"
@@ -22,10 +23,11 @@
#include "video_core/surface.h"
#include "video_core/textures/astc.h"
#include "video_core/textures/decoders.h"
-#include "video_core/utils.h"
namespace OpenGL {
+using VideoCore::MortonSwizzle;
+using VideoCore::MortonSwizzleMode;
using VideoCore::Surface::ComponentTypeFromDepthFormat;
using VideoCore::Surface::ComponentTypeFromRenderTarget;
using VideoCore::Surface::ComponentTypeFromTexture;
@@ -265,11 +267,11 @@ static constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex
{GL_COMPRESSED_RG_RGTC2, GL_RG, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
true}, // DXN2UNORM
{GL_COMPRESSED_SIGNED_RG_RGTC2, GL_RG, GL_INT, ComponentType::SNorm, true}, // DXN2SNORM
- {GL_COMPRESSED_RGBA_BPTC_UNORM_ARB, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
+ {GL_COMPRESSED_RGBA_BPTC_UNORM, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
true}, // BC7U
- {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT_ARB, GL_RGB, GL_UNSIGNED_INT_8_8_8_8,
- ComponentType::Float, true}, // BC6H_UF16
- {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT_ARB, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::Float,
+ {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::Float,
+ true}, // BC6H_UF16
+ {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::Float,
true}, // BC6H_SF16
{GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_4X4
{GL_RG8, GL_RG, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // G8R8U
@@ -306,8 +308,8 @@ static constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex
true}, // DXT23_SRGB
{GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
true}, // DXT45_SRGB
- {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM_ARB, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8,
- ComponentType::UNorm, true}, // BC7U_SRGB
+ {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
+ true}, // BC7U_SRGB
{GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_4X4_SRGB
{GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_8X8_SRGB
{GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_8X5_SRGB
@@ -346,7 +348,7 @@ static GLenum SurfaceTargetToGL(SurfaceTarget target) {
case SurfaceTarget::TextureCubemap:
return GL_TEXTURE_CUBE_MAP;
case SurfaceTarget::TextureCubeArray:
- return GL_TEXTURE_CUBE_MAP_ARRAY_ARB;
+ return GL_TEXTURE_CUBE_MAP_ARRAY;
}
LOG_CRITICAL(Render_OpenGL, "Unimplemented texture target={}", static_cast<u32>(target));
UNREACHABLE();
@@ -370,174 +372,7 @@ MathUtil::Rectangle<u32> SurfaceParams::GetRect(u32 mip_level) const {
return {0, actual_height, MipWidth(mip_level), 0};
}
-template <bool morton_to_gl, PixelFormat format>
-void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth, u32 depth, u8* gl_buffer,
- std::size_t gl_buffer_size, VAddr addr) {
- constexpr u32 bytes_per_pixel = GetBytesPerPixel(format);
-
- // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual
- // pixel values.
- const u32 tile_size_x{GetDefaultBlockWidth(format)};
- const u32 tile_size_y{GetDefaultBlockHeight(format)};
-
- if (morton_to_gl) {
- Tegra::Texture::UnswizzleTexture(gl_buffer, addr, tile_size_x, tile_size_y, bytes_per_pixel,
- stride, height, depth, block_height, block_depth);
- } else {
- Tegra::Texture::CopySwizzledData((stride + tile_size_x - 1) / tile_size_x,
- (height + tile_size_y - 1) / tile_size_y, depth,
- bytes_per_pixel, bytes_per_pixel, Memory::GetPointer(addr),
- gl_buffer, false, block_height, block_depth);
- }
-}
-
-using GLConversionArray = std::array<void (*)(u32, u32, u32, u32, u32, u8*, std::size_t, VAddr),
- VideoCore::Surface::MaxPixelFormat>;
-
-static constexpr GLConversionArray morton_to_gl_fns = {
- // clang-format off
- MortonCopy<true, PixelFormat::ABGR8U>,
- MortonCopy<true, PixelFormat::ABGR8S>,
- MortonCopy<true, PixelFormat::ABGR8UI>,
- MortonCopy<true, PixelFormat::B5G6R5U>,
- MortonCopy<true, PixelFormat::A2B10G10R10U>,
- MortonCopy<true, PixelFormat::A1B5G5R5U>,
- MortonCopy<true, PixelFormat::R8U>,
- MortonCopy<true, PixelFormat::R8UI>,
- MortonCopy<true, PixelFormat::RGBA16F>,
- MortonCopy<true, PixelFormat::RGBA16U>,
- MortonCopy<true, PixelFormat::RGBA16UI>,
- MortonCopy<true, PixelFormat::R11FG11FB10F>,
- MortonCopy<true, PixelFormat::RGBA32UI>,
- MortonCopy<true, PixelFormat::DXT1>,
- MortonCopy<true, PixelFormat::DXT23>,
- MortonCopy<true, PixelFormat::DXT45>,
- MortonCopy<true, PixelFormat::DXN1>,
- MortonCopy<true, PixelFormat::DXN2UNORM>,
- MortonCopy<true, PixelFormat::DXN2SNORM>,
- MortonCopy<true, PixelFormat::BC7U>,
- MortonCopy<true, PixelFormat::BC6H_UF16>,
- MortonCopy<true, PixelFormat::BC6H_SF16>,
- MortonCopy<true, PixelFormat::ASTC_2D_4X4>,
- MortonCopy<true, PixelFormat::G8R8U>,
- MortonCopy<true, PixelFormat::G8R8S>,
- MortonCopy<true, PixelFormat::BGRA8>,
- MortonCopy<true, PixelFormat::RGBA32F>,
- MortonCopy<true, PixelFormat::RG32F>,
- MortonCopy<true, PixelFormat::R32F>,
- MortonCopy<true, PixelFormat::R16F>,
- MortonCopy<true, PixelFormat::R16U>,
- MortonCopy<true, PixelFormat::R16S>,
- MortonCopy<true, PixelFormat::R16UI>,
- MortonCopy<true, PixelFormat::R16I>,
- MortonCopy<true, PixelFormat::RG16>,
- MortonCopy<true, PixelFormat::RG16F>,
- MortonCopy<true, PixelFormat::RG16UI>,
- MortonCopy<true, PixelFormat::RG16I>,
- MortonCopy<true, PixelFormat::RG16S>,
- MortonCopy<true, PixelFormat::RGB32F>,
- MortonCopy<true, PixelFormat::RGBA8_SRGB>,
- MortonCopy<true, PixelFormat::RG8U>,
- MortonCopy<true, PixelFormat::RG8S>,
- MortonCopy<true, PixelFormat::RG32UI>,
- MortonCopy<true, PixelFormat::R32UI>,
- MortonCopy<true, PixelFormat::ASTC_2D_8X8>,
- MortonCopy<true, PixelFormat::ASTC_2D_8X5>,
- MortonCopy<true, PixelFormat::ASTC_2D_5X4>,
- MortonCopy<true, PixelFormat::BGRA8_SRGB>,
- MortonCopy<true, PixelFormat::DXT1_SRGB>,
- MortonCopy<true, PixelFormat::DXT23_SRGB>,
- MortonCopy<true, PixelFormat::DXT45_SRGB>,
- MortonCopy<true, PixelFormat::BC7U_SRGB>,
- MortonCopy<true, PixelFormat::ASTC_2D_4X4_SRGB>,
- MortonCopy<true, PixelFormat::ASTC_2D_8X8_SRGB>,
- MortonCopy<true, PixelFormat::ASTC_2D_8X5_SRGB>,
- MortonCopy<true, PixelFormat::ASTC_2D_5X4_SRGB>,
- MortonCopy<true, PixelFormat::ASTC_2D_5X5>,
- MortonCopy<true, PixelFormat::ASTC_2D_5X5_SRGB>,
- MortonCopy<true, PixelFormat::ASTC_2D_10X8>,
- MortonCopy<true, PixelFormat::ASTC_2D_10X8_SRGB>,
- MortonCopy<true, PixelFormat::Z32F>,
- MortonCopy<true, PixelFormat::Z16>,
- MortonCopy<true, PixelFormat::Z24S8>,
- MortonCopy<true, PixelFormat::S8Z24>,
- MortonCopy<true, PixelFormat::Z32FS8>,
- // clang-format on
-};
-
-static constexpr GLConversionArray gl_to_morton_fns = {
- // clang-format off
- MortonCopy<false, PixelFormat::ABGR8U>,
- MortonCopy<false, PixelFormat::ABGR8S>,
- MortonCopy<false, PixelFormat::ABGR8UI>,
- MortonCopy<false, PixelFormat::B5G6R5U>,
- MortonCopy<false, PixelFormat::A2B10G10R10U>,
- MortonCopy<false, PixelFormat::A1B5G5R5U>,
- MortonCopy<false, PixelFormat::R8U>,
- MortonCopy<false, PixelFormat::R8UI>,
- MortonCopy<false, PixelFormat::RGBA16F>,
- MortonCopy<false, PixelFormat::RGBA16U>,
- MortonCopy<false, PixelFormat::RGBA16UI>,
- MortonCopy<false, PixelFormat::R11FG11FB10F>,
- MortonCopy<false, PixelFormat::RGBA32UI>,
- MortonCopy<false, PixelFormat::DXT1>,
- MortonCopy<false, PixelFormat::DXT23>,
- MortonCopy<false, PixelFormat::DXT45>,
- MortonCopy<false, PixelFormat::DXN1>,
- MortonCopy<false, PixelFormat::DXN2UNORM>,
- MortonCopy<false, PixelFormat::DXN2SNORM>,
- MortonCopy<false, PixelFormat::BC7U>,
- MortonCopy<false, PixelFormat::BC6H_UF16>,
- MortonCopy<false, PixelFormat::BC6H_SF16>,
- // TODO(Subv): Swizzling ASTC formats are not supported
- nullptr,
- MortonCopy<false, PixelFormat::G8R8U>,
- MortonCopy<false, PixelFormat::G8R8S>,
- MortonCopy<false, PixelFormat::BGRA8>,
- MortonCopy<false, PixelFormat::RGBA32F>,
- MortonCopy<false, PixelFormat::RG32F>,
- MortonCopy<false, PixelFormat::R32F>,
- MortonCopy<false, PixelFormat::R16F>,
- MortonCopy<false, PixelFormat::R16U>,
- MortonCopy<false, PixelFormat::R16S>,
- MortonCopy<false, PixelFormat::R16UI>,
- MortonCopy<false, PixelFormat::R16I>,
- MortonCopy<false, PixelFormat::RG16>,
- MortonCopy<false, PixelFormat::RG16F>,
- MortonCopy<false, PixelFormat::RG16UI>,
- MortonCopy<false, PixelFormat::RG16I>,
- MortonCopy<false, PixelFormat::RG16S>,
- MortonCopy<false, PixelFormat::RGB32F>,
- MortonCopy<false, PixelFormat::RGBA8_SRGB>,
- MortonCopy<false, PixelFormat::RG8U>,
- MortonCopy<false, PixelFormat::RG8S>,
- MortonCopy<false, PixelFormat::RG32UI>,
- MortonCopy<false, PixelFormat::R32UI>,
- nullptr,
- nullptr,
- nullptr,
- MortonCopy<false, PixelFormat::BGRA8_SRGB>,
- MortonCopy<false, PixelFormat::DXT1_SRGB>,
- MortonCopy<false, PixelFormat::DXT23_SRGB>,
- MortonCopy<false, PixelFormat::DXT45_SRGB>,
- MortonCopy<false, PixelFormat::BC7U_SRGB>,
- nullptr,
- nullptr,
- nullptr,
- nullptr,
- nullptr,
- nullptr,
- nullptr,
- nullptr,
- MortonCopy<false, PixelFormat::Z32F>,
- MortonCopy<false, PixelFormat::Z16>,
- MortonCopy<false, PixelFormat::Z24S8>,
- MortonCopy<false, PixelFormat::S8Z24>,
- MortonCopy<false, PixelFormat::Z32FS8>,
- // clang-format on
-};
-
-void SwizzleFunc(const GLConversionArray& functions, const SurfaceParams& params,
+void SwizzleFunc(const MortonSwizzleMode& mode, const SurfaceParams& params,
std::vector<u8>& gl_buffer, u32 mip_level) {
u32 depth = params.MipDepth(mip_level);
if (params.target == SurfaceTarget::Texture2D) {
@@ -550,19 +385,19 @@ void SwizzleFunc(const GLConversionArray& functions, const SurfaceParams& params
const u64 layer_size = params.LayerMemorySize();
const u64 gl_size = params.LayerSizeGL(mip_level);
for (u32 i = 0; i < params.depth; i++) {
- functions[static_cast<std::size_t>(params.pixel_format)](
- params.MipWidth(mip_level), params.MipBlockHeight(mip_level),
- params.MipHeight(mip_level), params.MipBlockDepth(mip_level), 1,
- gl_buffer.data() + offset_gl, gl_size, params.addr + offset);
+ MortonSwizzle(mode, params.pixel_format, params.MipWidth(mip_level),
+ params.MipBlockHeight(mip_level), params.MipHeight(mip_level),
+ params.MipBlockDepth(mip_level), 1, gl_buffer.data() + offset_gl, gl_size,
+ params.addr + offset);
offset += layer_size;
offset_gl += gl_size;
}
} else {
const u64 offset = params.GetMipmapLevelOffset(mip_level);
- functions[static_cast<std::size_t>(params.pixel_format)](
- params.MipWidth(mip_level), params.MipBlockHeight(mip_level),
- params.MipHeight(mip_level), params.MipBlockDepth(mip_level), depth, gl_buffer.data(),
- gl_buffer.size(), params.addr + offset);
+ MortonSwizzle(mode, params.pixel_format, params.MipWidth(mip_level),
+ params.MipBlockHeight(mip_level), params.MipHeight(mip_level),
+ params.MipBlockDepth(mip_level), depth, gl_buffer.data(), gl_buffer.size(),
+ params.addr + offset);
}
}
@@ -726,7 +561,7 @@ static void CopySurface(const Surface& src_surface, const Surface& dst_surface,
const std::size_t buffer_size = std::max(src_params.size_in_bytes, dst_params.size_in_bytes);
glBindBuffer(GL_PIXEL_PACK_BUFFER, copy_pbo_handle);
- glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_DRAW_ARB);
+ glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_DRAW);
if (source_format.compressed) {
glGetCompressedTextureImage(src_surface->Texture().handle, src_attachment,
static_cast<GLsizei>(src_params.size_in_bytes), nullptr);
@@ -996,7 +831,7 @@ void CachedSurface::LoadGLBuffer() {
ASSERT_MSG(params.block_width == 1, "Block width is defined as {} on texture type {}",
params.block_width, static_cast<u32>(params.target));
for (u32 i = 0; i < params.max_mip_level; i++)
- SwizzleFunc(morton_to_gl_fns, params, gl_buffer[i], i);
+ SwizzleFunc(MortonSwizzleMode::MortonToLinear, params, gl_buffer[i], i);
} else {
const auto texture_src_data{Memory::GetPointer(params.addr)};
const auto texture_src_data_end{texture_src_data + params.size_in_bytes_gl};
@@ -1035,7 +870,7 @@ void CachedSurface::FlushGLBuffer() {
ASSERT_MSG(params.block_width == 1, "Block width is defined as {} on texture type {}",
params.block_width, static_cast<u32>(params.target));
- SwizzleFunc(gl_to_morton_fns, params, gl_buffer[0], 0);
+ SwizzleFunc(MortonSwizzleMode::LinearToMorton, params, gl_buffer[0], 0);
} else {
std::memcpy(Memory::GetPointer(GetAddr()), gl_buffer[0].data(), GetSizeInBytes());
}
@@ -1275,6 +1110,31 @@ Surface RasterizerCacheOpenGL::GetUncachedSurface(const SurfaceParams& params) {
return surface;
}
+void RasterizerCacheOpenGL::FastLayeredCopySurface(const Surface& src_surface,
+ const Surface& dst_surface) {
+ const auto& init_params{src_surface->GetSurfaceParams()};
+ const auto& dst_params{dst_surface->GetSurfaceParams()};
+ VAddr address = init_params.addr;
+ const std::size_t layer_size = dst_params.LayerMemorySize();
+ for (u32 layer = 0; layer < dst_params.depth; layer++) {
+ for (u32 mipmap = 0; mipmap < dst_params.max_mip_level; mipmap++) {
+ const VAddr sub_address = address + dst_params.GetMipmapLevelOffset(mipmap);
+ const Surface& copy = TryGet(sub_address);
+ if (!copy)
+ continue;
+ const auto& src_params{copy->GetSurfaceParams()};
+ const u32 width{std::min(src_params.width, dst_params.MipWidth(mipmap))};
+ const u32 height{std::min(src_params.height, dst_params.MipHeight(mipmap))};
+
+ glCopyImageSubData(copy->Texture().handle, SurfaceTargetToGL(src_params.target), 0, 0,
+ 0, 0, dst_surface->Texture().handle,
+ SurfaceTargetToGL(dst_params.target), mipmap, 0, 0, layer, width,
+ height, 1);
+ }
+ address += layer_size;
+ }
+}
+
void RasterizerCacheOpenGL::FermiCopySurface(
const Tegra::Engines::Fermi2D::Regs::Surface& src_config,
const Tegra::Engines::Fermi2D::Regs::Surface& dst_config) {
@@ -1340,11 +1200,13 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface,
CopySurface(old_surface, new_surface, copy_pbo.handle);
}
break;
- case SurfaceTarget::TextureCubemap:
case SurfaceTarget::Texture3D:
+ AccurateCopySurface(old_surface, new_surface);
+ break;
+ case SurfaceTarget::TextureCubemap:
case SurfaceTarget::Texture2DArray:
case SurfaceTarget::TextureCubeArray:
- AccurateCopySurface(old_surface, new_surface);
+ FastLayeredCopySurface(old_surface, new_surface);
break;
default:
LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index 494f6b903..9ac79c5a4 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -350,6 +350,7 @@ private:
/// Performs a slow but accurate surface copy, flushing to RAM and reinterpreting the data
void AccurateCopySurface(const Surface& src_surface, const Surface& dst_surface);
+ void FastLayeredCopySurface(const Surface& src_surface, const Surface& dst_surface);
/// The surface reserve is a "backup" cache, this is where we put unique surfaces that have
/// previously been used. This is to prevent surfaces from being constantly created and
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index a85a7c0c5..038b25c75 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -84,6 +84,7 @@ CachedShader::CachedShader(VAddr addr, Maxwell::ShaderProgram program_type)
}
entries = program_result.second;
+ shader_length = entries.shader_length;
if (program_type != Maxwell::ShaderProgram::Geometry) {
OGLShader shader;
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index ffbf21831..08f470de3 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -30,7 +30,7 @@ public:
}
std::size_t GetSizeInBytes() const override {
- return GLShader::MAX_PROGRAM_CODE_LENGTH * sizeof(u64);
+ return shader_length;
}
// We do not have to flush this cache as things in it are never modified by us.
@@ -82,6 +82,7 @@ private:
u32 max_vertices, const std::string& debug_name);
VAddr addr;
+ std::size_t shader_length;
Maxwell::ShaderProgram program_type;
GLShader::ShaderSetup setup;
GLShader::ShaderEntries entries;
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 30050b585..3a75f9d16 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -34,6 +34,17 @@ constexpr u32 PROGRAM_HEADER_SIZE = sizeof(Tegra::Shader::Header);
constexpr u32 MAX_GEOMETRY_BUFFERS = 6;
constexpr u32 MAX_ATTRIBUTES = 0x100; // Size in vec4s, this value is untested
+static const char* INTERNAL_FLAG_NAMES[] = {"zero_flag", "sign_flag", "carry_flag",
+ "overflow_flag"};
+
+enum class InternalFlag : u64 {
+ ZeroFlag = 0,
+ SignFlag = 1,
+ CarryFlag = 2,
+ OverflowFlag = 3,
+ Amount
+};
+
class DecompileFail : public std::runtime_error {
public:
using std::runtime_error::runtime_error;
@@ -49,8 +60,7 @@ static std::string GetTopologyName(Tegra::Shader::OutputTopology topology) {
case Tegra::Shader::OutputTopology::TriangleStrip:
return "triangle_strip";
default:
- LOG_CRITICAL(Render_OpenGL, "Unknown output topology {}", static_cast<u32>(topology));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unknown output topology: {}", static_cast<u32>(topology));
return "points";
}
}
@@ -85,7 +95,8 @@ struct Subroutine {
class ControlFlowAnalyzer {
public:
ControlFlowAnalyzer(const ProgramCode& program_code, u32 main_offset, const std::string& suffix)
- : program_code(program_code) {
+ : program_code(program_code), shader_coverage_begin(main_offset),
+ shader_coverage_end(main_offset + 1) {
// Recursively finds all subroutines.
const Subroutine& program_main = AddSubroutine(main_offset, PROGRAM_END, suffix);
@@ -97,10 +108,16 @@ public:
return std::move(subroutines);
}
+ std::size_t GetShaderLength() const {
+ return shader_coverage_end * sizeof(u64);
+ }
+
private:
const ProgramCode& program_code;
std::set<Subroutine> subroutines;
std::map<std::pair<u32, u32>, ExitMethod> exit_method_map;
+ u32 shader_coverage_begin;
+ u32 shader_coverage_end;
/// Adds and analyzes a new subroutine if it is not added yet.
const Subroutine& AddSubroutine(u32 begin, u32 end, const std::string& suffix) {
@@ -142,6 +159,9 @@ private:
return exit_method;
for (u32 offset = begin; offset != end && offset != PROGRAM_END; ++offset) {
+ shader_coverage_begin = std::min(shader_coverage_begin, offset);
+ shader_coverage_end = std::max(shader_coverage_end, offset + 1);
+
const Instruction instr = {program_code[offset]};
if (const auto opcode = OpCode::Decode(instr)) {
switch (opcode->get().GetId()) {
@@ -167,8 +187,8 @@ private:
case OpCode::Id::SSY:
case OpCode::Id::PBK: {
// The SSY and PBK use a similar encoding as the BRA instruction.
- ASSERT_MSG(instr.bra.constant_buffer == 0,
- "Constant buffer branching is not supported");
+ UNIMPLEMENTED_IF_MSG(instr.bra.constant_buffer != 0,
+ "Constant buffer branching is not supported");
const u32 target = offset + instr.bra.GetBranchTarget();
labels.insert(target);
// Continue scanning for an exit method.
@@ -258,14 +278,6 @@ private:
const std::string& suffix;
};
-enum class InternalFlag : u64 {
- ZeroFlag = 0,
- CarryFlag = 1,
- OverflowFlag = 2,
- NaNFlag = 3,
- Amount
-};
-
/**
* Used to manage shader registers that are emulated with GLSL. This class keeps track of the state
* of all registers (e.g. whether they are currently being used as Floats or Integers), and
@@ -299,8 +311,7 @@ public:
// Default - do nothing
return value;
default:
- LOG_CRITICAL(HW_GPU, "Unimplemented conversion size {}", static_cast<u32>(size));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unimplemented conversion size: {}", static_cast<u32>(size));
}
}
@@ -363,7 +374,7 @@ public:
u64 value_num_components, bool is_saturated = false,
u64 dest_elem = 0, Register::Size size = Register::Size::Word,
bool sets_cc = false) {
- ASSERT_MSG(!is_saturated, "Unimplemented");
+ UNIMPLEMENTED_IF(is_saturated);
const std::string func{is_signed ? "intBitsToFloat" : "uintBitsToFloat"};
@@ -373,7 +384,7 @@ public:
if (sets_cc) {
const std::string zero_condition = "( " + ConvertIntegerSize(value, size) + " == 0 )";
SetInternalFlag(InternalFlag::ZeroFlag, zero_condition);
- LOG_WARNING(HW_GPU, "Control Codes Imcomplete.");
+ LOG_WARNING(HW_GPU, "Condition codes implementation is incomplete.");
}
}
@@ -392,7 +403,7 @@ public:
Tegra::Shader::HalfMerge merge, u64 dest_num_components,
u64 value_num_components, bool is_saturated = false,
u64 dest_elem = 0) {
- ASSERT_MSG(!is_saturated, "Unimplemented");
+ UNIMPLEMENTED_IF(is_saturated);
const std::string result = [&]() {
switch (merge) {
@@ -456,24 +467,25 @@ public:
shader.AddLine("lmem[" + index + "] = " + func + '(' + value + ");");
}
- std::string GetControlCode(const Tegra::Shader::ControlCode cc) const {
+ std::string GetConditionCode(const Tegra::Shader::ConditionCode cc) const {
switch (cc) {
- case Tegra::Shader::ControlCode::NEU:
+ case Tegra::Shader::ConditionCode::NEU:
return "!(" + GetInternalFlag(InternalFlag::ZeroFlag) + ')';
default:
- LOG_CRITICAL(HW_GPU, "Unimplemented Control Code {}", static_cast<u32>(cc));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unimplemented condition code: {}", static_cast<u32>(cc));
return "false";
}
}
- std::string GetInternalFlag(const InternalFlag ii) const {
- const u32 code = static_cast<u32>(ii);
- return "internalFlag_" + std::to_string(code) + suffix;
+ std::string GetInternalFlag(const InternalFlag flag) const {
+ const auto index = static_cast<u32>(flag);
+ ASSERT(index < static_cast<u32>(InternalFlag::Amount));
+
+ return std::string(INTERNAL_FLAG_NAMES[index]) + '_' + suffix;
}
- void SetInternalFlag(const InternalFlag ii, const std::string& value) const {
- shader.AddLine(GetInternalFlag(ii) + " = " + value + ';');
+ void SetInternalFlag(const InternalFlag flag, const std::string& value) const {
+ shader.AddLine(GetInternalFlag(flag) + " = " + value + ';');
}
/**
@@ -488,27 +500,42 @@ public:
const Register& buf_reg) {
const std::string dest = GetOutputAttribute(attribute);
const std::string src = GetRegisterAsFloat(val_reg);
+ if (dest.empty())
+ return;
- if (!dest.empty()) {
- // Can happen with unknown/unimplemented output attributes, in which case we ignore the
- // instruction for now.
- if (stage == Maxwell3D::Regs::ShaderStage::Geometry) {
- // TODO(Rodrigo): nouveau sets some attributes after setting emitting a geometry
- // shader. These instructions use a dirty register as buffer index, to avoid some
- // drivers from complaining about out of boundary writes, guard them.
- const std::string buf_index{"((" + GetRegisterAsInteger(buf_reg) + ") % " +
- std::to_string(MAX_GEOMETRY_BUFFERS) + ')'};
- shader.AddLine("amem[" + buf_index + "][" +
- std::to_string(static_cast<u32>(attribute)) + ']' +
- GetSwizzle(elem) + " = " + src + ';');
- } else {
- if (attribute == Attribute::Index::PointSize) {
- fixed_pipeline_output_attributes_used.insert(attribute);
- shader.AddLine(dest + " = " + src + ';');
- } else {
- shader.AddLine(dest + GetSwizzle(elem) + " = " + src + ';');
- }
- }
+ // Can happen with unknown/unimplemented output attributes, in which case we ignore the
+ // instruction for now.
+ if (stage == Maxwell3D::Regs::ShaderStage::Geometry) {
+ // TODO(Rodrigo): nouveau sets some attributes after setting emitting a geometry
+ // shader. These instructions use a dirty register as buffer index, to avoid some
+ // drivers from complaining about out of boundary writes, guard them.
+ const std::string buf_index{"((" + GetRegisterAsInteger(buf_reg) + ") % " +
+ std::to_string(MAX_GEOMETRY_BUFFERS) + ')'};
+ shader.AddLine("amem[" + buf_index + "][" +
+ std::to_string(static_cast<u32>(attribute)) + ']' + GetSwizzle(elem) +
+ " = " + src + ';');
+ return;
+ }
+
+ switch (attribute) {
+ case Attribute::Index::ClipDistances0123:
+ case Attribute::Index::ClipDistances4567: {
+ const u64 index = attribute == Attribute::Index::ClipDistances4567 ? 4 : 0 + elem;
+ UNIMPLEMENTED_IF_MSG(
+ ((header.vtg.clip_distances >> index) & 1) == 0,
+ "Shader is setting gl_ClipDistance{} without enabling it in the header", index);
+
+ fixed_pipeline_output_attributes_used.insert(attribute);
+ shader.AddLine(dest + '[' + std::to_string(index) + "] = " + src + ';');
+ break;
+ }
+ case Attribute::Index::PointSize:
+ fixed_pipeline_output_attributes_used.insert(attribute);
+ shader.AddLine(dest + " = " + src + ';');
+ break;
+ default:
+ shader.AddLine(dest + GetSwizzle(elem) + " = " + src + ';');
+ break;
}
}
@@ -624,8 +651,8 @@ private:
/// Generates declarations for internal flags.
void GenerateInternalFlags() {
- for (u32 ii = 0; ii < static_cast<u64>(InternalFlag::Amount); ii++) {
- const InternalFlag code = static_cast<InternalFlag>(ii);
+ for (u32 flag = 0; flag < static_cast<u32>(InternalFlag::Amount); flag++) {
+ const InternalFlag code = static_cast<InternalFlag>(flag);
declarations.AddLine("bool " + GetInternalFlag(code) + " = false;");
}
declarations.AddNewLine();
@@ -728,12 +755,19 @@ private:
void GenerateVertex() {
if (stage != Maxwell3D::Regs::ShaderStage::Vertex)
return;
+ bool clip_distances_declared = false;
+
declarations.AddLine("out gl_PerVertex {");
++declarations.scope;
declarations.AddLine("vec4 gl_Position;");
for (auto& o : fixed_pipeline_output_attributes_used) {
if (o == Attribute::Index::PointSize)
declarations.AddLine("float gl_PointSize;");
+ if (!clip_distances_declared && (o == Attribute::Index::ClipDistances0123 ||
+ o == Attribute::Index::ClipDistances4567)) {
+ declarations.AddLine("float gl_ClipDistance[];");
+ clip_distances_declared = true;
+ }
}
--declarations.scope;
declarations.AddLine("};");
@@ -761,8 +795,7 @@ private:
u64 dest_num_components, u64 value_num_components, u64 dest_elem,
bool precise) {
if (reg == Register::ZeroIndex) {
- LOG_CRITICAL(HW_GPU, "Cannot set Register::ZeroIndex");
- UNREACHABLE();
+ // Setting RZ is a nop in hardware.
return;
}
@@ -847,16 +880,13 @@ private:
if (declr_input_attribute.count(attribute) == 0) {
declr_input_attribute[attribute] = input_mode;
} else {
- if (declr_input_attribute[attribute] != input_mode) {
- LOG_CRITICAL(HW_GPU, "Same Input multiple input modes");
- UNREACHABLE();
- }
+ UNIMPLEMENTED_IF_MSG(declr_input_attribute[attribute] != input_mode,
+ "Multiple input modes for the same attribute");
}
return GeometryPass("input_attribute_" + std::to_string(index));
}
- LOG_CRITICAL(HW_GPU, "Unhandled input attribute: {}", static_cast<u32>(attribute));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled input attribute: {}", static_cast<u32>(attribute));
}
return "vec4(0, 0, 0, 0)";
@@ -882,24 +912,20 @@ private:
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled Ipa InterpMode: {}", static_cast<u32>(interp_mode));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled IPA interp mode: {}", static_cast<u32>(interp_mode));
}
}
switch (sample_mode) {
- case Tegra::Shader::IpaSampleMode::Centroid: {
- // Note not implemented, it can be implemented with the "centroid " keyword in glsl;
- LOG_CRITICAL(HW_GPU, "Ipa Sampler Mode: centroid, not implemented");
- UNREACHABLE();
+ case Tegra::Shader::IpaSampleMode::Centroid:
+ // It can be implemented with the "centroid " keyword in glsl
+ UNIMPLEMENTED_MSG("Unimplemented IPA sampler mode centroid");
break;
- }
- case Tegra::Shader::IpaSampleMode::Default: {
+ case Tegra::Shader::IpaSampleMode::Default:
// Default, n/a
break;
- }
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled Ipa SampleMode: {}", static_cast<u32>(sample_mode));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unimplemented IPA sampler mode: {}", static_cast<u32>(sample_mode));
+ break;
}
}
return out;
@@ -912,6 +938,10 @@ private:
return "gl_PointSize";
case Attribute::Index::Position:
return "position";
+ case Attribute::Index::ClipDistances0123:
+ case Attribute::Index::ClipDistances4567: {
+ return "gl_ClipDistance";
+ }
default:
const u32 index{static_cast<u32>(attribute) -
static_cast<u32>(Attribute::Index::Attribute_0)};
@@ -920,8 +950,7 @@ private:
return "output_attribute_" + std::to_string(index);
}
- LOG_CRITICAL(HW_GPU, "Unhandled output attribute: {}", index);
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled output attribute={}", index);
return {};
}
}
@@ -951,9 +980,10 @@ private:
class GLSLGenerator {
public:
GLSLGenerator(const std::set<Subroutine>& subroutines, const ProgramCode& program_code,
- u32 main_offset, Maxwell3D::Regs::ShaderStage stage, const std::string& suffix)
+ u32 main_offset, Maxwell3D::Regs::ShaderStage stage, const std::string& suffix,
+ std::size_t shader_length)
: subroutines(subroutines), program_code(program_code), main_offset(main_offset),
- stage(stage), suffix(suffix) {
+ stage(stage), suffix(suffix), shader_length(shader_length) {
std::memcpy(&header, program_code.data(), sizeof(Tegra::Shader::Header));
local_memory_size = header.GetLocalMemorySize();
regs.SetLocalMemory(local_memory_size);
@@ -966,7 +996,7 @@ public:
/// Returns entries in the shader that are useful for external functions
ShaderEntries GetEntries() const {
- return {regs.GetConstBuffersDeclarations(), regs.GetSamplers()};
+ return {regs.GetConstBuffersDeclarations(), regs.GetSamplers(), shader_length};
}
private:
@@ -1071,19 +1101,26 @@ private:
const std::string& op_a, const std::string& op_b) const {
using Tegra::Shader::PredCondition;
static const std::unordered_map<PredCondition, const char*> PredicateComparisonStrings = {
- {PredCondition::LessThan, "<"}, {PredCondition::Equal, "=="},
- {PredCondition::LessEqual, "<="}, {PredCondition::GreaterThan, ">"},
- {PredCondition::NotEqual, "!="}, {PredCondition::GreaterEqual, ">="},
- {PredCondition::LessThanWithNan, "<"}, {PredCondition::NotEqualWithNan, "!="},
- {PredCondition::GreaterThanWithNan, ">"}, {PredCondition::GreaterEqualWithNan, ">="}};
+ {PredCondition::LessThan, "<"},
+ {PredCondition::Equal, "=="},
+ {PredCondition::LessEqual, "<="},
+ {PredCondition::GreaterThan, ">"},
+ {PredCondition::NotEqual, "!="},
+ {PredCondition::GreaterEqual, ">="},
+ {PredCondition::LessThanWithNan, "<"},
+ {PredCondition::NotEqualWithNan, "!="},
+ {PredCondition::LessEqualWithNan, "<="},
+ {PredCondition::GreaterThanWithNan, ">"},
+ {PredCondition::GreaterEqualWithNan, ">="}};
const auto& comparison{PredicateComparisonStrings.find(condition)};
- ASSERT_MSG(comparison != PredicateComparisonStrings.end(),
- "Unknown predicate comparison operation");
+ UNIMPLEMENTED_IF_MSG(comparison == PredicateComparisonStrings.end(),
+ "Unknown predicate comparison operation");
std::string predicate{'(' + op_a + ") " + comparison->second + " (" + op_b + ')'};
if (condition == PredCondition::LessThanWithNan ||
condition == PredCondition::NotEqualWithNan ||
+ condition == PredCondition::LessEqualWithNan ||
condition == PredCondition::GreaterThanWithNan ||
condition == PredCondition::GreaterEqualWithNan) {
predicate += " || isnan(" + op_a + ") || isnan(" + op_b + ')';
@@ -1107,7 +1144,7 @@ private:
};
auto op = PredicateOperationStrings.find(operation);
- ASSERT_MSG(op != PredicateOperationStrings.end(), "Unknown predicate operation");
+ UNIMPLEMENTED_IF_MSG(op == PredicateOperationStrings.end(), "Unknown predicate operation");
return op->second;
}
@@ -1205,8 +1242,7 @@ private:
break;
}
default:
- LOG_CRITICAL(HW_GPU, "Unimplemented logic operation: {}", static_cast<u32>(logic_op));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unimplemented logic operation={}", static_cast<u32>(logic_op));
}
if (dest != Tegra::Shader::Register::ZeroIndex) {
@@ -1224,9 +1260,8 @@ private:
SetPredicate(static_cast<u64>(predicate), '(' + result + ") != 0");
break;
default:
- LOG_CRITICAL(HW_GPU, "Unimplemented predicate result mode: {}",
- static_cast<u32>(predicate_mode));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unimplemented predicate result mode: {}",
+ static_cast<u32>(predicate_mode));
}
}
@@ -1264,6 +1299,7 @@ private:
shader.AddLine('{');
++shader.scope;
shader.AddLine(coord);
+ shader.AddLine("vec4 texture_tmp = " + texture + ';');
// TEXS has two destination registers and a swizzle. The first two elements in the swizzle
// go into gpr0+0 and gpr0+1, and the rest goes into gpr28+0 and gpr28+1
@@ -1276,37 +1312,32 @@ private:
if (written_components < 2) {
// Write the first two swizzle components to gpr0 and gpr0+1
- regs.SetRegisterToFloat(instr.gpr0, component, texture, 1, 4, false,
+ regs.SetRegisterToFloat(instr.gpr0, component, "texture_tmp", 1, 4, false,
written_components % 2);
} else {
ASSERT(instr.texs.HasTwoDestinations());
// Write the rest of the swizzle components to gpr28 and gpr28+1
- regs.SetRegisterToFloat(instr.gpr28, component, texture, 1, 4, false,
+ regs.SetRegisterToFloat(instr.gpr28, component, "texture_tmp", 1, 4, false,
written_components % 2);
}
++written_components;
}
-
--shader.scope;
shader.AddLine('}');
}
static u32 TextureCoordinates(Tegra::Shader::TextureType texture_type) {
switch (texture_type) {
- case Tegra::Shader::TextureType::Texture1D: {
+ case Tegra::Shader::TextureType::Texture1D:
return 1;
- }
- case Tegra::Shader::TextureType::Texture2D: {
+ case Tegra::Shader::TextureType::Texture2D:
return 2;
- }
case Tegra::Shader::TextureType::Texture3D:
- case Tegra::Shader::TextureType::TextureCube: {
+ case Tegra::Shader::TextureType::TextureCube:
return 3;
- }
default:
- LOG_CRITICAL(HW_GPU, "Unhandled texture type {}", static_cast<u32>(texture_type));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled texture type: {}", static_cast<u32>(texture_type));
return 0;
}
}
@@ -1342,7 +1373,7 @@ private:
void EmitFragmentOutputsWrite() {
ASSERT(stage == Maxwell3D::Regs::ShaderStage::Fragment);
- ASSERT_MSG(header.ps.omap.sample_mask == 0, "Samplemask write is unimplemented");
+ UNIMPLEMENTED_IF_MSG(header.ps.omap.sample_mask != 0, "Samplemask write is unimplemented");
shader.AddLine("if (alpha_test[0] != 0) {");
++shader.scope;
@@ -1408,7 +1439,7 @@ private:
case Tegra::Shader::VideoType::Size32:
// TODO(Rodrigo): From my hardware tests it becomes a bit "mad" when
// this type is used (1 * 1 + 0 == 0x5b800000). Until a better
- // explanation is found: assert.
+ // explanation is found: abort.
UNIMPLEMENTED();
return zero;
case Tegra::Shader::VideoType::Invalid:
@@ -1464,8 +1495,7 @@ private:
// Decoding failure
if (!opcode) {
- LOG_CRITICAL(HW_GPU, "Unhandled instruction: {0:x}", instr.value);
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled instruction: {0:x}", instr.value);
return offset + 1;
}
@@ -1473,8 +1503,8 @@ private:
fmt::format("// {}: {} (0x{:016x})", offset, opcode->get().GetName(), instr.value));
using Tegra::Shader::Pred;
- ASSERT_MSG(instr.pred.full_pred != Pred::NeverExecute,
- "NeverExecute predicate not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.pred.full_pred == Pred::NeverExecute,
+ "NeverExecute predicate not implemented");
// Some instructions (like SSY) don't have a predicate field, they are always
// unconditionally executed.
@@ -1517,37 +1547,36 @@ private:
case OpCode::Id::FMUL_R:
case OpCode::Id::FMUL_IMM: {
// FMUL does not have 'abs' bits and only the second operand has a 'neg' bit.
- ASSERT_MSG(instr.fmul.tab5cb8_2 == 0, "FMUL tab5cb8_2({}) is not implemented",
- instr.fmul.tab5cb8_2.Value());
- ASSERT_MSG(instr.fmul.tab5c68_1 == 0, "FMUL tab5cb8_1({}) is not implemented",
- instr.fmul.tab5c68_1.Value());
- ASSERT_MSG(instr.fmul.tab5c68_0 == 1, "FMUL tab5cb8_0({}) is not implemented",
- instr.fmul.tab5c68_0
- .Value()); // SMO typical sends 1 here which seems to be the default
- ASSERT_MSG(instr.fmul.cc == 0, "FMUL cc is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.fmul.tab5cb8_2 != 0,
+ "FMUL tab5cb8_2({}) is not implemented",
+ instr.fmul.tab5cb8_2.Value());
+ UNIMPLEMENTED_IF_MSG(instr.fmul.tab5c68_1 != 0,
+ "FMUL tab5cb8_1({}) is not implemented",
+ instr.fmul.tab5c68_1.Value());
+ UNIMPLEMENTED_IF_MSG(
+ instr.fmul.tab5c68_0 != 1, "FMUL tab5cb8_0({}) is not implemented",
+ instr.fmul.tab5c68_0
+ .Value()); // SMO typical sends 1 here which seems to be the default
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in FMUL is not implemented");
op_b = GetOperandAbsNeg(op_b, false, instr.fmul.negate_b);
regs.SetRegisterToFloat(instr.gpr0, 0, op_a + " * " + op_b, 1, 1,
instr.alu.saturate_d, 0, true);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "FMUL Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::FADD_C:
case OpCode::Id::FADD_R:
case OpCode::Id::FADD_IMM: {
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in FADD is not implemented");
+
op_a = GetOperandAbsNeg(op_a, instr.alu.abs_a, instr.alu.negate_a);
op_b = GetOperandAbsNeg(op_b, instr.alu.abs_b, instr.alu.negate_b);
regs.SetRegisterToFloat(instr.gpr0, 0, op_a + " + " + op_b, 1, 1,
instr.alu.saturate_d, 0, true);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "FADD Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::MUFU: {
@@ -1582,15 +1611,17 @@ private:
instr.alu.saturate_d, 0, true);
break;
default:
- LOG_CRITICAL(HW_GPU, "Unhandled MUFU sub op: {0:x}",
- static_cast<unsigned>(instr.sub_op.Value()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled MUFU sub op={0:x}",
+ static_cast<unsigned>(instr.sub_op.Value()));
}
break;
}
case OpCode::Id::FMNMX_C:
case OpCode::Id::FMNMX_R:
case OpCode::Id::FMNMX_IMM: {
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in FMNMX is not implemented");
+
op_a = GetOperandAbsNeg(op_a, instr.alu.abs_a, instr.alu.negate_a);
op_b = GetOperandAbsNeg(op_b, instr.alu.abs_b, instr.alu.negate_b);
@@ -1601,10 +1632,6 @@ private:
'(' + condition + ") ? min(" + parameters + ") : max(" +
parameters + ')',
1, 1, false, 0, true);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "FMNMX Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::RRO_C:
@@ -1617,9 +1644,7 @@ private:
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled arithmetic instruction: {}",
- opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled arithmetic instruction: {}", opcode->get().GetName());
}
}
break;
@@ -1631,17 +1656,19 @@ private:
break;
}
case OpCode::Id::FMUL32_IMM: {
+ UNIMPLEMENTED_IF_MSG(instr.op_32.generates_cc,
+ "Condition codes generation in FMUL32 is not implemented");
+
regs.SetRegisterToFloat(instr.gpr0, 0,
regs.GetRegisterAsFloat(instr.gpr8) + " * " +
GetImmediate32(instr),
1, 1, instr.fmul32.saturate, 0, true);
- if (instr.op_32.generates_cc) {
- LOG_CRITICAL(HW_GPU, "FMUL32 Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::FADD32I: {
+ UNIMPLEMENTED_IF_MSG(instr.op_32.generates_cc,
+ "Condition codes generation in FADD32I is not implemented");
+
std::string op_a = regs.GetRegisterAsFloat(instr.gpr8);
std::string op_b = GetImmediate32(instr);
@@ -1662,23 +1689,22 @@ private:
}
regs.SetRegisterToFloat(instr.gpr0, 0, op_a + " + " + op_b, 1, 1, false, 0, true);
- if (instr.op_32.generates_cc) {
- LOG_CRITICAL(HW_GPU, "FADD32 Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
}
break;
}
case OpCode::Type::Bfe: {
- ASSERT_MSG(!instr.bfe.negate_b, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.bfe.negate_b);
std::string op_a = instr.bfe.negate_a ? "-" : "";
op_a += regs.GetRegisterAsInteger(instr.gpr8);
switch (opcode->get().GetId()) {
case OpCode::Id::BFE_IMM: {
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in BFE is not implemented");
+
std::string inner_shift =
'(' + op_a + " << " + std::to_string(instr.bfe.GetLeftShiftValue()) + ')';
std::string outer_shift =
@@ -1686,15 +1712,10 @@ private:
std::to_string(instr.bfe.GetLeftShiftValue() + instr.bfe.shift_position) + ')';
regs.SetRegisterToInteger(instr.gpr0, true, 0, outer_shift, 1, 1);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "BFE Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled BFE instruction: {}", opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled BFE instruction: {}", opcode->get().GetName());
}
}
@@ -1719,6 +1740,9 @@ private:
case OpCode::Id::SHR_C:
case OpCode::Id::SHR_R:
case OpCode::Id::SHR_IMM: {
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in SHR is not implemented");
+
if (!instr.shift.is_signed) {
// Logical shift right
op_a = "uint(" + op_a + ')';
@@ -1727,24 +1751,17 @@ private:
// Cast to int is superfluous for arithmetic shift, it's only for a logical shift
regs.SetRegisterToInteger(instr.gpr0, true, 0, "int(" + op_a + " >> " + op_b + ')',
1, 1);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "SHR Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::SHL_C:
case OpCode::Id::SHL_R:
case OpCode::Id::SHL_IMM:
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in SHL is not implemented");
regs.SetRegisterToInteger(instr.gpr0, true, 0, op_a + " << " + op_b, 1, 1);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "SHL Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled shift instruction: {}", opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled shift instruction: {}", opcode->get().GetName());
}
}
break;
@@ -1755,17 +1772,19 @@ private:
switch (opcode->get().GetId()) {
case OpCode::Id::IADD32I:
+ UNIMPLEMENTED_IF_MSG(instr.op_32.generates_cc,
+ "Condition codes generation in IADD32I is not implemented");
+
if (instr.iadd32i.negate_a)
op_a = "-(" + op_a + ')';
regs.SetRegisterToInteger(instr.gpr0, true, 0, op_a + " + " + op_b, 1, 1,
instr.iadd32i.saturate != 0);
- if (instr.op_32.generates_cc) {
- LOG_CRITICAL(HW_GPU, "IADD32 Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
case OpCode::Id::LOP32I: {
+ UNIMPLEMENTED_IF_MSG(instr.op_32.generates_cc,
+ "Condition codes generation in LOP32I is not implemented");
+
if (instr.alu.lop32i.invert_a)
op_a = "~(" + op_a + ')';
@@ -1775,16 +1794,11 @@ private:
WriteLogicOperation(instr.gpr0, instr.alu.lop32i.operation, op_a, op_b,
Tegra::Shader::PredicateResultMode::None,
Tegra::Shader::Pred::UnusedIndex);
- if (instr.op_32.generates_cc) {
- LOG_CRITICAL(HW_GPU, "LOP32I Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled ArithmeticIntegerImmediate instruction: {}",
- opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled ArithmeticIntegerImmediate instruction: {}",
+ opcode->get().GetName());
}
}
break;
@@ -1807,6 +1821,9 @@ private:
case OpCode::Id::IADD_C:
case OpCode::Id::IADD_R:
case OpCode::Id::IADD_IMM: {
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in IADD is not implemented");
+
if (instr.alu_integer.negate_a)
op_a = "-(" + op_a + ')';
@@ -1815,15 +1832,14 @@ private:
regs.SetRegisterToInteger(instr.gpr0, true, 0, op_a + " + " + op_b, 1, 1,
instr.alu.saturate_d);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "IADD Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::IADD3_C:
case OpCode::Id::IADD3_R:
case OpCode::Id::IADD3_IMM: {
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in IADD3 is not implemented");
+
std::string op_c = regs.GetRegisterAsInteger(instr.gpr39);
auto apply_height = [](auto height, auto& oprand) {
@@ -1837,9 +1853,8 @@ private:
oprand = "((" + oprand + ") >> 16)";
break;
default:
- LOG_CRITICAL(HW_GPU, "Unhandled IADD3 height: {}",
- static_cast<u32>(height.Value()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled IADD3 height: {}",
+ static_cast<u32>(height.Value()));
}
};
@@ -1880,16 +1895,14 @@ private:
}
regs.SetRegisterToInteger(instr.gpr0, true, 0, result, 1, 1);
-
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "IADD3 Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::ISCADD_C:
case OpCode::Id::ISCADD_R:
case OpCode::Id::ISCADD_IMM: {
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in ISCADD is not implemented");
+
if (instr.alu_integer.negate_a)
op_a = "-(" + op_a + ')';
@@ -1900,10 +1913,6 @@ private:
regs.SetRegisterToInteger(instr.gpr0, true, 0,
"((" + op_a + " << " + shift + ") + " + op_b + ')', 1, 1);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "ISCADD Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::POPC_C:
@@ -1927,6 +1936,9 @@ private:
case OpCode::Id::LOP_C:
case OpCode::Id::LOP_R:
case OpCode::Id::LOP_IMM: {
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in LOP is not implemented");
+
if (instr.alu.lop.invert_a)
op_a = "~(" + op_a + ')';
@@ -1935,15 +1947,14 @@ private:
WriteLogicOperation(instr.gpr0, instr.alu.lop.operation, op_a, op_b,
instr.alu.lop.pred_result_mode, instr.alu.lop.pred48);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "LOP Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::LOP3_C:
case OpCode::Id::LOP3_R:
case OpCode::Id::LOP3_IMM: {
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in LOP3 is not implemented");
+
const std::string op_c = regs.GetRegisterAsInteger(instr.gpr39);
std::string lut;
@@ -1954,17 +1965,15 @@ private:
}
WriteLop3Instruction(instr.gpr0, op_a, op_b, op_c, lut);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "LOP3 Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::IMNMX_C:
case OpCode::Id::IMNMX_R:
case OpCode::Id::IMNMX_IMM: {
- ASSERT_MSG(instr.imnmx.exchange == Tegra::Shader::IMinMaxExchange::None,
- "Unimplemented");
+ UNIMPLEMENTED_IF(instr.imnmx.exchange != Tegra::Shader::IMinMaxExchange::None);
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in IMNMX is not implemented");
+
const std::string condition =
GetPredicateCondition(instr.imnmx.pred, instr.imnmx.negate_pred != 0);
const std::string parameters = op_a + ',' + op_b;
@@ -1972,10 +1981,6 @@ private:
'(' + condition + ") ? min(" + parameters + ") : max(" +
parameters + ')',
1, 1);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "IMNMX Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::LEA_R2:
@@ -2030,24 +2035,19 @@ private:
op_b = regs.GetRegisterAsInteger(instr.gpr8);
op_a = std::to_string(instr.lea.imm.entry_a);
op_c = std::to_string(instr.lea.imm.entry_b);
- LOG_CRITICAL(HW_GPU, "Unhandled LEA subinstruction: {}",
- opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled LEA subinstruction: {}", opcode->get().GetName());
}
}
- if (instr.lea.pred48 != static_cast<u64>(Pred::UnusedIndex)) {
- LOG_ERROR(HW_GPU, "Unhandled LEA Predicate");
- UNREACHABLE();
- }
+ UNIMPLEMENTED_IF_MSG(instr.lea.pred48 != static_cast<u64>(Pred::UnusedIndex),
+ "Unhandled LEA Predicate");
const std::string value = '(' + op_a + " + (" + op_b + "*(1 << " + op_c + ")))";
regs.SetRegisterToInteger(instr.gpr0, true, 0, value, 1, 1);
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled ArithmeticInteger instruction: {}",
- opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled ArithmeticInteger instruction: {}",
+ opcode->get().GetName());
}
}
@@ -2056,7 +2056,7 @@ private:
case OpCode::Type::ArithmeticHalf: {
if (opcode->get().GetId() == OpCode::Id::HADD2_C ||
opcode->get().GetId() == OpCode::Id::HADD2_R) {
- ASSERT_MSG(instr.alu_half.ftz == 0, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.alu_half.ftz != 0);
}
const bool negate_a =
opcode->get().GetId() != OpCode::Id::HMUL2_R && instr.alu_half.negate_a != 0;
@@ -2094,9 +2094,8 @@ private:
case OpCode::Id::HMUL2_R:
return '(' + op_a + " * " + op_b + ')';
default:
- LOG_CRITICAL(HW_GPU, "Unhandled half float instruction: {}",
- opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled half float instruction: {}",
+ opcode->get().GetName());
return std::string("0");
}
}();
@@ -2107,10 +2106,10 @@ private:
}
case OpCode::Type::ArithmeticHalfImmediate: {
if (opcode->get().GetId() == OpCode::Id::HADD2_IMM) {
- ASSERT_MSG(instr.alu_half_imm.ftz == 0, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.alu_half_imm.ftz != 0);
} else {
- ASSERT_MSG(instr.alu_half_imm.precision == Tegra::Shader::HalfPrecision::None,
- "Unimplemented");
+ UNIMPLEMENTED_IF(instr.alu_half_imm.precision !=
+ Tegra::Shader::HalfPrecision::None);
}
const std::string op_a = GetHalfFloat(
@@ -2140,11 +2139,14 @@ private:
std::string op_b = instr.ffma.negate_b ? "-" : "";
std::string op_c = instr.ffma.negate_c ? "-" : "";
- ASSERT_MSG(instr.ffma.cc == 0, "FFMA cc not implemented");
- ASSERT_MSG(instr.ffma.tab5980_0 == 1, "FFMA tab5980_0({}) not implemented",
- instr.ffma.tab5980_0.Value()); // Seems to be 1 by default based on SMO
- ASSERT_MSG(instr.ffma.tab5980_1 == 0, "FFMA tab5980_1({}) not implemented",
- instr.ffma.tab5980_1.Value());
+ UNIMPLEMENTED_IF_MSG(instr.ffma.cc != 0, "FFMA cc not implemented");
+ UNIMPLEMENTED_IF_MSG(
+ instr.ffma.tab5980_0 != 1, "FFMA tab5980_0({}) not implemented",
+ instr.ffma.tab5980_0.Value()); // Seems to be 1 by default based on SMO
+ UNIMPLEMENTED_IF_MSG(instr.ffma.tab5980_1 != 0, "FFMA tab5980_1({}) not implemented",
+ instr.ffma.tab5980_1.Value());
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in FFMA is not implemented");
switch (opcode->get().GetId()) {
case OpCode::Id::FFMA_CR: {
@@ -2170,27 +2172,19 @@ private:
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled FFMA instruction: {}", opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled FFMA instruction: {}", opcode->get().GetName());
}
}
regs.SetRegisterToFloat(instr.gpr0, 0, "fma(" + op_a + ", " + op_b + ", " + op_c + ')',
1, 1, instr.alu.saturate_d, 0, true);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "FFMA Generates an unhandled Control Code");
- UNREACHABLE();
- }
-
break;
}
case OpCode::Type::Hfma2: {
if (opcode->get().GetId() == OpCode::Id::HFMA2_RR) {
- ASSERT_MSG(instr.hfma2.rr.precision == Tegra::Shader::HalfPrecision::None,
- "Unimplemented");
+ UNIMPLEMENTED_IF(instr.hfma2.rr.precision != Tegra::Shader::HalfPrecision::None);
} else {
- ASSERT_MSG(instr.hfma2.precision == Tegra::Shader::HalfPrecision::None,
- "Unimplemented");
+ UNIMPLEMENTED_IF(instr.hfma2.precision != Tegra::Shader::HalfPrecision::None);
}
const bool saturate = opcode->get().GetId() == OpCode::Id::HFMA2_RR
? instr.hfma2.rr.saturate != 0
@@ -2240,7 +2234,7 @@ private:
case OpCode::Type::Conversion: {
switch (opcode->get().GetId()) {
case OpCode::Id::I2I_R: {
- ASSERT_MSG(!instr.conversion.selector, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.conversion.selector);
std::string op_a = regs.GetRegisterAsInteger(
instr.gpr20, 0, instr.conversion.is_input_signed, instr.conversion.src_size);
@@ -2260,8 +2254,10 @@ private:
}
case OpCode::Id::I2F_R:
case OpCode::Id::I2F_C: {
- ASSERT_MSG(instr.conversion.dest_size == Register::Size::Word, "Unimplemented");
- ASSERT_MSG(!instr.conversion.selector, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.conversion.dest_size != Register::Size::Word);
+ UNIMPLEMENTED_IF(instr.conversion.selector);
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in I2F is not implemented");
std::string op_a{};
@@ -2286,16 +2282,13 @@ private:
}
regs.SetRegisterToFloat(instr.gpr0, 0, op_a, 1, 1);
-
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "I2F Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::F2F_R: {
- ASSERT_MSG(instr.conversion.dest_size == Register::Size::Word, "Unimplemented");
- ASSERT_MSG(instr.conversion.src_size == Register::Size::Word, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.conversion.dest_size != Register::Size::Word);
+ UNIMPLEMENTED_IF(instr.conversion.src_size != Register::Size::Word);
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in F2F is not implemented");
std::string op_a = regs.GetRegisterAsFloat(instr.gpr20);
if (instr.conversion.abs_a) {
@@ -2322,23 +2315,19 @@ private:
op_a = "trunc(" + op_a + ')';
break;
default:
- LOG_CRITICAL(HW_GPU, "Unimplemented f2f rounding mode {}",
- static_cast<u32>(instr.conversion.f2f.rounding.Value()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unimplemented F2F rounding mode {}",
+ static_cast<u32>(instr.conversion.f2f.rounding.Value()));
break;
}
regs.SetRegisterToFloat(instr.gpr0, 0, op_a, 1, 1, instr.alu.saturate_d);
-
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "F2F Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
case OpCode::Id::F2I_R:
case OpCode::Id::F2I_C: {
- ASSERT_MSG(instr.conversion.src_size == Register::Size::Word, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.conversion.src_size != Register::Size::Word);
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in F2I is not implemented");
std::string op_a{};
if (instr.is_b_gpr) {
@@ -2369,9 +2358,8 @@ private:
op_a = "trunc(" + op_a + ')';
break;
default:
- LOG_CRITICAL(HW_GPU, "Unimplemented f2i rounding mode {}",
- static_cast<u32>(instr.conversion.f2i.rounding.Value()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unimplemented F2I rounding mode {}",
+ static_cast<u32>(instr.conversion.f2i.rounding.Value()));
break;
}
@@ -2383,16 +2371,10 @@ private:
regs.SetRegisterToInteger(instr.gpr0, instr.conversion.is_output_signed, 0, op_a, 1,
1, false, 0, instr.conversion.dest_size);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "F2I Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled conversion instruction: {}",
- opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled conversion instruction: {}", opcode->get().GetName());
}
}
break;
@@ -2401,10 +2383,10 @@ private:
switch (opcode->get().GetId()) {
case OpCode::Id::LD_A: {
// Note: Shouldn't this be interp mode flat? As in no interpolation made.
- ASSERT_MSG(instr.gpr8.Value() == Register::ZeroIndex,
- "Indirect attribute loads are not supported");
- ASSERT_MSG((instr.attribute.fmt20.immediate.Value() % sizeof(u32)) == 0,
- "Unaligned attribute loads are not supported");
+ UNIMPLEMENTED_IF_MSG(instr.gpr8.Value() != Register::ZeroIndex,
+ "Indirect attribute loads are not supported");
+ UNIMPLEMENTED_IF_MSG((instr.attribute.fmt20.immediate.Value() % sizeof(u32)) != 0,
+ "Unaligned attribute loads are not supported");
Tegra::Shader::IpaMode input_mode{Tegra::Shader::IpaInterpMode::Perspective,
Tegra::Shader::IpaSampleMode::Default};
@@ -2431,7 +2413,7 @@ private:
break;
}
case OpCode::Id::LD_C: {
- ASSERT_MSG(instr.ld_c.unknown == 0, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.ld_c.unknown != 0);
// Add an extra scope and declare the index register inside to prevent
// overwriting it in case it is used as an output of the LD instruction.
@@ -2459,9 +2441,8 @@ private:
break;
}
default:
- LOG_CRITICAL(HW_GPU, "Unhandled type: {}",
- static_cast<unsigned>(instr.ld_c.type.Value()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled type: {}",
+ static_cast<unsigned>(instr.ld_c.type.Value()));
}
--shader.scope;
@@ -2469,6 +2450,9 @@ private:
break;
}
case OpCode::Id::LD_L: {
+ UNIMPLEMENTED_IF_MSG(instr.ld_l.unknown == 1, "LD_L Unhandled mode: {}",
+ static_cast<unsigned>(instr.ld_l.unknown.Value()));
+
// Add an extra scope and declare the index register inside to prevent
// overwriting it in case it is used as an output of the LD instruction.
shader.AddLine('{');
@@ -2481,20 +2465,13 @@ private:
const std::string op_a = regs.GetLocalMemoryAsFloat("index");
- if (instr.ld_l.unknown != 1) {
- LOG_CRITICAL(HW_GPU, "LD_L Unhandled mode: {}",
- static_cast<unsigned>(instr.ld_l.unknown.Value()));
- UNREACHABLE();
- }
-
switch (instr.ldst_sl.type.Value()) {
case Tegra::Shader::StoreType::Bytes32:
regs.SetRegisterToFloat(instr.gpr0, 0, op_a, 1, 1);
break;
default:
- LOG_CRITICAL(HW_GPU, "LD_L Unhandled type: {}",
- static_cast<unsigned>(instr.ldst_sl.type.Value()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("LD_L Unhandled type: {}",
+ static_cast<unsigned>(instr.ldst_sl.type.Value()));
}
--shader.scope;
@@ -2502,10 +2479,10 @@ private:
break;
}
case OpCode::Id::ST_A: {
- ASSERT_MSG(instr.gpr8.Value() == Register::ZeroIndex,
- "Indirect attribute loads are not supported");
- ASSERT_MSG((instr.attribute.fmt20.immediate.Value() % sizeof(u32)) == 0,
- "Unaligned attribute loads are not supported");
+ UNIMPLEMENTED_IF_MSG(instr.gpr8.Value() != Register::ZeroIndex,
+ "Indirect attribute loads are not supported");
+ UNIMPLEMENTED_IF_MSG((instr.attribute.fmt20.immediate.Value() % sizeof(u32)) != 0,
+ "Unaligned attribute loads are not supported");
u64 next_element = instr.attribute.fmt20.element;
u64 next_index = static_cast<u64>(instr.attribute.fmt20.index.Value());
@@ -2530,6 +2507,9 @@ private:
break;
}
case OpCode::Id::ST_L: {
+ UNIMPLEMENTED_IF_MSG(instr.st_l.unknown == 0, "ST_L Unhandled mode: {}",
+ static_cast<unsigned>(instr.st_l.unknown.Value()));
+
// Add an extra scope and declare the index register inside to prevent
// overwriting it in case it is used as an output of the LD instruction.
shader.AddLine('{');
@@ -2540,20 +2520,13 @@ private:
shader.AddLine("uint index = (" + op + " / 4);");
- if (instr.st_l.unknown != 0) {
- LOG_CRITICAL(HW_GPU, "ST_L Unhandled mode: {}",
- static_cast<unsigned>(instr.st_l.unknown.Value()));
- UNREACHABLE();
- }
-
switch (instr.ldst_sl.type.Value()) {
case Tegra::Shader::StoreType::Bytes32:
regs.SetLocalMemoryAsFloat("index", regs.GetRegisterAsFloat(instr.gpr0));
break;
default:
- LOG_CRITICAL(HW_GPU, "ST_L Unhandled type: {}",
- static_cast<unsigned>(instr.ldst_sl.type.Value()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("ST_L Unhandled type: {}",
+ static_cast<unsigned>(instr.ldst_sl.type.Value()));
}
--shader.scope;
@@ -2565,78 +2538,99 @@ private:
std::string coord;
const bool is_array = instr.tex.array != 0;
- ASSERT_MSG(!instr.tex.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
- "NODEP is not implemented");
- ASSERT_MSG(!instr.tex.UsesMiscMode(Tegra::Shader::TextureMiscMode::AOFFI),
- "AOFFI is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tex.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
+ "NODEP is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tex.UsesMiscMode(Tegra::Shader::TextureMiscMode::AOFFI),
+ "AOFFI is not implemented");
const bool depth_compare =
instr.tex.UsesMiscMode(Tegra::Shader::TextureMiscMode::DC);
u32 num_coordinates = TextureCoordinates(texture_type);
- if (depth_compare)
- num_coordinates += 1;
+ u32 start_index = 0;
+ std::string array_elem;
+ if (is_array) {
+ array_elem = regs.GetRegisterAsInteger(instr.gpr8);
+ start_index = 1;
+ }
+ const auto process_mode = instr.tex.GetTextureProcessMode();
+ u32 start_index_b = 0;
+ std::string lod_value;
+ if (process_mode != Tegra::Shader::TextureProcessMode::LZ &&
+ process_mode != Tegra::Shader::TextureProcessMode::None) {
+ start_index_b = 1;
+ lod_value = regs.GetRegisterAsFloat(instr.gpr20);
+ }
+
+ std::string depth_value;
+ if (depth_compare) {
+ depth_value = regs.GetRegisterAsFloat(instr.gpr20.Value() + start_index_b);
+ }
+
+ bool depth_compare_extra = false;
switch (num_coordinates) {
case 1: {
+ const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + start_index);
if (is_array) {
- const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- coord = "vec2 coords = vec2(" + x + ", " + index + ");";
+ if (depth_compare) {
+ coord = "vec3 coords = vec3(" + x + ", " + depth_value + ", " +
+ array_elem + ");";
+ } else {
+ coord = "vec2 coords = vec2(" + x + ", " + array_elem + ");";
+ }
} else {
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
- coord = "float coords = " + x + ';';
+ if (depth_compare) {
+ coord = "vec2 coords = vec2(" + x + ", " + depth_value + ");";
+ } else {
+ coord = "float coords = " + x + ';';
+ }
}
break;
}
case 2: {
+ const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + start_index);
+ const std::string y =
+ regs.GetRegisterAsFloat(instr.gpr8.Value() + start_index + 1);
if (is_array) {
- const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 2);
- coord = "vec3 coords = vec3(" + x + ", " + y + ", " + index + ");";
+ if (depth_compare) {
+ coord = "vec4 coords = vec4(" + x + ", " + y + ", " + depth_value +
+ ", " + array_elem + ");";
+ } else {
+ coord = "vec3 coords = vec3(" + x + ", " + y + ", " + array_elem + ");";
+ }
} else {
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
- const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+ if (depth_compare) {
+ coord =
+ "vec3 coords = vec3(" + x + ", " + y + ", " + depth_value + ");";
+ } else {
+ coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+ }
}
break;
}
case 3: {
- if (depth_compare) {
- if (is_array) {
- const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- const std::string y = regs.GetRegisterAsFloat(instr.gpr20);
- const std::string z = regs.GetRegisterAsFloat(instr.gpr20.Value() + 1);
- coord = "vec4 coords = vec4(" + x + ", " + y + ", " + z + ", " + index +
- ");";
- } else {
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
- const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- const std::string z = regs.GetRegisterAsFloat(instr.gpr20);
- coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
- }
+ const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + start_index);
+ const std::string y =
+ regs.GetRegisterAsFloat(instr.gpr8.Value() + start_index + 1);
+ const std::string z =
+ regs.GetRegisterAsFloat(instr.gpr8.Value() + start_index + 2);
+ if (is_array) {
+ depth_compare_extra = depth_compare;
+ coord = "vec4 coords = vec4(" + x + ", " + y + ", " + z + ", " +
+ array_elem + ");";
} else {
- if (is_array) {
- const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 2);
- const std::string z = regs.GetRegisterAsFloat(instr.gpr8.Value() + 3);
- coord = "vec4 coords = vec4(" + x + ", " + y + ", " + z + ", " + index +
- ");";
+ if (depth_compare) {
+ coord = "vec4 coords = vec4(" + x + ", " + y + ", " + z + ", " +
+ depth_value + ");";
} else {
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
- const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- const std::string z = regs.GetRegisterAsFloat(instr.gpr8.Value() + 2);
coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
}
}
break;
}
default:
- LOG_CRITICAL(HW_GPU, "Unhandled coordinates number {}",
- static_cast<u32>(num_coordinates));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled coordinates number {}",
+ static_cast<u32>(num_coordinates));
// Fallback to interpreting as a 2D texture for now
const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
@@ -2644,126 +2638,172 @@ private:
coord = "vec2 coords = vec2(" + x + ", " + y + ");";
texture_type = Tegra::Shader::TextureType::Texture2D;
}
- // TODO: make sure coordinates are always indexed to gpr8 and gpr20 is always bias
- // or lod.
- std::string op_c;
const std::string sampler =
GetSampler(instr.sampler, texture_type, is_array, depth_compare);
// Add an extra scope and declare the texture coords inside to prevent
// overwriting them in case they are used as outputs of the texs instruction.
- shader.AddLine("{");
+ shader.AddLine('{');
++shader.scope;
shader.AddLine(coord);
std::string texture;
switch (instr.tex.GetTextureProcessMode()) {
case Tegra::Shader::TextureProcessMode::None: {
- texture = "texture(" + sampler + ", coords)";
+ if (!depth_compare_extra) {
+ texture = "texture(" + sampler + ", coords)";
+ } else {
+ texture = "texture(" + sampler + ", coords, " + depth_value + ')';
+ }
break;
}
case Tegra::Shader::TextureProcessMode::LZ: {
- texture = "textureLod(" + sampler + ", coords, 0.0)";
+ if (!depth_compare_extra) {
+ texture = "textureLod(" + sampler + ", coords, 0.0)";
+ } else {
+ texture = "texture(" + sampler + ", coords, " + depth_value + ')';
+ }
break;
}
case Tegra::Shader::TextureProcessMode::LB:
case Tegra::Shader::TextureProcessMode::LBA: {
- if (depth_compare) {
- if (is_array)
- op_c = regs.GetRegisterAsFloat(instr.gpr20.Value() + 2);
- else
- op_c = regs.GetRegisterAsFloat(instr.gpr20.Value() + 1);
+ // TODO: Figure if A suffix changes the equation at all.
+ if (!depth_compare_extra) {
+ texture = "texture(" + sampler + ", coords, " + lod_value + ')';
} else {
- op_c = regs.GetRegisterAsFloat(instr.gpr20);
+ texture = "texture(" + sampler + ", coords, " + depth_value + ')';
+ LOG_WARNING(HW_GPU,
+ "OpenGL Limitation: can't set bias value along depth compare");
}
- // TODO: Figure if A suffix changes the equation at all.
- texture = "texture(" + sampler + ", coords, " + op_c + ')';
break;
}
case Tegra::Shader::TextureProcessMode::LL:
case Tegra::Shader::TextureProcessMode::LLA: {
- if (num_coordinates <= 2) {
- op_c = regs.GetRegisterAsFloat(instr.gpr20);
+ // TODO: Figure if A suffix changes the equation at all.
+ if (!depth_compare_extra) {
+ texture = "textureLod(" + sampler + ", coords, " + lod_value + ')';
} else {
- op_c = regs.GetRegisterAsFloat(instr.gpr20.Value() + 1);
+ texture = "texture(" + sampler + ", coords, " + depth_value + ')';
+ LOG_WARNING(HW_GPU,
+ "OpenGL Limitation: can't set lod value along depth compare");
}
- // TODO: Figure if A suffix changes the equation at all.
- texture = "textureLod(" + sampler + ", coords, " + op_c + ')';
break;
}
default: {
- texture = "texture(" + sampler + ", coords)";
- LOG_CRITICAL(HW_GPU, "Unhandled texture process mode {}",
- static_cast<u32>(instr.tex.GetTextureProcessMode()));
- UNREACHABLE();
+ if (!depth_compare_extra) {
+ texture = "texture(" + sampler + ", coords)";
+ } else {
+ texture = "texture(" + sampler + ", coords, " + depth_value + ')';
+ }
+ UNIMPLEMENTED_MSG("Unhandled texture process mode {}",
+ static_cast<u32>(instr.tex.GetTextureProcessMode()));
}
}
if (!depth_compare) {
+ shader.AddLine("vec4 texture_tmp = " + texture + ';');
std::size_t dest_elem{};
for (std::size_t elem = 0; elem < 4; ++elem) {
if (!instr.tex.IsComponentEnabled(elem)) {
// Skip disabled components
continue;
}
- regs.SetRegisterToFloat(instr.gpr0, elem, texture, 1, 4, false, dest_elem);
+ regs.SetRegisterToFloat(instr.gpr0, elem, "texture_tmp", 1, 4, false,
+ dest_elem);
++dest_elem;
}
} else {
regs.SetRegisterToFloat(instr.gpr0, 0, texture, 1, 1, false);
}
--shader.scope;
- shader.AddLine("}");
+ shader.AddLine('}');
break;
}
case OpCode::Id::TEXS: {
- std::string coord;
Tegra::Shader::TextureType texture_type{instr.texs.GetTextureType()};
bool is_array{instr.texs.IsArrayTexture()};
- ASSERT_MSG(!instr.texs.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
- "NODEP is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.texs.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
+ "NODEP is not implemented");
const bool depth_compare =
instr.texs.UsesMiscMode(Tegra::Shader::TextureMiscMode::DC);
u32 num_coordinates = TextureCoordinates(texture_type);
- if (depth_compare)
- num_coordinates += 1;
+ const auto process_mode = instr.texs.GetTextureProcessMode();
+ std::string lod_value;
+ std::string coord;
+ u32 lod_offset = 0;
+ if (process_mode == Tegra::Shader::TextureProcessMode::LL) {
+ if (num_coordinates > 2) {
+ lod_value = regs.GetRegisterAsFloat(instr.gpr20.Value() + 1);
+ lod_offset = 2;
+ } else {
+ lod_value = regs.GetRegisterAsFloat(instr.gpr20);
+ lod_offset = 1;
+ }
+ }
switch (num_coordinates) {
+ case 1: {
+ coord = "float coords = " + regs.GetRegisterAsFloat(instr.gpr8) + ';';
+ break;
+ }
case 2: {
if (is_array) {
- const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- const std::string y = regs.GetRegisterAsFloat(instr.gpr20);
- coord = "vec3 coords = vec3(" + x + ", " + y + ", " + index + ");";
+ if (depth_compare) {
+ const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
+ const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+ const std::string y = regs.GetRegisterAsFloat(instr.gpr20);
+ const std::string z = regs.GetRegisterAsFloat(instr.gpr20.Value() + 1);
+ coord = "vec4 coords = vec4(" + x + ", " + y + ", " + z + ", " + index +
+ ");";
+ } else {
+ const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
+ const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+ const std::string y = regs.GetRegisterAsFloat(instr.gpr20);
+ coord = "vec3 coords = vec3(" + x + ", " + y + ", " + index + ");";
+ }
} else {
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
- const std::string y = regs.GetRegisterAsFloat(instr.gpr20);
- coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+ if (lod_offset != 0) {
+ if (depth_compare) {
+ const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+ const std::string y =
+ regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+ const std::string z =
+ regs.GetRegisterAsFloat(instr.gpr20.Value() + lod_offset);
+ coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
+ } else {
+ const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+ const std::string y =
+ regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+ coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+ }
+ } else {
+ if (depth_compare) {
+ const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+ const std::string y =
+ regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+ const std::string z = regs.GetRegisterAsFloat(instr.gpr20);
+ coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
+ } else {
+ const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+ const std::string y = regs.GetRegisterAsFloat(instr.gpr20);
+ coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+ }
+ }
}
break;
}
case 3: {
- if (is_array) {
- const std::string index = regs.GetRegisterAsInteger(instr.gpr8);
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 2);
- const std::string z = regs.GetRegisterAsFloat(instr.gpr20);
- coord =
- "vec4 coords = vec4(" + x + ", " + y + ", " + z + ", " + index + ");";
- } else {
- const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
- const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- const std::string z = regs.GetRegisterAsFloat(instr.gpr20);
- coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
- }
+ const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
+ const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+ const std::string z = regs.GetRegisterAsFloat(instr.gpr20);
+ coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
break;
}
default:
- LOG_CRITICAL(HW_GPU, "Unhandled coordinates number {}",
- static_cast<u32>(num_coordinates));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled coordinates number {}",
+ static_cast<u32>(num_coordinates));
// Fallback to interpreting as a 2D texture for now
const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
@@ -2775,7 +2815,7 @@ private:
const std::string sampler =
GetSampler(instr.sampler, texture_type, is_array, depth_compare);
std::string texture;
- switch (instr.texs.GetTextureProcessMode()) {
+ switch (process_mode) {
case Tegra::Shader::TextureProcessMode::None: {
texture = "texture(" + sampler + ", coords)";
break;
@@ -2789,15 +2829,13 @@ private:
break;
}
case Tegra::Shader::TextureProcessMode::LL: {
- const std::string op_c = regs.GetRegisterAsFloat(instr.gpr20.Value() + 1);
- texture = "textureLod(" + sampler + ", coords, " + op_c + ')';
+ texture = "textureLod(" + sampler + ", coords, " + lod_value + ')';
break;
}
default: {
texture = "texture(" + sampler + ", coords)";
- LOG_CRITICAL(HW_GPU, "Unhandled texture process mode {}",
- static_cast<u32>(instr.texs.GetTextureProcessMode()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled texture process mode {}",
+ static_cast<u32>(instr.texs.GetTextureProcessMode()));
}
}
if (!depth_compare) {
@@ -2805,47 +2843,48 @@ private:
} else {
WriteTexsInstruction(instr, coord, "vec4(" + texture + ')');
}
+
break;
}
case OpCode::Id::TLDS: {
- std::string coord;
const Tegra::Shader::TextureType texture_type{instr.tlds.GetTextureType()};
const bool is_array{instr.tlds.IsArrayTexture()};
ASSERT(texture_type == Tegra::Shader::TextureType::Texture2D);
ASSERT(is_array == false);
- ASSERT_MSG(!instr.tlds.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
- "NODEP is not implemented");
- ASSERT_MSG(!instr.tlds.UsesMiscMode(Tegra::Shader::TextureMiscMode::AOFFI),
- "AOFFI is not implemented");
- ASSERT_MSG(!instr.tlds.UsesMiscMode(Tegra::Shader::TextureMiscMode::MZ),
- "MZ is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
+ "NODEP is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(Tegra::Shader::TextureMiscMode::AOFFI),
+ "AOFFI is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(Tegra::Shader::TextureMiscMode::MZ),
+ "MZ is not implemented");
- u32 op_c_offset = 0;
+ u32 extra_op_offset = 0;
+
+ // Scope to avoid variable name overlaps.
+ shader.AddLine('{');
+ ++shader.scope;
+ std::string coords;
switch (texture_type) {
case Tegra::Shader::TextureType::Texture1D: {
const std::string x = regs.GetRegisterAsInteger(instr.gpr8);
- coord = "int coords = " + x + ';';
+ coords = "float coords = " + x + ';';
break;
}
case Tegra::Shader::TextureType::Texture2D: {
- if (is_array) {
- LOG_CRITICAL(HW_GPU, "Unhandled 2d array texture");
- UNREACHABLE();
- } else {
- const std::string x = regs.GetRegisterAsInteger(instr.gpr8);
- const std::string y = regs.GetRegisterAsInteger(instr.gpr20);
- coord = "ivec2 coords = ivec2(" + x + ", " + y + ");";
- op_c_offset = 1;
- }
+ UNIMPLEMENTED_IF_MSG(is_array, "Unhandled 2d array texture");
+
+ const std::string x = regs.GetRegisterAsInteger(instr.gpr8);
+ const std::string y = regs.GetRegisterAsInteger(instr.gpr20);
+ // shader.AddLine("ivec2 coords = ivec2(" + x + ", " + y + ");");
+ coords = "ivec2 coords = ivec2(" + x + ", " + y + ");";
+ extra_op_offset = 1;
break;
}
default:
- LOG_CRITICAL(HW_GPU, "Unhandled texture type {}",
- static_cast<u32>(texture_type));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled texture type {}", static_cast<u32>(texture_type));
}
const std::string sampler =
GetSampler(instr.sampler, texture_type, is_array, false);
@@ -2856,19 +2895,22 @@ private:
break;
}
case Tegra::Shader::TextureProcessMode::LL: {
- const std::string op_c =
- regs.GetRegisterAsInteger(instr.gpr20.Value() + op_c_offset);
- texture = "texelFetch(" + sampler + ", coords, " + op_c + ')';
+ shader.AddLine(
+ "float lod = " +
+ regs.GetRegisterAsInteger(instr.gpr20.Value() + extra_op_offset) + ';');
+ texture = "texelFetch(" + sampler + ", coords, lod)";
break;
}
default: {
texture = "texelFetch(" + sampler + ", coords, 0)";
- LOG_CRITICAL(HW_GPU, "Unhandled texture process mode {}",
- static_cast<u32>(instr.tlds.GetTextureProcessMode()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled texture process mode {}",
+ static_cast<u32>(instr.tlds.GetTextureProcessMode()));
}
}
- WriteTexsInstruction(instr, coord, texture);
+ WriteTexsInstruction(instr, coords, texture);
+
+ --shader.scope;
+ shader.AddLine('}');
break;
}
case OpCode::Id::TLD4: {
@@ -2876,14 +2918,14 @@ private:
ASSERT(instr.tld4.array == 0);
std::string coord;
- ASSERT_MSG(!instr.tld4.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
- "NODEP is not implemented");
- ASSERT_MSG(!instr.tld4.UsesMiscMode(Tegra::Shader::TextureMiscMode::AOFFI),
- "AOFFI is not implemented");
- ASSERT_MSG(!instr.tld4.UsesMiscMode(Tegra::Shader::TextureMiscMode::NDV),
- "NDV is not implemented");
- ASSERT_MSG(!instr.tld4.UsesMiscMode(Tegra::Shader::TextureMiscMode::PTP),
- "PTP is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
+ "NODEP is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(Tegra::Shader::TextureMiscMode::AOFFI),
+ "AOFFI is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(Tegra::Shader::TextureMiscMode::NDV),
+ "NDV is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(Tegra::Shader::TextureMiscMode::PTP),
+ "PTP is not implemented");
const bool depth_compare =
instr.tld4.UsesMiscMode(Tegra::Shader::TextureMiscMode::DC);
auto texture_type = instr.tld4.texture_type.Value();
@@ -2891,61 +2933,71 @@ private:
if (depth_compare)
num_coordinates += 1;
+ // Add an extra scope and declare the texture coords inside to prevent
+ // overwriting them in case they are used as outputs of the texs instruction.
+ shader.AddLine('{');
+ ++shader.scope;
+
switch (num_coordinates) {
case 2: {
const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+ shader.AddLine("vec2 coords = vec2(" + x + ", " + y + ");");
break;
}
case 3: {
const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
const std::string z = regs.GetRegisterAsFloat(instr.gpr8.Value() + 2);
- coord = "vec3 coords = vec3(" + x + ", " + y + ", " + z + ");";
+ shader.AddLine("vec3 coords = vec3(" + x + ", " + y + ", " + z + ");");
break;
}
default:
- LOG_CRITICAL(HW_GPU, "Unhandled coordinates number {}",
- static_cast<u32>(num_coordinates));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled coordinates number {}",
+ static_cast<u32>(num_coordinates));
const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- coord = "vec2 coords = vec2(" + x + ", " + y + ");";
+ shader.AddLine("vec2 coords = vec2(" + x + ", " + y + ");");
texture_type = Tegra::Shader::TextureType::Texture2D;
}
const std::string sampler =
GetSampler(instr.sampler, texture_type, false, depth_compare);
- // Add an extra scope and declare the texture coords inside to prevent
- // overwriting them in case they are used as outputs of the texs instruction.
- shader.AddLine("{");
- ++shader.scope;
- shader.AddLine(coord);
+
const std::string texture = "textureGather(" + sampler + ", coords, " +
std::to_string(instr.tld4.component) + ')';
+
if (!depth_compare) {
+ shader.AddLine("vec4 texture_tmp = " + texture + ';');
std::size_t dest_elem{};
for (std::size_t elem = 0; elem < 4; ++elem) {
if (!instr.tex.IsComponentEnabled(elem)) {
// Skip disabled components
continue;
}
- regs.SetRegisterToFloat(instr.gpr0, elem, texture, 1, 4, false, dest_elem);
+ regs.SetRegisterToFloat(instr.gpr0, elem, "texture_tmp", 1, 4, false,
+ dest_elem);
++dest_elem;
}
} else {
regs.SetRegisterToFloat(instr.gpr0, 0, texture, 1, 1, false);
}
--shader.scope;
- shader.AddLine("}");
+ shader.AddLine('}');
break;
}
case OpCode::Id::TLD4S: {
- ASSERT_MSG(!instr.tld4s.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
- "NODEP is not implemented");
- ASSERT_MSG(!instr.tld4s.UsesMiscMode(Tegra::Shader::TextureMiscMode::AOFFI),
- "AOFFI is not implemented");
+ UNIMPLEMENTED_IF_MSG(
+ instr.tld4s.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
+ "NODEP is not implemented");
+ UNIMPLEMENTED_IF_MSG(
+ instr.tld4s.UsesMiscMode(Tegra::Shader::TextureMiscMode::AOFFI),
+ "AOFFI is not implemented");
+
+ // Scope to avoid variable name overlaps.
+ shader.AddLine('{');
+ ++shader.scope;
+ std::string coords;
const bool depth_compare =
instr.tld4s.UsesMiscMode(Tegra::Shader::TextureMiscMode::DC);
@@ -2954,28 +3006,32 @@ private:
// TODO(Subv): Figure out how the sampler type is encoded in the TLD4S instruction.
const std::string sampler = GetSampler(
instr.sampler, Tegra::Shader::TextureType::Texture2D, false, depth_compare);
- std::string coord;
if (!depth_compare) {
- coord = "vec2 coords = vec2(" + op_a + ", " + op_b + ");";
+ coords = "vec2 coords = vec2(" + op_a + ", " + op_b + ");";
} else {
// Note: TLD4S coordinate encoding works just like TEXS's
- const std::string op_c = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
- coord = "vec3 coords = vec3(" + op_a + ", " + op_c + ", " + op_b + ");";
+ const std::string op_y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
+ coords = "vec3 coords = vec3(" + op_a + ", " + op_y + ", " + op_b + ");";
}
const std::string texture = "textureGather(" + sampler + ", coords, " +
std::to_string(instr.tld4s.component) + ')';
if (!depth_compare) {
- WriteTexsInstruction(instr, coord, texture);
+ WriteTexsInstruction(instr, coords, texture);
} else {
- WriteTexsInstruction(instr, coord, "vec4(" + texture + ')');
+ WriteTexsInstruction(instr, coords, "vec4(" + texture + ')');
}
+
+ --shader.scope;
+ shader.AddLine('}');
break;
}
case OpCode::Id::TXQ: {
- ASSERT_MSG(!instr.txq.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
- "NODEP is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.txq.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
+ "NODEP is not implemented");
+ ++shader.scope;
+ shader.AddLine('{');
// TODO: the new commits on the texture refactor, change the way samplers work.
// Sadly, not all texture instructions specify the type of texture their sampler
// uses. This must be fixed at a later instance.
@@ -2983,23 +3039,30 @@ private:
GetSampler(instr.sampler, Tegra::Shader::TextureType::Texture2D, false, false);
switch (instr.txq.query_type) {
case Tegra::Shader::TextureQueryType::Dimension: {
- const std::string texture = "textureQueryLevels(" + sampler + ')';
- regs.SetRegisterToInteger(instr.gpr0, true, 0, texture, 1, 1);
+ const std::string texture = "textureSize(" + sampler + ", " +
+ regs.GetRegisterAsInteger(instr.gpr8) + ')';
+ const std::string mip_level = "textureQueryLevels(" + sampler + ')';
+ shader.AddLine("ivec2 sizes = " + texture + ';');
+ regs.SetRegisterToInteger(instr.gpr0, true, 0, "sizes.x", 1, 1);
+ regs.SetRegisterToInteger(instr.gpr0.Value() + 1, true, 0, "sizes.y", 1, 1);
+ regs.SetRegisterToInteger(instr.gpr0.Value() + 2, true, 0, "0", 1, 1);
+ regs.SetRegisterToInteger(instr.gpr0.Value() + 3, true, 0, mip_level, 1, 1);
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled texture query type: {}",
- static_cast<u32>(instr.txq.query_type.Value()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled texture query type: {}",
+ static_cast<u32>(instr.txq.query_type.Value()));
}
}
+ --shader.scope;
+ shader.AddLine('}');
break;
}
case OpCode::Id::TMML: {
- ASSERT_MSG(!instr.tmml.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
- "NODEP is not implemented");
- ASSERT_MSG(!instr.tmml.UsesMiscMode(Tegra::Shader::TextureMiscMode::NDV),
- "NDV is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tmml.UsesMiscMode(Tegra::Shader::TextureMiscMode::NODEP),
+ "NODEP is not implemented");
+ UNIMPLEMENTED_IF_MSG(instr.tmml.UsesMiscMode(Tegra::Shader::TextureMiscMode::NDV),
+ "NDV is not implemented");
const std::string x = regs.GetRegisterAsFloat(instr.gpr8);
const bool is_array = instr.tmml.array != 0;
@@ -3021,9 +3084,7 @@ private:
break;
}
default:
- LOG_CRITICAL(HW_GPU, "Unhandled texture type {}",
- static_cast<u32>(texture_type));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled texture type {}", static_cast<u32>(texture_type));
// Fallback to interpreting as a 2D texture for now
const std::string y = regs.GetRegisterAsFloat(instr.gpr8.Value() + 1);
@@ -3046,8 +3107,7 @@ private:
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled memory instruction: {}", opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled memory instruction: {}", opcode->get().GetName());
}
}
break;
@@ -3133,7 +3193,7 @@ private:
break;
}
case OpCode::Type::HalfSetPredicate: {
- ASSERT_MSG(instr.hsetp2.ftz == 0, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.hsetp2.ftz != 0);
const std::string op_a =
GetHalfFloat(regs.GetRegisterAsInteger(instr.gpr8, 0, false), instr.hsetp2.type_a,
@@ -3178,6 +3238,9 @@ private:
break;
}
case OpCode::Type::PredicateSetRegister: {
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in PSET is not implemented");
+
const std::string op_a =
GetPredicateCondition(instr.pset.pred12, instr.pset.neg_pred12 != 0);
const std::string op_b =
@@ -3198,12 +3261,6 @@ private:
const std::string value = '(' + result + ") ? 1.0 : 0.0";
regs.SetRegisterToFloat(instr.gpr0, 0, value, 1, 1);
}
-
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "PSET Generates an unhandled Control Code");
- UNREACHABLE();
- }
-
break;
}
case OpCode::Type::PredicateSetPredicate: {
@@ -3241,21 +3298,19 @@ private:
const std::string pred =
GetPredicateCondition(instr.csetp.pred39, instr.csetp.neg_pred39 != 0);
const std::string combiner = GetPredicateCombiner(instr.csetp.op);
- const std::string control_code = regs.GetControlCode(instr.csetp.cc);
+ const std::string condition_code = regs.GetConditionCode(instr.csetp.cc);
if (instr.csetp.pred3 != static_cast<u64>(Pred::UnusedIndex)) {
SetPredicate(instr.csetp.pred3,
- '(' + control_code + ") " + combiner + " (" + pred + ')');
+ '(' + condition_code + ") " + combiner + " (" + pred + ')');
}
if (instr.csetp.pred0 != static_cast<u64>(Pred::UnusedIndex)) {
SetPredicate(instr.csetp.pred0,
- "!(" + control_code + ") " + combiner + " (" + pred + ')');
+ "!(" + condition_code + ") " + combiner + " (" + pred + ')');
}
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled predicate instruction: {}",
- opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled predicate instruction: {}", opcode->get().GetName());
}
}
break;
@@ -3363,7 +3418,7 @@ private:
break;
}
case OpCode::Type::HalfSet: {
- ASSERT_MSG(instr.hset2.ftz == 0, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.hset2.ftz != 0);
const std::string op_a =
GetHalfFloat(regs.GetRegisterAsInteger(instr.gpr8, 0, false), instr.hset2.type_a,
@@ -3407,15 +3462,17 @@ private:
break;
}
case OpCode::Type::Xmad: {
- ASSERT_MSG(!instr.xmad.sign_a, "Unimplemented");
- ASSERT_MSG(!instr.xmad.sign_b, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.xmad.sign_a);
+ UNIMPLEMENTED_IF(instr.xmad.sign_b);
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in XMAD is not implemented");
std::string op_a{regs.GetRegisterAsInteger(instr.gpr8, 0, instr.xmad.sign_a)};
std::string op_b;
std::string op_c;
// TODO(bunnei): Needs to be fixed once op_a or op_b is signed
- ASSERT_MSG(instr.xmad.sign_a == instr.xmad.sign_b, "Unimplemented");
+ UNIMPLEMENTED_IF(instr.xmad.sign_a != instr.xmad.sign_b);
const bool is_signed{instr.xmad.sign_a == 1};
bool is_merge{};
@@ -3448,8 +3505,7 @@ private:
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled XMAD instruction: {}", opcode->get().GetName());
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled XMAD instruction: {}", opcode->get().GetName());
}
}
@@ -3485,9 +3541,8 @@ private:
op_c = "((" + op_c + ") + (" + src2 + "<< 16))";
break;
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled XMAD mode: {}",
- static_cast<u32>(instr.xmad.mode.Value()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled XMAD mode: {}",
+ static_cast<u32>(instr.xmad.mode.Value()));
}
}
@@ -3497,25 +3552,19 @@ private:
}
regs.SetRegisterToInteger(instr.gpr0, is_signed, 0, sum, 1, 1);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "XMAD Generates an unhandled Control Code");
- UNREACHABLE();
- }
break;
}
default: {
switch (opcode->get().GetId()) {
case OpCode::Id::EXIT: {
+ const Tegra::Shader::ConditionCode cc = instr.flow_condition_code;
+ UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T,
+ "EXIT condition code used: {}", static_cast<u32>(cc));
+
if (stage == Maxwell3D::Regs::ShaderStage::Fragment) {
EmitFragmentOutputsWrite();
}
- const Tegra::Shader::ControlCode cc = instr.flow_control_code;
- if (cc != Tegra::Shader::ControlCode::T) {
- LOG_CRITICAL(HW_GPU, "EXIT Control Code used: {}", static_cast<u32>(cc));
- UNREACHABLE();
- }
-
switch (instr.flow.cond) {
case Tegra::Shader::FlowCondition::Always:
shader.AddLine("return true;");
@@ -3530,26 +3579,24 @@ private:
case Tegra::Shader::FlowCondition::Fcsm_Tr:
// TODO(bunnei): What is this used for? If we assume this conditon is not
// satisifed, dual vertex shaders in Farming Simulator make more sense
- LOG_CRITICAL(HW_GPU, "Skipping unknown FlowCondition::Fcsm_Tr");
+ UNIMPLEMENTED_MSG("Skipping unknown FlowCondition::Fcsm_Tr");
break;
default:
- LOG_CRITICAL(HW_GPU, "Unhandled flow condition: {}",
- static_cast<u32>(instr.flow.cond.Value()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled flow condition: {}",
+ static_cast<u32>(instr.flow.cond.Value()));
}
break;
}
case OpCode::Id::KIL: {
- ASSERT(instr.flow.cond == Tegra::Shader::FlowCondition::Always);
+ UNIMPLEMENTED_IF(instr.flow.cond != Tegra::Shader::FlowCondition::Always);
+
+ const Tegra::Shader::ConditionCode cc = instr.flow_condition_code;
+ UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T,
+ "KIL condition code used: {}", static_cast<u32>(cc));
// Enclose "discard" in a conditional, so that GLSL compilation does not complain
// about unexecuted instructions that may follow this.
- const Tegra::Shader::ControlCode cc = instr.flow_control_code;
- if (cc != Tegra::Shader::ControlCode::T) {
- LOG_CRITICAL(HW_GPU, "KIL Control Code used: {}", static_cast<u32>(cc));
- UNREACHABLE();
- }
shader.AddLine("if (true) {");
++shader.scope;
shader.AddLine("discard;");
@@ -3559,7 +3606,8 @@ private:
break;
}
case OpCode::Id::OUT_R: {
- ASSERT(instr.gpr20.Value() == Register::ZeroIndex);
+ UNIMPLEMENTED_IF_MSG(instr.gpr20.Value() != Register::ZeroIndex,
+ "Stream buffer is not supported");
ASSERT_MSG(stage == Maxwell3D::Regs::ShaderStage::Geometry,
"OUT is expected to be used in a geometry shader.");
@@ -3586,18 +3634,17 @@ private:
break;
}
default: {
- LOG_CRITICAL(HW_GPU, "Unhandled system move: {}",
- static_cast<u32>(instr.sys20.Value()));
- UNREACHABLE();
+ UNIMPLEMENTED_MSG("Unhandled system move: {}",
+ static_cast<u32>(instr.sys20.Value()));
}
}
break;
}
case OpCode::Id::ISBERD: {
- ASSERT(instr.isberd.o == 0);
- ASSERT(instr.isberd.skew == 0);
- ASSERT(instr.isberd.shift == Tegra::Shader::IsberdShift::None);
- ASSERT(instr.isberd.mode == Tegra::Shader::IsberdMode::None);
+ UNIMPLEMENTED_IF(instr.isberd.o != 0);
+ UNIMPLEMENTED_IF(instr.isberd.skew != 0);
+ UNIMPLEMENTED_IF(instr.isberd.shift != Tegra::Shader::IsberdShift::None);
+ UNIMPLEMENTED_IF(instr.isberd.mode != Tegra::Shader::IsberdMode::None);
ASSERT_MSG(stage == Maxwell3D::Regs::ShaderStage::Geometry,
"ISBERD is expected to be used in a geometry shader.");
LOG_WARNING(HW_GPU, "ISBERD instruction is incomplete");
@@ -3605,13 +3652,13 @@ private:
break;
}
case OpCode::Id::BRA: {
- ASSERT_MSG(instr.bra.constant_buffer == 0,
- "BRA with constant buffers are not implemented");
- const Tegra::Shader::ControlCode cc = instr.flow_control_code;
- if (cc != Tegra::Shader::ControlCode::T) {
- LOG_CRITICAL(HW_GPU, "BRA Control Code used: {}", static_cast<u32>(cc));
- UNREACHABLE();
- }
+ UNIMPLEMENTED_IF_MSG(instr.bra.constant_buffer != 0,
+ "BRA with constant buffers are not implemented");
+
+ const Tegra::Shader::ConditionCode cc = instr.flow_condition_code;
+ UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T,
+ "BRA condition code used: {}", static_cast<u32>(cc));
+
const u32 target = offset + instr.bra.GetBranchTarget();
shader.AddLine("{ jmp_to = " + std::to_string(target) + "u; break; }");
break;
@@ -3634,7 +3681,8 @@ private:
// The SSY opcode tells the GPU where to re-converge divergent execution paths, it
// sets the target of the jump that the SYNC instruction will make. The SSY opcode
// has a similar structure to the BRA opcode.
- ASSERT_MSG(instr.bra.constant_buffer == 0, "Constant buffer flow is not supported");
+ UNIMPLEMENTED_IF_MSG(instr.bra.constant_buffer != 0,
+ "Constant buffer flow is not supported");
const u32 target = offset + instr.bra.GetBranchTarget();
EmitPushToFlowStack(target);
@@ -3644,29 +3692,28 @@ private:
// PBK pushes to a stack the address where BRK will jump to. This shares stack with
// SSY but using SYNC on a PBK address will kill the shader execution. We don't
// emulate this because it's very unlikely a driver will emit such invalid shader.
- ASSERT_MSG(instr.bra.constant_buffer == 0, "Constant buffer PBK is not supported");
+ UNIMPLEMENTED_IF_MSG(instr.bra.constant_buffer != 0,
+ "Constant buffer PBK is not supported");
const u32 target = offset + instr.bra.GetBranchTarget();
EmitPushToFlowStack(target);
break;
}
case OpCode::Id::SYNC: {
+ const Tegra::Shader::ConditionCode cc = instr.flow_condition_code;
+ UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T,
+ "SYNC condition code used: {}", static_cast<u32>(cc));
+
// The SYNC opcode jumps to the address previously set by the SSY opcode
- const Tegra::Shader::ControlCode cc = instr.flow_control_code;
- if (cc != Tegra::Shader::ControlCode::T) {
- LOG_CRITICAL(HW_GPU, "SYNC Control Code used: {}", static_cast<u32>(cc));
- UNREACHABLE();
- }
EmitPopFromFlowStack();
break;
}
case OpCode::Id::BRK: {
// The BRK opcode jumps to the address previously set by the PBK opcode
- const Tegra::Shader::ControlCode cc = instr.flow_control_code;
- if (cc != Tegra::Shader::ControlCode::T) {
- LOG_CRITICAL(HW_GPU, "BRK Control Code used: {}", static_cast<u32>(cc));
- UNREACHABLE();
- }
+ const Tegra::Shader::ConditionCode cc = instr.flow_condition_code;
+ UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T,
+ "BRK condition code used: {}", static_cast<u32>(cc));
+
EmitPopFromFlowStack();
break;
}
@@ -3677,6 +3724,9 @@ private:
break;
}
case OpCode::Id::VMAD: {
+ UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+ "Condition codes generation in VMAD is not implemented");
+
const bool result_signed = instr.video.signed_a == 1 || instr.video.signed_b == 1;
const std::string op_a = GetVideoOperandA(instr);
const std::string op_b = GetVideoOperandB(instr);
@@ -3696,11 +3746,6 @@ private:
regs.SetRegisterToInteger(instr.gpr0, result_signed, 1, result, 1, 1,
instr.vmad.saturate == 1, 0, Register::Size::Word,
instr.vmad.cc);
- if (instr.generates_cc) {
- LOG_CRITICAL(HW_GPU, "VMAD Generates an unhandled Control Code");
- UNREACHABLE();
- }
-
break;
}
case OpCode::Id::VSETP: {
@@ -3727,10 +3772,7 @@ private:
}
break;
}
- default: {
- LOG_CRITICAL(HW_GPU, "Unhandled instruction: {}", opcode->get().GetName());
- UNREACHABLE();
- }
+ default: { UNIMPLEMENTED_MSG("Unhandled instruction: {}", opcode->get().GetName()); }
}
break;
@@ -3855,6 +3897,7 @@ private:
Maxwell3D::Regs::ShaderStage stage;
const std::string& suffix;
u64 local_memory_size;
+ std::size_t shader_length;
ShaderWriter shader;
ShaderWriter declarations;
@@ -3873,9 +3916,10 @@ std::optional<ProgramResult> DecompileProgram(const ProgramCode& program_code, u
Maxwell3D::Regs::ShaderStage stage,
const std::string& suffix) {
try {
- const auto subroutines =
- ControlFlowAnalyzer(program_code, main_offset, suffix).GetSubroutines();
- GLSLGenerator generator(subroutines, program_code, main_offset, stage, suffix);
+ ControlFlowAnalyzer analyzer(program_code, main_offset, suffix);
+ const auto subroutines = analyzer.GetSubroutines();
+ GLSLGenerator generator(subroutines, program_code, main_offset, stage, suffix,
+ analyzer.GetShaderLength());
return ProgramResult{generator.GetShaderCode(), generator.GetEntries()};
} catch (const DecompileFail& exception) {
LOG_ERROR(HW_GPU, "Shader decompilation failed: {}", exception.what());
diff --git a/src/video_core/renderer_opengl/gl_shader_gen.h b/src/video_core/renderer_opengl/gl_shader_gen.h
index 520b9d4e3..b425d98ae 100644
--- a/src/video_core/renderer_opengl/gl_shader_gen.h
+++ b/src/video_core/renderer_opengl/gl_shader_gen.h
@@ -163,6 +163,7 @@ private:
struct ShaderEntries {
std::vector<ConstBufferEntry> const_buffer_entries;
std::vector<SamplerEntry> texture_samplers;
+ std::size_t shader_length;
};
using ProgramResult = std::pair<std::string, ShaderEntries>;
diff --git a/src/video_core/renderer_opengl/gl_state.cpp b/src/video_core/renderer_opengl/gl_state.cpp
index d9910c6e8..934f4db78 100644
--- a/src/video_core/renderer_opengl/gl_state.cpp
+++ b/src/video_core/renderer_opengl/gl_state.cpp
@@ -233,6 +233,28 @@ void OpenGLState::ApplyStencilTest() const {
config_stencil(GL_BACK, stencil.back, cur_state.stencil.back);
}
}
+// Viewport does not affects glClearBuffer so emulate viewport using scissor test
+void OpenGLState::EmulateViewportWithScissor() {
+ auto& current = viewports[0];
+ if (current.scissor.enabled) {
+ const GLint left = std::max(current.x, current.scissor.x);
+ const GLint right =
+ std::max(current.x + current.width, current.scissor.x + current.scissor.width);
+ const GLint bottom = std::max(current.y, current.scissor.y);
+ const GLint top =
+ std::max(current.y + current.height, current.scissor.y + current.scissor.height);
+ current.scissor.x = std::max(left, 0);
+ current.scissor.y = std::max(bottom, 0);
+ current.scissor.width = std::max(right - left, 0);
+ current.scissor.height = std::max(top - bottom, 0);
+ } else {
+ current.scissor.enabled = true;
+ current.scissor.x = current.x;
+ current.scissor.y = current.y;
+ current.scissor.width = current.width;
+ current.scissor.height = current.height;
+ }
+}
void OpenGLState::ApplyViewport() const {
if (GLAD_GL_ARB_viewport_array && geometry_shaders.enabled) {
@@ -242,7 +264,9 @@ void OpenGLState::ApplyViewport() const {
const auto& updated = viewports[i];
if (updated.x != current.x || updated.y != current.y ||
updated.width != current.width || updated.height != current.height) {
- glViewportIndexedf(i, updated.x, updated.y, updated.width, updated.height);
+ glViewportIndexedf(
+ i, static_cast<GLfloat>(updated.x), static_cast<GLfloat>(updated.y),
+ static_cast<GLfloat>(updated.width), static_cast<GLfloat>(updated.height));
}
if (updated.depth_range_near != current.depth_range_near ||
updated.depth_range_far != current.depth_range_far) {
@@ -270,8 +294,7 @@ void OpenGLState::ApplyViewport() const {
const auto& updated = viewports[0];
if (updated.x != current.x || updated.y != current.y || updated.width != current.width ||
updated.height != current.height) {
- glViewport(static_cast<GLint>(updated.x), static_cast<GLint>(updated.y),
- static_cast<GLsizei>(updated.width), static_cast<GLsizei>(updated.height));
+ glViewport(updated.x, updated.y, updated.width, updated.height);
}
if (updated.depth_range_near != current.depth_range_near ||
updated.depth_range_far != current.depth_range_far) {
@@ -339,14 +362,14 @@ void OpenGLState::ApplyTargetBlending(std::size_t target, bool force) const {
if (blend_changed || updated.src_rgb_func != current.src_rgb_func ||
updated.dst_rgb_func != current.dst_rgb_func || updated.src_a_func != current.src_a_func ||
updated.dst_a_func != current.dst_a_func) {
- glBlendFuncSeparateiARB(static_cast<GLuint>(target), updated.src_rgb_func,
- updated.dst_rgb_func, updated.src_a_func, updated.dst_a_func);
+ glBlendFuncSeparatei(static_cast<GLuint>(target), updated.src_rgb_func,
+ updated.dst_rgb_func, updated.src_a_func, updated.dst_a_func);
}
if (blend_changed || updated.rgb_equation != current.rgb_equation ||
updated.a_equation != current.a_equation) {
- glBlendEquationSeparateiARB(static_cast<GLuint>(target), updated.rgb_equation,
- updated.a_equation);
+ glBlendEquationSeparatei(static_cast<GLuint>(target), updated.rgb_equation,
+ updated.a_equation);
}
}
diff --git a/src/video_core/renderer_opengl/gl_state.h b/src/video_core/renderer_opengl/gl_state.h
index bdc743b0f..032fc43f0 100644
--- a/src/video_core/renderer_opengl/gl_state.h
+++ b/src/video_core/renderer_opengl/gl_state.h
@@ -156,10 +156,10 @@ public:
} draw;
struct viewport {
- GLfloat x;
- GLfloat y;
- GLfloat width;
- GLfloat height;
+ GLint x;
+ GLint y;
+ GLint width;
+ GLint height;
GLfloat depth_range_near; // GL_DEPTH_RANGE
GLfloat depth_range_far; // GL_DEPTH_RANGE
struct {
@@ -206,6 +206,7 @@ public:
OpenGLState& ResetBuffer(GLuint handle);
OpenGLState& ResetVertexArray(GLuint handle);
OpenGLState& ResetFramebuffer(GLuint handle);
+ void EmulateViewportWithScissor();
private:
static OpenGLState cur_state;
diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h
index 065b3929c..a8833c06e 100644
--- a/src/video_core/renderer_opengl/maxwell_to_gl.h
+++ b/src/video_core/renderer_opengl/maxwell_to_gl.h
@@ -218,14 +218,19 @@ inline GLenum DepthCompareFunc(Tegra::Texture::DepthCompareFunc func) {
inline GLenum BlendEquation(Maxwell::Blend::Equation equation) {
switch (equation) {
case Maxwell::Blend::Equation::Add:
+ case Maxwell::Blend::Equation::AddGL:
return GL_FUNC_ADD;
case Maxwell::Blend::Equation::Subtract:
+ case Maxwell::Blend::Equation::SubtractGL:
return GL_FUNC_SUBTRACT;
case Maxwell::Blend::Equation::ReverseSubtract:
+ case Maxwell::Blend::Equation::ReverseSubtractGL:
return GL_FUNC_REVERSE_SUBTRACT;
case Maxwell::Blend::Equation::Min:
+ case Maxwell::Blend::Equation::MinGL:
return GL_MIN;
case Maxwell::Blend::Equation::Max:
+ case Maxwell::Blend::Equation::MaxGL:
return GL_MAX;
}
LOG_ERROR(Render_OpenGL, "Unimplemented blend equation={}", static_cast<u32>(equation));
diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp
index ea38da932..4fd0d66c5 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.cpp
+++ b/src/video_core/renderer_opengl/renderer_opengl.cpp
@@ -19,9 +19,9 @@
#include "core/settings.h"
#include "core/telemetry_session.h"
#include "core/tracer/recorder.h"
+#include "video_core/morton.h"
#include "video_core/renderer_opengl/gl_rasterizer.h"
#include "video_core/renderer_opengl/renderer_opengl.h"
-#include "video_core/utils.h"
namespace OpenGL {
@@ -304,6 +304,12 @@ void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture,
gl_framebuffer_data.resize(texture.width * texture.height * 4);
break;
default:
+ internal_format = GL_RGBA;
+ texture.gl_format = GL_RGBA;
+ texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+ gl_framebuffer_data.resize(texture.width * texture.height * 4);
+ LOG_CRITICAL(Render_OpenGL, "Unknown framebuffer pixel format: {}",
+ static_cast<u32>(framebuffer.pixel_format));
UNREACHABLE();
}
@@ -484,7 +490,7 @@ bool RendererOpenGL::Init() {
Core::Telemetry().AddField(Telemetry::FieldType::UserSystem, "GPU_Model", gpu_model);
Core::Telemetry().AddField(Telemetry::FieldType::UserSystem, "GPU_OpenGL_Version", gl_version);
- if (!GLAD_GL_VERSION_3_3) {
+ if (!GLAD_GL_VERSION_4_3) {
return false;
}
diff --git a/src/video_core/utils.h b/src/video_core/utils.h
deleted file mode 100644
index e0a14d48f..000000000
--- a/src/video_core/utils.h
+++ /dev/null
@@ -1,164 +0,0 @@
-// Copyright 2014 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include "common/common_types.h"
-
-namespace VideoCore {
-
-// 8x8 Z-Order coordinate from 2D coordinates
-static inline u32 MortonInterleave(u32 x, u32 y) {
- static const u32 xlut[] = {0x00, 0x01, 0x04, 0x05, 0x10, 0x11, 0x14, 0x15};
- static const u32 ylut[] = {0x00, 0x02, 0x08, 0x0a, 0x20, 0x22, 0x28, 0x2a};
- return xlut[x % 8] + ylut[y % 8];
-}
-
-/**
- * Calculates the offset of the position of the pixel in Morton order
- */
-static inline u32 GetMortonOffset(u32 x, u32 y, u32 bytes_per_pixel) {
- // Images are split into 8x8 tiles. Each tile is composed of four 4x4 subtiles each
- // of which is composed of four 2x2 subtiles each of which is composed of four texels.
- // Each structure is embedded into the next-bigger one in a diagonal pattern, e.g.
- // texels are laid out in a 2x2 subtile like this:
- // 2 3
- // 0 1
- //
- // The full 8x8 tile has the texels arranged like this:
- //
- // 42 43 46 47 58 59 62 63
- // 40 41 44 45 56 57 60 61
- // 34 35 38 39 50 51 54 55
- // 32 33 36 37 48 49 52 53
- // 10 11 14 15 26 27 30 31
- // 08 09 12 13 24 25 28 29
- // 02 03 06 07 18 19 22 23
- // 00 01 04 05 16 17 20 21
- //
- // This pattern is what's called Z-order curve, or Morton order.
-
- const unsigned int block_height = 8;
- const unsigned int coarse_x = x & ~7;
-
- u32 i = VideoCore::MortonInterleave(x, y);
-
- const unsigned int offset = coarse_x * block_height;
-
- return (i + offset) * bytes_per_pixel;
-}
-
-static inline u32 MortonInterleave128(u32 x, u32 y) {
- // 128x128 Z-Order coordinate from 2D coordinates
- static constexpr u32 xlut[] = {
- 0x0000, 0x0001, 0x0002, 0x0003, 0x0008, 0x0009, 0x000a, 0x000b, 0x0040, 0x0041, 0x0042,
- 0x0043, 0x0048, 0x0049, 0x004a, 0x004b, 0x0800, 0x0801, 0x0802, 0x0803, 0x0808, 0x0809,
- 0x080a, 0x080b, 0x0840, 0x0841, 0x0842, 0x0843, 0x0848, 0x0849, 0x084a, 0x084b, 0x1000,
- 0x1001, 0x1002, 0x1003, 0x1008, 0x1009, 0x100a, 0x100b, 0x1040, 0x1041, 0x1042, 0x1043,
- 0x1048, 0x1049, 0x104a, 0x104b, 0x1800, 0x1801, 0x1802, 0x1803, 0x1808, 0x1809, 0x180a,
- 0x180b, 0x1840, 0x1841, 0x1842, 0x1843, 0x1848, 0x1849, 0x184a, 0x184b, 0x2000, 0x2001,
- 0x2002, 0x2003, 0x2008, 0x2009, 0x200a, 0x200b, 0x2040, 0x2041, 0x2042, 0x2043, 0x2048,
- 0x2049, 0x204a, 0x204b, 0x2800, 0x2801, 0x2802, 0x2803, 0x2808, 0x2809, 0x280a, 0x280b,
- 0x2840, 0x2841, 0x2842, 0x2843, 0x2848, 0x2849, 0x284a, 0x284b, 0x3000, 0x3001, 0x3002,
- 0x3003, 0x3008, 0x3009, 0x300a, 0x300b, 0x3040, 0x3041, 0x3042, 0x3043, 0x3048, 0x3049,
- 0x304a, 0x304b, 0x3800, 0x3801, 0x3802, 0x3803, 0x3808, 0x3809, 0x380a, 0x380b, 0x3840,
- 0x3841, 0x3842, 0x3843, 0x3848, 0x3849, 0x384a, 0x384b, 0x0000, 0x0001, 0x0002, 0x0003,
- 0x0008, 0x0009, 0x000a, 0x000b, 0x0040, 0x0041, 0x0042, 0x0043, 0x0048, 0x0049, 0x004a,
- 0x004b, 0x0800, 0x0801, 0x0802, 0x0803, 0x0808, 0x0809, 0x080a, 0x080b, 0x0840, 0x0841,
- 0x0842, 0x0843, 0x0848, 0x0849, 0x084a, 0x084b, 0x1000, 0x1001, 0x1002, 0x1003, 0x1008,
- 0x1009, 0x100a, 0x100b, 0x1040, 0x1041, 0x1042, 0x1043, 0x1048, 0x1049, 0x104a, 0x104b,
- 0x1800, 0x1801, 0x1802, 0x1803, 0x1808, 0x1809, 0x180a, 0x180b, 0x1840, 0x1841, 0x1842,
- 0x1843, 0x1848, 0x1849, 0x184a, 0x184b, 0x2000, 0x2001, 0x2002, 0x2003, 0x2008, 0x2009,
- 0x200a, 0x200b, 0x2040, 0x2041, 0x2042, 0x2043, 0x2048, 0x2049, 0x204a, 0x204b, 0x2800,
- 0x2801, 0x2802, 0x2803, 0x2808, 0x2809, 0x280a, 0x280b, 0x2840, 0x2841, 0x2842, 0x2843,
- 0x2848, 0x2849, 0x284a, 0x284b, 0x3000, 0x3001, 0x3002, 0x3003, 0x3008, 0x3009, 0x300a,
- 0x300b, 0x3040, 0x3041, 0x3042, 0x3043, 0x3048, 0x3049, 0x304a, 0x304b, 0x3800, 0x3801,
- 0x3802, 0x3803, 0x3808, 0x3809, 0x380a, 0x380b, 0x3840, 0x3841, 0x3842, 0x3843, 0x3848,
- 0x3849, 0x384a, 0x384b, 0x0000, 0x0001, 0x0002, 0x0003, 0x0008, 0x0009, 0x000a, 0x000b,
- 0x0040, 0x0041, 0x0042, 0x0043, 0x0048, 0x0049, 0x004a, 0x004b, 0x0800, 0x0801, 0x0802,
- 0x0803, 0x0808, 0x0809, 0x080a, 0x080b, 0x0840, 0x0841, 0x0842, 0x0843, 0x0848, 0x0849,
- 0x084a, 0x084b, 0x1000, 0x1001, 0x1002, 0x1003, 0x1008, 0x1009, 0x100a, 0x100b, 0x1040,
- 0x1041, 0x1042, 0x1043, 0x1048, 0x1049, 0x104a, 0x104b, 0x1800, 0x1801, 0x1802, 0x1803,
- 0x1808, 0x1809, 0x180a, 0x180b, 0x1840, 0x1841, 0x1842, 0x1843, 0x1848, 0x1849, 0x184a,
- 0x184b, 0x2000, 0x2001, 0x2002, 0x2003, 0x2008, 0x2009, 0x200a, 0x200b, 0x2040, 0x2041,
- 0x2042, 0x2043, 0x2048, 0x2049, 0x204a, 0x204b, 0x2800, 0x2801, 0x2802, 0x2803, 0x2808,
- 0x2809, 0x280a, 0x280b, 0x2840, 0x2841, 0x2842, 0x2843, 0x2848, 0x2849, 0x284a, 0x284b,
- 0x3000, 0x3001, 0x3002, 0x3003, 0x3008, 0x3009, 0x300a, 0x300b, 0x3040, 0x3041, 0x3042,
- 0x3043, 0x3048, 0x3049, 0x304a, 0x304b, 0x3800, 0x3801, 0x3802, 0x3803, 0x3808, 0x3809,
- 0x380a, 0x380b, 0x3840, 0x3841, 0x3842, 0x3843, 0x3848, 0x3849, 0x384a, 0x384b,
- };
- static constexpr u32 ylut[] = {
- 0x0000, 0x0004, 0x0010, 0x0014, 0x0020, 0x0024, 0x0030, 0x0034, 0x0080, 0x0084, 0x0090,
- 0x0094, 0x00a0, 0x00a4, 0x00b0, 0x00b4, 0x0100, 0x0104, 0x0110, 0x0114, 0x0120, 0x0124,
- 0x0130, 0x0134, 0x0180, 0x0184, 0x0190, 0x0194, 0x01a0, 0x01a4, 0x01b0, 0x01b4, 0x0200,
- 0x0204, 0x0210, 0x0214, 0x0220, 0x0224, 0x0230, 0x0234, 0x0280, 0x0284, 0x0290, 0x0294,
- 0x02a0, 0x02a4, 0x02b0, 0x02b4, 0x0300, 0x0304, 0x0310, 0x0314, 0x0320, 0x0324, 0x0330,
- 0x0334, 0x0380, 0x0384, 0x0390, 0x0394, 0x03a0, 0x03a4, 0x03b0, 0x03b4, 0x0400, 0x0404,
- 0x0410, 0x0414, 0x0420, 0x0424, 0x0430, 0x0434, 0x0480, 0x0484, 0x0490, 0x0494, 0x04a0,
- 0x04a4, 0x04b0, 0x04b4, 0x0500, 0x0504, 0x0510, 0x0514, 0x0520, 0x0524, 0x0530, 0x0534,
- 0x0580, 0x0584, 0x0590, 0x0594, 0x05a0, 0x05a4, 0x05b0, 0x05b4, 0x0600, 0x0604, 0x0610,
- 0x0614, 0x0620, 0x0624, 0x0630, 0x0634, 0x0680, 0x0684, 0x0690, 0x0694, 0x06a0, 0x06a4,
- 0x06b0, 0x06b4, 0x0700, 0x0704, 0x0710, 0x0714, 0x0720, 0x0724, 0x0730, 0x0734, 0x0780,
- 0x0784, 0x0790, 0x0794, 0x07a0, 0x07a4, 0x07b0, 0x07b4, 0x0000, 0x0004, 0x0010, 0x0014,
- 0x0020, 0x0024, 0x0030, 0x0034, 0x0080, 0x0084, 0x0090, 0x0094, 0x00a0, 0x00a4, 0x00b0,
- 0x00b4, 0x0100, 0x0104, 0x0110, 0x0114, 0x0120, 0x0124, 0x0130, 0x0134, 0x0180, 0x0184,
- 0x0190, 0x0194, 0x01a0, 0x01a4, 0x01b0, 0x01b4, 0x0200, 0x0204, 0x0210, 0x0214, 0x0220,
- 0x0224, 0x0230, 0x0234, 0x0280, 0x0284, 0x0290, 0x0294, 0x02a0, 0x02a4, 0x02b0, 0x02b4,
- 0x0300, 0x0304, 0x0310, 0x0314, 0x0320, 0x0324, 0x0330, 0x0334, 0x0380, 0x0384, 0x0390,
- 0x0394, 0x03a0, 0x03a4, 0x03b0, 0x03b4, 0x0400, 0x0404, 0x0410, 0x0414, 0x0420, 0x0424,
- 0x0430, 0x0434, 0x0480, 0x0484, 0x0490, 0x0494, 0x04a0, 0x04a4, 0x04b0, 0x04b4, 0x0500,
- 0x0504, 0x0510, 0x0514, 0x0520, 0x0524, 0x0530, 0x0534, 0x0580, 0x0584, 0x0590, 0x0594,
- 0x05a0, 0x05a4, 0x05b0, 0x05b4, 0x0600, 0x0604, 0x0610, 0x0614, 0x0620, 0x0624, 0x0630,
- 0x0634, 0x0680, 0x0684, 0x0690, 0x0694, 0x06a0, 0x06a4, 0x06b0, 0x06b4, 0x0700, 0x0704,
- 0x0710, 0x0714, 0x0720, 0x0724, 0x0730, 0x0734, 0x0780, 0x0784, 0x0790, 0x0794, 0x07a0,
- 0x07a4, 0x07b0, 0x07b4, 0x0000, 0x0004, 0x0010, 0x0014, 0x0020, 0x0024, 0x0030, 0x0034,
- 0x0080, 0x0084, 0x0090, 0x0094, 0x00a0, 0x00a4, 0x00b0, 0x00b4, 0x0100, 0x0104, 0x0110,
- 0x0114, 0x0120, 0x0124, 0x0130, 0x0134, 0x0180, 0x0184, 0x0190, 0x0194, 0x01a0, 0x01a4,
- 0x01b0, 0x01b4, 0x0200, 0x0204, 0x0210, 0x0214, 0x0220, 0x0224, 0x0230, 0x0234, 0x0280,
- 0x0284, 0x0290, 0x0294, 0x02a0, 0x02a4, 0x02b0, 0x02b4, 0x0300, 0x0304, 0x0310, 0x0314,
- 0x0320, 0x0324, 0x0330, 0x0334, 0x0380, 0x0384, 0x0390, 0x0394, 0x03a0, 0x03a4, 0x03b0,
- 0x03b4, 0x0400, 0x0404, 0x0410, 0x0414, 0x0420, 0x0424, 0x0430, 0x0434, 0x0480, 0x0484,
- 0x0490, 0x0494, 0x04a0, 0x04a4, 0x04b0, 0x04b4, 0x0500, 0x0504, 0x0510, 0x0514, 0x0520,
- 0x0524, 0x0530, 0x0534, 0x0580, 0x0584, 0x0590, 0x0594, 0x05a0, 0x05a4, 0x05b0, 0x05b4,
- 0x0600, 0x0604, 0x0610, 0x0614, 0x0620, 0x0624, 0x0630, 0x0634, 0x0680, 0x0684, 0x0690,
- 0x0694, 0x06a0, 0x06a4, 0x06b0, 0x06b4, 0x0700, 0x0704, 0x0710, 0x0714, 0x0720, 0x0724,
- 0x0730, 0x0734, 0x0780, 0x0784, 0x0790, 0x0794, 0x07a0, 0x07a4, 0x07b0, 0x07b4,
- };
- return xlut[x % 128] + ylut[y % 128];
-}
-
-static inline u32 GetMortonOffset128(u32 x, u32 y, u32 bytes_per_pixel) {
- // Calculates the offset of the position of the pixel in Morton order
- // Framebuffer images are split into 128x128 tiles.
-
- const unsigned int block_height = 128;
- const unsigned int coarse_x = x & ~127;
-
- u32 i = MortonInterleave128(x, y);
-
- const unsigned int offset = coarse_x * block_height;
-
- return (i + offset) * bytes_per_pixel;
-}
-
-static inline void MortonCopyPixels128(u32 width, u32 height, u32 bytes_per_pixel,
- u32 gl_bytes_per_pixel, u8* morton_data, u8* gl_data,
- bool morton_to_gl) {
- u8* data_ptrs[2];
- for (unsigned y = 0; y < height; ++y) {
- for (unsigned x = 0; x < width; ++x) {
- const u32 coarse_y = y & ~127;
- u32 morton_offset =
- GetMortonOffset128(x, y, bytes_per_pixel) + coarse_y * width * bytes_per_pixel;
- u32 gl_pixel_index = (x + y * width) * gl_bytes_per_pixel;
-
- data_ptrs[morton_to_gl] = morton_data + morton_offset;
- data_ptrs[!morton_to_gl] = &gl_data[gl_pixel_index];
-
- memcpy(data_ptrs[0], data_ptrs[1], bytes_per_pixel);
- }
- }
-}
-
-} // namespace VideoCore