rsx: Restore shader disassembler to working state

This commit is contained in:
kd-11 2021-11-04 23:04:01 +03:00 committed by kd-11
parent 067b35217a
commit f7eacf70ec
3 changed files with 117 additions and 67 deletions

View File

@ -48,6 +48,13 @@ namespace gl
int find_count = 14;
int ext_count = 0;
glGetIntegerv(GL_NUM_EXTENSIONS, &ext_count);
if (!ext_count)
{
rsx_log.error("Coult not initialize GL driver capabilities. Is OpenGL initialized?");
return;
}
std::string vendor_string = reinterpret_cast<const char*>(glGetString(GL_VENDOR));
std::string version_string = reinterpret_cast<const char*>(glGetString(GL_VERSION));
std::string renderer_string = reinterpret_cast<const char*>(glGetString(GL_RENDERER));

View File

@ -232,11 +232,6 @@ void CgBinaryDisasm::TaskFP()
m_size = 0;
u32* data = reinterpret_cast<u32*>(&m_buffer[m_offset]);
ensure((m_buffer_size - m_offset) % sizeof(u32) == 0);
for (u32 i = 0; i < (m_buffer_size - m_offset) / sizeof(u32); i++)
{
// Get BE data
data[i] = std::bit_cast<u32, be_t<u32>>(data[i]);
}
enum
{

View File

@ -3,15 +3,21 @@
#include "Emu/Memory/vm.h"
#include "Emu/RSX/GL/GLVertexProgram.h"
#include "Emu/RSX/GL/GLFragmentProgram.h"
#include "Emu/RSX/Program/ProgramStateCache.h"
#include "Utilities/File.h"
using CGprofile = be_t<u32>;
using CGbool = be_t<s32>;
using CGresource = be_t<u32>;
using CGenum = be_t<u32>;
using CGtype = be_t<u32>;
using CGprofile = u32;
using CGbool = s32;
using CGresource = u32;
using CGenum = u32;
using CGtype = u32;
using CGbitfield = u32;
using CGbitfield16 = u16;
using CGint = s32;
using CGuint = u32;
using CgBinaryOffset = be_t<u32>;
using CgBinaryOffset = CGuint;
using CgBinarySize = CgBinaryOffset;
using CgBinaryEmbeddedConstantOffset = CgBinaryOffset;
using CgBinaryFloatOffset = CgBinaryOffset;
using CgBinaryStringOffset = CgBinaryOffset;
@ -36,13 +42,13 @@ struct CgBinaryParameter
CGtype type; // cgGetParameterType()
CGresource res; // cgGetParameterResource()
CGenum var; // cgGetParameterVariability()
be_t<s32> resIndex; // cgGetParameterResourceIndex()
CGint resIndex; // cgGetParameterResourceIndex()
CgBinaryStringOffset name; // cgGetParameterName()
CgBinaryFloatOffset defaultValue; // default constant value
CgBinaryEmbeddedConstantOffset embeddedConst; // embedded constant information
CgBinaryStringOffset semantic; // cgGetParameterSemantic()
CGenum direction; // cgGetParameterDirection()
be_t<s32> paramno; // 0..n: cgGetParameterIndex() -1: globals
CGint paramno; // 0..n: cgGetParameterIndex() -1: globals
CGbool isReferenced; // cgIsParameterReferenced()
CGbool isShared; // cgIsParameterShared()
};
@ -50,12 +56,12 @@ struct CgBinaryParameter
// attributes needed for vshaders
struct CgBinaryVertexProgram
{
be_t<u32> instructionCount; // #instructions
be_t<u32> instructionSlot; // load address (indexed reads!)
be_t<u32> registerCount; // R registers count
be_t<u32> attributeInputMask; // attributes vs reads from
be_t<u32> attributeOutputMask; // attributes vs writes (uses SET_VERTEX_ATTRIB_OUTPUT_MASK bits)
be_t<u32> userClipMask; // user clip plane enables (for SET_USER_CLIP_PLANE_CONTROL)
CgBinarySize instructionCount; // #instructions
CgBinarySize instructionSlot; // load address (indexed reads!)
CgBinarySize registerCount; // R registers count
CGbitfield attributeInputMask; // attributes vs reads from
CGbitfield attributeOutputMask; // attributes vs writes (uses SET_VERTEX_ATTRIB_OUTPUT_MASK bits)
CGbitfield userClipMask; // user clip plane enables (for SET_USER_CLIP_PLANE_CONTROL)
};
typedef enum
@ -68,12 +74,12 @@ typedef enum
// attributes needed for pshaders
struct CgBinaryFragmentProgram
{
be_t<u32> instructionCount; // #instructions
be_t<u32> attributeInputMask; // attributes fp reads (uses SET_VERTEX_ATTRIB_OUTPUT_MASK bits)
be_t<u32> partialTexType; // texid 0..15 use two bits each marking whether the texture format requires partial load: see CgBinaryPartialTexType
be_t<u16> texCoordsInputMask; // tex coords used by frag prog. (tex<n> is bit n)
be_t<u16> texCoords2D; // tex coords that are 2d (tex<n> is bit n)
be_t<u16> texCoordsCentroid; // tex coords that are centroid (tex<n> is bit n)
CgBinarySize instructionCount; // #instructions
CGbitfield attributeInputMask; // attributes fp reads (uses SET_VERTEX_ATTRIB_OUTPUT_MASK bits)
CGbitfield partialTexType; // texid 0..15 use two bits each marking whether the texture format requires partial load: see CgBinaryPartialTexType
CGbitfield16 texCoordsInputMask; // tex coords used by frag prog. (tex<n> is bit n)
CGbitfield16 texCoords2D; // tex coords that are 2d (tex<n> is bit n)
CGbitfield16 texCoordsCentroid; // tex coords that are centroid (tex<n> is bit n)
u8 registerCount; // R registers count
u8 outputFromH0; // final color from R0 or H0
u8 depthReplace; // fp generated z depth value
@ -86,20 +92,20 @@ struct CgBinaryProgram
CGprofile profile;
// binary revision (used to verify binary and driver structs match)
be_t<u32> binaryFormatRevision;
CgBinarySize binaryFormatRevision;
// total size of this struct including profile and totalSize field
be_t<u32> totalSize;
CgBinarySize totalSize;
// parameter usually queried using cgGet[First/Next]LeafParameter
be_t<u32> parameterCount;
CgBinarySize parameterCount;
CgBinaryParameterOffset parameterArray;
// depending on profile points to a CgBinaryVertexProgram or CgBinaryFragmentProgram struct
CgBinaryOffset program;
// raw ucode data
be_t<u32> ucodeSize;
CgBinarySize ucodeSize;
CgBinaryOffset ucode;
// variable length data follows
@ -255,12 +261,63 @@ public:
return reinterpret_cast<T&>(m_buffer[offset]);
}
void ConvertToLE(CgBinaryProgram& prog)
{
// BE payload, requires that data be swapped
const auto be_profile = prog.profile;
auto swap_be32 = [&](u32 start_offset, size_t size_bytes)
{
auto start = reinterpret_cast<u32*>(m_buffer + start_offset);
auto end = reinterpret_cast<u32*>(m_buffer + start_offset + size_bytes);
for (auto data = start; data < end; ++data)
{
*data = std::bit_cast<be_t<u32>>(*data);
}
};
// 1. Swap the header
swap_be32(0, sizeof(CgBinaryProgram));
// 2. Swap parameters
swap_be32(prog.parameterArray, sizeof(CgBinaryParameter) * prog.parameterCount);
// 3. Swap the ucode
swap_be32(prog.ucode, m_buffer_size - prog.ucode);
// 4. Swap the domain header
if (be_profile == 7004u)
{
// Need to swap each field individually
auto& fprog = GetCgRef<CgBinaryFragmentProgram>(prog.program);
fprog.instructionCount = std::bit_cast<be_t<u32>>(fprog.instructionCount);
fprog.attributeInputMask = std::bit_cast<be_t<u32>>(fprog.attributeInputMask);
fprog.partialTexType = std::bit_cast<be_t<u32>>(fprog.partialTexType);
fprog.texCoordsInputMask = std::bit_cast<be_t<u16>>(fprog.texCoordsInputMask);
fprog.texCoords2D = std::bit_cast<be_t<u16>>(fprog.texCoords2D);
fprog.texCoordsCentroid = std::bit_cast<be_t<u16>>(fprog.texCoordsCentroid);
}
else
{
// Swap entire header block as all fields are u32
swap_be32(prog.program, sizeof(CgBinaryVertexProgram));
}
}
void BuildShaderBody()
{
ParamArray param_array;
auto& prog = GetCgRef<CgBinaryProgram>(0);
if (const u32 be_profile = std::bit_cast<be_t<u32>>(prog.profile);
be_profile == 7003u || be_profile == 7004u)
{
ConvertToLE(prog);
ensure(be_profile == prog.profile);
}
if (prog.profile == 7004u)
{
auto& fprog = GetCgRef<CgBinaryFragmentProgram>(prog.program);
@ -292,34 +349,31 @@ public:
m_offset = prog.ucode;
TaskFP();
// reload binary data in the virtual memory, temporary solution
{
u32 ptr;
{
fs::file f(m_path);
if (!f) return;
u32 unused;
std::vector<u32> be_data;
usz size = f.size();
vm::init();
ptr = vm::alloc(static_cast<u32>(size), vm::main);
f.read(vm::base(ptr), size);
// Swap bytes. FP decompiler expects input in BE
for (u32* ptr = reinterpret_cast<u32*>(m_buffer + m_offset),
*end = reinterpret_cast<u32*>(m_buffer + m_buffer_size);
ptr < end; ++ptr)
{
be_data.push_back(std::bit_cast<be_t<u32>>(*ptr));
}
auto& vmprog = vm::_ref<CgBinaryProgram>(ptr);
auto& vmfprog = vm::_ref<CgBinaryFragmentProgram>(ptr + vmprog.program);
u32 size;
u32 ctrl = (vmfprog.outputFromH0 ? 0 : 0x40) | (vmfprog.depthReplace ? 0xe : 0);
std::vector<rsx::texture_dimension_extended> td;
RSXFragmentProgram prog;
prog.ucode_length = 0, prog.data = vm::base(ptr + vmprog.ucode), prog.offset = 0, prog.ctrl = ctrl;
GLFragmentDecompilerThread(m_glsl_shader, param_array, prog, size).Task();
vm::close();
}
auto metadata = program_hash_util::fragment_program_utils::analyse_fragment_program(be_data.data());
prog.ctrl = (fprog.outputFromH0 ? 0 : 0x40) | (fprog.depthReplace ? 0xe : 0);
prog.offset = metadata.program_start_offset;
prog.ucode_length = metadata.program_ucode_length;
prog.total_length = metadata.program_ucode_length + metadata.program_start_offset;
prog.data = reinterpret_cast<u8*>(be_data.data()) + metadata.program_start_offset;
for (u32 i = 0; i < 16; ++i) prog.texture_state.set_dimension(rsx::texture_dimension_extended::texture_dimension_2d, i);
GLFragmentDecompilerThread(m_glsl_shader, param_array, prog, unused).Task();
}
else
{
auto& vprog = GetCgRef<CgBinaryVertexProgram>(prog.program);
const auto& vprog = GetCgRef<CgBinaryVertexProgram>(prog.program);
m_arb_shader += "\n";
m_arb_shader += fmt::format("# binaryFormatRevision 0x%x\n", prog.binaryFormatRevision);
m_arb_shader += fmt::format("# profile sce_vp_rsx\n");
@ -347,22 +401,16 @@ public:
m_arb_shader += "\n";
m_offset = prog.ucode;
ensure((m_buffer_size - m_offset) % sizeof(u32) == 0);
u32* vdata = reinterpret_cast<u32*>(&m_buffer[m_offset]);
ensure((m_buffer_size - m_offset) % sizeof(u32) == 0);
for (u32 i = 0; i < (m_buffer_size - m_offset) / sizeof(u32); i++)
{
vdata[i] = std::bit_cast<u32, be_t<u32>>(vdata[i]);
}
for (u32 i = 0; i < prog.ucodeSize / sizeof(u32); i++)
{
m_data.push_back(vdata[i]);
}
m_data.resize(prog.ucodeSize / sizeof(u32));
std::memcpy(m_data.data(), vdata, prog.ucodeSize);
TaskVP();
RSXVertexProgram prog;
prog.data = m_data;
program_hash_util::vertex_program_utils::analyse_vertex_program(vdata, 0, prog);
for (u32 i = 0; i < 4; ++i) prog.texture_state.set_dimension(rsx::texture_dimension_extended::texture_dimension_2d, i);
GLVertexDecompilerThread(prog, m_glsl_shader, param_array).Task();
}
}