summaryrefslogtreecommitdiffstats
path: root/src/video_core/engines
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core/engines')
-rw-r--r--src/video_core/engines/maxwell_3d.cpp36
-rw-r--r--src/video_core/engines/maxwell_3d.h25
-rw-r--r--src/video_core/engines/shader_bytecode.h12
3 files changed, 52 insertions, 21 deletions
diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp
index 27ef865a2..d79c50919 100644
--- a/src/video_core/engines/maxwell_3d.cpp
+++ b/src/video_core/engines/maxwell_3d.cpp
@@ -43,15 +43,17 @@ void Maxwell3D::CallMacroMethod(u32 method, std::vector<u32> parameters) {
// Reset the current macro.
executing_macro = 0;
- // The requested macro must have been uploaded already.
- auto macro_code = uploaded_macros.find(method);
- if (macro_code == uploaded_macros.end()) {
- LOG_ERROR(HW_GPU, "Macro {:04X} was not uploaded", method);
+ // Lookup the macro offset
+ const u32 entry{(method - MacroRegistersStart) >> 1};
+ const auto& search{macro_offsets.find(entry)};
+ if (search == macro_offsets.end()) {
+ LOG_CRITICAL(HW_GPU, "macro not found for method 0x{:X}!", method);
+ UNREACHABLE();
return;
}
// Execute the current macro.
- macro_interpreter.Execute(macro_code->second, std::move(parameters));
+ macro_interpreter.Execute(search->second, std::move(parameters));
}
void Maxwell3D::WriteReg(u32 method, u32 value, u32 remaining_params) {
@@ -97,6 +99,10 @@ void Maxwell3D::WriteReg(u32 method, u32 value, u32 remaining_params) {
ProcessMacroUpload(value);
break;
}
+ case MAXWELL3D_REG_INDEX(macros.bind): {
+ ProcessMacroBind(value);
+ break;
+ }
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[0]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[1]):
case MAXWELL3D_REG_INDEX(const_buffer.cb_data[2]):
@@ -158,16 +164,20 @@ void Maxwell3D::WriteReg(u32 method, u32 value, u32 remaining_params) {
}
void Maxwell3D::ProcessMacroUpload(u32 data) {
- // Store the uploaded macro code to interpret them when they're called.
- auto& macro = uploaded_macros[regs.macros.entry * 2 + MacroRegistersStart];
- macro.push_back(data);
+ ASSERT_MSG(regs.macros.upload_address < macro_memory.size(),
+ "upload_address exceeded macro_memory size!");
+ macro_memory[regs.macros.upload_address++] = data;
+}
+
+void Maxwell3D::ProcessMacroBind(u32 data) {
+ macro_offsets[regs.macros.entry] = data;
}
void Maxwell3D::ProcessQueryGet() {
GPUVAddr sequence_address = regs.query.QueryAddress();
// Since the sequence address is given as a GPU VAddr, we have to convert it to an application
// VAddr before writing.
- boost::optional<VAddr> address = memory_manager.GpuToCpuAddress(sequence_address);
+ std::optional<VAddr> address = memory_manager.GpuToCpuAddress(sequence_address);
// TODO(Subv): Support the other query units.
ASSERT_MSG(regs.query.query_get.unit == Regs::QueryUnit::Crop,
@@ -285,7 +295,7 @@ void Maxwell3D::ProcessCBData(u32 value) {
// Don't allow writing past the end of the buffer.
ASSERT(regs.const_buffer.cb_pos + sizeof(u32) <= regs.const_buffer.cb_size);
- boost::optional<VAddr> address =
+ std::optional<VAddr> address =
memory_manager.GpuToCpuAddress(buffer_address + regs.const_buffer.cb_pos);
Memory::Write32(*address, value);
@@ -298,7 +308,7 @@ Texture::TICEntry Maxwell3D::GetTICEntry(u32 tic_index) const {
GPUVAddr tic_base_address = regs.tic.TICAddress();
GPUVAddr tic_address_gpu = tic_base_address + tic_index * sizeof(Texture::TICEntry);
- boost::optional<VAddr> tic_address_cpu = memory_manager.GpuToCpuAddress(tic_address_gpu);
+ std::optional<VAddr> tic_address_cpu = memory_manager.GpuToCpuAddress(tic_address_gpu);
Texture::TICEntry tic_entry;
Memory::ReadBlock(*tic_address_cpu, &tic_entry, sizeof(Texture::TICEntry));
@@ -322,7 +332,7 @@ Texture::TSCEntry Maxwell3D::GetTSCEntry(u32 tsc_index) const {
GPUVAddr tsc_base_address = regs.tsc.TSCAddress();
GPUVAddr tsc_address_gpu = tsc_base_address + tsc_index * sizeof(Texture::TSCEntry);
- boost::optional<VAddr> tsc_address_cpu = memory_manager.GpuToCpuAddress(tsc_address_gpu);
+ std::optional<VAddr> tsc_address_cpu = memory_manager.GpuToCpuAddress(tsc_address_gpu);
Texture::TSCEntry tsc_entry;
Memory::ReadBlock(*tsc_address_cpu, &tsc_entry, sizeof(Texture::TSCEntry));
@@ -386,7 +396,7 @@ Texture::FullTextureInfo Maxwell3D::GetStageTexture(Regs::ShaderStage stage,
ASSERT(tex_info_address < tex_info_buffer.address + tex_info_buffer.size);
- boost::optional<VAddr> tex_address_cpu = memory_manager.GpuToCpuAddress(tex_info_address);
+ std::optional<VAddr> tex_address_cpu = memory_manager.GpuToCpuAddress(tex_info_address);
Texture::TextureHandle tex_handle{Memory::Read32(*tex_address_cpu)};
Texture::FullTextureInfo tex_info{};
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index 443affc36..50873813e 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -475,12 +475,13 @@ public:
INSERT_PADDING_WORDS(0x45);
struct {
- INSERT_PADDING_WORDS(1);
+ u32 upload_address;
u32 data;
u32 entry;
+ u32 bind;
} macros;
- INSERT_PADDING_WORDS(0x189);
+ INSERT_PADDING_WORDS(0x188);
u32 tfb_enabled;
@@ -994,12 +995,25 @@ public:
/// Returns the texture information for a specific texture in a specific shader stage.
Texture::FullTextureInfo GetStageTexture(Regs::ShaderStage stage, std::size_t offset) const;
+ /// Memory for macro code - it's undetermined how big this is, however 1MB is much larger than
+ /// we've seen used.
+ using MacroMemory = std::array<u32, 0x40000>;
+
+ /// Gets a reference to macro memory.
+ const MacroMemory& GetMacroMemory() const {
+ return macro_memory;
+ }
+
private:
void InitializeRegisterDefaults();
VideoCore::RasterizerInterface& rasterizer;
- std::unordered_map<u32, std::vector<u32>> uploaded_macros;
+ /// Start offsets of each macro in macro_memory
+ std::unordered_map<u32, u32> macro_offsets;
+
+ /// Memory for macro code
+ MacroMemory macro_memory;
/// Macro method that is currently being executed / being fed parameters.
u32 executing_macro = 0;
@@ -1022,9 +1036,12 @@ private:
*/
void CallMacroMethod(u32 method, std::vector<u32> parameters);
- /// Handles writes to the macro uploading registers.
+ /// Handles writes to the macro uploading register.
void ProcessMacroUpload(u32 data);
+ /// Handles writes to the macro bind register.
+ void ProcessMacroBind(u32 data);
+
/// Handles a write to the CLEAR_BUFFERS register.
void ProcessClearBuffers();
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 55763332e..83a6fd875 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -5,12 +5,11 @@
#pragma once
#include <bitset>
+#include <optional>
#include <string>
#include <tuple>
#include <vector>
-#include <boost/optional.hpp>
-
#include "common/assert.h"
#include "common/bit_field.h"
#include "common/common_types.h"
@@ -579,6 +578,10 @@ union Instruction {
} fmul32;
union {
+ BitField<52, 1, u64> generates_cc;
+ } op_32;
+
+ union {
BitField<48, 1, u64> is_signed;
} shift;
@@ -1457,7 +1460,7 @@ public:
Type type;
};
- static boost::optional<const Matcher&> Decode(Instruction instr) {
+ static std::optional<std::reference_wrapper<const Matcher>> Decode(Instruction instr) {
static const auto table{GetDecodeTable()};
const auto matches_instruction = [instr](const auto& matcher) {
@@ -1465,7 +1468,8 @@ public:
};
auto iter = std::find_if(table.begin(), table.end(), matches_instruction);
- return iter != table.end() ? boost::optional<const Matcher&>(*iter) : boost::none;
+ return iter != table.end() ? std::optional<std::reference_wrapper<const Matcher>>(*iter)
+ : std::nullopt;
}
private:
generated by cgit v1.2.3 (git 2.25.1) at 2024-10-11 06:46:26 +0200