2015-05-24 04:55:12 +00:00
|
|
|
// Copyright 2008 Dolphin Emulator Project
|
2015-05-17 23:08:10 +00:00
|
|
|
// Licensed under GPLv2+
|
2013-04-18 03:09:55 +00:00
|
|
|
// Refer to the license.txt file included.
|
2009-03-07 08:35:01 +00:00
|
|
|
|
2010-03-06 02:07:48 +00:00
|
|
|
#include <algorithm>
|
2014-06-05 15:55:21 +00:00
|
|
|
#include <memory>
|
2014-08-25 03:53:28 +00:00
|
|
|
#include <mutex>
|
2012-12-10 06:40:28 +00:00
|
|
|
#include <unordered_map>
|
2014-08-15 14:17:06 +00:00
|
|
|
#include <utility>
|
2009-03-07 08:35:01 +00:00
|
|
|
#include <vector>
|
|
|
|
|
2014-09-19 04:17:41 +00:00
|
|
|
#include "Common/CommonFuncs.h"
|
2014-02-17 10:18:15 +00:00
|
|
|
#include "Core/HW/Memmap.h"
|
2009-03-07 08:35:01 +00:00
|
|
|
|
2014-08-05 02:51:42 +00:00
|
|
|
#include "VideoCommon/BPMemory.h"
|
2014-08-02 06:42:15 +00:00
|
|
|
#include "VideoCommon/IndexGenerator.h"
|
2014-02-17 10:18:15 +00:00
|
|
|
#include "VideoCommon/Statistics.h"
|
2014-12-13 00:51:14 +00:00
|
|
|
#include "VideoCommon/VertexLoaderBase.h"
|
2014-02-17 10:18:15 +00:00
|
|
|
#include "VideoCommon/VertexLoaderManager.h"
|
2014-07-25 23:10:44 +00:00
|
|
|
#include "VideoCommon/VertexManagerBase.h"
|
2014-02-17 10:18:15 +00:00
|
|
|
#include "VideoCommon/VertexShaderManager.h"
|
|
|
|
#include "VideoCommon/VideoCommon.h"
|
2009-03-07 08:35:01 +00:00
|
|
|
|
2014-08-15 14:17:06 +00:00
|
|
|
|
2010-03-06 02:07:48 +00:00
|
|
|
|
2009-03-07 08:35:01 +00:00
|
|
|
namespace VertexLoaderManager
|
|
|
|
{
|
|
|
|
|
2014-12-12 07:53:48 +00:00
|
|
|
typedef std::unordered_map<PortableVertexDeclaration, std::unique_ptr<NativeVertexFormat>> NativeVertexFormatMap;
|
|
|
|
static NativeVertexFormatMap s_native_vertex_map;
|
|
|
|
static NativeVertexFormat* s_current_vtx_fmt;
|
|
|
|
|
2014-12-13 00:51:14 +00:00
|
|
|
typedef std::unordered_map<VertexLoaderUID, std::unique_ptr<VertexLoaderBase>> VertexLoaderMap;
|
2014-08-25 03:53:28 +00:00
|
|
|
static std::mutex s_vertex_loader_map_lock;
|
|
|
|
static VertexLoaderMap s_vertex_loader_map;
|
2009-03-07 08:35:01 +00:00
|
|
|
// TODO - change into array of pointers. Keep a map of all seen so far.
|
|
|
|
|
|
|
|
void Init()
|
|
|
|
{
|
|
|
|
MarkAllDirty();
|
2014-08-27 17:38:00 +00:00
|
|
|
for (auto& map_entry : g_main_cp_state.vertex_loaders)
|
|
|
|
map_entry = nullptr;
|
|
|
|
for (auto& map_entry : g_preprocess_cp_state.vertex_loaders)
|
2014-08-25 03:53:28 +00:00
|
|
|
map_entry = nullptr;
|
2009-04-24 16:40:58 +00:00
|
|
|
RecomputeCachedArraybases();
|
2015-02-12 23:15:57 +00:00
|
|
|
SETSTAT(stats.numVertexLoaders, 0);
|
2009-03-07 08:35:01 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void Shutdown()
|
|
|
|
{
|
2014-08-25 03:53:28 +00:00
|
|
|
std::lock_guard<std::mutex> lk(s_vertex_loader_map_lock);
|
|
|
|
s_vertex_loader_map.clear();
|
2014-12-12 07:53:48 +00:00
|
|
|
s_native_vertex_map.clear();
|
2009-03-07 08:35:01 +00:00
|
|
|
}
|
|
|
|
|
2013-04-24 13:21:54 +00:00
|
|
|
namespace
|
|
|
|
{
|
|
|
|
struct entry
|
|
|
|
{
|
2009-03-07 08:35:01 +00:00
|
|
|
std::string text;
|
|
|
|
u64 num_verts;
|
2013-04-24 13:21:54 +00:00
|
|
|
bool operator < (const entry &other) const
|
|
|
|
{
|
2009-03-07 08:35:01 +00:00
|
|
|
return num_verts > other.num_verts;
|
|
|
|
}
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
void AppendListToString(std::string *dest)
|
|
|
|
{
|
2014-08-25 03:53:28 +00:00
|
|
|
std::lock_guard<std::mutex> lk(s_vertex_loader_map_lock);
|
2009-03-07 08:35:01 +00:00
|
|
|
std::vector<entry> entries;
|
|
|
|
|
|
|
|
size_t total_size = 0;
|
2014-08-25 03:53:28 +00:00
|
|
|
for (const auto& map_entry : s_vertex_loader_map)
|
2009-03-07 08:35:01 +00:00
|
|
|
{
|
|
|
|
entry e;
|
2014-08-25 03:53:28 +00:00
|
|
|
map_entry.second->AppendToString(&e.text);
|
2014-12-12 23:23:54 +00:00
|
|
|
e.num_verts = map_entry.second->m_numLoadedVertices;
|
2009-03-07 08:35:01 +00:00
|
|
|
entries.push_back(e);
|
|
|
|
total_size += e.text.size() + 1;
|
|
|
|
}
|
|
|
|
sort(entries.begin(), entries.end());
|
|
|
|
dest->reserve(dest->size() + total_size);
|
2014-02-12 15:00:34 +00:00
|
|
|
for (const entry& entry : entries)
|
2013-04-24 13:21:54 +00:00
|
|
|
{
|
2014-02-12 15:00:34 +00:00
|
|
|
dest->append(entry.text);
|
2009-03-07 08:35:01 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void MarkAllDirty()
|
|
|
|
{
|
2014-10-22 00:42:55 +00:00
|
|
|
g_main_cp_state.attr_dirty = BitSet32::AllTrue(8);
|
|
|
|
g_preprocess_cp_state.attr_dirty = BitSet32::AllTrue(8);
|
2009-03-07 08:35:01 +00:00
|
|
|
}
|
|
|
|
|
2015-01-02 20:05:36 +00:00
|
|
|
static VertexLoaderBase* RefreshLoader(int vtx_attr_group, bool preprocess = false)
|
2014-08-15 14:17:06 +00:00
|
|
|
{
|
2015-01-02 20:05:36 +00:00
|
|
|
CPState* state = preprocess ? &g_preprocess_cp_state : &g_main_cp_state;
|
|
|
|
|
2014-12-13 00:51:14 +00:00
|
|
|
VertexLoaderBase* loader;
|
2014-10-22 00:42:55 +00:00
|
|
|
if (state->attr_dirty[vtx_attr_group])
|
2014-08-15 14:17:06 +00:00
|
|
|
{
|
2015-01-02 20:05:36 +00:00
|
|
|
// We are not allowed to create a native vertex format on preprocessing as this is on the wrong thread
|
|
|
|
bool check_for_native_format = !preprocess;
|
|
|
|
|
2014-08-27 17:38:00 +00:00
|
|
|
VertexLoaderUID uid(state->vtx_desc, state->vtx_attr[vtx_attr_group]);
|
2014-08-25 03:53:28 +00:00
|
|
|
std::lock_guard<std::mutex> lk(s_vertex_loader_map_lock);
|
|
|
|
VertexLoaderMap::iterator iter = s_vertex_loader_map.find(uid);
|
|
|
|
if (iter != s_vertex_loader_map.end())
|
2014-08-15 14:17:06 +00:00
|
|
|
{
|
2014-08-25 03:53:28 +00:00
|
|
|
loader = iter->second.get();
|
2015-01-02 20:05:36 +00:00
|
|
|
check_for_native_format &= !loader->m_native_vertex_format;
|
2014-08-15 14:17:06 +00:00
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2014-12-13 00:51:14 +00:00
|
|
|
loader = VertexLoaderBase::CreateVertexLoader(state->vtx_desc, state->vtx_attr[vtx_attr_group]);
|
|
|
|
s_vertex_loader_map[uid] = std::unique_ptr<VertexLoaderBase>(loader);
|
2015-01-02 20:05:36 +00:00
|
|
|
INCSTAT(stats.numVertexLoaders);
|
|
|
|
}
|
|
|
|
if (check_for_native_format)
|
|
|
|
{
|
2014-12-12 07:53:48 +00:00
|
|
|
// search for a cached native vertex format
|
2014-12-12 23:23:54 +00:00
|
|
|
const PortableVertexDeclaration& format = loader->m_native_vtx_decl;
|
2015-01-08 09:44:16 +00:00
|
|
|
std::unique_ptr<NativeVertexFormat>& native = s_native_vertex_map[format];
|
2014-12-12 07:53:48 +00:00
|
|
|
if (!native)
|
|
|
|
{
|
2015-01-08 09:44:16 +00:00
|
|
|
native.reset(g_vertex_manager->CreateNativeVertexFormat());
|
2014-12-12 07:53:48 +00:00
|
|
|
native->Initialize(format);
|
2014-12-12 23:23:54 +00:00
|
|
|
native->m_components = loader->m_native_components;
|
2014-12-12 07:53:48 +00:00
|
|
|
}
|
|
|
|
loader->m_native_vertex_format = native.get();
|
2014-08-15 14:17:06 +00:00
|
|
|
}
|
2014-08-27 17:38:00 +00:00
|
|
|
state->vertex_loaders[vtx_attr_group] = loader;
|
2014-10-22 00:42:55 +00:00
|
|
|
state->attr_dirty[vtx_attr_group] = false;
|
2014-08-25 03:53:28 +00:00
|
|
|
} else {
|
2014-08-27 17:38:00 +00:00
|
|
|
loader = state->vertex_loaders[vtx_attr_group];
|
2014-08-15 14:17:06 +00:00
|
|
|
}
|
2014-08-25 03:53:28 +00:00
|
|
|
return loader;
|
2014-08-15 14:17:06 +00:00
|
|
|
}
|
|
|
|
|
2015-01-31 08:23:50 +00:00
|
|
|
int RunVertices(int vtx_attr_group, int primitive, int count, DataReader src, bool skip_drawing, bool is_preprocess)
|
2009-03-07 08:35:01 +00:00
|
|
|
{
|
|
|
|
if (!count)
|
2014-12-09 07:35:04 +00:00
|
|
|
return 0;
|
2014-08-27 17:38:00 +00:00
|
|
|
|
2015-01-31 08:23:50 +00:00
|
|
|
VertexLoaderBase* loader = RefreshLoader(vtx_attr_group, is_preprocess);
|
2014-07-25 23:10:44 +00:00
|
|
|
|
2014-12-12 23:23:54 +00:00
|
|
|
int size = count * loader->m_VertexSize;
|
2014-12-09 07:35:04 +00:00
|
|
|
if ((int)src.size() < size)
|
|
|
|
return -1;
|
Refactor opcode decoding a bit to kill FifoCommandRunnable.
Separated out from my gpu-determinism branch by request. It's not a big
commit; I just like to write long commit messages.
The main reason to kill it is hopefully a slight performance improvement
from avoiding the double switch (especially in single core mode);
however, this also improves cycle calculation, as described below.
- FifoCommandRunnable is removed; in its stead, Decode returns the
number of cycles (which only matters for "sync" GPU mode), or 0 if there
was not enough data, and is also responsible for unknown opcode alerts.
Decode and DecodeSemiNop are almost identical, so the latter is replaced
with a skipped_frame parameter to Decode. Doesn't mean we can't improve
skipped_frame mode to do less work; if, at such a point, branching on it
has too much overhead (it certainly won't now), it can always be changed
to a template parameter.
- FifoCommandRunnable used a fixed, large cycle count for display lists,
regardless of the contents. Presumably the actual hardware's processing
time is mostly the processing time of whatever commands are in the list,
and with this change InterpretDisplayList can just return the list's
cycle count to be added to the total. (Since the calculation for this
is part of Decode, it didn't seem easy to split this change up.)
To facilitate this, Decode also gains an explicit 'end' parameter in
lieu of FifoCommandRunnable's call to GetVideoBufferEndPtr, which can
point to there or to the end of a display list (or elsewhere in
gpu-determinism, but that's another story). Also, as a small
optimization, InterpretDisplayList now calls OpcodeDecoder_Run rather
than having its own Decode loop, to allow Decode to be inlined (haven't
checked whether this actually happens though).
skipped_frame mode still does not traverse display lists and uses the
old fake value of 45 cycles. degasus has suggested that this hack is
not essential for performance and can be removed, but I want to separate
any potential performance impact of that from this commit.
2014-09-01 05:11:32 +00:00
|
|
|
|
2015-01-31 08:23:50 +00:00
|
|
|
if (skip_drawing || is_preprocess)
|
2014-12-09 07:35:04 +00:00
|
|
|
return size;
|
2014-08-05 02:51:42 +00:00
|
|
|
|
2014-07-25 23:10:44 +00:00
|
|
|
// If the native vertex format changed, force a flush.
|
2014-12-12 07:53:48 +00:00
|
|
|
if (loader->m_native_vertex_format != s_current_vtx_fmt)
|
2014-07-25 23:10:44 +00:00
|
|
|
VertexManager::Flush();
|
2014-12-12 07:53:48 +00:00
|
|
|
s_current_vtx_fmt = loader->m_native_vertex_format;
|
2014-07-25 23:10:44 +00:00
|
|
|
|
2015-01-24 01:37:20 +00:00
|
|
|
// if cull mode is CULL_ALL, tell VertexManager to skip triangles and quads.
|
|
|
|
// They still need to go through vertex loading, because we need to calculate a zfreeze refrence slope.
|
|
|
|
bool cullall = (bpmem.genMode.cullmode == GenMode::CULL_ALL && primitive < 5);
|
|
|
|
|
2014-12-09 07:35:04 +00:00
|
|
|
DataReader dst = VertexManager::PrepareForAdditionalData(primitive, count,
|
2015-01-24 01:37:20 +00:00
|
|
|
loader->m_native_vtx_decl.stride, cullall);
|
2014-08-02 06:42:15 +00:00
|
|
|
|
2015-04-06 09:44:13 +00:00
|
|
|
count = loader->RunVertices(src, dst, count);
|
2014-08-02 06:42:15 +00:00
|
|
|
|
|
|
|
IndexGenerator::AddIndices(primitive, count);
|
|
|
|
|
2014-12-12 23:23:54 +00:00
|
|
|
VertexManager::FlushData(count, loader->m_native_vtx_decl.stride);
|
2014-12-09 07:35:04 +00:00
|
|
|
|
2014-08-02 06:42:15 +00:00
|
|
|
ADDSTAT(stats.thisFrame.numPrims, count);
|
|
|
|
INCSTAT(stats.thisFrame.numPrimitiveJoins);
|
2014-12-09 07:35:04 +00:00
|
|
|
return size;
|
2014-01-30 13:48:23 +00:00
|
|
|
}
|
2009-08-08 01:39:56 +00:00
|
|
|
|
2014-07-25 23:10:44 +00:00
|
|
|
NativeVertexFormat* GetCurrentVertexFormat()
|
2014-06-05 15:55:21 +00:00
|
|
|
{
|
2014-07-25 23:10:44 +00:00
|
|
|
return s_current_vtx_fmt;
|
2014-06-05 15:55:21 +00:00
|
|
|
}
|
|
|
|
|
2009-03-07 08:35:01 +00:00
|
|
|
} // namespace
|
|
|
|
|
2014-08-27 17:38:00 +00:00
|
|
|
void LoadCPReg(u32 sub_cmd, u32 value, bool is_preprocess)
|
2009-03-07 08:35:01 +00:00
|
|
|
{
|
2014-08-27 17:38:00 +00:00
|
|
|
bool update_global_state = !is_preprocess;
|
|
|
|
CPState* state = is_preprocess ? &g_preprocess_cp_state : &g_main_cp_state;
|
2009-03-07 08:35:01 +00:00
|
|
|
switch (sub_cmd & 0xF0)
|
|
|
|
{
|
|
|
|
case 0x30:
|
2014-08-27 17:38:00 +00:00
|
|
|
if (update_global_state)
|
|
|
|
VertexShaderManager::SetTexMatrixChangedA(value);
|
2009-03-07 08:35:01 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
case 0x40:
|
2014-08-27 17:38:00 +00:00
|
|
|
if (update_global_state)
|
|
|
|
VertexShaderManager::SetTexMatrixChangedB(value);
|
2009-03-07 08:35:01 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
case 0x50:
|
2014-08-27 17:38:00 +00:00
|
|
|
state->vtx_desc.Hex &= ~0x1FFFF; // keep the Upper bits
|
|
|
|
state->vtx_desc.Hex |= value;
|
2014-10-22 00:42:55 +00:00
|
|
|
state->attr_dirty = BitSet32::AllTrue(8);
|
2009-03-07 08:35:01 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
case 0x60:
|
2014-08-27 17:38:00 +00:00
|
|
|
state->vtx_desc.Hex &= 0x1FFFF; // keep the lower 17Bits
|
|
|
|
state->vtx_desc.Hex |= (u64)value << 17;
|
2014-10-22 00:42:55 +00:00
|
|
|
state->attr_dirty = BitSet32::AllTrue(8);
|
2009-03-07 08:35:01 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
case 0x70:
|
|
|
|
_assert_((sub_cmd & 0x0F) < 8);
|
2014-08-27 17:38:00 +00:00
|
|
|
state->vtx_attr[sub_cmd & 7].g0.Hex = value;
|
2014-10-22 00:42:55 +00:00
|
|
|
state->attr_dirty[sub_cmd & 7] = true;
|
2009-03-07 08:35:01 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
case 0x80:
|
|
|
|
_assert_((sub_cmd & 0x0F) < 8);
|
2014-08-27 17:38:00 +00:00
|
|
|
state->vtx_attr[sub_cmd & 7].g1.Hex = value;
|
2014-10-22 00:42:55 +00:00
|
|
|
state->attr_dirty[sub_cmd & 7] = true;
|
2009-03-07 08:35:01 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
case 0x90:
|
|
|
|
_assert_((sub_cmd & 0x0F) < 8);
|
2014-08-27 17:38:00 +00:00
|
|
|
state->vtx_attr[sub_cmd & 7].g2.Hex = value;
|
2014-10-22 00:42:55 +00:00
|
|
|
state->attr_dirty[sub_cmd & 7] = true;
|
2009-03-07 08:35:01 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
// Pointers to vertex arrays in GC RAM
|
|
|
|
case 0xA0:
|
2014-08-27 17:38:00 +00:00
|
|
|
state->array_bases[sub_cmd & 0xF] = value;
|
|
|
|
if (update_global_state)
|
|
|
|
cached_arraybases[sub_cmd & 0xF] = Memory::GetPointer(value);
|
2009-03-07 08:35:01 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
case 0xB0:
|
2014-08-27 17:38:00 +00:00
|
|
|
state->array_strides[sub_cmd & 0xF] = value & 0xFF;
|
2009-03-07 08:35:01 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-03-27 02:55:08 +00:00
|
|
|
void FillCPMemoryArray(u32 *memory)
|
|
|
|
{
|
2014-08-27 17:26:06 +00:00
|
|
|
memory[0x30] = g_main_cp_state.matrix_index_a.Hex;
|
|
|
|
memory[0x40] = g_main_cp_state.matrix_index_b.Hex;
|
|
|
|
memory[0x50] = (u32)g_main_cp_state.vtx_desc.Hex;
|
|
|
|
memory[0x60] = (u32)(g_main_cp_state.vtx_desc.Hex >> 17);
|
2011-03-27 02:55:08 +00:00
|
|
|
|
|
|
|
for (int i = 0; i < 8; ++i)
|
|
|
|
{
|
2014-08-27 17:26:06 +00:00
|
|
|
memory[0x70 + i] = g_main_cp_state.vtx_attr[i].g0.Hex;
|
|
|
|
memory[0x80 + i] = g_main_cp_state.vtx_attr[i].g1.Hex;
|
|
|
|
memory[0x90 + i] = g_main_cp_state.vtx_attr[i].g2.Hex;
|
2011-03-27 02:55:08 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
for (int i = 0; i < 16; ++i)
|
|
|
|
{
|
2014-08-27 17:26:06 +00:00
|
|
|
memory[0xA0 + i] = g_main_cp_state.array_bases[i];
|
|
|
|
memory[0xB0 + i] = g_main_cp_state.array_strides[i];
|
2011-03-27 02:55:08 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-03-07 08:35:01 +00:00
|
|
|
void RecomputeCachedArraybases()
|
|
|
|
{
|
|
|
|
for (int i = 0; i < 16; i++)
|
|
|
|
{
|
2014-08-27 17:26:06 +00:00
|
|
|
cached_arraybases[i] = Memory::GetPointer(g_main_cp_state.array_bases[i]);
|
2009-03-07 08:35:01 +00:00
|
|
|
}
|
|
|
|
}
|