dolphin/Source/Core/VideoCommon/BPStructs.cpp

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1278 lines
47 KiB
C++
Raw Normal View History

// Copyright 2009 Dolphin Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#include "VideoCommon/BPStructs.h"
#include <algorithm>
#include <cmath>
#include <cstring>
#include <string>
#include <fmt/format.h>
#include "Common/CommonTypes.h"
#include "Common/EnumMap.h"
#include "Common/Logging/Log.h"
2014-09-08 22:24:49 -06:00
#include "Core/ConfigManager.h"
#include "Core/CoreTiming.h"
#include "Core/DolphinAnalytics.h"
#include "Core/FifoPlayer/FifoPlayer.h"
#include "Core/FifoPlayer/FifoRecorder.h"
#include "Core/HW/Memmap.h"
#include "Core/HW/VideoInterface.h"
#include "VideoCommon/BPFunctions.h"
#include "VideoCommon/BPMemory.h"
2014-09-16 19:04:37 -06:00
#include "VideoCommon/BoundingBox.h"
2014-07-08 08:49:33 -06:00
#include "VideoCommon/Fifo.h"
#include "VideoCommon/FramebufferManager.h"
#include "VideoCommon/GeometryShaderManager.h"
#include "VideoCommon/OpcodeDecoding.h"
#include "VideoCommon/PerfQueryBase.h"
#include "VideoCommon/PixelEngine.h"
#include "VideoCommon/PixelShaderManager.h"
#include "VideoCommon/RenderBase.h"
#include "VideoCommon/TMEM.h"
#include "VideoCommon/TextureCacheBase.h"
#include "VideoCommon/TextureDecoder.h"
#include "VideoCommon/VertexShaderManager.h"
#include "VideoCommon/VideoBackendBase.h"
#include "VideoCommon/VideoCommon.h"
#include "VideoCommon/VideoConfig.h"
using namespace BPFunctions;
static constexpr Common::EnumMap<float, GammaCorrection::Invalid2_2> s_gammaLUT = {1.0f, 1.7f, 2.2f,
2.2f};
void BPInit()
{
memset(reinterpret_cast<u8*>(&bpmem), 0, sizeof(bpmem));
bpmem.bpMask = 0xFFFFFF;
}
static void BPWritten(const BPCmd& bp, int cycles_into_future)
{
/*
----------------------------------------------------------------------------------------------------------------
Purpose: Writes to the BP registers
Called: At the end of every: OpcodeDecoding.cpp ExecuteDisplayList > Decode() > LoadBPReg
How It Works: First the pipeline is flushed then update the bpmem with the new value.
Some of the BP cases have to call certain functions while others just update the bpmem.
some bp cases check the changes variable, because they might not have to be updated all
the time
NOTE: it seems not all bp cases like checking changes, so calling if (bp.changes == 0 ? false :
true)
had to be ditched and the games seem to work fine with out it.
NOTE2: Yet Another GameCube Documentation calls them Bypass Raster State Registers but possibly
completely wrong
NOTE3: This controls the register groups: RAS1/2, SU, TF, TEV, C/Z, PEC
TODO: Turn into function table. The (future) DisplayList (DL) jit can then call the functions
directly,
getting rid of dynamic dispatch. Unfortunately, few games use DLs properly - most\
just stuff geometry in them and don't put state changes there
----------------------------------------------------------------------------------------------------------------
*/
if (((s32*)&bpmem)[bp.address] == bp.newvalue)
{
if (!(bp.address == BPMEM_TRIGGER_EFB_COPY || bp.address == BPMEM_CLEARBBOX1 ||
bp.address == BPMEM_CLEARBBOX2 || bp.address == BPMEM_SETDRAWDONE ||
bp.address == BPMEM_PE_TOKEN_ID || bp.address == BPMEM_PE_TOKEN_INT_ID ||
bp.address == BPMEM_LOADTLUT0 || bp.address == BPMEM_LOADTLUT1 ||
bp.address == BPMEM_TEXINVALIDATE || bp.address == BPMEM_PRELOAD_MODE ||
bp.address == BPMEM_CLEAR_PIXEL_PERF))
{
return;
}
}
FlushPipeline();
((u32*)&bpmem)[bp.address] = bp.newvalue;
switch (bp.address)
{
case BPMEM_GENMODE: // Set the Generation Mode
PRIM_LOG("genmode: texgen={}, col={}, multisampling={}, tev={}, cullmode={}, ind={}, zfeeze={}",
bpmem.genMode.numtexgens, bpmem.genMode.numcolchans, bpmem.genMode.multisampling,
bpmem.genMode.numtevstages + 1, bpmem.genMode.cullmode, bpmem.genMode.numindstages,
bpmem.genMode.zfreeze);
2017-07-19 23:25:24 -06:00
if (bp.changes)
PixelShaderManager::SetGenModeChanged();
2014-05-20 09:28:08 -06:00
// Only call SetGenerationMode when cull mode changes.
if (bp.changes & 0xC000)
SetGenerationMode();
2014-05-20 09:23:47 -06:00
return;
case BPMEM_IND_MTXA: // Index Matrix Changed
case BPMEM_IND_MTXB:
case BPMEM_IND_MTXC:
case BPMEM_IND_MTXA + 3:
case BPMEM_IND_MTXB + 3:
case BPMEM_IND_MTXC + 3:
case BPMEM_IND_MTXA + 6:
case BPMEM_IND_MTXB + 6:
case BPMEM_IND_MTXC + 6:
if (bp.changes)
PixelShaderManager::SetIndMatrixChanged((bp.address - BPMEM_IND_MTXA) / 3);
2014-05-20 09:23:47 -06:00
return;
case BPMEM_RAS1_SS0: // Index Texture Coordinate Scale 0
if (bp.changes)
PixelShaderManager::SetIndTexScaleChanged(false);
2014-05-20 09:23:47 -06:00
return;
case BPMEM_RAS1_SS1: // Index Texture Coordinate Scale 1
if (bp.changes)
PixelShaderManager::SetIndTexScaleChanged(true);
2014-05-20 09:23:47 -06:00
return;
// ----------------
// Scissor Control
// ----------------
case BPMEM_SCISSORTL: // Scissor Rectable Top, Left
case BPMEM_SCISSORBR: // Scissor Rectable Bottom, Right
case BPMEM_SCISSOROFFSET: // Scissor Offset
VertexShaderManager::SetViewportChanged();
GeometryShaderManager::SetViewportChanged();
2014-05-20 09:23:47 -06:00
return;
case BPMEM_LINEPTWIDTH: // Line Width
GeometryShaderManager::SetLinePtWidthChanged();
2014-05-20 09:23:47 -06:00
return;
case BPMEM_ZMODE: // Depth Control
PRIM_LOG("zmode: test={}, func={}, upd={}", bpmem.zmode.testenable, bpmem.zmode.func,
bpmem.zmode.updateenable);
SetDepthMode();
PixelShaderManager::SetZModeControl();
2014-05-20 09:23:47 -06:00
return;
case BPMEM_BLENDMODE: // Blending Control
2014-05-20 09:28:08 -06:00
if (bp.changes & 0xFFFF)
{
PRIM_LOG("blendmode: en={}, open={}, colupd={}, alphaupd={}, dst={}, src={}, sub={}, mode={}",
bpmem.blendmode.blendenable, bpmem.blendmode.logicopenable,
bpmem.blendmode.colorupdate, bpmem.blendmode.alphaupdate, bpmem.blendmode.dstfactor,
bpmem.blendmode.srcfactor, bpmem.blendmode.subtract, bpmem.blendmode.logicmode);
SetBlendMode();
2017-07-19 23:25:24 -06:00
PixelShaderManager::SetBlendModeChanged();
}
2014-05-20 09:23:47 -06:00
return;
2014-05-20 09:28:08 -06:00
case BPMEM_CONSTANTALPHA: // Set Destination Alpha
PRIM_LOG("constalpha: alp={}, en={}", bpmem.dstalpha.alpha, bpmem.dstalpha.enable);
2017-07-19 23:25:24 -06:00
if (bp.changes)
{
PixelShaderManager::SetAlpha();
PixelShaderManager::SetDestAlphaChanged();
}
2014-05-20 09:28:08 -06:00
if (bp.changes & 0x100)
SetBlendMode();
2014-05-20 09:23:47 -06:00
return;
// This is called when the game is done drawing the new frame (eg: like in DX: Begin(); Draw();
// End();)
// Triggers an interrupt on the PPC side so that the game knows when the GPU has finished drawing.
// Tokens are similar.
case BPMEM_SETDRAWDONE:
switch (bp.newvalue & 0xFF)
{
case 0x02:
g_texture_cache->FlushEFBCopies();
g_framebuffer_manager->InvalidatePeekCache(false);
if (!Fifo::UseDeterministicGPUThread())
PixelEngine::SetFinish(cycles_into_future); // may generate interrupt
DEBUG_LOG_FMT(VIDEO, "GXSetDrawDone SetPEFinish (value: {:#04X})", bp.newvalue & 0xFFFF);
2014-05-20 09:23:47 -06:00
return;
default:
WARN_LOG_FMT(VIDEO, "GXSetDrawDone ??? (value {:#04X})", bp.newvalue & 0xFFFF);
2014-05-20 09:23:47 -06:00
return;
}
2014-05-20 09:23:47 -06:00
return;
case BPMEM_PE_TOKEN_ID: // Pixel Engine Token ID
g_texture_cache->FlushEFBCopies();
g_framebuffer_manager->InvalidatePeekCache(false);
if (!Fifo::UseDeterministicGPUThread())
PixelEngine::SetToken(static_cast<u16>(bp.newvalue & 0xFFFF), false, cycles_into_future);
DEBUG_LOG_FMT(VIDEO, "SetPEToken {:#06X}", bp.newvalue & 0xFFFF);
2014-05-20 09:23:47 -06:00
return;
case BPMEM_PE_TOKEN_INT_ID: // Pixel Engine Interrupt Token ID
g_texture_cache->FlushEFBCopies();
g_framebuffer_manager->InvalidatePeekCache(false);
if (!Fifo::UseDeterministicGPUThread())
PixelEngine::SetToken(static_cast<u16>(bp.newvalue & 0xFFFF), true, cycles_into_future);
DEBUG_LOG_FMT(VIDEO, "SetPEToken + INT {:#06X}", bp.newvalue & 0xFFFF);
2014-05-20 09:23:47 -06:00
return;
// ------------------------
// EFB copy command. This copies a rectangle from the EFB to either RAM in a texture format or to
// XFB as YUYV.
// It can also optionally clear the EFB while copying from it. To emulate this, we of course copy
// first and clear afterwards.
case BPMEM_TRIGGER_EFB_COPY: // Copy EFB Region or Render to the XFB or Clear the screen.
{
// The bottom right is within the rectangle
// The values in bpmem.copyTexSrcXY and bpmem.copyTexSrcWH are updated in case 0x49 and 0x4a in
// this function
u32 destAddr = bpmem.copyTexDest << 5;
u32 destStride = bpmem.copyMipMapStrideChannels << 5;
MathUtil::Rectangle<s32> srcRect;
srcRect.left = bpmem.copyTexSrcXY.x;
srcRect.top = bpmem.copyTexSrcXY.y;
// Here Width+1 like Height, otherwise some textures are corrupted already since the native
// resolution.
srcRect.right = bpmem.copyTexSrcXY.x + bpmem.copyTexSrcWH.x + 1;
srcRect.bottom = bpmem.copyTexSrcXY.y + bpmem.copyTexSrcWH.y + 1;
const UPE_Copy PE_copy = bpmem.triggerEFBCopy;
// Since the copy X and Y coordinates/sizes are 10-bit, the game can configure a copy region up
// to 1024x1024. Hardware tests have found that the number of bytes written does not depend on
// the configured stride, instead it is based on the size registers, writing beyond the length
// of a single row. The data written for the pixels which lie outside the EFB bounds does not
// wrap around instead returning different colors based on the pixel format of the EFB. This
// suggests it's not based on coordinates, but instead on memory addresses. The effect of a
// within-bounds size but out-of-bounds offset (e.g. offset 320,0, size 640,480) are the same.
// As it would be difficult to emulate the exact behavior of out-of-bounds reads, instead of
// writing the junk data, we don't write anything to RAM at all for over-sized copies, and clamp
// to the EFB borders for over-offset copies. The arcade virtual console games (e.g. 1942) are
// known for configuring these out-of-range copies.
if (u32(srcRect.right) > EFB_WIDTH || u32(srcRect.bottom) > EFB_HEIGHT)
{
WARN_LOG_FMT(VIDEO, "Oversized EFB copy: {}x{} (offset {},{} stride {})", srcRect.GetWidth(),
srcRect.GetHeight(), srcRect.left, srcRect.top, destStride);
2021-09-16 18:39:06 -06:00
if (u32(srcRect.left) >= EFB_WIDTH || u32(srcRect.top) >= EFB_HEIGHT)
{
// This is not a sane src rectangle, it doesn't touch any valid image data at all
// Just ignore it
// Apparently Mario Kart Wii in wifi mode can generate a deformed EFB copy of size 4x4
// at offset (328,1020)
if (PE_copy.copy_to_xfb == 1)
{
// Make sure we disable Bounding box to match the side effects of the non-failure path
g_renderer->BBoxDisable();
}
2021-09-16 18:39:06 -06:00
return;
}
// Clamp the copy region to fit within EFB. So that we don't end up with a stretched image.
srcRect.right = std::clamp<int>(srcRect.right, 0, EFB_WIDTH);
srcRect.bottom = std::clamp<int>(srcRect.bottom, 0, EFB_HEIGHT);
}
const u32 copy_width = srcRect.GetWidth();
const u32 copy_height = srcRect.GetHeight();
// Check if we are to copy from the EFB or draw to the XFB
if (PE_copy.copy_to_xfb == 0)
{
// bpmem.zcontrol.pixel_format to PixelFormat::Z24 is when the game wants to copy from ZBuffer
// (Zbuffer uses 24-bit Format)
bool is_depth_copy = bpmem.zcontrol.pixel_format == PixelFormat::Z24;
g_texture_cache->CopyRenderTargetToTexture(
destAddr, PE_copy.tp_realFormat(), copy_width, copy_height, destStride, is_depth_copy,
srcRect, PE_copy.intensity_fmt && PE_copy.auto_conv, PE_copy.half_scale, 1.0f,
s_gammaLUT[PE_copy.gamma], bpmem.triggerEFBCopy.clamp_top,
bpmem.triggerEFBCopy.clamp_bottom, bpmem.copyfilter.GetCoefficients());
}
else
{
// We should be able to get away with deactivating the current bbox tracking
// here. Not sure if there's a better spot to put this.
// the number of lines copied is determined by the y scale * source efb height
g_renderer->BBoxDisable();
float yScale;
if (PE_copy.scale_invert)
yScale = 256.0f / static_cast<float>(bpmem.dispcopyyscale);
else
yScale = static_cast<float>(bpmem.dispcopyyscale) / 256.0f;
2015-09-04 07:32:30 -06:00
float num_xfb_lines = 1.0f + bpmem.copyTexSrcWH.y * yScale;
u32 height = static_cast<u32>(num_xfb_lines);
DEBUG_LOG_FMT(VIDEO,
"RenderToXFB: destAddr: {:08x} | srcRect [{} {} {} {}] | fbWidth: {} | "
"fbStride: {} | fbHeight: {} | yScale: {}",
destAddr, srcRect.left, srcRect.top, srcRect.right, srcRect.bottom,
bpmem.copyTexSrcWH.x + 1, destStride, height, yScale);
2017-05-29 16:02:09 -06:00
bool is_depth_copy = bpmem.zcontrol.pixel_format == PixelFormat::Z24;
g_texture_cache->CopyRenderTargetToTexture(
destAddr, EFBCopyFormat::XFB, copy_width, height, destStride, is_depth_copy, srcRect,
false, false, yScale, s_gammaLUT[PE_copy.gamma], bpmem.triggerEFBCopy.clamp_top,
bpmem.triggerEFBCopy.clamp_bottom, bpmem.copyfilter.GetCoefficients());
2017-05-29 16:02:09 -06:00
// This stays in to signal end of a "frame"
g_renderer->RenderToXFB(destAddr, srcRect, destStride, height, s_gammaLUT[PE_copy.gamma]);
if (g_ActiveConfig.bImmediateXFB)
{
// below div two to convert from bytes to pixels - it expects width, not stride
g_renderer->Swap(destAddr, destStride / 2, destStride, height, CoreTiming::GetTicks());
}
else
{
if (FifoPlayer::GetInstance().IsRunningWithFakeVideoInterfaceUpdates())
{
VideoInterface::FakeVIUpdate(destAddr, srcRect.GetWidth(), destStride, height);
}
}
}
// Clear the rectangular region after copying it.
if (PE_copy.clear)
{
ClearScreen(srcRect);
}
2014-05-20 09:23:47 -06:00
return;
}
case BPMEM_LOADTLUT0: // This one updates bpmem.tlutXferSrc, no need to do anything here.
2014-05-20 09:23:47 -06:00
return;
case BPMEM_LOADTLUT1: // Load a Texture Look Up Table
{
u32 tlutTMemAddr = (bp.newvalue & 0x3FF) << 9;
u32 tlutXferCount = (bp.newvalue & 0x1FFC00) >> 5;
u32 addr = bpmem.tmem_config.tlut_src << 5;
// The GameCube ignores the upper bits of this address. Some games (WW, MKDD) set them.
if (!SConfig::GetInstance().bWii)
addr = addr & 0x01FFFFFF;
Memory::CopyFromEmu(texMem + tlutTMemAddr, addr, tlutXferCount);
if (OpcodeDecoder::g_record_fifo_data)
FifoRecorder::GetInstance().UseMemory(addr, tlutXferCount, MemoryUpdate::TMEM);
TMEM::InvalidateAll();
2014-05-20 09:23:47 -06:00
return;
}
case BPMEM_FOGRANGE: // Fog Settings Control
case BPMEM_FOGRANGE + 1:
case BPMEM_FOGRANGE + 2:
case BPMEM_FOGRANGE + 3:
case BPMEM_FOGRANGE + 4:
case BPMEM_FOGRANGE + 5:
if (bp.changes)
PixelShaderManager::SetFogRangeAdjustChanged();
2014-05-20 09:23:47 -06:00
return;
case BPMEM_FOGPARAM0:
case BPMEM_FOGBMAGNITUDE:
case BPMEM_FOGBEXPONENT:
case BPMEM_FOGPARAM3:
if (bp.changes)
PixelShaderManager::SetFogParamChanged();
2014-05-20 09:23:47 -06:00
return;
case BPMEM_FOGCOLOR: // Fog Color
if (bp.changes)
PixelShaderManager::SetFogColorChanged();
2014-05-20 09:23:47 -06:00
return;
case BPMEM_ALPHACOMPARE: // Compare Alpha Values
PRIM_LOG("alphacmp: ref0={}, ref1={}, comp0={}, comp1={}, logic={}", bpmem.alpha_test.ref0,
bpmem.alpha_test.ref1, bpmem.alpha_test.comp0, bpmem.alpha_test.comp1,
bpmem.alpha_test.logic);
if (bp.changes & 0xFFFF)
PixelShaderManager::SetAlpha();
if (bp.changes)
{
2017-07-19 23:25:24 -06:00
PixelShaderManager::SetAlphaTestChanged();
SetBlendMode();
}
2014-05-20 09:23:47 -06:00
return;
case BPMEM_BIAS: // BIAS
PRIM_LOG("ztex bias={:#x}", bpmem.ztex1.bias);
if (bp.changes)
PixelShaderManager::SetZTextureBias();
2014-05-20 09:23:47 -06:00
return;
case BPMEM_ZTEX2: // Z Texture type
{
if (bp.changes & 3)
PixelShaderManager::SetZTextureTypeChanged();
if (bp.changes & 12)
2017-07-19 23:25:24 -06:00
PixelShaderManager::SetZTextureOpChanged();
PRIM_LOG("ztex op={}, type={}", bpmem.ztex2.op, bpmem.ztex2.type);
}
2014-05-20 09:23:47 -06:00
return;
// ----------------------------------
// Display Copy Filtering Control - GX_SetCopyFilter(u8 aa,u8 sample_pattern[12][2],u8 vf,u8
// vfilter[7])
// Fields: Destination, Frame2Field, Gamma, Source
// ----------------------------------
2014-05-03 13:53:42 -06:00
case BPMEM_DISPLAYCOPYFILTER: // if (aa) { use sample_pattern } else { use 666666 }
case BPMEM_DISPLAYCOPYFILTER + 1: // if (aa) { use sample_pattern } else { use 666666 }
case BPMEM_DISPLAYCOPYFILTER + 2: // if (aa) { use sample_pattern } else { use 666666 }
case BPMEM_DISPLAYCOPYFILTER + 3: // if (aa) { use sample_pattern } else { use 666666 }
2014-05-20 09:23:47 -06:00
case BPMEM_COPYFILTER0: // if (vf) { use vfilter } else { use 595000 }
case BPMEM_COPYFILTER1: // if (vf) { use vfilter } else { use 000015 }
return;
// -----------------------------------
// Interlacing Control
// -----------------------------------
case BPMEM_FIELDMASK: // GX_SetFieldMask(u8 even_mask,u8 odd_mask)
case BPMEM_FIELDMODE: // GX_SetFieldMode(u8 field_mode,u8 half_aspect_ratio)
2014-05-20 09:23:47 -06:00
// TODO
return;
// ----------------------------------------
// Unimportant regs (Clock, Perf, ...)
// ----------------------------------------
case BPMEM_BUSCLOCK0: // TB Bus Clock ?
case BPMEM_BUSCLOCK1: // TB Bus Clock ?
case BPMEM_PERF0_TRI: // Perf: Triangles
case BPMEM_PERF0_QUAD: // Perf: Quads
case BPMEM_PERF1: // Perf: Some Clock, Texels, TX, TC
return;
// ----------------
// EFB Copy config
// ----------------
case BPMEM_EFB_TL: // EFB Source Rect. Top, Left
2021-02-07 13:32:45 -07:00
case BPMEM_EFB_WH: // EFB Source Rect. Width, Height - 1
case BPMEM_EFB_ADDR: // EFB Target Address
2014-05-20 09:23:47 -06:00
return;
// --------------
// Clear Config
// --------------
case BPMEM_CLEAR_AR: // Alpha and Red Components
case BPMEM_CLEAR_GB: // Green and Blue Components
case BPMEM_CLEAR_Z: // Z Components (24-bit Zbuffer)
2014-05-20 09:23:47 -06:00
return;
// -------------------------
// Bounding Box Control
// -------------------------
case BPMEM_CLEARBBOX1:
case BPMEM_CLEARBBOX2:
2016-10-07 20:55:47 -06:00
{
const u8 offset = bp.address & 2;
g_renderer->BBoxEnable();
g_renderer->BBoxWrite(offset, bp.newvalue & 0x3ff);
g_renderer->BBoxWrite(offset + 1, bp.newvalue >> 10);
2016-10-07 20:55:47 -06:00
}
2014-05-20 09:23:47 -06:00
return;
2012-08-07 10:57:53 -06:00
case BPMEM_TEXINVALIDATE:
TMEM::Invalidate(bp.newvalue);
2014-05-20 09:23:47 -06:00
return;
case BPMEM_ZCOMPARE: // Set the Z-Compare and EFB pixel format
OnPixelFormatChange();
if (bp.changes & 7)
SetBlendMode(); // dual source could be activated by changing to PIXELFMT_RGBA6_Z24
PixelShaderManager::SetZModeControl();
2014-05-20 09:23:47 -06:00
return;
case BPMEM_MIPMAP_STRIDE: // MipMap Stride Channel
case BPMEM_COPYYSCALE: // Display Copy Y Scale
2014-05-20 09:28:08 -06:00
/* 24 RID
* 21 BC3 - Ind. Tex Stage 3 NTexCoord
* 18 BI3 - Ind. Tex Stage 3 NTexMap
* 15 BC2 - Ind. Tex Stage 2 NTexCoord
* 12 BI2 - Ind. Tex Stage 2 NTexMap
* 9 BC1 - Ind. Tex Stage 1 NTexCoord
* 6 BI1 - Ind. Tex Stage 1 NTexMap
* 3 BC0 - Ind. Tex Stage 0 NTexCoord
* 0 BI0 - Ind. Tex Stage 0 NTexMap */
case BPMEM_IREF:
2017-07-19 23:25:24 -06:00
{
if (bp.changes)
PixelShaderManager::SetTevIndirectChanged();
return;
}
2014-05-20 09:28:08 -06:00
case BPMEM_TEV_KSEL: // Texture Environment Swap Mode Table 0
case BPMEM_TEV_KSEL + 1: // Texture Environment Swap Mode Table 1
case BPMEM_TEV_KSEL + 2: // Texture Environment Swap Mode Table 2
case BPMEM_TEV_KSEL + 3: // Texture Environment Swap Mode Table 3
case BPMEM_TEV_KSEL + 4: // Texture Environment Swap Mode Table 4
case BPMEM_TEV_KSEL + 5: // Texture Environment Swap Mode Table 5
case BPMEM_TEV_KSEL + 6: // Texture Environment Swap Mode Table 6
case BPMEM_TEV_KSEL + 7: // Texture Environment Swap Mode Table 7
2017-07-19 23:25:24 -06:00
PixelShaderManager::SetTevKSel(bp.address - BPMEM_TEV_KSEL, bp.newvalue);
return;
/* This Register can be used to limit to which bits of BP registers is
* actually written to. The mask is only valid for the next BP write,
* and will reset itself afterwards. It's handled as a special case in
* LoadBPReg. */
case BPMEM_BP_MASK:
case BPMEM_IND_IMASK: // Index Mask ?
case BPMEM_REVBITS: // Always set to 0x0F when GX_InitRevBits() is called.
2014-05-20 09:23:47 -06:00
return;
case BPMEM_CLEAR_PIXEL_PERF:
// GXClearPixMetric writes 0xAAA here, Sunshine alternates this register between values 0x000
// and 0xAAA
if (PerfQueryBase::ShouldEmulate())
g_perf_query->ResetQuery();
2014-05-20 09:23:47 -06:00
return;
2012-01-29 13:17:22 -07:00
case BPMEM_PRELOAD_ADDR:
case BPMEM_PRELOAD_TMEMEVEN:
case BPMEM_PRELOAD_TMEMODD: // Used when PRELOAD_MODE is set
2014-05-20 09:23:47 -06:00
return;
2012-01-29 13:17:22 -07:00
case BPMEM_PRELOAD_MODE: // Set to 0 when GX_TexModeSync() is called.
2012-08-07 10:57:53 -06:00
// if this is different from 0, manual TMEM management is used (GX_PreloadEntireTexture).
2012-01-29 13:17:22 -07:00
if (bp.newvalue != 0)
{
// TODO: Not quite sure if this is completely correct (likely not)
// NOTE: libogc's implementation of GX_PreloadEntireTexture seems flawed, so it's not
// necessarily a good reference for RE'ing this feature.
2012-01-29 13:17:22 -07:00
BPS_TmemConfig& tmem_cfg = bpmem.tmem_config;
u32 src_addr = tmem_cfg.preload_addr << 5; // TODO: Should we add mask here on GC?
u32 bytes_read = 0;
u32 tmem_addr_even = tmem_cfg.preload_tmem_even * TMEM_LINE_SIZE;
if (tmem_cfg.preload_tile_info.type != 3)
{
bytes_read = tmem_cfg.preload_tile_info.count * TMEM_LINE_SIZE;
if (tmem_addr_even + bytes_read > TMEM_SIZE)
bytes_read = TMEM_SIZE - tmem_addr_even;
Memory::CopyFromEmu(texMem + tmem_addr_even, src_addr, bytes_read);
}
else // RGBA8 tiles (and CI14, but that might just be stupid libogc!)
{
u8* src_ptr = Memory::GetPointer(src_addr);
// AR and GB tiles are stored in separate TMEM banks => can't use a single memcpy for
// everything
u32 tmem_addr_odd = tmem_cfg.preload_tmem_odd * TMEM_LINE_SIZE;
for (u32 i = 0; i < tmem_cfg.preload_tile_info.count; ++i)
{
if (tmem_addr_even + TMEM_LINE_SIZE > TMEM_SIZE ||
tmem_addr_odd + TMEM_LINE_SIZE > TMEM_SIZE)
break;
memcpy(texMem + tmem_addr_even, src_ptr + bytes_read, TMEM_LINE_SIZE);
memcpy(texMem + tmem_addr_odd, src_ptr + bytes_read + TMEM_LINE_SIZE, TMEM_LINE_SIZE);
tmem_addr_even += TMEM_LINE_SIZE;
tmem_addr_odd += TMEM_LINE_SIZE;
bytes_read += TMEM_LINE_SIZE * 2;
}
}
if (OpcodeDecoder::g_record_fifo_data)
FifoRecorder::GetInstance().UseMemory(src_addr, bytes_read, MemoryUpdate::TMEM);
TMEM::InvalidateAll();
2012-01-29 13:17:22 -07:00
}
2014-05-20 09:23:47 -06:00
return;
// ---------------------------------------------------
// Set the TEV Color
// ---------------------------------------------------
//
// NOTE: Each of these registers actually maps to two variables internally.
// There's a bit that specifies which one is currently written to.
//
// NOTE: Some games write only to the RA register (or only to the BG register).
// We may not assume that the unwritten register holds a valid value, hence
// both component pairs need to be loaded individually.
case BPMEM_TEV_COLOR_RA:
case BPMEM_TEV_COLOR_RA + 2:
case BPMEM_TEV_COLOR_RA + 4:
case BPMEM_TEV_COLOR_RA + 6:
{
int num = (bp.address >> 1) & 0x3;
if (bpmem.tevregs[num].ra.type == TevRegType::Constant)
{
PixelShaderManager::SetTevKonstColor(num, 0, bpmem.tevregs[num].ra.red);
PixelShaderManager::SetTevKonstColor(num, 3, bpmem.tevregs[num].ra.alpha);
}
else
{
PixelShaderManager::SetTevColor(num, 0, bpmem.tevregs[num].ra.red);
PixelShaderManager::SetTevColor(num, 3, bpmem.tevregs[num].ra.alpha);
}
return;
}
case BPMEM_TEV_COLOR_BG:
case BPMEM_TEV_COLOR_BG + 2:
case BPMEM_TEV_COLOR_BG + 4:
case BPMEM_TEV_COLOR_BG + 6:
{
int num = (bp.address >> 1) & 0x3;
if (bpmem.tevregs[num].bg.type == TevRegType::Constant)
{
PixelShaderManager::SetTevKonstColor(num, 1, bpmem.tevregs[num].bg.green);
PixelShaderManager::SetTevKonstColor(num, 2, bpmem.tevregs[num].bg.blue);
}
else
{
PixelShaderManager::SetTevColor(num, 1, bpmem.tevregs[num].bg.green);
PixelShaderManager::SetTevColor(num, 2, bpmem.tevregs[num].bg.blue);
}
return;
}
2014-05-20 09:23:47 -06:00
default:
break;
2014-05-20 09:23:47 -06:00
}
2014-05-20 09:23:47 -06:00
switch (bp.address & 0xFC) // Texture sampler filter
{
// -------------------------
// Texture Environment Order
// -------------------------
case BPMEM_TREF:
case BPMEM_TREF + 4:
2017-07-19 23:25:24 -06:00
PixelShaderManager::SetTevOrder(bp.address - BPMEM_TREF, bp.newvalue);
2014-05-20 09:23:47 -06:00
return;
// ----------------------
// Set wrap size
// ----------------------
case BPMEM_SU_SSIZE: // Matches BPMEM_SU_TSIZE too
2014-05-20 09:23:47 -06:00
case BPMEM_SU_SSIZE + 4:
case BPMEM_SU_SSIZE + 8:
case BPMEM_SU_SSIZE + 12:
if (bp.changes)
{
2014-05-20 09:23:47 -06:00
PixelShaderManager::SetTexCoordChanged((bp.address - BPMEM_SU_SSIZE) >> 1);
GeometryShaderManager::SetTexCoordChanged((bp.address - BPMEM_SU_SSIZE) >> 1);
}
2014-05-20 09:23:47 -06:00
return;
}
if ((bp.address & 0xc0) == 0x80)
{
auto tex_address = TexUnitAddress::FromBPAddress(bp.address);
switch (tex_address.Reg)
{
// ------------------------
// BPMEM_TX_SETMODE0 - (Texture lookup and filtering mode) LOD/BIAS Clamp, MaxAnsio, LODBIAS,
// DiagLoad, Min Filter, Mag Filter, Wrap T, S
// BPMEM_TX_SETMODE1 - (LOD Stuff) - Max LOD, Min LOD
// ------------------------
case TexUnitAddress::Register::SETMODE0:
case TexUnitAddress::Register::SETMODE1:
TMEM::ConfigurationChanged(tex_address, bp.newvalue);
return;
// --------------------------------------------
// BPMEM_TX_SETIMAGE0 - Texture width, height, format
// BPMEM_TX_SETIMAGE1 - even LOD address in TMEM - Image Type, Cache Height, Cache Width,
// TMEM Offset
// BPMEM_TX_SETIMAGE2 - odd LOD address in TMEM - Cache Height, Cache Width, TMEM Offset
// BPMEM_TX_SETIMAGE3 - Address of Texture in main memory
// --------------------------------------------
case TexUnitAddress::Register::SETIMAGE0:
case TexUnitAddress::Register::SETIMAGE1:
case TexUnitAddress::Register::SETIMAGE2:
case TexUnitAddress::Register::SETIMAGE3:
TMEM::ConfigurationChanged(tex_address, bp.newvalue);
return;
// -------------------------------
// Set a TLUT
// BPMEM_TX_SETTLUT - Format, TMEM Offset (offset of TLUT from start of TMEM high bank > > 5)
// -------------------------------
case TexUnitAddress::Register::SETTLUT:
TMEM::ConfigurationChanged(tex_address, bp.newvalue);
return;
case TexUnitAddress::Register::UNKNOWN:
break; // Not handled
}
}
2014-05-20 09:23:47 -06:00
switch (bp.address & 0xF0)
{
// --------------
// Indirect Tev
// --------------
2017-07-19 23:25:24 -06:00
case BPMEM_IND_CMD:
PixelShaderManager::SetTevIndirectChanged();
2014-05-20 09:23:47 -06:00
return;
// --------------------------------------------------
// Set Color/Alpha of a Tev
// BPMEM_TEV_COLOR_ENV - Dest, Shift, Clamp, Sub, Bias, Sel A, Sel B, Sel C, Sel D
// BPMEM_TEV_ALPHA_ENV - Dest, Shift, Clamp, Sub, Bias, Sel A, Sel B, Sel C, Sel D, T Swap, R Swap
// --------------------------------------------------
2017-07-19 23:25:24 -06:00
case BPMEM_TEV_COLOR_ENV: // Texture Environment 1
case BPMEM_TEV_COLOR_ENV + 16:
PixelShaderManager::SetTevCombiner((bp.address - BPMEM_TEV_COLOR_ENV) >> 1,
(bp.address - BPMEM_TEV_COLOR_ENV) & 1, bp.newvalue);
2014-05-20 09:23:47 -06:00
return;
default:
break;
}
DolphinAnalytics::Instance().ReportGameQuirk(GameQuirk::USES_UNKNOWN_BP_COMMAND);
WARN_LOG_FMT(VIDEO, "Unknown BP opcode: address = {:#010x} value = {:#010x}", bp.address,
bp.newvalue);
}
// Call browser: OpcodeDecoding.cpp RunCallback::OnBP()
void LoadBPReg(u8 reg, u32 value, int cycles_into_future)
2014-05-03 14:36:29 -06:00
{
int oldval = ((u32*)&bpmem)[reg];
int newval = (oldval & ~bpmem.bpMask) | (value & bpmem.bpMask);
2014-05-03 14:36:29 -06:00
int changes = (oldval ^ newval) & 0xFFFFFF;
BPCmd bp = {reg, changes, newval};
2014-05-03 14:36:29 -06:00
// Reset the mask register if we're not trying to set it ourselves.
if (reg != BPMEM_BP_MASK)
2014-05-03 14:36:29 -06:00
bpmem.bpMask = 0xFFFFFF;
BPWritten(bp, cycles_into_future);
2014-05-03 14:36:29 -06:00
}
void LoadBPRegPreprocess(u8 reg, u32 value, int cycles_into_future)
Add the 'desynced GPU thread' mode. It's a relatively big commit (less big with -w), but it's hard to test any of this separately... The basic problem is that in netplay or movies, the state of the CPU must be deterministic, including when the game receives notification that the GPU has processed FIFO data. Dual core mode notifies the game whenever the GPU thread actually gets around to doing the work, so it isn't deterministic. Single core mode is because it notifies the game 'instantly' (after processing the data synchronously), but it's too slow for many systems and games. My old dc-netplay branch worked as follows: everything worked as normal except the state of the CP registers was a lie, and the CPU thread only delivered results when idle detection triggered (waiting for the GPU if they weren't ready at that point). Usually, a game is idle iff all the work for the frame has been done, except for a small amount of work depending on the GPU result, so neither the CPU or the GPU waiting on the other affected performance much. However, it's possible that the game could be waiting for some earlier interrupt, and any of several games which, for whatever reason, never went into a detectable idle (even when I tried to improve the detection) would never receive results at all. (The current method should have better compatibility, but it also has slightly higher overhead and breaks some other things, so I want to reimplement this, hopefully with less impact on the code, in the future.) With this commit, the basic idea is that the CPU thread acts as if the work has been done instantly, like single core mode, but actually hands it off asynchronously to the GPU thread (after backing up some data that the game might change in memory before it's actually done). Since the work isn't done, any feedback from the GPU to the CPU, such as real XFB/EFB copies (virtual are OK), EFB pokes, performance queries, etc. is broken; but most games work with these options disabled, and there is no need to try to detect what the CPU thread is doing. Technically: when the flag g_use_deterministic_gpu_thread (currently stuck on) is on, the CPU thread calls RunGpu like in single core mode. This function synchronously copies the data from the FIFO to the internal video buffer and updates the CP registers, interrupts, etc. However, instead of the regular ReadDataFromFifo followed by running the opcode decoder, it runs ReadDataFromFifoOnCPU -> OpcodeDecoder_Preprocess, which relatively quickly scans through the FIFO data, detects SetFinish calls etc., which are immediately fired, and saves certain associated data from memory (e.g. display lists) in AuxBuffers (a parallel stream to the main FIFO, which is a bit slow at the moment), before handing the data off to the GPU thread to actually render. That makes up the bulk of this commit. In various circumstances, including the aforementioned EFB pokes and performance queries as well as swap requests (i.e. the end of a frame - we don't want the CPU potentially pumping out frames too quickly and the GPU falling behind*), SyncGPU is called to wait for actual completion. The overhead mainly comes from OpcodeDecoder_Preprocess (which is, again, synchronous), as well as the actual copying. Currently, display lists and such are escrowed from main memory even though they usually won't change over the course of a frame, and textures are not even though they might, resulting in a small chance of graphical glitches. When the texture locking (i.e. fault on write) code lands, I can make this all correct and maybe a little faster. * This suggests an alternate determinism method of just delaying results until a short time before the end of each frame. For all I know this might mostly work - I haven't tried it - but if any significant work hinges on the competion of render to texture etc., the frame will be missed.
2014-08-27 20:56:19 -06:00
{
// masking via BPMEM_BP_MASK could hypothetically be a problem
u32 newval = value & 0xffffff;
switch (reg)
Add the 'desynced GPU thread' mode. It's a relatively big commit (less big with -w), but it's hard to test any of this separately... The basic problem is that in netplay or movies, the state of the CPU must be deterministic, including when the game receives notification that the GPU has processed FIFO data. Dual core mode notifies the game whenever the GPU thread actually gets around to doing the work, so it isn't deterministic. Single core mode is because it notifies the game 'instantly' (after processing the data synchronously), but it's too slow for many systems and games. My old dc-netplay branch worked as follows: everything worked as normal except the state of the CP registers was a lie, and the CPU thread only delivered results when idle detection triggered (waiting for the GPU if they weren't ready at that point). Usually, a game is idle iff all the work for the frame has been done, except for a small amount of work depending on the GPU result, so neither the CPU or the GPU waiting on the other affected performance much. However, it's possible that the game could be waiting for some earlier interrupt, and any of several games which, for whatever reason, never went into a detectable idle (even when I tried to improve the detection) would never receive results at all. (The current method should have better compatibility, but it also has slightly higher overhead and breaks some other things, so I want to reimplement this, hopefully with less impact on the code, in the future.) With this commit, the basic idea is that the CPU thread acts as if the work has been done instantly, like single core mode, but actually hands it off asynchronously to the GPU thread (after backing up some data that the game might change in memory before it's actually done). Since the work isn't done, any feedback from the GPU to the CPU, such as real XFB/EFB copies (virtual are OK), EFB pokes, performance queries, etc. is broken; but most games work with these options disabled, and there is no need to try to detect what the CPU thread is doing. Technically: when the flag g_use_deterministic_gpu_thread (currently stuck on) is on, the CPU thread calls RunGpu like in single core mode. This function synchronously copies the data from the FIFO to the internal video buffer and updates the CP registers, interrupts, etc. However, instead of the regular ReadDataFromFifo followed by running the opcode decoder, it runs ReadDataFromFifoOnCPU -> OpcodeDecoder_Preprocess, which relatively quickly scans through the FIFO data, detects SetFinish calls etc., which are immediately fired, and saves certain associated data from memory (e.g. display lists) in AuxBuffers (a parallel stream to the main FIFO, which is a bit slow at the moment), before handing the data off to the GPU thread to actually render. That makes up the bulk of this commit. In various circumstances, including the aforementioned EFB pokes and performance queries as well as swap requests (i.e. the end of a frame - we don't want the CPU potentially pumping out frames too quickly and the GPU falling behind*), SyncGPU is called to wait for actual completion. The overhead mainly comes from OpcodeDecoder_Preprocess (which is, again, synchronous), as well as the actual copying. Currently, display lists and such are escrowed from main memory even though they usually won't change over the course of a frame, and textures are not even though they might, resulting in a small chance of graphical glitches. When the texture locking (i.e. fault on write) code lands, I can make this all correct and maybe a little faster. * This suggests an alternate determinism method of just delaying results until a short time before the end of each frame. For all I know this might mostly work - I haven't tried it - but if any significant work hinges on the competion of render to texture etc., the frame will be missed.
2014-08-27 20:56:19 -06:00
{
case BPMEM_SETDRAWDONE:
if ((newval & 0xff) == 0x02)
PixelEngine::SetFinish(cycles_into_future);
Add the 'desynced GPU thread' mode. It's a relatively big commit (less big with -w), but it's hard to test any of this separately... The basic problem is that in netplay or movies, the state of the CPU must be deterministic, including when the game receives notification that the GPU has processed FIFO data. Dual core mode notifies the game whenever the GPU thread actually gets around to doing the work, so it isn't deterministic. Single core mode is because it notifies the game 'instantly' (after processing the data synchronously), but it's too slow for many systems and games. My old dc-netplay branch worked as follows: everything worked as normal except the state of the CP registers was a lie, and the CPU thread only delivered results when idle detection triggered (waiting for the GPU if they weren't ready at that point). Usually, a game is idle iff all the work for the frame has been done, except for a small amount of work depending on the GPU result, so neither the CPU or the GPU waiting on the other affected performance much. However, it's possible that the game could be waiting for some earlier interrupt, and any of several games which, for whatever reason, never went into a detectable idle (even when I tried to improve the detection) would never receive results at all. (The current method should have better compatibility, but it also has slightly higher overhead and breaks some other things, so I want to reimplement this, hopefully with less impact on the code, in the future.) With this commit, the basic idea is that the CPU thread acts as if the work has been done instantly, like single core mode, but actually hands it off asynchronously to the GPU thread (after backing up some data that the game might change in memory before it's actually done). Since the work isn't done, any feedback from the GPU to the CPU, such as real XFB/EFB copies (virtual are OK), EFB pokes, performance queries, etc. is broken; but most games work with these options disabled, and there is no need to try to detect what the CPU thread is doing. Technically: when the flag g_use_deterministic_gpu_thread (currently stuck on) is on, the CPU thread calls RunGpu like in single core mode. This function synchronously copies the data from the FIFO to the internal video buffer and updates the CP registers, interrupts, etc. However, instead of the regular ReadDataFromFifo followed by running the opcode decoder, it runs ReadDataFromFifoOnCPU -> OpcodeDecoder_Preprocess, which relatively quickly scans through the FIFO data, detects SetFinish calls etc., which are immediately fired, and saves certain associated data from memory (e.g. display lists) in AuxBuffers (a parallel stream to the main FIFO, which is a bit slow at the moment), before handing the data off to the GPU thread to actually render. That makes up the bulk of this commit. In various circumstances, including the aforementioned EFB pokes and performance queries as well as swap requests (i.e. the end of a frame - we don't want the CPU potentially pumping out frames too quickly and the GPU falling behind*), SyncGPU is called to wait for actual completion. The overhead mainly comes from OpcodeDecoder_Preprocess (which is, again, synchronous), as well as the actual copying. Currently, display lists and such are escrowed from main memory even though they usually won't change over the course of a frame, and textures are not even though they might, resulting in a small chance of graphical glitches. When the texture locking (i.e. fault on write) code lands, I can make this all correct and maybe a little faster. * This suggests an alternate determinism method of just delaying results until a short time before the end of each frame. For all I know this might mostly work - I haven't tried it - but if any significant work hinges on the competion of render to texture etc., the frame will be missed.
2014-08-27 20:56:19 -06:00
break;
case BPMEM_PE_TOKEN_ID:
PixelEngine::SetToken(newval & 0xffff, false, cycles_into_future);
Add the 'desynced GPU thread' mode. It's a relatively big commit (less big with -w), but it's hard to test any of this separately... The basic problem is that in netplay or movies, the state of the CPU must be deterministic, including when the game receives notification that the GPU has processed FIFO data. Dual core mode notifies the game whenever the GPU thread actually gets around to doing the work, so it isn't deterministic. Single core mode is because it notifies the game 'instantly' (after processing the data synchronously), but it's too slow for many systems and games. My old dc-netplay branch worked as follows: everything worked as normal except the state of the CP registers was a lie, and the CPU thread only delivered results when idle detection triggered (waiting for the GPU if they weren't ready at that point). Usually, a game is idle iff all the work for the frame has been done, except for a small amount of work depending on the GPU result, so neither the CPU or the GPU waiting on the other affected performance much. However, it's possible that the game could be waiting for some earlier interrupt, and any of several games which, for whatever reason, never went into a detectable idle (even when I tried to improve the detection) would never receive results at all. (The current method should have better compatibility, but it also has slightly higher overhead and breaks some other things, so I want to reimplement this, hopefully with less impact on the code, in the future.) With this commit, the basic idea is that the CPU thread acts as if the work has been done instantly, like single core mode, but actually hands it off asynchronously to the GPU thread (after backing up some data that the game might change in memory before it's actually done). Since the work isn't done, any feedback from the GPU to the CPU, such as real XFB/EFB copies (virtual are OK), EFB pokes, performance queries, etc. is broken; but most games work with these options disabled, and there is no need to try to detect what the CPU thread is doing. Technically: when the flag g_use_deterministic_gpu_thread (currently stuck on) is on, the CPU thread calls RunGpu like in single core mode. This function synchronously copies the data from the FIFO to the internal video buffer and updates the CP registers, interrupts, etc. However, instead of the regular ReadDataFromFifo followed by running the opcode decoder, it runs ReadDataFromFifoOnCPU -> OpcodeDecoder_Preprocess, which relatively quickly scans through the FIFO data, detects SetFinish calls etc., which are immediately fired, and saves certain associated data from memory (e.g. display lists) in AuxBuffers (a parallel stream to the main FIFO, which is a bit slow at the moment), before handing the data off to the GPU thread to actually render. That makes up the bulk of this commit. In various circumstances, including the aforementioned EFB pokes and performance queries as well as swap requests (i.e. the end of a frame - we don't want the CPU potentially pumping out frames too quickly and the GPU falling behind*), SyncGPU is called to wait for actual completion. The overhead mainly comes from OpcodeDecoder_Preprocess (which is, again, synchronous), as well as the actual copying. Currently, display lists and such are escrowed from main memory even though they usually won't change over the course of a frame, and textures are not even though they might, resulting in a small chance of graphical glitches. When the texture locking (i.e. fault on write) code lands, I can make this all correct and maybe a little faster. * This suggests an alternate determinism method of just delaying results until a short time before the end of each frame. For all I know this might mostly work - I haven't tried it - but if any significant work hinges on the competion of render to texture etc., the frame will be missed.
2014-08-27 20:56:19 -06:00
break;
case BPMEM_PE_TOKEN_INT_ID: // Pixel Engine Interrupt Token ID
PixelEngine::SetToken(newval & 0xffff, true, cycles_into_future);
Add the 'desynced GPU thread' mode. It's a relatively big commit (less big with -w), but it's hard to test any of this separately... The basic problem is that in netplay or movies, the state of the CPU must be deterministic, including when the game receives notification that the GPU has processed FIFO data. Dual core mode notifies the game whenever the GPU thread actually gets around to doing the work, so it isn't deterministic. Single core mode is because it notifies the game 'instantly' (after processing the data synchronously), but it's too slow for many systems and games. My old dc-netplay branch worked as follows: everything worked as normal except the state of the CP registers was a lie, and the CPU thread only delivered results when idle detection triggered (waiting for the GPU if they weren't ready at that point). Usually, a game is idle iff all the work for the frame has been done, except for a small amount of work depending on the GPU result, so neither the CPU or the GPU waiting on the other affected performance much. However, it's possible that the game could be waiting for some earlier interrupt, and any of several games which, for whatever reason, never went into a detectable idle (even when I tried to improve the detection) would never receive results at all. (The current method should have better compatibility, but it also has slightly higher overhead and breaks some other things, so I want to reimplement this, hopefully with less impact on the code, in the future.) With this commit, the basic idea is that the CPU thread acts as if the work has been done instantly, like single core mode, but actually hands it off asynchronously to the GPU thread (after backing up some data that the game might change in memory before it's actually done). Since the work isn't done, any feedback from the GPU to the CPU, such as real XFB/EFB copies (virtual are OK), EFB pokes, performance queries, etc. is broken; but most games work with these options disabled, and there is no need to try to detect what the CPU thread is doing. Technically: when the flag g_use_deterministic_gpu_thread (currently stuck on) is on, the CPU thread calls RunGpu like in single core mode. This function synchronously copies the data from the FIFO to the internal video buffer and updates the CP registers, interrupts, etc. However, instead of the regular ReadDataFromFifo followed by running the opcode decoder, it runs ReadDataFromFifoOnCPU -> OpcodeDecoder_Preprocess, which relatively quickly scans through the FIFO data, detects SetFinish calls etc., which are immediately fired, and saves certain associated data from memory (e.g. display lists) in AuxBuffers (a parallel stream to the main FIFO, which is a bit slow at the moment), before handing the data off to the GPU thread to actually render. That makes up the bulk of this commit. In various circumstances, including the aforementioned EFB pokes and performance queries as well as swap requests (i.e. the end of a frame - we don't want the CPU potentially pumping out frames too quickly and the GPU falling behind*), SyncGPU is called to wait for actual completion. The overhead mainly comes from OpcodeDecoder_Preprocess (which is, again, synchronous), as well as the actual copying. Currently, display lists and such are escrowed from main memory even though they usually won't change over the course of a frame, and textures are not even though they might, resulting in a small chance of graphical glitches. When the texture locking (i.e. fault on write) code lands, I can make this all correct and maybe a little faster. * This suggests an alternate determinism method of just delaying results until a short time before the end of each frame. For all I know this might mostly work - I haven't tried it - but if any significant work hinges on the competion of render to texture etc., the frame will be missed.
2014-08-27 20:56:19 -06:00
break;
}
}
std::pair<std::string, std::string> GetBPRegInfo(u8 cmd, u32 cmddata)
2014-05-03 14:36:29 -06:00
{
// Macro to set the register name and make sure it was written correctly via compile time assertion
#define RegName(reg) ((void)(reg), #reg)
#define DescriptionlessReg(reg) std::make_pair(RegName(reg), "");
switch (cmd)
2014-05-03 14:36:29 -06:00
{
case BPMEM_GENMODE: // 0x00
return std::make_pair(RegName(BPMEM_GENMODE), fmt::to_string(GenMode{.hex = cmddata}));
2014-05-03 14:36:29 -06:00
case BPMEM_DISPLAYCOPYFILTER: // 0x01
case BPMEM_DISPLAYCOPYFILTER + 1:
case BPMEM_DISPLAYCOPYFILTER + 2:
case BPMEM_DISPLAYCOPYFILTER + 3:
2014-05-03 14:36:29 -06:00
// TODO: This is actually the sample pattern used for copies from an antialiased EFB
return DescriptionlessReg(BPMEM_DISPLAYCOPYFILTER);
2014-05-03 14:36:29 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_IND_MTXA: // 0x06
case BPMEM_IND_MTXA + 3:
case BPMEM_IND_MTXA + 6:
return std::make_pair(fmt::format("BPMEM_IND_MTXA Matrix {}", (cmd - BPMEM_IND_MTXA) / 3),
fmt::format("Matrix {} column A\n{}", (cmd - BPMEM_IND_MTXA) / 3,
IND_MTXA{.hex = cmddata}));
case BPMEM_IND_MTXB: // 0x07
case BPMEM_IND_MTXB + 3:
case BPMEM_IND_MTXB + 6:
return std::make_pair(fmt::format("BPMEM_IND_MTXB Matrix {}", (cmd - BPMEM_IND_MTXB) / 3),
fmt::format("Matrix {} column B\n{}", (cmd - BPMEM_IND_MTXB) / 3,
IND_MTXB{.hex = cmddata}));
case BPMEM_IND_MTXC: // 0x08
case BPMEM_IND_MTXC + 3:
2014-06-01 23:51:52 -06:00
case BPMEM_IND_MTXC + 6:
return std::make_pair(fmt::format("BPMEM_IND_MTXC Matrix {}", (cmd - BPMEM_IND_MTXC) / 3),
fmt::format("Matrix {} column C\n{}", (cmd - BPMEM_IND_MTXC) / 3,
IND_MTXC{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_IND_IMASK: // 0x0F
return DescriptionlessReg(BPMEM_IND_IMASK);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_IND_CMD: // 0x10
case BPMEM_IND_CMD + 1:
case BPMEM_IND_CMD + 2:
case BPMEM_IND_CMD + 3:
case BPMEM_IND_CMD + 4:
case BPMEM_IND_CMD + 5:
case BPMEM_IND_CMD + 6:
case BPMEM_IND_CMD + 7:
case BPMEM_IND_CMD + 8:
case BPMEM_IND_CMD + 9:
case BPMEM_IND_CMD + 10:
case BPMEM_IND_CMD + 11:
case BPMEM_IND_CMD + 12:
case BPMEM_IND_CMD + 13:
case BPMEM_IND_CMD + 14:
case BPMEM_IND_CMD + 15:
return std::make_pair(fmt::format("BPMEM_IND_CMD number {}", cmd - BPMEM_IND_CMD),
fmt::to_string(TevStageIndirect{.fullhex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_SCISSORTL: // 0x20
return std::make_pair(RegName(BPMEM_SCISSORTL), fmt::to_string(ScissorPos{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_SCISSORBR: // 0x21
return std::make_pair(RegName(BPMEM_SCISSORBR), fmt::to_string(ScissorPos{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_LINEPTWIDTH: // 0x22
return std::make_pair(RegName(BPMEM_LINEPTWIDTH), fmt::to_string(LPSize{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_PERF0_TRI: // 0x23
return DescriptionlessReg(BPMEM_PERF0_TRI);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_PERF0_QUAD: // 0x24
return DescriptionlessReg(BPMEM_PERF0_QUAD);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_RAS1_SS0: // 0x25
return std::make_pair(RegName(BPMEM_RAS1_SS0),
fmt::to_string(std::make_pair(cmd, TEXSCALE{.hex = cmddata})));
2014-06-01 23:51:52 -06:00
case BPMEM_RAS1_SS1: // 0x26
return std::make_pair(RegName(BPMEM_RAS1_SS1),
fmt::to_string(std::make_pair(cmd, TEXSCALE{.hex = cmddata})));
2014-06-01 23:51:52 -06:00
case BPMEM_IREF: // 0x27
return std::make_pair(RegName(BPMEM_IREF), fmt::to_string(RAS1_IREF{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_TREF: // 0x28
case BPMEM_TREF + 1:
case BPMEM_TREF + 2:
case BPMEM_TREF + 3:
case BPMEM_TREF + 4:
case BPMEM_TREF + 5:
case BPMEM_TREF + 6:
case BPMEM_TREF + 7:
return std::make_pair(fmt::format("BPMEM_TREF number {}", cmd - BPMEM_TREF),
fmt::to_string(std::make_pair(cmd, TwoTevStageOrders{.hex = cmddata})));
2014-06-01 23:51:52 -06:00
case BPMEM_SU_SSIZE: // 0x30
case BPMEM_SU_SSIZE + 2:
case BPMEM_SU_SSIZE + 4:
case BPMEM_SU_SSIZE + 6:
case BPMEM_SU_SSIZE + 8:
case BPMEM_SU_SSIZE + 10:
case BPMEM_SU_SSIZE + 12:
case BPMEM_SU_SSIZE + 14:
return std::make_pair(fmt::format("BPMEM_SU_SSIZE number {}", (cmd - BPMEM_SU_SSIZE) / 2),
fmt::to_string(std::make_pair(true, TCInfo{.hex = cmddata})));
2014-06-01 23:51:52 -06:00
case BPMEM_SU_TSIZE: // 0x31
case BPMEM_SU_TSIZE + 2:
case BPMEM_SU_TSIZE + 4:
case BPMEM_SU_TSIZE + 6:
case BPMEM_SU_TSIZE + 8:
case BPMEM_SU_TSIZE + 10:
case BPMEM_SU_TSIZE + 12:
case BPMEM_SU_TSIZE + 14:
return std::make_pair(fmt::format("BPMEM_SU_TSIZE number {}", (cmd - BPMEM_SU_TSIZE) / 2),
fmt::to_string(std::make_pair(false, TCInfo{.hex = cmddata})));
2014-06-01 23:51:52 -06:00
case BPMEM_ZMODE: // 0x40
return std::make_pair(RegName(BPMEM_ZMODE), fmt::format("Z mode: {}", ZMode{.hex = cmddata}));
2014-05-03 14:36:29 -06:00
case BPMEM_BLENDMODE: // 0x41
return std::make_pair(RegName(BPMEM_BLENDMODE), fmt::to_string(BlendMode{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_CONSTANTALPHA: // 0x42
return std::make_pair(RegName(BPMEM_CONSTANTALPHA),
fmt::to_string(ConstantAlpha{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_ZCOMPARE: // 0x43
return std::make_pair(RegName(BPMEM_ZCOMPARE), fmt::to_string(PEControl{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_FIELDMASK: // 0x44
return std::make_pair(RegName(BPMEM_FIELDMASK), fmt::to_string(FieldMask{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_SETDRAWDONE: // 0x45
return DescriptionlessReg(BPMEM_SETDRAWDONE);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_BUSCLOCK0: // 0x46
return DescriptionlessReg(BPMEM_BUSCLOCK0);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_PE_TOKEN_ID: // 0x47
return DescriptionlessReg(BPMEM_PE_TOKEN_ID);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_PE_TOKEN_INT_ID: // 0x48
return DescriptionlessReg(BPMEM_PE_TOKEN_INT_ID);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_EFB_TL: // 0x49
{
const X10Y10 left_top{.hex = cmddata};
return std::make_pair(RegName(BPMEM_EFB_TL),
fmt::format("EFB Left: {}\nEFB Top: {}", left_top.x, left_top.y));
2014-06-01 23:51:52 -06:00
}
2021-02-07 13:32:45 -07:00
case BPMEM_EFB_WH: // 0x4A
2014-05-03 14:36:29 -06:00
{
const X10Y10 width_height{.hex = cmddata};
return std::make_pair(
RegName(BPMEM_EFB_WH),
fmt::format("EFB Width: {}\nEFB Height: {}", width_height.x + 1, width_height.y + 1));
2014-05-03 14:36:29 -06:00
}
2014-05-03 14:36:29 -06:00
case BPMEM_EFB_ADDR: // 0x4B
return std::make_pair(
RegName(BPMEM_EFB_ADDR),
fmt::format("EFB Target address (32 byte aligned): 0x{:06X}", cmddata << 5));
2014-06-01 23:51:52 -06:00
case BPMEM_MIPMAP_STRIDE: // 0x4D
return DescriptionlessReg(BPMEM_MIPMAP_STRIDE);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-05-03 14:36:29 -06:00
case BPMEM_COPYYSCALE: // 0x4E
return std::make_pair(
RegName(BPMEM_COPYYSCALE),
fmt::format("Y scaling factor (XFB copy only): 0x{:X} ({}, reciprocal {})", cmddata,
static_cast<float>(cmddata) / 256.f, 256.f / static_cast<float>(cmddata)));
2014-05-03 14:36:29 -06:00
case BPMEM_CLEAR_AR: // 0x4F
return std::make_pair(RegName(BPMEM_CLEAR_AR),
fmt::format("Clear color alpha: 0x{:02X}\nClear color red: 0x{:02X}",
(cmddata & 0xFF00) >> 8, cmddata & 0xFF));
2014-05-03 14:36:29 -06:00
case BPMEM_CLEAR_GB: // 0x50
return std::make_pair(RegName(BPMEM_CLEAR_GB),
fmt::format("Clear color green: 0x{:02X}\nClear color blue: 0x{:02X}",
(cmddata & 0xFF00) >> 8, cmddata & 0xFF));
2014-05-03 14:36:29 -06:00
case BPMEM_CLEAR_Z: // 0x51
return std::make_pair(RegName(BPMEM_CLEAR_Z), fmt::format("Clear Z value: 0x{:06X}", cmddata));
2014-05-03 14:36:29 -06:00
case BPMEM_TRIGGER_EFB_COPY: // 0x52
return std::make_pair(RegName(BPMEM_TRIGGER_EFB_COPY),
fmt::to_string(UPE_Copy{.Hex = cmddata}));
2014-05-03 14:36:29 -06:00
case BPMEM_COPYFILTER0: // 0x53
{
const u32 w0 = (cmddata & 0x00003f);
const u32 w1 = (cmddata & 0x000fc0) >> 6;
const u32 w2 = (cmddata & 0x03f000) >> 12;
const u32 w3 = (cmddata & 0xfc0000) >> 18;
return std::make_pair(RegName(BPMEM_COPYFILTER0),
fmt::format("w0: {}\nw1: {}\nw2: {}\nw3: {}", w0, w1, w2, w3));
}
2014-05-03 14:36:29 -06:00
case BPMEM_COPYFILTER1: // 0x54
{
const u32 w4 = (cmddata & 0x00003f);
const u32 w5 = (cmddata & 0x000fc0) >> 6;
const u32 w6 = (cmddata & 0x03f000) >> 12;
// There is no w7
return std::make_pair(RegName(BPMEM_COPYFILTER1),
fmt::format("w4: {}\nw5: {}\nw6: {}", w4, w5, w6));
}
2014-06-01 23:51:52 -06:00
case BPMEM_CLEARBBOX1: // 0x55
return DescriptionlessReg(BPMEM_CLEARBBOX1);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_CLEARBBOX2: // 0x56
return DescriptionlessReg(BPMEM_CLEARBBOX2);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_CLEAR_PIXEL_PERF: // 0x57
return DescriptionlessReg(BPMEM_CLEAR_PIXEL_PERF);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_REVBITS: // 0x58
return DescriptionlessReg(BPMEM_REVBITS);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_SCISSOROFFSET: // 0x59
return std::make_pair(RegName(BPMEM_SCISSOROFFSET),
fmt::to_string(ScissorOffset{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_PRELOAD_ADDR: // 0x60
return DescriptionlessReg(BPMEM_PRELOAD_ADDR);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_PRELOAD_TMEMEVEN: // 0x61
return DescriptionlessReg(BPMEM_PRELOAD_TMEMEVEN);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_PRELOAD_TMEMODD: // 0x62
return DescriptionlessReg(BPMEM_PRELOAD_TMEMODD);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_PRELOAD_MODE: // 0x63
return std::make_pair(RegName(BPMEM_PRELOAD_MODE),
fmt::to_string(BPU_PreloadTileInfo{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_LOADTLUT0: // 0x64
return DescriptionlessReg(BPMEM_LOADTLUT0);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_LOADTLUT1: // 0x65
return DescriptionlessReg(BPMEM_LOADTLUT1);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_TEXINVALIDATE: // 0x66
return DescriptionlessReg(BPMEM_TEXINVALIDATE);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_PERF1: // 0x67
return DescriptionlessReg(BPMEM_PERF1);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_FIELDMODE: // 0x68
return std::make_pair(RegName(BPMEM_FIELDMODE), fmt::to_string(FieldMode{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_BUSCLOCK1: // 0x69
return DescriptionlessReg(BPMEM_BUSCLOCK1);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_TX_SETMODE0: // 0x80
case BPMEM_TX_SETMODE0 + 1:
case BPMEM_TX_SETMODE0 + 2:
case BPMEM_TX_SETMODE0 + 3:
return std::make_pair(fmt::format("BPMEM_TX_SETMODE0 Texture Unit {}", cmd - BPMEM_TX_SETMODE0),
fmt::to_string(TexMode0{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_TX_SETMODE1: // 0x84
case BPMEM_TX_SETMODE1 + 1:
case BPMEM_TX_SETMODE1 + 2:
case BPMEM_TX_SETMODE1 + 3:
return std::make_pair(fmt::format("BPMEM_TX_SETMODE1 Texture Unit {}", cmd - BPMEM_TX_SETMODE1),
fmt::to_string(TexMode1{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_TX_SETIMAGE0: // 0x88
case BPMEM_TX_SETIMAGE0 + 1:
case BPMEM_TX_SETIMAGE0 + 2:
case BPMEM_TX_SETIMAGE0 + 3:
return std::make_pair(
fmt::format("BPMEM_TX_SETIMAGE0 Texture Unit {}", cmd - BPMEM_TX_SETIMAGE0),
fmt::to_string(TexImage0{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_TX_SETIMAGE1: // 0x8C
case BPMEM_TX_SETIMAGE1 + 1:
case BPMEM_TX_SETIMAGE1 + 2:
case BPMEM_TX_SETIMAGE1 + 3:
return std::make_pair(
fmt::format("BPMEM_TX_SETIMAGE1 Texture Unit {}", cmd - BPMEM_TX_SETIMAGE1),
fmt::to_string(TexImage1{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_TX_SETIMAGE2: // 0x90
case BPMEM_TX_SETIMAGE2 + 1:
case BPMEM_TX_SETIMAGE2 + 2:
case BPMEM_TX_SETIMAGE2 + 3:
return std::make_pair(
fmt::format("BPMEM_TX_SETIMAGE2 Texture Unit {}", cmd - BPMEM_TX_SETIMAGE2),
fmt::to_string(TexImage2{.hex = cmddata}));
2014-05-03 14:36:29 -06:00
case BPMEM_TX_SETIMAGE3: // 0x94
case BPMEM_TX_SETIMAGE3 + 1:
case BPMEM_TX_SETIMAGE3 + 2:
case BPMEM_TX_SETIMAGE3 + 3:
return std::make_pair(
fmt::format("BPMEM_TX_SETIMAGE3 Texture Unit {}", cmd - BPMEM_TX_SETIMAGE3),
fmt::to_string(TexImage3{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_TX_SETTLUT: // 0x98
case BPMEM_TX_SETTLUT + 1:
case BPMEM_TX_SETTLUT + 2:
case BPMEM_TX_SETTLUT + 3:
return std::make_pair(fmt::format("BPMEM_TX_SETTLUT Texture Unit {}", cmd - BPMEM_TX_SETTLUT),
fmt::to_string(TexTLUT{.hex = cmddata}));
case BPMEM_TX_SETMODE0_4: // 0xA0
case BPMEM_TX_SETMODE0_4 + 1:
case BPMEM_TX_SETMODE0_4 + 2:
case BPMEM_TX_SETMODE0_4 + 3:
return std::make_pair(
fmt::format("BPMEM_TX_SETMODE0_4 Texture Unit {}", cmd - BPMEM_TX_SETMODE0_4 + 4),
fmt::to_string(TexMode0{.hex = cmddata}));
case BPMEM_TX_SETMODE1_4: // 0xA4
case BPMEM_TX_SETMODE1_4 + 1:
case BPMEM_TX_SETMODE1_4 + 2:
case BPMEM_TX_SETMODE1_4 + 3:
return std::make_pair(
fmt::format("BPMEM_TX_SETMODE1_4 Texture Unit {}", cmd - BPMEM_TX_SETMODE1_4 + 4),
fmt::to_string(TexMode1{.hex = cmddata}));
case BPMEM_TX_SETIMAGE0_4: // 0xA8
case BPMEM_TX_SETIMAGE0_4 + 1:
case BPMEM_TX_SETIMAGE0_4 + 2:
case BPMEM_TX_SETIMAGE0_4 + 3:
return std::make_pair(
fmt::format("BPMEM_TX_SETIMAGE0_4 Texture Unit {}", cmd - BPMEM_TX_SETIMAGE0_4 + 4),
fmt::to_string(TexImage0{.hex = cmddata}));
case BPMEM_TX_SETIMAGE1_4: // 0xAC
case BPMEM_TX_SETIMAGE1_4 + 1:
case BPMEM_TX_SETIMAGE1_4 + 2:
case BPMEM_TX_SETIMAGE1_4 + 3:
return std::make_pair(
fmt::format("BPMEM_TX_SETIMAGE1_4 Texture Unit {}", cmd - BPMEM_TX_SETIMAGE1_4 + 4),
fmt::to_string(TexImage1{.hex = cmddata}));
case BPMEM_TX_SETIMAGE2_4: // 0xB0
case BPMEM_TX_SETIMAGE2_4 + 1:
case BPMEM_TX_SETIMAGE2_4 + 2:
case BPMEM_TX_SETIMAGE2_4 + 3:
return std::make_pair(
fmt::format("BPMEM_TX_SETIMAGE2_4 Texture Unit {}", cmd - BPMEM_TX_SETIMAGE2_4 + 4),
fmt::to_string(TexImage2{.hex = cmddata}));
case BPMEM_TX_SETIMAGE3_4: // 0xB4
case BPMEM_TX_SETIMAGE3_4 + 1:
case BPMEM_TX_SETIMAGE3_4 + 2:
case BPMEM_TX_SETIMAGE3_4 + 3:
return std::make_pair(
fmt::format("BPMEM_TX_SETIMAGE3_4 Texture Unit {}", cmd - BPMEM_TX_SETIMAGE3_4 + 4),
fmt::to_string(TexImage3{.hex = cmddata}));
case BPMEM_TX_SETTLUT_4: // 0xB8
case BPMEM_TX_SETTLUT_4 + 1:
case BPMEM_TX_SETTLUT_4 + 2:
case BPMEM_TX_SETTLUT_4 + 3:
return std::make_pair(
fmt::format("BPMEM_TX_SETTLUT_4 Texture Unit {}", cmd - BPMEM_TX_SETTLUT_4 + 4),
fmt::to_string(TexTLUT{.hex = cmddata}));
2014-05-03 14:36:29 -06:00
case BPMEM_TEV_COLOR_ENV: // 0xC0
case BPMEM_TEV_COLOR_ENV + 2:
case BPMEM_TEV_COLOR_ENV + 4:
case BPMEM_TEV_COLOR_ENV + 6:
2014-05-03 14:36:29 -06:00
case BPMEM_TEV_COLOR_ENV + 8:
case BPMEM_TEV_COLOR_ENV + 10:
case BPMEM_TEV_COLOR_ENV + 12:
case BPMEM_TEV_COLOR_ENV + 14:
case BPMEM_TEV_COLOR_ENV + 16:
case BPMEM_TEV_COLOR_ENV + 18:
case BPMEM_TEV_COLOR_ENV + 20:
case BPMEM_TEV_COLOR_ENV + 22:
case BPMEM_TEV_COLOR_ENV + 24:
case BPMEM_TEV_COLOR_ENV + 26:
case BPMEM_TEV_COLOR_ENV + 28:
case BPMEM_TEV_COLOR_ENV + 30:
return std::make_pair(
fmt::format("BPMEM_TEV_COLOR_ENV Tev stage {}", (cmd - BPMEM_TEV_COLOR_ENV) / 2),
fmt::to_string(TevStageCombiner::ColorCombiner{.hex = cmddata}));
2014-05-03 14:36:29 -06:00
case BPMEM_TEV_ALPHA_ENV: // 0xC1
case BPMEM_TEV_ALPHA_ENV + 2:
case BPMEM_TEV_ALPHA_ENV + 4:
case BPMEM_TEV_ALPHA_ENV + 6:
case BPMEM_TEV_ALPHA_ENV + 8:
case BPMEM_TEV_ALPHA_ENV + 10:
case BPMEM_TEV_ALPHA_ENV + 12:
case BPMEM_TEV_ALPHA_ENV + 14:
case BPMEM_TEV_ALPHA_ENV + 16:
case BPMEM_TEV_ALPHA_ENV + 18:
case BPMEM_TEV_ALPHA_ENV + 20:
case BPMEM_TEV_ALPHA_ENV + 22:
case BPMEM_TEV_ALPHA_ENV + 24:
case BPMEM_TEV_ALPHA_ENV + 26:
case BPMEM_TEV_ALPHA_ENV + 28:
case BPMEM_TEV_ALPHA_ENV + 30:
return std::make_pair(
fmt::format("BPMEM_TEV_ALPHA_ENV Tev stage {}", (cmd - BPMEM_TEV_ALPHA_ENV) / 2),
fmt::to_string(TevStageCombiner::AlphaCombiner{.hex = cmddata}));
case BPMEM_TEV_COLOR_RA: // 0xE0
case BPMEM_TEV_COLOR_RA + 2: // 0xE2
case BPMEM_TEV_COLOR_RA + 4: // 0xE4
case BPMEM_TEV_COLOR_RA + 6: // 0xE6
return std::make_pair(
fmt::format("BPMEM_TEV_COLOR_RA Tev register {}", (cmd - BPMEM_TEV_COLOR_RA) / 2),
fmt::to_string(TevReg::RA{.hex = cmddata}));
case BPMEM_TEV_COLOR_BG: // 0xE1
case BPMEM_TEV_COLOR_BG + 2: // 0xE3
case BPMEM_TEV_COLOR_BG + 4: // 0xE5
case BPMEM_TEV_COLOR_BG + 6: // 0xE7
return std::make_pair(
fmt::format("BPMEM_TEV_COLOR_BG Tev register {}", (cmd - BPMEM_TEV_COLOR_BG) / 2),
fmt::to_string(TevReg::BG{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_FOGRANGE: // 0xE8
return std::make_pair("BPMEM_FOGRANGE Base",
fmt::to_string(FogRangeParams::RangeBase{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_FOGRANGE + 1:
case BPMEM_FOGRANGE + 2:
case BPMEM_FOGRANGE + 3:
case BPMEM_FOGRANGE + 4:
case BPMEM_FOGRANGE + 5:
return std::make_pair(fmt::format("BPMEM_FOGRANGE K element {}", cmd - BPMEM_FOGRANGE),
fmt::to_string(FogRangeKElement{.HEX = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_FOGPARAM0: // 0xEE
return std::make_pair(RegName(BPMEM_FOGPARAM0), fmt::to_string(FogParam0{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_FOGBMAGNITUDE: // 0xEF
return DescriptionlessReg(BPMEM_FOGBMAGNITUDE);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_FOGBEXPONENT: // 0xF0
return DescriptionlessReg(BPMEM_FOGBEXPONENT);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_FOGPARAM3: // 0xF1
return std::make_pair(RegName(BPMEM_FOGPARAM3), fmt::to_string(FogParam3{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_FOGCOLOR: // 0xF2
return std::make_pair(RegName(BPMEM_FOGCOLOR),
fmt::to_string(FogParams::FogColor{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_ALPHACOMPARE: // 0xF3
return std::make_pair(RegName(BPMEM_ALPHACOMPARE), fmt::to_string(AlphaTest{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_BIAS: // 0xF4
return DescriptionlessReg(BPMEM_BIAS);
2014-06-01 23:51:52 -06:00
// TODO: Description
2014-06-01 23:51:52 -06:00
case BPMEM_ZTEX2: // 0xF5
return std::make_pair(RegName(BPMEM_ZTEX2), fmt::to_string(ZTex2{.hex = cmddata}));
2014-06-01 23:51:52 -06:00
case BPMEM_TEV_KSEL: // 0xF6
case BPMEM_TEV_KSEL + 1:
case BPMEM_TEV_KSEL + 2:
case BPMEM_TEV_KSEL + 3:
case BPMEM_TEV_KSEL + 4:
case BPMEM_TEV_KSEL + 5:
case BPMEM_TEV_KSEL + 6:
case BPMEM_TEV_KSEL + 7:
return std::make_pair(fmt::format("BPMEM_TEV_KSEL number {}", cmd - BPMEM_TEV_KSEL),
fmt::to_string(std::make_pair(cmd, TevKSel{.hex = cmddata})));
case BPMEM_BP_MASK: // 0xFE
return std::make_pair(RegName(BPMEM_BP_MASK),
fmt::format("The next BP command will only update these bits; others "
"will retain their prior values: {:06x}",
cmddata));
default:
return std::make_pair(fmt::format("Unknown BP Reg: {:02x}={:06x}", cmd, cmddata), "");
2014-05-03 14:36:29 -06:00
#undef DescriptionlessReg
#undef RegName
2014-05-03 14:36:29 -06:00
}
}
// Called when loading a saved state.
void BPReload()
{
// restore anything that goes straight to the renderer.
// let's not risk actually replaying any writes.
// note that PixelShaderManager is already covered since it has its own DoState.
SetGenerationMode();
SetScissorAndViewport();
SetDepthMode();
SetBlendMode();
OnPixelFormatChange();
}