cpmadevs/cnq3
0
0
Fork 0
mirror of https://bitbucket.org/CPMADevs/cnq3 synced 2024-11-10 06:31:48 +00:00
Code Issues Projects Releases Packages Wiki Activity

D3D11 tweaks, added r_d3d11_presentMode, removed r_d3d11_maxQueuedFrames

Browse source 
the D3D11 tweaks:
- better error message formatting
- D3DDDIERR_DEVICEREMOVED from Present is a fatal error too
- synchronized offsets is always the automatic behavior for now
  it turns out that split mode isn't always the fastest for nVidia
This commit is contained in:
myT 2020-02-24 01:48:43 +01:00
parent 85583acc9c
commit 36f363b47a
7 changed files with 3197 additions and 56 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

13
changelog.txt
View file

@ -62,13 +62,22 @@ add: r_gl3_geoStream <0 to 4> (default: 0) sets the geometry upload strategy
r_gl3_geoStream 2 = Map Unsychronized
r_gl3_geoStream 3 = Map Persistent when possible, else automatic
r_gl3_geoStream 4 = AMD pinned memory when available, else automatic
this only applies to the GL3 rendering back-end
add: r_d3d11_syncOffsets <0|1|2> (default: 2) synchronizes vertex buffer offsets
r_d3d11_syncOffsets 0 = split buffer offsets (more data transferred)
r_d3d11_syncOffsets 1 = sync'd buffer offsets (more API calls)
r_d3d11_syncOffsets 2 = decide automatically
this only applies to the D3D11 rendering back-end
add: r_d3d11_maxQueuedFrames <1 to 16> (default: 3) sets the maximum pre-rendered frame count
add: r_d3d11_presentMode <0|1> (default: 0) sets the Direct3D presentation mode
r_d3d11_presentMode 0 = blit + discard
r_d3d11_presentMode 1 = flip + discard when available, else blit + discard
about the 'flip + discard' presentation mode:
- Windows 10 is required and tearing must be supported
- it should help performance by sharing buffer data (less copying)
- it allows for V-Sync to be disabled with variable refresh rate displays
- performance issues were observed on some machines, so it's off by default
add: r_khr_debug <0|1|2> (default: 2) decides whether to create an OpenGL debug context
r_khr_debug 0 = forced OFF
@ -126,7 +135,7 @@ chg: r_lightmap is now latched again
chg: r_fullbright is now latched again
it now also supports surfaces referencing q3map2's "external lightmap" atlas images
chg: negative r_swapInterval values will request adaptive v-sync when using an OpenGL back-end
chg: negative r_swapInterval values will request adaptive V-Sync when using an OpenGL back-end
fix: crash due to lack of memory for the sound system

1494
code/renderer/dxgi/dxgi1_4.h Normal file
View file

File diff suppressed because it is too large Load diff

1545
code/renderer/dxgi/dxgi1_5.h Normal file
View file

File diff suppressed because it is too large Load diff

168
code/renderer/tr_backend_d3d11.cpp
View file

@ -27,6 +27,30 @@ along with Challenge Quake 3. If not, see <https://www.gnu.org/licenses/>.
#include <Windows.h>
#include <d3d11.h>
#include <dxgi.h>
#include <dxgi1_2.h>
#include <dxgi1_3.h>
#pragma region Windows 10 SDK
#if !defined(__dxgicommon_h__)
enum DXGI_COLOR_SPACE_TYPE;
#endif
#include "dxgi/dxgi1_4.h"
#include "dxgi/dxgi1_5.h"
#if !defined(DXGI_PRESENT_ALLOW_TEARING)
#define DXGI_PRESENT_ALLOW_TEARING 0x00000200UL
#endif
#if !defined(DXGI_SWAP_EFFECT_FLIP_DISCARD)
#define DXGI_SWAP_EFFECT_FLIP_DISCARD ((DXGI_SWAP_EFFECT)4)
#endif
#if !defined(DXGI_SWAP_CHAIN_FLAG_ALLOW_TEARING)
#define DXGI_SWAP_CHAIN_FLAG_ALLOW_TEARING ((DXGI_SWAP_CHAIN_FLAG)2048)
#endif
#pragma endregion
#include "hlsl/generic_vs.h"
#include "hlsl/generic_ps.h"
@ -66,6 +90,10 @@ static ShaderDesc genericPixelShaders[4] =
# undef far
#endif
#if !defined(D3DDDIERR_DEVICEREMOVED)
# define D3DDDIERR_DEVICEREMOVED ((HRESULT)0x88760870L)
#endif
#define MAX_GPU_TEXTURE_SIZE 2048 // instead of D3D11_REQ_TEXTURE2D_U_OR_V_DIMENSION
#define BLEND_STATE_COUNT (D3D11_BLEND_SRC_ALPHA_SAT + 1)
@ -80,7 +108,9 @@ Current info:
To look at:
- near clip plane seems to be further in the GL2 back-end in 3D land
- find/use D3DDDIERR_DEVICEREMOVED? are the docs correct about it?
Known issues:
- device re-creation isn't handled by OBS' capture plug-in
*/
@ -358,6 +388,7 @@ struct Direct3DStatic
IDXGISwapChain* swapChain;
HMODULE library;
qbool flipAndTear;
AdapterInfo adapterInfo;
};
@ -383,8 +414,25 @@ static const char* GetSystemErrorString(HRESULT hr)
systemErrorStr, sizeof(systemErrorStr) - 1, NULL);
if(written == 0)
{
// we have nothing valid
Q_strncpyz(systemErrorStr, "???", sizeof(systemErrorStr));
}
else
{
// remove the trailing whitespace
char* s = systemErrorStr + strlen(systemErrorStr) - 1;
while(s >= systemErrorStr)
{
if(*s == '\r' || *s == '\n' || *s == '\t' || *s == ' ')
{
*s-- = '\0';
}
else
{
break;
}
}
}
return systemErrorStr;
}
@ -398,7 +446,7 @@ static qbool Check(HRESULT hr, const char* function)
if(d3d.errorMode == EM_FATAL)
{
ri.Error(ERR_FATAL, va("'%s' failed with code 0x%08X (%s)", function, (unsigned int)hr, GetSystemErrorString(hr)));
ri.Error(ERR_FATAL, "'%s' failed with code 0x%08X (%s)\n", function, (unsigned int)hr, GetSystemErrorString(hr));
}
return qfalse;
}
@ -413,7 +461,7 @@ static qbool CheckAndName(HRESULT hr, const char* function, ID3D11DeviceChild* r
if(d3d.errorMode == EM_FATAL)
{
ri.Error(ERR_FATAL, va("'%s' failed to create '%s' with code 0x%08X (%s)", function, resourceName, (unsigned int)hr, GetSystemErrorString(hr)));
ri.Error(ERR_FATAL, "'%s' failed to create '%s' with code 0x%08X (%s)\n", function, resourceName, (unsigned int)hr, GetSystemErrorString(hr));
}
return qfalse;
}
@ -574,12 +622,8 @@ static DXGI_FORMAT GetRenderTargetColorFormat(int format)
static void ResetShaderData(ID3D11Resource* buffer, const void* data, size_t bytes)
{
D3D11_MAPPED_SUBRESOURCE ms;
HRESULT hr = d3ds.context->Map(buffer, 0, D3D11_MAP_WRITE_DISCARD, NULL, &ms);
if(FAILED(hr))
{
ri.Error(ERR_FATAL, "Map failed");
return;
}
const HRESULT hr = d3ds.context->Map(buffer, 0, D3D11_MAP_WRITE_DISCARD, NULL, &ms);
Check(hr, "Map on shader data");
memcpy(ms.pData, data, bytes);
d3ds.context->Unmap(buffer, NULL);
}
@ -597,12 +641,8 @@ static void AppendVertexData(VertexBuffer* buffer, const void* data, int itemCou
if(data != NULL || mapType == D3D11_MAP_WRITE_DISCARD)
{
D3D11_MAPPED_SUBRESOURCE ms;
HRESULT hr = d3ds.context->Map(buffer->buffer, 0, mapType, NULL, &ms);
if(FAILED(hr))
{
ri.Error(ERR_FATAL, "Map failed");
return;
}
const HRESULT hr = d3ds.context->Map(buffer->buffer, 0, mapType, NULL, &ms);
Check(hr, "Map on vertex data");
if(data != NULL)
{
memcpy((byte*)ms.pData + buffer->writeIndex * buffer->itemSize, data, itemCount * buffer->itemSize);
@ -1043,6 +1083,52 @@ static void FindBestAvailableAA(DXGI_SAMPLE_DESC* sampleDesc)
}
}
static qbool CheckFlipAndTearSupport()
{
if(r_d3d11_presentMode->integer != DXGIPM_FLIPDISCARD)
{
return qfalse;
}
HMODULE library = LoadLibraryA("DXGI.dll");
if(library == NULL)
{
ri.Printf(PRINT_WARNING, "CheckTearingSupport: DXGI.dll couldn't be found or opened\n");
return qfalse;
}
typedef HRESULT (WINAPI *PFN_CreateDXGIFactory)(REFIID riid, _Out_ void **ppFactory);
PFN_CreateDXGIFactory pCreateDXGIFactory = (PFN_CreateDXGIFactory)GetProcAddress(library, "CreateDXGIFactory");
if(pCreateDXGIFactory == NULL)
{
FreeLibrary(library);
ri.Printf(PRINT_WARNING, "CheckTearingSupport: Failed to locate CreateDXGIFactory in DXGI.dll\n");
return qfalse;
}
HRESULT hr;
BOOL enabled = FALSE;
IDXGIFactory5* pFactory;
hr = (*pCreateDXGIFactory)(__uuidof(IDXGIFactory5), (void**)&pFactory);
if(FAILED(hr))
{
FreeLibrary(library);
ri.Printf(PRINT_WARNING, "CheckTearingSupport: 'CreateDXGIFactory' failed with code 0x%08X (%s)\n", (unsigned int)hr, GetSystemErrorString(hr));
return qfalse;
}
hr = pFactory->CheckFeatureSupport(DXGI_FEATURE_PRESENT_ALLOW_TEARING, &enabled, sizeof(enabled));
pFactory->Release();
FreeLibrary(library);
if(FAILED(hr))
{
ri.Printf(PRINT_WARNING, "CheckTearingSupport: 'IDXGIFactory5::CheckFeatureSupport' failed with code 0x%08X (%s)\n", (unsigned int)hr, GetSystemErrorString(hr));
return qfalse;
}
return enabled != 0;
}
static qbool GAL_Init()
{
Sys_V_Init(GAL_D3D11);
@ -1071,8 +1157,9 @@ static qbool GAL_Init()
flags |= D3D11_CREATE_DEVICE_DEBUG;
#endif
d3ds.flipAndTear = CheckFlipAndTearSupport();
ZeroMemory(&swapChainDesc, sizeof(swapChainDesc));
swapChainDesc.BufferCount = 1;
swapChainDesc.BufferDesc.Width = glInfo.winWidth;
swapChainDesc.BufferDesc.Height = glInfo.winHeight;
swapChainDesc.BufferDesc.Format = DXGI_FORMAT_R8G8B8A8_UNORM;
@ -1085,11 +1172,22 @@ static qbool GAL_Init()
swapChainDesc.Windowed = TRUE;
swapChainDesc.BufferDesc.ScanlineOrdering = DXGI_MODE_SCANLINE_ORDER_UNSPECIFIED;
swapChainDesc.BufferDesc.Scaling = DXGI_MODE_SCALING_UNSPECIFIED;
if(d3ds.flipAndTear)
{
// flip and tear, until it is done
swapChainDesc.BufferCount = 2;
swapChainDesc.SwapEffect = DXGI_SWAP_EFFECT_FLIP_DISCARD;
swapChainDesc.Flags = DXGI_SWAP_CHAIN_FLAG_ALLOW_TEARING;
}
else
{
swapChainDesc.BufferCount = 1;
swapChainDesc.SwapEffect = DXGI_SWAP_EFFECT_DISCARD;
swapChainDesc.Flags = 0;
}
create_device:
HRESULT hr = (*pD3D11CreateDeviceAndSwapChain)(
hr = (*pD3D11CreateDeviceAndSwapChain)(
NULL, D3D_DRIVER_TYPE_HARDWARE, NULL, flags, featureLevels, ARRAY_LEN(featureLevels), D3D11_SDK_VERSION,
&swapChainDesc, &d3ds.swapChain, &d3ds.device, NULL, &d3ds.context);
if(hr == DXGI_ERROR_SDK_COMPONENT_MISSING)
@ -1542,7 +1640,6 @@ static qbool GAL_Init()
textureDesc.MiscFlags = 0;
mipGenOK &= D3D11_CreateTexture2D(&textureDesc, 0, &d3d.mipGenTextures[i].texture, va("mip-map generation texture #%d", i + 1));
D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc;
ZeroMemory(&srvDesc, sizeof(srvDesc));
srvDesc.Format = textureDesc.Format;
srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D;
@ -1630,30 +1727,15 @@ static qbool GAL_Init()
}
}
qbool maxFrameLatencySet = qfalse;
IDXGIDevice1* dxgiDevice1;
if(SUCCEEDED(d3ds.device->QueryInterface(__uuidof(IDXGIDevice1), (void**)&dxgiDevice1)))
{
if(SUCCEEDED(dxgiDevice1->SetMaximumFrameLatency(r_d3d11_maxQueuedFrames->integer)))
{
maxFrameLatencySet = qtrue;
}
COM_RELEASE(dxgiDevice1);
}
if(maxFrameLatencySet)
{
ri.Printf(PRINT_ALL, "Max. queued frames: %d\n", r_d3d11_maxQueuedFrames->integer);
}
else
{
ri.Printf(PRINT_ERROR, "Failed to set the max. number of queued frames\n");
}
if(r_d3d11_syncOffsets->integer == D3D11SO_AUTO)
{
#if 0
// only nVidia's drivers seem to consistently handle the extra IASetVertexBuffers calls well enough
d3d.splitBufferOffsets = Q_stristr(glConfig.renderer_string, "NVIDIA") != NULL;
#else
// however, we'll just treat all drivers as equally dumb by default for now
d3d.splitBufferOffsets = D3D11SO_SYNCEDOFFSETS;
#endif
}
else
{
@ -1870,10 +1952,9 @@ static void GAL_EndFrame()
EndQueries();
// Present interval flags
// flags: DXGI_PRESENT_DO_NOT_SEQUENCE DXGI_PRESENT_ALLOW_TEARING
const HRESULT hr = d3ds.swapChain->Present(abs(r_swapInterval->integer), 0);
if(hr == DXGI_ERROR_DEVICE_REMOVED)
const UINT presentFlags = d3ds.flipAndTear && r_swapInterval->integer == 0 ? DXGI_PRESENT_ALLOW_TEARING : 0;
const HRESULT hr = d3ds.swapChain->Present(abs(r_swapInterval->integer), presentFlags);
if(hr == DXGI_ERROR_DEVICE_REMOVED || hr == D3DDDIERR_DEVICEREMOVED)
{
ri.Error(ERR_FATAL, "Direct3D device was removed! Reason: %s", GetDeviceRemovedReason());
}
@ -1917,7 +1998,7 @@ static void WriteInvalidImage(int w, int h, int alignment, colorSpace_t colorSpa
memset(out, 0x7F, PAD(w * 3, alignment) * h);
}
static void GAL_ReadPixels(int x, int y, int w, int h, int alignment, colorSpace_t colorSpace, void* out)
static void GAL_ReadPixels(int, int, int w, int h, int alignment, colorSpace_t colorSpace, void* out)
{
if(d3d.readbackTexture == NULL)
{
@ -2445,6 +2526,7 @@ static void GAL_PrintInfo()
{
ri.Printf(PRINT_ALL, "Direct3D device feature level: %s\n", d3ds.device->GetFeatureLevel() == D3D_FEATURE_LEVEL_11_0 ? "11.0" : "10.1");
ri.Printf(PRINT_ALL, "Direct3D vertex buffer upload strategy: %s\n", d3d.splitBufferOffsets ? "split offsets" : "sync'd offsets");
ri.Printf(PRINT_ALL, "DXGI presentation model: %s\n", d3ds.flipAndTear ? "flip + discard" : "blit + discard");
if(d3ds.adapterInfo.valid)
{
ri.Printf(PRINT_ALL, "%6d MB of dedicated GPU memory\n", d3ds.adapterInfo.dedicatedVideoMemoryMB);

12
code/renderer/tr_help.h
View file

@ -20,9 +20,15 @@ S_COLOR_VAL " 0 " S_COLOR_HELP "= Split buffer offsets (more API calls)\n" \
S_COLOR_VAL " 1 " S_COLOR_HELP "= Sync'd buffer offsets (more data transferred)\n" \
S_COLOR_VAL " 2 " S_COLOR_HELP "= Decide automatically"
#define help_r_d3d11_maxQueuedFrames \
"max. pre-rendered frames\n" \
"The number of frames that the system is allowed to queue for rendering."
#define help_r_d3d11_presentMode \
"presentation mode\n" \
S_COLOR_VAL " 0 " S_COLOR_HELP "= Blit + Discard\n" \
S_COLOR_VAL " 1 " S_COLOR_HELP "= Flip + Discard when available, else Blit + Discard\n" \
"About the 'Flip + Discard' presentation mode:\n" \
"- Windows 10 is required and tearing must be supported.\n" \
"- It should help performance by sharing buffer data (less copying).\n" \
"- It allows for V-Sync to be disabled with variable refresh rate displays.\n" \
"- Performance issues were observed on some machines, so it's off by default."
#define help_r_ext_max_anisotropy \
"max. allowed anisotropy ratio\n" \

4
code/renderer/tr_init.cpp
View file

@ -75,7 +75,7 @@ cvar_t *r_mipGenGamma;
cvar_t *r_noiseScale;
cvar_t *r_gl3_geoStream;
cvar_t *r_d3d11_syncOffsets;
cvar_t *r_d3d11_maxQueuedFrames;
cvar_t *r_d3d11_presentMode;
cvar_t *r_ext_max_anisotropy;
cvar_t *r_msaa;
@ -367,7 +367,7 @@ static const cvarTableItem_t r_cvars[] =
{ &r_mipGenGamma, "r_mipGenGamma", "1.8", CVAR_ARCHIVE | CVAR_LATCH, CVART_FLOAT, "1.0", "3.0", help_r_mipGenGamma },
{ &r_gl3_geoStream, "r_gl3_geoStream", "0", CVAR_ARCHIVE | CVAR_LATCH, CVART_INTEGER, "0", XSTRING(GL3MAP_MAX), help_r_gl3_geoStream },
{ &r_d3d11_syncOffsets, "r_d3d11_syncOffsets", "2", CVAR_ARCHIVE | CVAR_LATCH, CVART_INTEGER, "0", XSTRING(D3D11SO_MAX), help_r_d3d11_syncOffsets },
{ &r_d3d11_maxQueuedFrames, "r_d3d11_maxQueuedFrames", "3", CVAR_ARCHIVE | CVAR_LATCH, CVART_INTEGER, "1", "16", help_r_d3d11_maxQueuedFrames },
{ &r_d3d11_presentMode, "r_d3d11_presentMode", "0", CVAR_ARCHIVE | CVAR_LATCH, CVART_INTEGER, "0", XSTRING(DXGIPM_MAX), help_r_d3d11_presentMode },
{ &r_ext_max_anisotropy, "r_ext_max_anisotropy", "16", CVAR_ARCHIVE | CVAR_LATCH, CVART_INTEGER, "0", "16", help_r_ext_max_anisotropy },
{ &r_msaa, "r_msaa", "0", CVAR_ARCHIVE | CVAR_LATCH, CVART_INTEGER, "0", "32", "anti-aliasing sample count, " S_COLOR_VAL "0" S_COLOR_HELP "=off" },
{ &r_picmip, "r_picmip", "0", CVAR_ARCHIVE | CVAR_LATCH, CVART_INTEGER, "0", "16", help_r_picmip },

13
code/renderer/tr_local.h
View file

@ -961,6 +961,11 @@ extern trGlobals_t tr;
#define D3D11SO_AUTO 2
#define D3D11SO_MAX 2
// r_d3d11_presentMode
#define DXGIPM_BLITDISCARD 0
#define DXGIPM_FLIPDISCARD 1
#define DXGIPM_MAX 1
extern cvar_t *r_backend;
extern cvar_t *r_verbose; // used for verbose debug spew
@ -997,13 +1002,13 @@ extern cvar_t *r_softSprites; // draws certain surfaces as depth particles
extern cvar_t *r_gpuMipGen; // uses GPU-side mip-map generation
extern cvar_t *r_alphaToCoverage; // enables A2C on alpha-tested surfaces
extern cvar_t *r_dither; // enables dithering
extern cvar_t *r_rtColorFormat; // see RTCF_*
extern cvar_t *r_rtColorFormat; // color render target format, see RTCF_*
extern cvar_t *r_mipGenFilter; // if the string is invalid, Lanczos 4 is used
extern cvar_t *r_mipGenGamma; // what gamma-space do we consider the textures to be in
extern cvar_t *r_gl3_geoStream; // vertex/index streaming strategy
extern cvar_t *r_d3d11_syncOffsets; // vertex attribute streaming strategy
extern cvar_t *r_d3d11_maxQueuedFrames; // max. pre-rendered frames
extern cvar_t *r_gl3_geoStream; // vertex/index streaming strategy, see GL3MAP_*
extern cvar_t *r_d3d11_syncOffsets; // vertex attribute streaming strategy, see D3D11SO_*
extern cvar_t *r_d3d11_presentMode; // DXGI presentation model, see DXGIPM_*
extern cvar_t *r_ext_max_anisotropy;
extern cvar_t *r_msaa;