gecko-dev/gfx/thebes/gfxWindowsPlatform.cpp

2928 строки
90 KiB
C++
Исходник Обычный вид История

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*- */
/* vim: set ts=8 sts=4 et sw=4 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
2012-05-21 15:12:37 +04:00
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
2006-02-01 05:35:38 +03:00
#include "gfxWindowsPlatform.h"
#include "cairo.h"
#include "mozilla/ArrayUtils.h"
2006-02-01 05:35:38 +03:00
#include "gfxImageSurface.h"
#include "gfxWindowsSurface.h"
#include "nsUnicharUtils.h"
#include "mozilla/Preferences.h"
#include "mozilla/Services.h"
#include "mozilla/Snprintf.h"
#include "mozilla/WindowsVersion.h"
#include "nsServiceManagerUtils.h"
#include "nsTArray.h"
#include "mozilla/Telemetry.h"
#include "GeckoProfiler.h"
#include "nsIWindowsRegKey.h"
#include "nsIFile.h"
#include "plbase64.h"
#include "nsIXULRuntime.h"
#include "imgLoader.h"
#include "nsIGfxInfo.h"
#include "GfxDriverInfo.h"
#include "gfxCrashReporterUtils.h"
#include "gfxGDIFontList.h"
#include "gfxGDIFont.h"
#include "mozilla/layers/CompositorParent.h" // for CompositorParent::IsInCompositorThread
#include "DeviceManagerD3D9.h"
#include "mozilla/layers/ReadbackManagerD3D11.h"
#include "WinUtils.h"
#include "gfxDWriteFontList.h"
#include "gfxDWriteFonts.h"
#include "gfxDWriteCommon.h"
#include <dwrite.h>
#include "gfxTextRun.h"
#include "gfxUserFontSet.h"
#include "nsWindowsHelpers.h"
#include "gfx2DGlue.h"
#include <string>
#include <d3d10_1.h>
#include "mozilla/gfx/2D.h"
#include "nsMemory.h"
#include <d3d11.h>
#include "nsIMemoryReporter.h"
#include <winternl.h>
#include "d3dkmtQueryStatistics.h"
#include "SurfaceCache.h"
#include "gfxPrefs.h"
#include "VsyncSource.h"
#include "DriverCrashGuard.h"
#include "mozilla/dom/ContentParent.h"
using namespace mozilla;
using namespace mozilla::gfx;
using namespace mozilla::layers;
using namespace mozilla::widget;
using namespace mozilla::image;
DCFromDrawTarget::DCFromDrawTarget(DrawTarget& aDrawTarget)
{
mDC = nullptr;
if (aDrawTarget.GetBackendType() == BackendType::CAIRO) {
cairo_t* ctx = static_cast<cairo_t*>
(aDrawTarget.GetNativeSurface(NativeSurfaceType::CAIRO_CONTEXT));
if (ctx) {
cairo_surface_t* surf = cairo_get_group_target(ctx);
if (surf) {
cairo_surface_type_t surfaceType = cairo_surface_get_type(surf);
if (surfaceType == CAIRO_SURFACE_TYPE_WIN32 ||
surfaceType == CAIRO_SURFACE_TYPE_WIN32_PRINTING) {
mDC = cairo_win32_surface_get_dc(surf);
mNeedsRelease = false;
SaveDC(mDC);
cairo_scaled_font_t* scaled = cairo_get_scaled_font(ctx);
cairo_win32_scaled_font_select_font(scaled, mDC);
}
}
}
}
if (!mDC) {
// Get the whole screen DC:
mDC = GetDC(nullptr);
SetGraphicsMode(mDC, GM_ADVANCED);
mNeedsRelease = true;
}
}
static const char *kFeatureLevelPref =
"gfx.direct3d.last_used_feature_level_idx";
static const int kSupportedFeatureLevels[] =
{ D3D10_FEATURE_LEVEL_10_1, D3D10_FEATURE_LEVEL_10_0 };
class GfxD2DVramReporter final : public nsIMemoryReporter
{
~GfxD2DVramReporter() {}
public:
NS_DECL_ISUPPORTS
NS_IMETHOD CollectReports(nsIHandleReportCallback* aHandleReport,
nsISupports* aData, bool aAnonymize)
{
nsresult rv;
rv = MOZ_COLLECT_REPORT(
"gfx-d2d-vram-draw-target", KIND_OTHER, UNITS_BYTES,
Factory::GetD2DVRAMUsageDrawTarget(),
"Video memory used by D2D DrawTargets.");
NS_ENSURE_SUCCESS(rv, rv);
rv = MOZ_COLLECT_REPORT(
"gfx-d2d-vram-source-surface", KIND_OTHER, UNITS_BYTES,
Factory::GetD2DVRAMUsageSourceSurface(),
"Video memory used by D2D SourceSurfaces.");
NS_ENSURE_SUCCESS(rv, rv);
return NS_OK;
}
};
NS_IMPL_ISUPPORTS(GfxD2DVramReporter, nsIMemoryReporter)
#define GFX_USE_CLEARTYPE_ALWAYS "gfx.font_rendering.cleartype.always_use_for_content"
#define GFX_DOWNLOADABLE_FONTS_USE_CLEARTYPE "gfx.font_rendering.cleartype.use_for_downloadable_fonts"
#define GFX_CLEARTYPE_PARAMS "gfx.font_rendering.cleartype_params."
#define GFX_CLEARTYPE_PARAMS_GAMMA "gfx.font_rendering.cleartype_params.gamma"
#define GFX_CLEARTYPE_PARAMS_CONTRAST "gfx.font_rendering.cleartype_params.enhanced_contrast"
#define GFX_CLEARTYPE_PARAMS_LEVEL "gfx.font_rendering.cleartype_params.cleartype_level"
#define GFX_CLEARTYPE_PARAMS_STRUCTURE "gfx.font_rendering.cleartype_params.pixel_structure"
#define GFX_CLEARTYPE_PARAMS_MODE "gfx.font_rendering.cleartype_params.rendering_mode"
class GPUAdapterReporter final : public nsIMemoryReporter
{
// Callers must Release the DXGIAdapter after use or risk mem-leak
static bool GetDXGIAdapter(IDXGIAdapter **aDXGIAdapter)
{
ID3D11Device *d3d11Device;
IDXGIDevice *dxgiDevice;
bool result = false;
if ((d3d11Device = mozilla::gfx::Factory::GetDirect3D11Device())) {
if (d3d11Device->QueryInterface(__uuidof(IDXGIDevice), (void **)&dxgiDevice) == S_OK) {
result = (dxgiDevice->GetAdapter(aDXGIAdapter) == S_OK);
dxgiDevice->Release();
}
}
return result;
}
~GPUAdapterReporter() {}
public:
NS_DECL_ISUPPORTS
NS_IMETHOD
CollectReports(nsIMemoryReporterCallback* aCb,
nsISupports* aClosure, bool aAnonymize)
{
HANDLE ProcessHandle = GetCurrentProcess();
int64_t dedicatedBytesUsed = 0;
int64_t sharedBytesUsed = 0;
int64_t committedBytesUsed = 0;
IDXGIAdapter *DXGIAdapter;
HMODULE gdi32Handle;
PFND3DKMTQS queryD3DKMTStatistics = nullptr;
// GPU memory reporting is not available before Windows 7
if (!IsWin7OrLater())
return NS_OK;
if ((gdi32Handle = LoadLibrary(TEXT("gdi32.dll"))))
queryD3DKMTStatistics = (PFND3DKMTQS)GetProcAddress(gdi32Handle, "D3DKMTQueryStatistics");
if (queryD3DKMTStatistics && GetDXGIAdapter(&DXGIAdapter)) {
// Most of this block is understood thanks to wj32's work on Process Hacker
DXGI_ADAPTER_DESC adapterDesc;
D3DKMTQS queryStatistics;
DXGIAdapter->GetDesc(&adapterDesc);
DXGIAdapter->Release();
memset(&queryStatistics, 0, sizeof(D3DKMTQS));
queryStatistics.Type = D3DKMTQS_PROCESS;
queryStatistics.AdapterLuid = adapterDesc.AdapterLuid;
queryStatistics.hProcess = ProcessHandle;
if (NT_SUCCESS(queryD3DKMTStatistics(&queryStatistics))) {
committedBytesUsed = queryStatistics.QueryResult.ProcessInfo.SystemMemory.BytesAllocated;
}
memset(&queryStatistics, 0, sizeof(D3DKMTQS));
queryStatistics.Type = D3DKMTQS_ADAPTER;
queryStatistics.AdapterLuid = adapterDesc.AdapterLuid;
if (NT_SUCCESS(queryD3DKMTStatistics(&queryStatistics))) {
ULONG i;
ULONG segmentCount = queryStatistics.QueryResult.AdapterInfo.NbSegments;
for (i = 0; i < segmentCount; i++) {
memset(&queryStatistics, 0, sizeof(D3DKMTQS));
queryStatistics.Type = D3DKMTQS_SEGMENT;
queryStatistics.AdapterLuid = adapterDesc.AdapterLuid;
queryStatistics.QuerySegment.SegmentId = i;
if (NT_SUCCESS(queryD3DKMTStatistics(&queryStatistics))) {
bool aperture;
// SegmentInformation has a different definition in Win7 than later versions
if (!IsWin8OrLater())
aperture = queryStatistics.QueryResult.SegmentInfoWin7.Aperture;
else
aperture = queryStatistics.QueryResult.SegmentInfoWin8.Aperture;
memset(&queryStatistics, 0, sizeof(D3DKMTQS));
queryStatistics.Type = D3DKMTQS_PROCESS_SEGMENT;
queryStatistics.AdapterLuid = adapterDesc.AdapterLuid;
queryStatistics.hProcess = ProcessHandle;
queryStatistics.QueryProcessSegment.SegmentId = i;
if (NT_SUCCESS(queryD3DKMTStatistics(&queryStatistics))) {
ULONGLONG bytesCommitted;
if (!IsWin8OrLater())
bytesCommitted = queryStatistics.QueryResult.ProcessSegmentInfo.Win7.BytesCommitted;
else
bytesCommitted = queryStatistics.QueryResult.ProcessSegmentInfo.Win8.BytesCommitted;
if (aperture)
sharedBytesUsed += bytesCommitted;
else
dedicatedBytesUsed += bytesCommitted;
}
}
}
}
}
FreeLibrary(gdi32Handle);
#define REPORT(_path, _amount, _desc) \
do { \
nsresult rv; \
rv = aCb->Callback(EmptyCString(), NS_LITERAL_CSTRING(_path), \
KIND_OTHER, UNITS_BYTES, _amount, \
NS_LITERAL_CSTRING(_desc), aClosure); \
NS_ENSURE_SUCCESS(rv, rv); \
} while (0)
REPORT("gpu-committed", committedBytesUsed,
"Memory committed by the Windows graphics system.");
REPORT("gpu-dedicated", dedicatedBytesUsed,
"Out-of-process memory allocated for this process in a "
"physical GPU adapter's memory.");
REPORT("gpu-shared", sharedBytesUsed,
"In-process memory that is shared with the GPU.");
#undef REPORT
return NS_OK;
}
};
NS_IMPL_ISUPPORTS(GPUAdapterReporter, nsIMemoryReporter)
Atomic<size_t> gfxWindowsPlatform::sD3D11MemoryUsed;
class D3D11TextureReporter final : public nsIMemoryReporter
{
~D3D11TextureReporter() {}
public:
NS_DECL_ISUPPORTS
NS_IMETHOD CollectReports(nsIHandleReportCallback *aHandleReport,
nsISupports* aData, bool aAnonymize) override
{
return MOZ_COLLECT_REPORT("d3d11-shared-textures", KIND_OTHER, UNITS_BYTES,
gfxWindowsPlatform::sD3D11MemoryUsed,
"Memory used for D3D11 shared textures");
}
};
NS_IMPL_ISUPPORTS(D3D11TextureReporter, nsIMemoryReporter)
Atomic<size_t> gfxWindowsPlatform::sD3D9MemoryUsed;
class D3D9TextureReporter final : public nsIMemoryReporter
{
~D3D9TextureReporter() {}
public:
NS_DECL_ISUPPORTS
NS_IMETHOD CollectReports(nsIHandleReportCallback *aHandleReport,
nsISupports* aData, bool aAnonymize) override
{
return MOZ_COLLECT_REPORT("d3d9-shared-textures", KIND_OTHER, UNITS_BYTES,
gfxWindowsPlatform::sD3D9MemoryUsed,
"Memory used for D3D9 shared textures");
}
};
NS_IMPL_ISUPPORTS(D3D9TextureReporter, nsIMemoryReporter)
Atomic<size_t> gfxWindowsPlatform::sD3D9SharedTextureUsed;
class D3D9SharedTextureReporter final : public nsIMemoryReporter
{
~D3D9SharedTextureReporter() {}
public:
NS_DECL_ISUPPORTS
NS_IMETHOD CollectReports(nsIHandleReportCallback *aHandleReport,
nsISupports* aData, bool aAnonymize) override
{
return MOZ_COLLECT_REPORT("d3d9-shared-texture", KIND_OTHER, UNITS_BYTES,
gfxWindowsPlatform::sD3D9SharedTextureUsed,
"Memory used for D3D9 shared textures");
}
};
NS_IMPL_ISUPPORTS(D3D9SharedTextureReporter, nsIMemoryReporter)
// Device init data should only be used on child processes, so we protect it
// behind a getter here.
static DeviceInitData sDeviceInitDataDoNotUseDirectly;
static inline DeviceInitData&
GetParentDevicePrefs()
{
MOZ_ASSERT(XRE_IsContentProcess());
return sDeviceInitDataDoNotUseDirectly;
}
2006-02-01 05:35:38 +03:00
gfxWindowsPlatform::gfxWindowsPlatform()
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
: mRenderMode(RENDER_GDI)
, mIsWARP(false)
, mHasDeviceReset(false)
, mHasFakeDeviceReset(false)
, mCompositorD3D11TextureSharingWorks(false)
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
, mAcceleration(FeatureStatus::Unused)
, mD3D11Status(FeatureStatus::Unused)
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
, mD2D1Status(FeatureStatus::Unused)
2006-02-01 05:35:38 +03:00
{
mUseClearTypeForDownloadableFonts = UNINITIALIZED_VALUE;
mUseClearTypeAlways = UNINITIALIZED_VALUE;
/*
* Initialize COM
*/
CoInitialize(nullptr);
RegisterStrongMemoryReporter(new GfxD2DVramReporter());
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
// Set up the D3D11 feature levels we can ask for.
if (IsWin8OrLater()) {
mFeatureLevels.AppendElement(D3D_FEATURE_LEVEL_11_1);
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
mFeatureLevels.AppendElement(D3D_FEATURE_LEVEL_11_0);
mFeatureLevels.AppendElement(D3D_FEATURE_LEVEL_10_1);
mFeatureLevels.AppendElement(D3D_FEATURE_LEVEL_10_0);
mFeatureLevels.AppendElement(D3D_FEATURE_LEVEL_9_3);
UpdateDeviceInitData();
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
InitializeDevices();
UpdateRenderMode();
RegisterStrongMemoryReporter(new GPUAdapterReporter());
RegisterStrongMemoryReporter(new D3D11TextureReporter());
RegisterStrongMemoryReporter(new D3D9TextureReporter());
RegisterStrongMemoryReporter(new D3D9SharedTextureReporter());
}
gfxWindowsPlatform::~gfxWindowsPlatform()
{
mDeviceManager = nullptr;
mD3D11Device = nullptr;
mD3D11ContentDevice = nullptr;
mD3D11ImageBridgeDevice = nullptr;
mozilla::gfx::Factory::D2DCleanup();
mAdapter = nullptr;
/*
* Uninitialize COM
*/
CoUninitialize();
}
bool
gfxWindowsPlatform::CanUseHardwareVideoDecoding()
{
if (!gfxPrefs::LayersPreferD3D9() && !mCompositorD3D11TextureSharingWorks) {
return false;
}
return !IsWARP() && gfxPlatform::CanUseHardwareVideoDecoding();
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
bool
gfxWindowsPlatform::InitDWriteSupport()
{
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
MOZ_ASSERT(!mDWriteFactory && IsVistaOrLater());
mozilla::ScopedGfxFeatureReporter reporter("DWrite");
decltype(DWriteCreateFactory)* createDWriteFactory = (decltype(DWriteCreateFactory)*)
GetProcAddress(LoadLibraryW(L"dwrite.dll"), "DWriteCreateFactory");
if (!createDWriteFactory) {
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
return false;
}
// I need a direct pointer to be able to cast to IUnknown**, I also need to
// remember to release this because the nsRefPtr will AddRef it.
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<IDWriteFactory> factory;
HRESULT hr = createDWriteFactory(
DWRITE_FACTORY_TYPE_SHARED,
__uuidof(IDWriteFactory),
(IUnknown **)((IDWriteFactory **)getter_AddRefs(factory)));
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
if (FAILED(hr) || !factory) {
return false;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
mDWriteFactory = factory;
SetupClearTypeParams();
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
reporter.SetSuccessful();
return true;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
bool
gfxWindowsPlatform::HandleDeviceReset()
{
DeviceResetReason resetReason = DeviceResetReason::OK;
if (!DidRenderingDeviceReset(&resetReason)) {
return false;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
if (mHasFakeDeviceReset) {
if (XRE_IsParentProcess()) {
// Notify child processes that we got a device reset.
nsTArray<dom::ContentParent*> processes;
dom::ContentParent::GetAll(processes);
for (size_t i = 0; i < processes.Length(); i++) {
processes[i]->SendTestGraphicsDeviceReset(uint32_t(resetReason));
}
}
} else {
Telemetry::Accumulate(Telemetry::DEVICE_RESET_REASON, uint32_t(resetReason));
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
// Remove devices and adapters.
ResetD3D11Devices();
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
mAdapter = nullptr;
// Reset local state. Note: we leave feature status variables as-is. They
// will be recomputed by InitializeDevices().
mHasDeviceReset = false;
mHasFakeDeviceReset = false;
mCompositorD3D11TextureSharingWorks = false;
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
mDeviceResetReason = DeviceResetReason::OK;
imgLoader::Singleton()->ClearCache(true);
imgLoader::Singleton()->ClearCache(false);
gfxAlphaBoxBlur::ShutdownBlurCache();
// Since we got a device reset, we must ask the parent process for an updated
// list of which devices to create.
UpdateDeviceInitData();
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
InitializeDevices();
return true;
}
static const BackendType SOFTWARE_BACKEND = BackendType::CAIRO;
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
void
gfxWindowsPlatform::UpdateBackendPrefs()
{
uint32_t canvasMask = BackendTypeBit(SOFTWARE_BACKEND);
uint32_t contentMask = BackendTypeBit(SOFTWARE_BACKEND);
BackendType defaultBackend = SOFTWARE_BACKEND;
if (GetD2D1Status() == FeatureStatus::Available) {
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
mRenderMode = RENDER_DIRECT2D;
contentMask |= BackendTypeBit(BackendType::DIRECT2D1_1);
canvasMask |= BackendTypeBit(BackendType::DIRECT2D1_1);
defaultBackend = BackendType::DIRECT2D1_1;
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
} else {
mRenderMode = RENDER_GDI;
canvasMask |= BackendTypeBit(BackendType::SKIA);
}
contentMask |= BackendTypeBit(BackendType::SKIA);
InitBackendPrefs(canvasMask, defaultBackend, contentMask, defaultBackend);
}
void
gfxWindowsPlatform::UpdateRenderMode()
{
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
bool didReset = HandleDeviceReset();
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
UpdateBackendPrefs();
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
if (didReset) {
mScreenReferenceDrawTarget =
CreateOffscreenContentDrawTarget(IntSize(1, 1), SurfaceFormat::B8G8R8A8);
}
}
void
gfxWindowsPlatform::ForceDeviceReset(ForcedDeviceResetReason aReason)
{
Telemetry::Accumulate(Telemetry::FORCED_DEVICE_RESET_REASON, uint32_t(aReason));
mDeviceResetReason = DeviceResetReason::FORCED_RESET;
mHasDeviceReset = true;
}
mozilla::gfx::BackendType
gfxWindowsPlatform::GetContentBackendFor(mozilla::layers::LayersBackend aLayers)
{
if (aLayers == LayersBackend::LAYERS_D3D11) {
return gfxPlatform::GetDefaultContentBackend();
}
// If we're not accelerated with D3D11, never use D2D.
return SOFTWARE_BACKEND;
}
gfxPlatformFontList*
gfxWindowsPlatform::CreatePlatformFontList()
{
gfxPlatformFontList *pfl;
// bug 630201 - older pre-RTM versions of Direct2D/DirectWrite cause odd
// crashers so blacklist them altogether
if (IsNotWin7PreRTM() && GetDWriteFactory()) {
pfl = new gfxDWriteFontList();
if (NS_SUCCEEDED(pfl->InitFontList())) {
return pfl;
}
// DWrite font initialization failed! Don't know why this would happen,
// but apparently it can - see bug 594865.
// So we're going to fall back to GDI fonts & rendering.
gfxPlatformFontList::Shutdown();
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
DisableD2D();
}
pfl = new gfxGDIFontList();
if (NS_SUCCEEDED(pfl->InitFontList())) {
return pfl;
}
gfxPlatformFontList::Shutdown();
return nullptr;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
// This function will permanently disable D2D for the session. It's intended to
// be used when, after initially chosing to use Direct2D, we encounter a
// scenario we can't support.
//
// This is called during gfxPlatform::Init() so at this point there should be no
// DrawTargetD2D/1 instances.
void
gfxWindowsPlatform::DisableD2D()
{
mD2D1Status = FeatureStatus::Failed;
Factory::SetDirect3D11Device(nullptr);
UpdateBackendPrefs();
}
already_AddRefed<gfxASurface>
gfxWindowsPlatform::CreateOffscreenSurface(const IntSize& aSize,
gfxImageFormat aFormat)
2006-02-01 05:35:38 +03:00
{
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<gfxASurface> surf = nullptr;
#ifdef CAIRO_HAS_WIN32_SURFACE
if (mRenderMode == RENDER_GDI || mRenderMode == RENDER_DIRECT2D)
surf = new gfxWindowsSurface(aSize, aFormat);
#endif
if (!surf || surf->CairoStatus()) {
surf = new gfxImageSurface(aSize, aFormat);
}
return surf.forget();
}
already_AddRefed<ScaledFont>
gfxWindowsPlatform::GetScaledFontForFont(DrawTarget* aTarget, gfxFont *aFont)
{
if (aFont->GetType() == gfxFont::FONT_TYPE_DWRITE) {
gfxDWriteFont *font = static_cast<gfxDWriteFont*>(aFont);
NativeFont nativeFont;
nativeFont.mType = NativeFontType::DWRITE_FONT_FACE;
nativeFont.mFont = font->GetFontFace();
if (aTarget->GetBackendType() == BackendType::CAIRO) {
return Factory::CreateScaledFontWithCairo(nativeFont,
font->GetAdjustedSize(),
font->GetCairoScaledFont());
}
return Factory::CreateScaledFontForNativeFont(nativeFont,
font->GetAdjustedSize());
}
NS_ASSERTION(aFont->GetType() == gfxFont::FONT_TYPE_GDI,
"Fonts on windows should be GDI or DWrite!");
NativeFont nativeFont;
nativeFont.mType = NativeFontType::GDI_FONT_FACE;
LOGFONT lf;
GetObject(static_cast<gfxGDIFont*>(aFont)->GetHFONT(), sizeof(LOGFONT), &lf);
nativeFont.mFont = &lf;
if (aTarget->GetBackendType() == BackendType::CAIRO) {
return Factory::CreateScaledFontWithCairo(nativeFont,
aFont->GetAdjustedSize(),
aFont->GetCairoScaledFont());
}
return Factory::CreateScaledFontForNativeFont(nativeFont, aFont->GetAdjustedSize());
}
2006-02-01 05:35:38 +03:00
nsresult
gfxWindowsPlatform::GetFontList(nsIAtom *aLangGroup,
2006-02-01 05:35:38 +03:00
const nsACString& aGenericFamily,
nsTArray<nsString>& aListOfFonts)
2006-02-01 05:35:38 +03:00
{
gfxPlatformFontList::PlatformFontList()->GetFontList(aLangGroup, aGenericFamily, aListOfFonts);
return NS_OK;
}
nsresult
gfxWindowsPlatform::UpdateFontList()
{
gfxPlatformFontList::PlatformFontList()->UpdateFontList();
return NS_OK;
}
static const char kFontAparajita[] = "Aparajita";
static const char kFontArabicTypesetting[] = "Arabic Typesetting";
static const char kFontArial[] = "Arial";
static const char kFontArialUnicodeMS[] = "Arial Unicode MS";
static const char kFontCambria[] = "Cambria";
static const char kFontCambriaMath[] = "Cambria Math";
static const char kFontEbrima[] = "Ebrima";
static const char kFontEstrangeloEdessa[] = "Estrangelo Edessa";
static const char kFontEuphemia[] = "Euphemia";
static const char kFontGabriola[] = "Gabriola";
static const char kFontJavaneseText[] = "Javanese Text";
static const char kFontKhmerUI[] = "Khmer UI";
static const char kFontLaoUI[] = "Lao UI";
static const char kFontLeelawadeeUI[] = "Leelawadee UI";
static const char kFontLucidaSansUnicode[] = "Lucida Sans Unicode";
static const char kFontMVBoli[] = "MV Boli";
static const char kFontMalgunGothic[] = "Malgun Gothic";
static const char kFontMicrosoftJhengHei[] = "Microsoft JhengHei";
static const char kFontMicrosoftNewTaiLue[] = "Microsoft New Tai Lue";
static const char kFontMicrosoftPhagsPa[] = "Microsoft PhagsPa";
static const char kFontMicrosoftTaiLe[] = "Microsoft Tai Le";
static const char kFontMicrosoftUighur[] = "Microsoft Uighur";
static const char kFontMicrosoftYaHei[] = "Microsoft YaHei";
static const char kFontMicrosoftYiBaiti[] = "Microsoft Yi Baiti";
static const char kFontMeiryo[] = "Meiryo";
static const char kFontMongolianBaiti[] = "Mongolian Baiti";
static const char kFontMyanmarText[] = "Myanmar Text";
static const char kFontNirmalaUI[] = "Nirmala UI";
static const char kFontNyala[] = "Nyala";
static const char kFontPlantagenetCherokee[] = "Plantagenet Cherokee";
static const char kFontSegoeUI[] = "Segoe UI";
static const char kFontSegoeUIEmoji[] = "Segoe UI Emoji";
static const char kFontSegoeUISymbol[] = "Segoe UI Symbol";
static const char kFontSylfaen[] = "Sylfaen";
static const char kFontTraditionalArabic[] = "Traditional Arabic";
static const char kFontUtsaah[] = "Utsaah";
static const char kFontYuGothic[] = "Yu Gothic";
void
gfxWindowsPlatform::GetCommonFallbackFonts(uint32_t aCh, uint32_t aNextCh,
int32_t aRunScript,
nsTArray<const char*>& aFontList)
{
if (aNextCh == 0xfe0fu) {
aFontList.AppendElement(kFontSegoeUIEmoji);
}
// Arial is used as the default fallback for system fallback
aFontList.AppendElement(kFontArial);
if (!IS_IN_BMP(aCh)) {
uint32_t p = aCh >> 16;
if (p == 1) { // SMP plane
if (aNextCh == 0xfe0eu) {
aFontList.AppendElement(kFontSegoeUISymbol);
aFontList.AppendElement(kFontSegoeUIEmoji);
} else {
if (aNextCh != 0xfe0fu) {
aFontList.AppendElement(kFontSegoeUIEmoji);
}
aFontList.AppendElement(kFontSegoeUISymbol);
}
aFontList.AppendElement(kFontEbrima);
aFontList.AppendElement(kFontNirmalaUI);
aFontList.AppendElement(kFontCambriaMath);
}
} else {
uint32_t b = (aCh >> 8) & 0xff;
switch (b) {
case 0x05:
aFontList.AppendElement(kFontEstrangeloEdessa);
aFontList.AppendElement(kFontCambria);
break;
case 0x06:
aFontList.AppendElement(kFontMicrosoftUighur);
break;
case 0x07:
aFontList.AppendElement(kFontEstrangeloEdessa);
aFontList.AppendElement(kFontMVBoli);
aFontList.AppendElement(kFontEbrima);
break;
case 0x09:
aFontList.AppendElement(kFontNirmalaUI);
aFontList.AppendElement(kFontUtsaah);
aFontList.AppendElement(kFontAparajita);
break;
case 0x0e:
aFontList.AppendElement(kFontLaoUI);
break;
case 0x10:
aFontList.AppendElement(kFontMyanmarText);
break;
case 0x11:
aFontList.AppendElement(kFontMalgunGothic);
break;
case 0x12:
case 0x13:
aFontList.AppendElement(kFontNyala);
aFontList.AppendElement(kFontPlantagenetCherokee);
break;
case 0x14:
case 0x15:
case 0x16:
aFontList.AppendElement(kFontEuphemia);
aFontList.AppendElement(kFontSegoeUISymbol);
break;
case 0x17:
aFontList.AppendElement(kFontKhmerUI);
break;
case 0x18: // Mongolian
aFontList.AppendElement(kFontMongolianBaiti);
aFontList.AppendElement(kFontEuphemia);
break;
case 0x19:
aFontList.AppendElement(kFontMicrosoftTaiLe);
aFontList.AppendElement(kFontMicrosoftNewTaiLue);
aFontList.AppendElement(kFontKhmerUI);
break;
break;
case 0x1a:
aFontList.AppendElement(kFontLeelawadeeUI);
break;
case 0x1c:
aFontList.AppendElement(kFontNirmalaUI);
break;
case 0x20: // Symbol ranges
case 0x21:
case 0x22:
case 0x23:
case 0x24:
case 0x25:
case 0x26:
case 0x27:
case 0x29:
case 0x2a:
case 0x2b:
case 0x2c:
aFontList.AppendElement(kFontSegoeUI);
aFontList.AppendElement(kFontSegoeUISymbol);
aFontList.AppendElement(kFontCambria);
aFontList.AppendElement(kFontMeiryo);
aFontList.AppendElement(kFontArial);
aFontList.AppendElement(kFontLucidaSansUnicode);
aFontList.AppendElement(kFontEbrima);
break;
case 0x2d:
case 0x2e:
case 0x2f:
aFontList.AppendElement(kFontEbrima);
aFontList.AppendElement(kFontNyala);
aFontList.AppendElement(kFontSegoeUI);
aFontList.AppendElement(kFontSegoeUISymbol);
aFontList.AppendElement(kFontMeiryo);
break;
case 0x28: // Braille
aFontList.AppendElement(kFontSegoeUISymbol);
break;
case 0x30:
case 0x31:
aFontList.AppendElement(kFontMicrosoftYaHei);
break;
case 0x32:
aFontList.AppendElement(kFontMalgunGothic);
break;
case 0x4d:
aFontList.AppendElement(kFontSegoeUISymbol);
break;
case 0x9f:
aFontList.AppendElement(kFontMicrosoftYaHei);
aFontList.AppendElement(kFontYuGothic);
break;
case 0xa0: // Yi
case 0xa1:
case 0xa2:
case 0xa3:
case 0xa4:
aFontList.AppendElement(kFontMicrosoftYiBaiti);
aFontList.AppendElement(kFontSegoeUI);
break;
case 0xa5:
case 0xa6:
case 0xa7:
aFontList.AppendElement(kFontEbrima);
aFontList.AppendElement(kFontSegoeUI);
aFontList.AppendElement(kFontCambriaMath);
break;
case 0xa8:
aFontList.AppendElement(kFontMicrosoftPhagsPa);
aFontList.AppendElement(kFontNirmalaUI);
break;
case 0xa9:
aFontList.AppendElement(kFontMalgunGothic);
aFontList.AppendElement(kFontJavaneseText);
break;
case 0xaa:
aFontList.AppendElement(kFontMyanmarText);
break;
case 0xab:
aFontList.AppendElement(kFontEbrima);
aFontList.AppendElement(kFontNyala);
break;
case 0xd7:
aFontList.AppendElement(kFontMalgunGothic);
break;
case 0xfb:
aFontList.AppendElement(kFontMicrosoftUighur);
aFontList.AppendElement(kFontGabriola);
aFontList.AppendElement(kFontSylfaen);
break;
case 0xfc:
case 0xfd:
aFontList.AppendElement(kFontTraditionalArabic);
aFontList.AppendElement(kFontArabicTypesetting);
break;
case 0xfe:
aFontList.AppendElement(kFontTraditionalArabic);
aFontList.AppendElement(kFontMicrosoftJhengHei);
break;
case 0xff:
aFontList.AppendElement(kFontMicrosoftJhengHei);
break;
default:
break;
}
}
// Arial Unicode MS has lots of glyphs for obscure characters,
// use it as a last resort
aFontList.AppendElement(kFontArialUnicodeMS);
}
nsresult
gfxWindowsPlatform::GetStandardFamilyName(const nsAString& aFontName, nsAString& aFamilyName)
{
gfxPlatformFontList::PlatformFontList()->GetStandardFamilyName(aFontName, aFamilyName);
return NS_OK;
2006-02-01 05:35:38 +03:00
}
gfxFontGroup *
gfxWindowsPlatform::CreateFontGroup(const FontFamilyList& aFontFamilyList,
const gfxFontStyle *aStyle,
gfxTextPerfMetrics* aTextPerf,
gfxUserFontSet *aUserFontSet,
gfxFloat aDevToCssSize)
{
return new gfxFontGroup(aFontFamilyList, aStyle, aTextPerf,
aUserFontSet, aDevToCssSize);
}
gfxFontEntry*
gfxWindowsPlatform::LookupLocalFont(const nsAString& aFontName,
uint16_t aWeight,
int16_t aStretch,
uint8_t aStyle)
{
return gfxPlatformFontList::PlatformFontList()->LookupLocalFont(aFontName,
aWeight,
aStretch,
aStyle);
}
gfxFontEntry*
gfxWindowsPlatform::MakePlatformFont(const nsAString& aFontName,
uint16_t aWeight,
int16_t aStretch,
uint8_t aStyle,
const uint8_t* aFontData,
uint32_t aLength)
{
return gfxPlatformFontList::PlatformFontList()->MakePlatformFont(aFontName,
aWeight,
aStretch,
aStyle,
aFontData,
aLength);
}
bool
gfxWindowsPlatform::IsFontFormatSupported(nsIURI *aFontURI, uint32_t aFormatFlags)
{
// check for strange format flags
NS_ASSERTION(!(aFormatFlags & gfxUserFontSet::FLAG_FORMAT_NOT_USED),
"strange font format hint set");
// accept supported formats
if (aFormatFlags & gfxUserFontSet::FLAG_FORMATS_COMMON) {
return true;
}
// reject all other formats, known and unknown
if (aFormatFlags != 0) {
return false;
}
// no format hint set, need to look at data
return true;
}
static DeviceResetReason HResultToResetReason(HRESULT hr)
{
switch (hr) {
case DXGI_ERROR_DEVICE_HUNG:
return DeviceResetReason::HUNG;
case DXGI_ERROR_DEVICE_REMOVED:
return DeviceResetReason::REMOVED;
case DXGI_ERROR_DEVICE_RESET:
return DeviceResetReason::RESET;
case DXGI_ERROR_DRIVER_INTERNAL_ERROR:
return DeviceResetReason::DRIVER_ERROR;
case DXGI_ERROR_INVALID_CALL:
return DeviceResetReason::INVALID_CALL;
case E_OUTOFMEMORY:
return DeviceResetReason::OUT_OF_MEMORY;
default:
MOZ_ASSERT(false);
}
return DeviceResetReason::UNKNOWN;
}
bool
gfxWindowsPlatform::IsDeviceReset(HRESULT hr, DeviceResetReason* aResetReason)
{
if (hr != S_OK) {
mDeviceResetReason = HResultToResetReason(hr);
mHasDeviceReset = true;
if (aResetReason) {
*aResetReason = mDeviceResetReason;
}
return true;
}
return false;
}
void
gfxWindowsPlatform::TestDeviceReset(DeviceResetReason aReason)
{
if (mHasDeviceReset) {
return;
}
mHasDeviceReset = true;
mHasFakeDeviceReset = true;
mDeviceResetReason = aReason;
}
bool
gfxWindowsPlatform::DidRenderingDeviceReset(DeviceResetReason* aResetReason)
{
if (mHasDeviceReset) {
if (aResetReason) {
*aResetReason = mDeviceResetReason;
}
return true;
}
if (aResetReason) {
*aResetReason = DeviceResetReason::OK;
}
if (mD3D11Device) {
HRESULT hr = mD3D11Device->GetDeviceRemovedReason();
if (IsDeviceReset(hr, aResetReason)) {
return true;
}
}
if (mD3D11ContentDevice) {
HRESULT hr = mD3D11ContentDevice->GetDeviceRemovedReason();
if (IsDeviceReset(hr, aResetReason)) {
return true;
}
}
if (XRE_IsParentProcess() && gfxPrefs::DeviceResetForTesting()) {
TestDeviceReset((DeviceResetReason)gfxPrefs::DeviceResetForTesting());
if (aResetReason) {
*aResetReason = mDeviceResetReason;
}
gfxPrefs::SetDeviceResetForTesting(0);
return true;
}
return false;
}
BOOL CALLBACK
InvalidateWindowForDeviceReset(HWND aWnd, LPARAM aMsg)
{
RedrawWindow(aWnd, nullptr, nullptr,
RDW_INVALIDATE|RDW_INTERNALPAINT|RDW_FRAME);
return TRUE;
}
bool
gfxWindowsPlatform::UpdateForDeviceReset()
{
PROFILER_LABEL_FUNC(js::ProfileEntry::Category::GRAPHICS);
if (!DidRenderingDeviceReset()) {
return false;
}
// Trigger an ::OnPaint for each window.
::EnumThreadWindows(GetCurrentThreadId(),
InvalidateWindowForDeviceReset,
0);
gfxCriticalNote << "Detected rendering device reset on refresh";
return true;
}
void
gfxWindowsPlatform::GetPlatformCMSOutputProfile(void* &mem, size_t &mem_size)
{
WCHAR str[MAX_PATH];
DWORD size = MAX_PATH;
BOOL res;
mem = nullptr;
mem_size = 0;
HDC dc = GetDC(nullptr);
if (!dc)
return;
MOZ_SEH_TRY {
res = GetICMProfileW(dc, &size, (LPWSTR)&str);
} MOZ_SEH_EXCEPT(GetExceptionCode() == EXCEPTION_ILLEGAL_INSTRUCTION) {
res = FALSE;
}
ReleaseDC(nullptr, dc);
if (!res)
return;
#ifdef _WIN32
qcms_data_from_unicode_path(str, &mem, &mem_size);
#ifdef DEBUG_tor
if (mem_size > 0)
fprintf(stderr,
"ICM profile read from %s successfully\n",
NS_ConvertUTF16toUTF8(str).get());
#endif // DEBUG_tor
#endif // _WIN32
}
bool
gfxWindowsPlatform::UseClearTypeForDownloadableFonts()
{
if (mUseClearTypeForDownloadableFonts == UNINITIALIZED_VALUE) {
mUseClearTypeForDownloadableFonts = Preferences::GetBool(GFX_DOWNLOADABLE_FONTS_USE_CLEARTYPE, true);
}
return mUseClearTypeForDownloadableFonts;
}
bool
gfxWindowsPlatform::UseClearTypeAlways()
{
if (mUseClearTypeAlways == UNINITIALIZED_VALUE) {
mUseClearTypeAlways = Preferences::GetBool(GFX_USE_CLEARTYPE_ALWAYS, false);
}
return mUseClearTypeAlways;
}
void
gfxWindowsPlatform::GetDLLVersion(char16ptr_t aDLLPath, nsAString& aVersion)
{
DWORD versInfoSize, vers[4] = {0};
// version info not available case
aVersion.AssignLiteral(MOZ_UTF16("0.0.0.0"));
versInfoSize = GetFileVersionInfoSizeW(aDLLPath, nullptr);
AutoTArray<BYTE,512> versionInfo;
if (versInfoSize == 0 ||
!versionInfo.AppendElements(uint32_t(versInfoSize)))
{
return;
}
if (!GetFileVersionInfoW(aDLLPath, 0, versInfoSize,
LPBYTE(versionInfo.Elements())))
{
return;
}
UINT len = 0;
VS_FIXEDFILEINFO *fileInfo = nullptr;
if (!VerQueryValue(LPBYTE(versionInfo.Elements()), TEXT("\\"),
(LPVOID *)&fileInfo, &len) ||
len == 0 ||
fileInfo == nullptr)
{
return;
}
DWORD fileVersMS = fileInfo->dwFileVersionMS;
DWORD fileVersLS = fileInfo->dwFileVersionLS;
vers[0] = HIWORD(fileVersMS);
vers[1] = LOWORD(fileVersMS);
vers[2] = HIWORD(fileVersLS);
vers[3] = LOWORD(fileVersLS);
char buf[256];
snprintf_literal(buf, "%u.%u.%u.%u", vers[0], vers[1], vers[2], vers[3]);
aVersion.Assign(NS_ConvertUTF8toUTF16(buf));
}
void
gfxWindowsPlatform::GetCleartypeParams(nsTArray<ClearTypeParameterInfo>& aParams)
{
HKEY hKey, subKey;
DWORD i, rv, size, type;
WCHAR displayName[256], subkeyName[256];
aParams.Clear();
// construct subkeys based on HKLM subkeys, assume they are same for HKCU
rv = RegOpenKeyExW(HKEY_LOCAL_MACHINE,
L"Software\\Microsoft\\Avalon.Graphics",
0, KEY_READ, &hKey);
if (rv != ERROR_SUCCESS) {
return;
}
// enumerate over subkeys
for (i = 0, rv = ERROR_SUCCESS; rv != ERROR_NO_MORE_ITEMS; i++) {
size = ArrayLength(displayName);
rv = RegEnumKeyExW(hKey, i, displayName, &size,
nullptr, nullptr, nullptr, nullptr);
if (rv != ERROR_SUCCESS) {
continue;
}
ClearTypeParameterInfo ctinfo;
ctinfo.displayName.Assign(displayName);
DWORD subrv, value;
bool foundData = false;
swprintf_s(subkeyName, ArrayLength(subkeyName),
L"Software\\Microsoft\\Avalon.Graphics\\%s", displayName);
// subkey for gamma, pixel structure
subrv = RegOpenKeyExW(HKEY_LOCAL_MACHINE,
subkeyName, 0, KEY_QUERY_VALUE, &subKey);
if (subrv == ERROR_SUCCESS) {
size = sizeof(value);
subrv = RegQueryValueExW(subKey, L"GammaLevel", nullptr, &type,
(LPBYTE)&value, &size);
if (subrv == ERROR_SUCCESS && type == REG_DWORD) {
foundData = true;
ctinfo.gamma = value;
}
size = sizeof(value);
subrv = RegQueryValueExW(subKey, L"PixelStructure", nullptr, &type,
(LPBYTE)&value, &size);
if (subrv == ERROR_SUCCESS && type == REG_DWORD) {
foundData = true;
ctinfo.pixelStructure = value;
}
RegCloseKey(subKey);
}
// subkey for cleartype level, enhanced contrast
subrv = RegOpenKeyExW(HKEY_CURRENT_USER,
subkeyName, 0, KEY_QUERY_VALUE, &subKey);
if (subrv == ERROR_SUCCESS) {
size = sizeof(value);
subrv = RegQueryValueExW(subKey, L"ClearTypeLevel", nullptr, &type,
(LPBYTE)&value, &size);
if (subrv == ERROR_SUCCESS && type == REG_DWORD) {
foundData = true;
ctinfo.clearTypeLevel = value;
}
size = sizeof(value);
subrv = RegQueryValueExW(subKey, L"EnhancedContrastLevel",
nullptr, &type, (LPBYTE)&value, &size);
if (subrv == ERROR_SUCCESS && type == REG_DWORD) {
foundData = true;
ctinfo.enhancedContrast = value;
}
RegCloseKey(subKey);
}
if (foundData) {
aParams.AppendElement(ctinfo);
}
}
RegCloseKey(hKey);
}
void
gfxWindowsPlatform::FontsPrefsChanged(const char *aPref)
{
bool clearTextFontCaches = true;
gfxPlatform::FontsPrefsChanged(aPref);
if (!aPref) {
mUseClearTypeForDownloadableFonts = UNINITIALIZED_VALUE;
mUseClearTypeAlways = UNINITIALIZED_VALUE;
} else if (!strcmp(GFX_DOWNLOADABLE_FONTS_USE_CLEARTYPE, aPref)) {
mUseClearTypeForDownloadableFonts = UNINITIALIZED_VALUE;
} else if (!strcmp(GFX_USE_CLEARTYPE_ALWAYS, aPref)) {
mUseClearTypeAlways = UNINITIALIZED_VALUE;
} else if (!strncmp(GFX_CLEARTYPE_PARAMS, aPref, strlen(GFX_CLEARTYPE_PARAMS))) {
SetupClearTypeParams();
} else {
clearTextFontCaches = false;
}
if (clearTextFontCaches) {
gfxFontCache *fc = gfxFontCache::GetCache();
if (fc) {
fc->Flush();
}
}
}
#define ENHANCED_CONTRAST_REGISTRY_KEY \
HKEY_CURRENT_USER, "Software\\Microsoft\\Avalon.Graphics\\DISPLAY1\\EnhancedContrastLevel"
void
gfxWindowsPlatform::SetupClearTypeParams()
{
if (GetDWriteFactory()) {
// any missing prefs will default to invalid (-1) and be ignored;
// out-of-range values will also be ignored
FLOAT gamma = -1.0;
FLOAT contrast = -1.0;
FLOAT level = -1.0;
int geometry = -1;
int mode = -1;
int32_t value;
if (NS_SUCCEEDED(Preferences::GetInt(GFX_CLEARTYPE_PARAMS_GAMMA, &value))) {
if (value >= 1000 && value <= 2200) {
gamma = FLOAT(value / 1000.0);
}
}
if (NS_SUCCEEDED(Preferences::GetInt(GFX_CLEARTYPE_PARAMS_CONTRAST, &value))) {
if (value >= 0 && value <= 1000) {
contrast = FLOAT(value / 100.0);
}
}
if (NS_SUCCEEDED(Preferences::GetInt(GFX_CLEARTYPE_PARAMS_LEVEL, &value))) {
if (value >= 0 && value <= 100) {
level = FLOAT(value / 100.0);
}
}
if (NS_SUCCEEDED(Preferences::GetInt(GFX_CLEARTYPE_PARAMS_STRUCTURE, &value))) {
if (value >= 0 && value <= 2) {
geometry = value;
}
}
if (NS_SUCCEEDED(Preferences::GetInt(GFX_CLEARTYPE_PARAMS_MODE, &value))) {
if (value >= 0 && value <= 5) {
mode = value;
}
}
cairo_dwrite_set_cleartype_params(gamma, contrast, level, geometry, mode);
switch (mode) {
case DWRITE_RENDERING_MODE_ALIASED:
case DWRITE_RENDERING_MODE_CLEARTYPE_GDI_CLASSIC:
mMeasuringMode = DWRITE_MEASURING_MODE_GDI_CLASSIC;
break;
case DWRITE_RENDERING_MODE_CLEARTYPE_GDI_NATURAL:
mMeasuringMode = DWRITE_MEASURING_MODE_GDI_NATURAL;
break;
default:
mMeasuringMode = DWRITE_MEASURING_MODE_NATURAL;
break;
}
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<IDWriteRenderingParams> defaultRenderingParams;
GetDWriteFactory()->CreateRenderingParams(getter_AddRefs(defaultRenderingParams));
// For EnhancedContrast, we override the default if the user has not set it
// in the registry (by using the ClearType Tuner).
if (contrast >= 0.0 && contrast <= 10.0) {
contrast = contrast;
} else {
HKEY hKey;
if (RegOpenKeyExA(ENHANCED_CONTRAST_REGISTRY_KEY,
0, KEY_READ, &hKey) == ERROR_SUCCESS)
{
contrast = defaultRenderingParams->GetEnhancedContrast();
RegCloseKey(hKey);
} else {
contrast = 1.0;
}
}
// For parameters that have not been explicitly set,
// we copy values from default params (or our overridden value for contrast)
if (gamma < 1.0 || gamma > 2.2) {
gamma = defaultRenderingParams->GetGamma();
}
if (level < 0.0 || level > 1.0) {
level = defaultRenderingParams->GetClearTypeLevel();
}
DWRITE_PIXEL_GEOMETRY dwriteGeometry =
static_cast<DWRITE_PIXEL_GEOMETRY>(geometry);
DWRITE_RENDERING_MODE renderMode =
static_cast<DWRITE_RENDERING_MODE>(mode);
if (dwriteGeometry < DWRITE_PIXEL_GEOMETRY_FLAT ||
dwriteGeometry > DWRITE_PIXEL_GEOMETRY_BGR) {
dwriteGeometry = defaultRenderingParams->GetPixelGeometry();
}
if (renderMode < DWRITE_RENDERING_MODE_DEFAULT ||
renderMode > DWRITE_RENDERING_MODE_CLEARTYPE_NATURAL_SYMMETRIC) {
renderMode = defaultRenderingParams->GetRenderingMode();
}
mRenderingParams[TEXT_RENDERING_NO_CLEARTYPE] = defaultRenderingParams;
GetDWriteFactory()->CreateCustomRenderingParams(gamma, contrast, level,
dwriteGeometry, renderMode,
getter_AddRefs(mRenderingParams[TEXT_RENDERING_NORMAL]));
GetDWriteFactory()->CreateCustomRenderingParams(gamma, contrast, level,
dwriteGeometry, DWRITE_RENDERING_MODE_CLEARTYPE_GDI_CLASSIC,
getter_AddRefs(mRenderingParams[TEXT_RENDERING_GDI_CLASSIC]));
}
}
void
gfxWindowsPlatform::OnDeviceManagerDestroy(DeviceManagerD3D9* aDeviceManager)
{
if (aDeviceManager == mDeviceManager) {
mDeviceManager = nullptr;
}
}
IDirect3DDevice9*
gfxWindowsPlatform::GetD3D9Device()
{
DeviceManagerD3D9* manager = GetD3D9DeviceManager();
return manager ? manager->device() : nullptr;
}
DeviceManagerD3D9*
gfxWindowsPlatform::GetD3D9DeviceManager()
{
// We should only create the d3d9 device on the compositor thread
// or we don't have a compositor thread.
if (!mDeviceManager &&
(!gfxPlatform::UsesOffMainThreadCompositing() ||
CompositorParent::IsInCompositorThread())) {
mDeviceManager = new DeviceManagerD3D9();
if (!mDeviceManager->Init()) {
gfxCriticalError() << "[D3D9] Could not Initialize the DeviceManagerD3D9";
mDeviceManager = nullptr;
}
}
return mDeviceManager;
}
ID3D11Device*
gfxWindowsPlatform::GetD3D11Device()
{
return mD3D11Device;
}
ID3D11Device*
gfxWindowsPlatform::GetD3D11ContentDevice()
{
return mD3D11ContentDevice;
}
ID3D11Device*
gfxWindowsPlatform::GetD3D11ImageBridgeDevice()
{
return mD3D11ImageBridgeDevice;
}
ID3D11Device*
gfxWindowsPlatform::GetD3D11DeviceForCurrentThread()
{
if (NS_IsMainThread()) {
return GetD3D11ContentDevice();
} else {
return GetD3D11ImageBridgeDevice();
}
}
ReadbackManagerD3D11*
gfxWindowsPlatform::GetReadbackManager()
{
if (!mD3D11ReadbackManager) {
mD3D11ReadbackManager = new ReadbackManagerD3D11();
}
return mD3D11ReadbackManager;
}
bool
gfxWindowsPlatform::IsOptimus()
{
static int knowIsOptimus = -1;
if (knowIsOptimus == -1) {
// other potential optimus -- nvd3d9wrapx.dll & nvdxgiwrap.dll
if (GetModuleHandleA("nvumdshim.dll") ||
GetModuleHandleA("nvumdshimx.dll"))
{
knowIsOptimus = 1;
} else {
knowIsOptimus = 0;
}
}
return knowIsOptimus;
}
IDXGIAdapter1*
gfxWindowsPlatform::GetDXGIAdapter()
{
if (mAdapter) {
return mAdapter;
}
nsModuleHandle dxgiModule(LoadLibrarySystem32(L"dxgi.dll"));
decltype(CreateDXGIFactory1)* createDXGIFactory1 = (decltype(CreateDXGIFactory1)*)
GetProcAddress(dxgiModule, "CreateDXGIFactory1");
if (!createDXGIFactory1) {
return nullptr;
}
// Try to use a DXGI 1.1 adapter in order to share resources
// across processes.
RefPtr<IDXGIFactory1> factory1;
HRESULT hr = createDXGIFactory1(__uuidof(IDXGIFactory1),
getter_AddRefs(factory1));
if (FAILED(hr) || !factory1) {
// This seems to happen with some people running the iZ3D driver.
// They won't get acceleration.
return nullptr;
}
if (!XRE_IsContentProcess()) {
// In the parent process, we pick the first adapter.
if (FAILED(factory1->EnumAdapters1(0, getter_AddRefs(mAdapter)))) {
return nullptr;
}
} else {
const DxgiAdapterDesc& parent = GetParentDevicePrefs().adapter();
// In the child process, we search for the adapter that matches the parent
// process. The first adapter can be mismatched on dual-GPU systems.
for (UINT index = 0; ; index++) {
RefPtr<IDXGIAdapter1> adapter;
if (FAILED(factory1->EnumAdapters1(index, getter_AddRefs(adapter)))) {
break;
}
DXGI_ADAPTER_DESC desc;
if (SUCCEEDED(adapter->GetDesc(&desc)) &&
desc.AdapterLuid.HighPart == parent.AdapterLuid.HighPart &&
desc.AdapterLuid.LowPart == parent.AdapterLuid.LowPart &&
desc.VendorId == parent.VendorId &&
desc.DeviceId == parent.DeviceId)
{
mAdapter = adapter.forget();
break;
}
}
}
if (!mAdapter) {
return nullptr;
}
// We leak this module everywhere, we might as well do so here as well.
dxgiModule.disown();
return mAdapter;
}
bool DoesD3D11DeviceWork()
{
static bool checked = false;
static bool result = false;
if (checked)
return result;
checked = true;
if (gfxPrefs::Direct2DForceEnabled() ||
gfxPrefs::LayersAccelerationForceEnabled())
{
result = true;
return true;
}
if (GetModuleHandleW(L"igd10umd32.dll")) {
const wchar_t* checkModules[] = {L"dlumd32.dll",
L"dlumd11.dll",
L"dlumd10.dll"};
for (int i=0; i<PR_ARRAY_SIZE(checkModules); i+=1) {
if (GetModuleHandleW(checkModules[i])) {
nsString displayLinkModuleVersionString;
gfxWindowsPlatform::GetDLLVersion(checkModules[i],
displayLinkModuleVersionString);
uint64_t displayLinkModuleVersion;
if (!ParseDriverVersion(displayLinkModuleVersionString,
&displayLinkModuleVersion)) {
gfxCriticalError() << "DisplayLink: could not parse version "
<< checkModules[i];
gANGLESupportsD3D11 = false;
return false;
}
if (displayLinkModuleVersion <= V(8,6,1,36484)) {
gfxCriticalError(CriticalLog::DefaultOptions(false)) << "DisplayLink: too old version " << displayLinkModuleVersionString.get();
gANGLESupportsD3D11 = false;
return false;
}
}
}
}
result = true;
return true;
}
static bool
GetDxgiDesc(ID3D11Device* device, DXGI_ADAPTER_DESC* out)
{
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<IDXGIDevice> dxgiDevice;
HRESULT hr = device->QueryInterface(__uuidof(IDXGIDevice), getter_AddRefs(dxgiDevice));
if (FAILED(hr)) {
return false;
}
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<IDXGIAdapter> dxgiAdapter;
if (FAILED(dxgiDevice->GetAdapter(getter_AddRefs(dxgiAdapter)))) {
return false;
}
return SUCCEEDED(dxgiAdapter->GetDesc(out));
}
static void
CheckForAdapterMismatch(ID3D11Device *device)
{
DXGI_ADAPTER_DESC desc;
PodZero(&desc);
GetDxgiDesc(device, &desc);
nsCOMPtr<nsIGfxInfo> gfxInfo = services::GetGfxInfo();
nsString vendorID;
gfxInfo->GetAdapterVendorID(vendorID);
nsresult ec;
int32_t vendor = vendorID.ToInteger(&ec, 16);
if (vendor != desc.VendorId) {
gfxCriticalNote << "VendorIDMismatch V " << hexa(vendor) << " " << hexa(desc.VendorId);
}
}
bool DoesRenderTargetViewNeedsRecreating(ID3D11Device *device)
{
bool result = false;
// CreateTexture2D is known to crash on lower feature levels, see bugs
// 1170211 and 1089413.
if (device->GetFeatureLevel() < D3D_FEATURE_LEVEL_10_0) {
return true;
}
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<ID3D11DeviceContext> deviceContext;
device->GetImmediateContext(getter_AddRefs(deviceContext));
int backbufferWidth = 32; int backbufferHeight = 32;
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<ID3D11Texture2D> offscreenTexture;
RefPtr<IDXGIKeyedMutex> keyedMutex;
D3D11_TEXTURE2D_DESC offscreenTextureDesc = { 0 };
offscreenTextureDesc.Width = backbufferWidth;
offscreenTextureDesc.Height = backbufferHeight;
offscreenTextureDesc.Format = DXGI_FORMAT_B8G8R8A8_UNORM;
offscreenTextureDesc.MipLevels = 0;
offscreenTextureDesc.ArraySize = 1;
offscreenTextureDesc.SampleDesc.Count = 1;
offscreenTextureDesc.SampleDesc.Quality = 0;
offscreenTextureDesc.Usage = D3D11_USAGE_DEFAULT;
offscreenTextureDesc.BindFlags = D3D11_BIND_RENDER_TARGET | D3D11_BIND_SHADER_RESOURCE;
offscreenTextureDesc.CPUAccessFlags = 0;
offscreenTextureDesc.MiscFlags = D3D11_RESOURCE_MISC_SHARED_KEYEDMUTEX;
HRESULT hr = device->CreateTexture2D(&offscreenTextureDesc, NULL, getter_AddRefs(offscreenTexture));
if (FAILED(hr)) {
gfxCriticalNote << "DoesRecreatingCreateTexture2DFail";
return false;
}
hr = offscreenTexture->QueryInterface(__uuidof(IDXGIKeyedMutex), (void**)getter_AddRefs(keyedMutex));
if (FAILED(hr)) {
gfxCriticalNote << "DoesRecreatingKeyedMutexFailed";
return false;
}
D3D11_RENDER_TARGET_VIEW_DESC offscreenRTVDesc;
offscreenRTVDesc.Format = DXGI_FORMAT_B8G8R8A8_UNORM;
offscreenRTVDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2D;
offscreenRTVDesc.Texture2D.MipSlice = 0;
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<ID3D11RenderTargetView> offscreenRTView;
hr = device->CreateRenderTargetView(offscreenTexture, &offscreenRTVDesc, getter_AddRefs(offscreenRTView));
if (FAILED(hr)) {
gfxCriticalNote << "DoesRecreatingCreateRenderTargetViewFailed";
return false;
}
// Acquire and clear
keyedMutex->AcquireSync(0, INFINITE);
FLOAT color1[4] = { 1, 1, 0.5, 1 };
deviceContext->ClearRenderTargetView(offscreenRTView, color1);
keyedMutex->ReleaseSync(0);
keyedMutex->AcquireSync(0, INFINITE);
FLOAT color2[4] = { 1, 1, 0, 1 };
deviceContext->ClearRenderTargetView(offscreenRTView, color2);
D3D11_TEXTURE2D_DESC desc;
offscreenTexture->GetDesc(&desc);
desc.Usage = D3D11_USAGE_STAGING;
desc.CPUAccessFlags = D3D11_CPU_ACCESS_READ;
desc.MiscFlags = 0;
desc.BindFlags = 0;
ID3D11Texture2D* cpuTexture;
hr = device->CreateTexture2D(&desc, NULL, &cpuTexture);
if (FAILED(hr)) {
gfxCriticalNote << "DoesRecreatingCreateCPUTextureFailed";
return false;
}
deviceContext->CopyResource(cpuTexture, offscreenTexture);
D3D11_MAPPED_SUBRESOURCE mapped;
hr = deviceContext->Map(cpuTexture, 0, D3D11_MAP_READ, 0, &mapped);
if (FAILED(hr)) {
gfxCriticalNote << "DoesRecreatingMapFailed " << hexa(hr);
return false;
}
int resultColor = *(int*)mapped.pData;
deviceContext->Unmap(cpuTexture, 0);
cpuTexture->Release();
// XXX on some drivers resultColor will not have changed to
// match the clear
if (resultColor != 0xffffff00) {
gfxCriticalNote << "RenderTargetViewNeedsRecreating";
result = true;
}
keyedMutex->ReleaseSync(0);
// It seems like this may only happen when we're using the NVIDIA gpu
CheckForAdapterMismatch(device);
return result;
}
static bool TryCreateTexture2D(ID3D11Device *device,
D3D11_TEXTURE2D_DESC* desc,
D3D11_SUBRESOURCE_DATA* data,
RefPtr<ID3D11Texture2D>& texture)
{
// Older Intel driver version (see bug 1221348 for version #s) crash when
// creating a texture with shared keyed mutex and data.
MOZ_SEH_TRY {
return !FAILED(device->CreateTexture2D(desc, data, getter_AddRefs(texture)));
} MOZ_SEH_EXCEPT(EXCEPTION_EXECUTE_HANDLER) {
// For now we want to aggregrate all the crash signature to a known crash.
MOZ_CRASH("Crash creating texture. See bug 1221348.");
return false;
}
}
// See bug 1083071. On some drivers, Direct3D 11 CreateShaderResourceView fails
// with E_OUTOFMEMORY.
bool DoesD3D11TextureSharingWorkInternal(ID3D11Device *device, DXGI_FORMAT format, UINT bindflags)
{
// CreateTexture2D is known to crash on lower feature levels, see bugs
// 1170211 and 1089413.
if (device->GetFeatureLevel() < D3D_FEATURE_LEVEL_10_0) {
return false;
}
if (gfxPrefs::Direct2DForceEnabled() ||
gfxPrefs::LayersAccelerationForceEnabled())
{
return true;
}
if (GetModuleHandleW(L"atidxx32.dll")) {
nsCOMPtr<nsIGfxInfo> gfxInfo = services::GetGfxInfo();
if (gfxInfo) {
nsString vendorID, vendorID2;
gfxInfo->GetAdapterVendorID(vendorID);
gfxInfo->GetAdapterVendorID2(vendorID2);
if (vendorID.EqualsLiteral("0x8086") && vendorID2.IsEmpty()) {
if (!gfxPrefs::LayersAMDSwitchableGfxEnabled()) {
return false;
}
gfxCriticalError(CriticalLog::DefaultOptions(false)) << "PossiblyBrokenSurfaceSharing_UnexpectedAMDGPU";
}
}
}
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<ID3D11Texture2D> texture;
D3D11_TEXTURE2D_DESC desc;
const int texture_size = 32;
desc.Width = texture_size;
desc.Height = texture_size;
desc.MipLevels = 1;
desc.ArraySize = 1;
desc.Format = format;
desc.SampleDesc.Count = 1;
desc.SampleDesc.Quality = 0;
desc.Usage = D3D11_USAGE_DEFAULT;
desc.CPUAccessFlags = 0;
desc.MiscFlags = D3D11_RESOURCE_MISC_SHARED_KEYEDMUTEX;
desc.BindFlags = bindflags;
uint32_t color[texture_size * texture_size];
for (size_t i = 0; i < sizeof(color)/sizeof(color[0]); i++) {
color[i] = 0xff00ffff;
}
// XXX If we pass the data directly at texture creation time we
// get a crash on Intel 8.5.10.[18xx-1994] drivers.
// We can work around this issue by doing UpdateSubresource.
if (!TryCreateTexture2D(device, &desc, nullptr, texture)) {
gfxCriticalError() << "DoesD3D11TextureSharingWork_TryCreateTextureFailure";
return false;
}
RefPtr<IDXGIKeyedMutex> sourceSharedMutex;
texture->QueryInterface(__uuidof(IDXGIKeyedMutex), (void**)getter_AddRefs(sourceSharedMutex));
if (FAILED(sourceSharedMutex->AcquireSync(0, 30*1000))) {
gfxCriticalError() << "DoesD3D11TextureSharingWork_SourceMutexTimeout";
// only wait for 30 seconds
return false;
}
RefPtr<ID3D11DeviceContext> deviceContext;
device->GetImmediateContext(getter_AddRefs(deviceContext));
int stride = texture_size * 4;
deviceContext->UpdateSubresource(texture, 0, nullptr, color, stride, stride * texture_size);
if (FAILED(sourceSharedMutex->ReleaseSync(0))) {
gfxCriticalError() << "DoesD3D11TextureSharingWork_SourceReleaseSyncTimeout";
return false;
}
HANDLE shareHandle;
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<IDXGIResource> otherResource;
if (FAILED(texture->QueryInterface(__uuidof(IDXGIResource),
getter_AddRefs(otherResource))))
{
gfxCriticalError() << "DoesD3D11TextureSharingWork_GetResourceFailure";
return false;
}
if (FAILED(otherResource->GetSharedHandle(&shareHandle))) {
gfxCriticalError() << "DoesD3D11TextureSharingWork_GetSharedTextureFailure";
return false;
}
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<ID3D11Resource> sharedResource;
RefPtr<ID3D11Texture2D> sharedTexture;
if (FAILED(device->OpenSharedResource(shareHandle, __uuidof(ID3D11Resource),
getter_AddRefs(sharedResource))))
{
gfxCriticalError(CriticalLog::DefaultOptions(false)) << "OpenSharedResource failed for format " << format;
return false;
}
if (FAILED(sharedResource->QueryInterface(__uuidof(ID3D11Texture2D),
getter_AddRefs(sharedTexture))))
{
gfxCriticalError() << "DoesD3D11TextureSharingWork_GetSharedTextureFailure";
return false;
}
// create a staging texture for readback
RefPtr<ID3D11Texture2D> cpuTexture;
desc.Usage = D3D11_USAGE_STAGING;
desc.CPUAccessFlags = D3D11_CPU_ACCESS_READ;
desc.MiscFlags = 0;
desc.BindFlags = 0;
if (FAILED(device->CreateTexture2D(&desc, nullptr, getter_AddRefs(cpuTexture)))) {
gfxCriticalError() << "DoesD3D11TextureSharingWork_CreateTextureFailure";
return false;
}
RefPtr<IDXGIKeyedMutex> sharedMutex;
sharedResource->QueryInterface(__uuidof(IDXGIKeyedMutex), (void**)getter_AddRefs(sharedMutex));
if (FAILED(sharedMutex->AcquireSync(0, 30*1000))) {
gfxCriticalError() << "DoesD3D11TextureSharingWork_AcquireSyncTimeout";
// only wait for 30 seconds
return false;
}
// Copy to the cpu texture so that we can readback
deviceContext->CopyResource(cpuTexture, sharedTexture);
D3D11_MAPPED_SUBRESOURCE mapped;
int resultColor = 0;
if (SUCCEEDED(deviceContext->Map(cpuTexture, 0, D3D11_MAP_READ, 0, &mapped))) {
// read the texture
resultColor = *(int*)mapped.pData;
deviceContext->Unmap(cpuTexture, 0);
} else {
gfxCriticalError() << "DoesD3D11TextureSharingWork_MapFailed";
return false;
}
sharedMutex->ReleaseSync(0);
// check that the color we put in is the color we get out
if (resultColor != color[0]) {
// Shared surfaces seem to be broken on dual AMD & Intel HW when using the
// AMD GPU
gfxCriticalNote << "DoesD3D11TextureSharingWork_ColorMismatch";
return false;
}
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<ID3D11ShaderResourceView> sharedView;
// This if(FAILED()) is the one that actually fails on systems affected by bug 1083071.
if (FAILED(device->CreateShaderResourceView(sharedTexture, NULL, getter_AddRefs(sharedView)))) {
gfxCriticalNote << "CreateShaderResourceView failed for format" << format;
return false;
}
return true;
}
bool DoesD3D11TextureSharingWork(ID3D11Device *device)
{
return DoesD3D11TextureSharingWorkInternal(device, DXGI_FORMAT_B8G8R8A8_UNORM, D3D11_BIND_RENDER_TARGET | D3D11_BIND_SHADER_RESOURCE);
}
bool DoesD3D11AlphaTextureSharingWork(ID3D11Device *device)
{
return DoesD3D11TextureSharingWorkInternal(device, DXGI_FORMAT_R8_UNORM, D3D11_BIND_SHADER_RESOURCE);
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
static inline bool
CanUseWARP()
{
if (gfxPrefs::LayersD3D11ForceWARP()) {
return true;
}
// The child process can only use WARP if the parent process is also using
// WARP.
if (XRE_IsContentProcess()) {
return GetParentDevicePrefs().useD3D11WARP();
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
// It seems like nvdxgiwrap makes a mess of WARP. See bug 1154703.
if (!IsWin8OrLater() ||
gfxPrefs::LayersD3D11DisableWARP() ||
GetModuleHandleA("nvdxgiwrap.dll"))
{
return false;
}
return true;
}
FeatureStatus
gfxWindowsPlatform::CheckD3D11Support(bool* aCanUseHardware)
{
// Don't revive D3D11 support after a failure.
if (IsFeatureStatusFailure(mD3D11Status)) {
return mD3D11Status;
}
if (XRE_IsContentProcess()) {
if (!GetParentDevicePrefs().useD3D11()) {
return FeatureStatus::Blocked;
}
*aCanUseHardware = !GetParentDevicePrefs().useD3D11WARP();
return FeatureStatus::Available;
}
if (gfxPrefs::LayersD3D11ForceWARP()) {
*aCanUseHardware = false;
return FeatureStatus::Available;
}
if (gfxPrefs::LayersAccelerationForceEnabled()) {
*aCanUseHardware = true;
return FeatureStatus::Available;
}
if (nsCOMPtr<nsIGfxInfo> gfxInfo = services::GetGfxInfo()) {
int32_t status;
if (NS_SUCCEEDED(gfxInfo->GetFeatureStatus(nsIGfxInfo::FEATURE_DIRECT3D_11_LAYERS, &status))) {
if (status != nsIGfxInfo::FEATURE_STATUS_OK) {
if (CanUseWARP()) {
*aCanUseHardware = false;
return FeatureStatus::Available;
}
return FeatureStatus::Blacklisted;
}
}
}
// If we've used WARP once, we continue to use it after device resets.
*aCanUseHardware = !mIsWARP;
return FeatureStatus::Available;
}
// We don't have access to the D3D11CreateDevice type in gfxWindowsPlatform.h,
// since it doesn't include d3d11.h, so we use a static here. It should only
// be used within InitializeD3D11.
decltype(D3D11CreateDevice)* sD3D11CreateDeviceFn = nullptr;
bool
gfxWindowsPlatform::AttemptD3D11DeviceCreationHelper(
IDXGIAdapter1* aAdapter, HRESULT& aResOut)
{
MOZ_SEH_TRY {
aResOut =
sD3D11CreateDeviceFn(
aAdapter, D3D_DRIVER_TYPE_UNKNOWN, nullptr,
// Use D3D11_CREATE_DEVICE_PREVENT_INTERNAL_THREADING_OPTIMIZATIONS
// to prevent bug 1092260. IE 11 also uses this flag.
D3D11_CREATE_DEVICE_BGRA_SUPPORT | D3D11_CREATE_DEVICE_PREVENT_INTERNAL_THREADING_OPTIMIZATIONS,
mFeatureLevels.Elements(), mFeatureLevels.Length(),
D3D11_SDK_VERSION, getter_AddRefs(mD3D11Device), nullptr, nullptr);
} MOZ_SEH_EXCEPT (EXCEPTION_EXECUTE_HANDLER) {
return false;
}
return true;
}
FeatureStatus
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
gfxWindowsPlatform::AttemptD3D11DeviceCreation()
{
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<IDXGIAdapter1> adapter = GetDXGIAdapter();
if (!adapter) {
return FeatureStatus::Unavailable;
}
HRESULT hr;
if (!AttemptD3D11DeviceCreationHelper(adapter, hr)) {
gfxCriticalError() << "Crash during D3D11 device creation";
return FeatureStatus::Crashed;
}
if (FAILED(hr) || !mD3D11Device) {
mD3D11Device = nullptr;
gfxCriticalError() << "D3D11 device creation failed: " << hexa(hr);
return FeatureStatus::Failed;
}
if (!DoesD3D11DeviceWork()) {
mD3D11Device = nullptr;
return FeatureStatus::Blocked;
}
if (!mD3D11Device) {
return FeatureStatus::Failed;
}
// Only test this when not using WARP since it can fail and cause
// GetDeviceRemovedReason to return weird values.
mCompositorD3D11TextureSharingWorks = ::DoesD3D11TextureSharingWork(mD3D11Device);
if (!mCompositorD3D11TextureSharingWorks || !DoesRenderTargetViewNeedsRecreating(mD3D11Device)) {
gANGLESupportsD3D11 = false;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
mD3D11Device->SetExceptionMode(0);
mIsWARP = false;
return FeatureStatus::Available;
}
bool
gfxWindowsPlatform::AttemptWARPDeviceCreationHelper(
ScopedGfxFeatureReporter& aReporterWARP, HRESULT& aResOut)
{
MOZ_SEH_TRY {
aResOut =
sD3D11CreateDeviceFn(
nullptr, D3D_DRIVER_TYPE_WARP, nullptr,
// Use D3D11_CREATE_DEVICE_PREVENT_INTERNAL_THREADING_OPTIMIZATIONS
// to prevent bug 1092260. IE 11 also uses this flag.
D3D11_CREATE_DEVICE_BGRA_SUPPORT,
mFeatureLevels.Elements(), mFeatureLevels.Length(),
D3D11_SDK_VERSION, getter_AddRefs(mD3D11Device), nullptr, nullptr);
aReporterWARP.SetSuccessful();
} MOZ_SEH_EXCEPT (EXCEPTION_EXECUTE_HANDLER) {
return false;
}
return true;
}
FeatureStatus
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
gfxWindowsPlatform::AttemptWARPDeviceCreation()
{
ScopedGfxFeatureReporter reporterWARP("D3D11-WARP", gfxPrefs::LayersD3D11ForceWARP());
HRESULT hr;
if (!AttemptWARPDeviceCreationHelper(reporterWARP, hr)) {
gfxCriticalError() << "Exception occurred initializing WARP D3D11 device!";
return FeatureStatus::Crashed;
}
if (FAILED(hr) || !mD3D11Device) {
// This should always succeed... in theory.
gfxCriticalError() << "Failed to initialize WARP D3D11 device! " << hexa(hr);
return FeatureStatus::Failed;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
// Only test for texture sharing on Windows 8 since it puts the device into
// an unusable state if used on Windows 7
if (IsWin8OrLater()) {
mCompositorD3D11TextureSharingWorks = ::DoesD3D11TextureSharingWork(mD3D11Device);
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
}
mD3D11Device->SetExceptionMode(0);
mIsWARP = true;
return FeatureStatus::Available;
}
bool
gfxWindowsPlatform::ContentAdapterIsParentAdapter(ID3D11Device* device)
{
DXGI_ADAPTER_DESC desc;
if (!GetDxgiDesc(device, &desc)) {
gfxCriticalNote << "Could not query device DXGI adapter info";
return false;
}
const DxgiAdapterDesc& parent = GetParentDevicePrefs().adapter();
if (desc.VendorId != parent.VendorId ||
desc.DeviceId != parent.DeviceId ||
desc.SubSysId != parent.SubSysId ||
desc.AdapterLuid.HighPart != parent.AdapterLuid.HighPart ||
desc.AdapterLuid.LowPart != parent.AdapterLuid.LowPart)
{
gfxCriticalNote << "VendorIDMismatch P " << hexa(parent.VendorId) << " " << hexa(desc.VendorId);
return false;
}
return true;
}
bool
gfxWindowsPlatform::AttemptD3D11ContentDeviceCreationHelper(
IDXGIAdapter1* aAdapter, HRESULT& aResOut)
{
MOZ_SEH_TRY {
aResOut =
sD3D11CreateDeviceFn(
aAdapter, mIsWARP ? D3D_DRIVER_TYPE_WARP : D3D_DRIVER_TYPE_UNKNOWN,
nullptr, D3D11_CREATE_DEVICE_BGRA_SUPPORT,
mFeatureLevels.Elements(), mFeatureLevels.Length(),
D3D11_SDK_VERSION, getter_AddRefs(mD3D11ContentDevice), nullptr, nullptr);
} MOZ_SEH_EXCEPT (EXCEPTION_EXECUTE_HANDLER) {
return false;
}
return true;
}
FeatureStatus
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
gfxWindowsPlatform::AttemptD3D11ContentDeviceCreation()
{
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<IDXGIAdapter1> adapter;
if (!mIsWARP) {
adapter = GetDXGIAdapter();
if (!adapter) {
return FeatureStatus::Unavailable;
}
}
HRESULT hr;
if (!AttemptD3D11ContentDeviceCreationHelper(adapter, hr)) {
return FeatureStatus::Crashed;
}
if (FAILED(hr) || !mD3D11ContentDevice) {
return FeatureStatus::Failed;
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
}
// InitializeD2D() will abort early if the compositor device did not support
// texture sharing. If we're in the content process, we can't rely on the
// parent device alone: some systems have dual GPUs that are capable of
// binding the parent and child processes to different GPUs. As a safety net,
// we re-check texture sharing against the newly created D3D11 content device.
// If it fails, we won't use Direct2D.
if (XRE_IsContentProcess()) {
if (!DoesD3D11TextureSharingWork(mD3D11ContentDevice)) {
mD3D11ContentDevice = nullptr;
return FeatureStatus::Failed;
}
DebugOnly<bool> ok = ContentAdapterIsParentAdapter(mD3D11ContentDevice);
MOZ_ASSERT(ok);
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
mD3D11ContentDevice->SetExceptionMode(0);
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<ID3D10Multithread> multi;
hr = mD3D11ContentDevice->QueryInterface(__uuidof(ID3D10Multithread), getter_AddRefs(multi));
if (SUCCEEDED(hr) && multi) {
multi->SetMultithreadProtected(TRUE);
}
return FeatureStatus::Available;
}
FeatureStatus
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
gfxWindowsPlatform::AttemptD3D11ImageBridgeDeviceCreation()
{
HRESULT hr = E_INVALIDARG;
MOZ_SEH_TRY{
hr =
sD3D11CreateDeviceFn(GetDXGIAdapter(), D3D_DRIVER_TYPE_UNKNOWN, nullptr,
D3D11_CREATE_DEVICE_BGRA_SUPPORT,
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
mFeatureLevels.Elements(), mFeatureLevels.Length(),
D3D11_SDK_VERSION, getter_AddRefs(mD3D11ImageBridgeDevice), nullptr, nullptr);
} MOZ_SEH_EXCEPT(EXCEPTION_EXECUTE_HANDLER) {
return FeatureStatus::Crashed;
}
if (FAILED(hr) || !mD3D11ImageBridgeDevice) {
return FeatureStatus::Failed;
}
mD3D11ImageBridgeDevice->SetExceptionMode(0);
if (!DoesD3D11AlphaTextureSharingWork(mD3D11ImageBridgeDevice)) {
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
mD3D11ImageBridgeDevice = nullptr;
return FeatureStatus::Failed;
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
}
if (XRE_IsContentProcess()) {
ContentAdapterIsParentAdapter(mD3D11ImageBridgeDevice);
}
return FeatureStatus::Available;
}
void
gfxWindowsPlatform::SetDeviceInitData(mozilla::gfx::DeviceInitData& aData)
{
MOZ_ASSERT(XRE_IsContentProcess());
sDeviceInitDataDoNotUseDirectly = aData;
}
void
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
gfxWindowsPlatform::InitializeDevices()
{
// If acceleration is disabled, we refuse to initialize anything.
mAcceleration = CheckAccelerationSupport();
if (IsFeatureStatusFailure(mAcceleration)) {
return;
}
// If we previously crashed initializing devices, bail out now. This is
// effectively a parent-process only check, since the content process
// cannot create a lock file.
D3D11LayersCrashGuard detectCrashes;
if (detectCrashes.Crashed()) {
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
mAcceleration = FeatureStatus::Blocked;
return;
}
// If we're going to prefer D3D9, stop here. The rest of this function
// attempts to use D3D11 features.
if (gfxPrefs::LayersPreferD3D9()) {
mD3D11Status = FeatureStatus::Disabled;
return;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
// First, initialize D3D11. If this succeeds we attempt to use Direct2D.
InitializeD3D11();
// Initialize Direct2D.
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
if (mD3D11Status == FeatureStatus::Available) {
InitializeD2D();
}
// Usually we want D2D in order to use DWrite, but if the users have it
// forced, we'll let them have it, as unsupported configuration.
if (gfxPrefs::DirectWriteFontRenderingForceEnabled() &&
IsFeatureStatusFailure(mD2D1Status) &&
!mDWriteFactory) {
gfxCriticalNote << "Attempting DWrite without D2D support";
InitDWriteSupport();
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
}
FeatureStatus
gfxWindowsPlatform::CheckAccelerationSupport()
{
// Don't retry acceleration if it failed earlier.
if (IsFeatureStatusFailure(mAcceleration)) {
return mAcceleration;
}
if (XRE_IsContentProcess()) {
return GetParentDevicePrefs().useAcceleration()
? FeatureStatus::Available
: FeatureStatus::Blocked;
}
if (InSafeMode()) {
return FeatureStatus::Blocked;
}
if (!ShouldUseLayersAcceleration()) {
return FeatureStatus::Disabled;
}
return FeatureStatus::Available;
}
bool
gfxWindowsPlatform::CanUseD3D11ImageBridge()
{
if (XRE_IsContentProcess()) {
if (!GetParentDevicePrefs().useD3D11ImageBridge()) {
return false;
}
}
return !mIsWARP;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
void
gfxWindowsPlatform::InitializeD3D11()
{
// This function attempts to initialize our D3D11 devices, if the hardware
// is not blacklisted for D3D11 layers. This first attempt will try to create
// a hardware accelerated device. If this creation fails or the hardware is
// blacklisted, then this function will abort if WARP is disabled, causing us
// to fallback to D3D9 or Basic layers. If WARP is not disabled it will use
// a WARP device which should always be available on Windows 7 and higher.
// Check if D3D11 is supported on this hardware.
bool canUseHardware = true;
mD3D11Status = CheckD3D11Support(&canUseHardware);
if (IsFeatureStatusFailure(mD3D11Status)) {
return;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
// Check if D3D11 is available on this system.
nsModuleHandle d3d11Module(LoadLibrarySystem32(L"d3d11.dll"));
sD3D11CreateDeviceFn =
(decltype(D3D11CreateDevice)*)GetProcAddress(d3d11Module, "D3D11CreateDevice");
if (!sD3D11CreateDeviceFn) {
// We should just be on Windows Vista or XP in this case.
mD3D11Status = FeatureStatus::Unavailable;
return;
}
// Check if a failure was injected for testing.
if (gfxPrefs::DeviceFailForTesting()) {
mD3D11Status = FeatureStatus::Failed;
return;
}
if (XRE_IsParentProcess()) {
// First try to create a hardware accelerated device.
if (canUseHardware) {
mD3D11Status = AttemptD3D11DeviceCreation();
if (mD3D11Status == FeatureStatus::Crashed) {
return;
}
}
// If that failed, see if we can use WARP.
if (!mD3D11Device) {
if (!CanUseWARP()) {
mD3D11Status = FeatureStatus::Blocked;
return;
}
mD3D11Status = AttemptWARPDeviceCreation();
}
// If we still have no device by now, exit.
if (!mD3D11Device) {
MOZ_ASSERT(IsFeatureStatusFailure(mD3D11Status));
return;
}
RefPtr<ID3D10Multithread> multi;
HRESULT hr = mD3D11Device->QueryInterface(__uuidof(ID3D10Multithread), getter_AddRefs(multi));
if (SUCCEEDED(hr) && multi) {
multi->SetMultithreadProtected(TRUE);
}
// Either device creation function should have returned Available.
MOZ_ASSERT(mD3D11Status == FeatureStatus::Available);
} else {
// Child processes do not need a compositor, but they do need to know
// whether the parent process is using WARP and whether or not texture
// sharing works.
mIsWARP = !canUseHardware;
mCompositorD3D11TextureSharingWorks = GetParentDevicePrefs().d3d11TextureSharingWorks();
mD3D11Status = FeatureStatus::Available;
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
}
if (CanUseD3D11ImageBridge()) {
if (AttemptD3D11ImageBridgeDeviceCreation() == FeatureStatus::Crashed) {
DisableD3D11AfterCrash();
return;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
}
if (AttemptD3D11ContentDeviceCreation() == FeatureStatus::Crashed) {
DisableD3D11AfterCrash();
return;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
// We leak these everywhere and we need them our entire runtime anyway, let's
// leak it here as well. We keep the pointer to sD3D11CreateDeviceFn around
// as well for D2D1 and device resets.
d3d11Module.disown();
}
void
gfxWindowsPlatform::DisableD3D11AfterCrash()
{
mD3D11Status = FeatureStatus::Crashed;
ResetD3D11Devices();
}
void
gfxWindowsPlatform::ResetD3D11Devices()
{
mD3D11Device = nullptr;
mD3D11ContentDevice = nullptr;
mD3D11ImageBridgeDevice = nullptr;
Factory::SetDirect3D11Device(nullptr);
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
static bool
IsD2DBlacklisted()
{
nsCOMPtr<nsIGfxInfo> gfxInfo = services::GetGfxInfo();
if (gfxInfo) {
int32_t status;
if (NS_SUCCEEDED(gfxInfo->GetFeatureStatus(nsIGfxInfo::FEATURE_DIRECT2D, &status))) {
if (status != nsIGfxInfo::FEATURE_STATUS_OK) {
return true;
}
}
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
return false;
}
// Check whether we can support Direct2D. Although some of these checks will
// not change after a TDR (like the OS version), we could find a driver change
// that runs us into the blacklist.
FeatureStatus
gfxWindowsPlatform::CheckD2D1Support()
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
{
// Don't revive D2D1 support after a failure.
if (IsFeatureStatusFailure(mD2D1Status)) {
return mD2D1Status;
}
if (!gfxPrefs::Direct2DForceEnabled() && IsD2DBlacklisted()) {
return FeatureStatus::Blacklisted;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
// Do not ever try to use D2D if it's explicitly disabled.
if (gfxPrefs::Direct2DDisabled()) {
return FeatureStatus::Disabled;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
// Direct2D is only Vista or higher, but we require a D3D11 compositor to
// use it. (This check may be implied by the fact that we do not get here
// without a D3D11 compositor device.)
if (!IsVistaOrLater()) {
return FeatureStatus::Unavailable;
}
// Normally we don't use D2D content drawing when using WARP. However if
// WARP is force-enabled, we will let Direct2D use WARP as well.
if (mIsWARP && !gfxPrefs::LayersD3D11ForceWARP()) {
return FeatureStatus::Blocked;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
if (!Factory::SupportsD2D1()) {
return FeatureStatus::Unavailable;
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
}
if (XRE_IsContentProcess()) {
return GetParentDevicePrefs().useD2D1()
? FeatureStatus::Available
: FeatureStatus::Blocked;
}
return FeatureStatus::Available;
}
void
gfxWindowsPlatform::InitializeD2D()
{
ScopedGfxFeatureReporter d2d1_1("D2D1.1");
mD2D1Status = CheckD2D1Support();
if (IsFeatureStatusFailure(mD2D1Status)) {
return;
}
if (!mCompositorD3D11TextureSharingWorks) {
mD2D1Status = FeatureStatus::Failed;
return;
}
// Using Direct2D depends on DWrite support.
if (!mDWriteFactory && !InitDWriteSupport()) {
mD2D1Status = FeatureStatus::Failed;
return;
}
if (!mD3D11ContentDevice) {
mD2D1Status = FeatureStatus::Failed;
return;
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
}
mD2D1Status = FeatureStatus::Available;
Factory::SetDirect3D11Device(mD3D11ContentDevice);
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
d2d1_1.SetSuccessful();
mD2D1Status = FeatureStatus::Available;
}
bool
gfxWindowsPlatform::CreateD3D11DecoderDeviceHelper(
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
IDXGIAdapter1* aAdapter, RefPtr<ID3D11Device>& aDevice, HRESULT& aResOut)
{
MOZ_SEH_TRY{
aResOut =
sD3D11CreateDeviceFn(
aAdapter, D3D_DRIVER_TYPE_UNKNOWN, nullptr,
D3D11_CREATE_DEVICE_VIDEO_SUPPORT,
mFeatureLevels.Elements(), mFeatureLevels.Length(),
D3D11_SDK_VERSION, getter_AddRefs(aDevice), nullptr, nullptr);
} MOZ_SEH_EXCEPT(EXCEPTION_EXECUTE_HANDLER) {
return false;
}
return true;
}
already_AddRefed<ID3D11Device>
gfxWindowsPlatform::CreateD3D11DecoderDevice()
{
if (!sD3D11CreateDeviceFn) {
// We should just be on Windows Vista or XP in this case.
return nullptr;
}
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<IDXGIAdapter1> adapter = GetDXGIAdapter();
if (!adapter) {
return nullptr;
}
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<ID3D11Device> device;
HRESULT hr;
if (!CreateD3D11DecoderDeviceHelper(adapter, device, hr)) {
return nullptr;
}
if (FAILED(hr) || !device || !DoesD3D11DeviceWork()) {
return nullptr;
}
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<ID3D10Multithread> multi;
device->QueryInterface(__uuidof(ID3D10Multithread), getter_AddRefs(multi));
multi->SetMultithreadProtected(TRUE);
return device.forget();
}
static bool
DwmCompositionEnabled()
{
MOZ_ASSERT(WinUtils::dwmIsCompositionEnabledPtr);
BOOL dwmEnabled = false;
WinUtils::dwmIsCompositionEnabledPtr(&dwmEnabled);
return dwmEnabled;
}
class D3DVsyncSource final : public VsyncSource
{
public:
class D3DVsyncDisplay final : public VsyncSource::Display
{
NS_INLINE_DECL_THREADSAFE_REFCOUNTING(D3DVsyncDisplay)
public:
D3DVsyncDisplay()
: mPrevVsync(TimeStamp::Now())
, mVsyncEnabledLock("D3DVsyncEnabledLock")
, mVsyncEnabled(false)
{
mVsyncThread = new base::Thread("WindowsVsyncThread");
const double rate = 1000 / 60.0;
mSoftwareVsyncRate = TimeDuration::FromMilliseconds(rate);
MOZ_RELEASE_ASSERT(mVsyncThread->Start(), "Could not start Windows vsync thread");
SetVsyncRate();
}
void SetVsyncRate()
{
if (!DwmCompositionEnabled()) {
mVsyncRate = TimeDuration::FromMilliseconds(1000.0 / 60.0);
return;
}
DWM_TIMING_INFO vblankTime;
// Make sure to init the cbSize, otherwise GetCompositionTiming will fail
vblankTime.cbSize = sizeof(DWM_TIMING_INFO);
HRESULT hr = WinUtils::dwmGetCompositionTimingInfoPtr(0, &vblankTime);
if (SUCCEEDED(hr)) {
UNSIGNED_RATIO refreshRate = vblankTime.rateRefresh;
// We get the rate in hertz / time, but we want the rate in ms.
float rate = ((float) refreshRate.uiDenominator
/ (float) refreshRate.uiNumerator) * 1000;
mVsyncRate = TimeDuration::FromMilliseconds(rate);
} else {
mVsyncRate = TimeDuration::FromMilliseconds(1000.0 / 60.0);
}
}
virtual void EnableVsync() override
{
MOZ_ASSERT(NS_IsMainThread());
MOZ_ASSERT(mVsyncThread->IsRunning());
{ // scope lock
MonitorAutoLock lock(mVsyncEnabledLock);
if (mVsyncEnabled) {
return;
}
mVsyncEnabled = true;
}
CancelableTask* vsyncStart = NewRunnableMethod(this,
&D3DVsyncDisplay::VBlankLoop);
mVsyncThread->message_loop()->PostTask(FROM_HERE, vsyncStart);
}
virtual void DisableVsync() override
{
MOZ_ASSERT(NS_IsMainThread());
MOZ_ASSERT(mVsyncThread->IsRunning());
MonitorAutoLock lock(mVsyncEnabledLock);
if (!mVsyncEnabled) {
return;
}
mVsyncEnabled = false;
}
virtual bool IsVsyncEnabled() override
{
MOZ_ASSERT(NS_IsMainThread());
MonitorAutoLock lock(mVsyncEnabledLock);
return mVsyncEnabled;
}
virtual TimeDuration GetVsyncRate() override
{
return mVsyncRate;
}
void ScheduleSoftwareVsync(TimeStamp aVsyncTimestamp)
{
MOZ_ASSERT(IsInVsyncThread());
NS_WARNING("DwmComposition dynamically disabled, falling back to software timers");
TimeStamp nextVsync = aVsyncTimestamp + mSoftwareVsyncRate;
TimeDuration delay = nextVsync - TimeStamp::Now();
if (delay.ToMilliseconds() < 0) {
delay = mozilla::TimeDuration::FromMilliseconds(0);
}
mVsyncThread->message_loop()->PostDelayedTask(FROM_HERE,
NewRunnableMethod(this, &D3DVsyncDisplay::VBlankLoop),
delay.ToMilliseconds());
}
TimeStamp GetAdjustedVsyncTimeStamp(LARGE_INTEGER& aFrequency,
QPC_TIME& aQpcVblankTime)
{
TimeStamp vsync = TimeStamp::Now();
LARGE_INTEGER qpcNow;
QueryPerformanceCounter(&qpcNow);
const int microseconds = 1000000;
int64_t adjust = qpcNow.QuadPart - aQpcVblankTime;
int64_t usAdjust = (adjust * microseconds) / aFrequency.QuadPart;
vsync -= TimeDuration::FromMicroseconds((double) usAdjust);
if (IsWin10OrLater()) {
// On Windows 10 and on, DWMGetCompositionTimingInfo, mostly
// reports the upcoming vsync time, which is in the future.
// It can also sometimes report a vblank time in the past.
// Since large parts of Gecko assume TimeStamps can't be in future,
// use the previous vsync.
// Windows 10 and Intel HD vsync timestamps are messy and
// all over the place once in a while. Most of the time,
// it reports the upcoming vsync. Sometimes, that upcoming
// vsync is in the past. Sometimes that upcoming vsync is before
// the previously seen vsync. Sometimes, the previous vsync
// is still in the future. In these error cases,
// we try to normalize to Now().
TimeStamp upcomingVsync = vsync;
if (upcomingVsync < mPrevVsync) {
// Windows can report a vsync that's before
// the previous one. So update it to sometime in the future.
upcomingVsync = TimeStamp::Now() + TimeDuration::FromMilliseconds(1);
}
vsync = mPrevVsync;
mPrevVsync = upcomingVsync;
}
// On Windows 7 and 8, DwmFlush wakes up AFTER qpcVBlankTime
// from DWMGetCompositionTimingInfo. We can return the adjusted vsync.
// Once in a while, the reported vsync timestamp can be in the future.
// Normalize the reported timestamp to now.
if (vsync >= TimeStamp::Now()) {
vsync = TimeStamp::Now();
}
return vsync;
}
void VBlankLoop()
{
MOZ_ASSERT(IsInVsyncThread());
MOZ_ASSERT(sizeof(int64_t) == sizeof(QPC_TIME));
DWM_TIMING_INFO vblankTime;
// Make sure to init the cbSize, otherwise GetCompositionTiming will fail
vblankTime.cbSize = sizeof(DWM_TIMING_INFO);
LARGE_INTEGER frequency;
QueryPerformanceFrequency(&frequency);
TimeStamp vsync = TimeStamp::Now();
// On Windows 10, DwmGetCompositionInfo returns the upcoming vsync.
// See GetAdjustedVsyncTimestamp.
// On start, set mPrevVsync to the "next" vsync
// So we'll use this timestamp on the 2nd loop iteration.
mPrevVsync = vsync + mSoftwareVsyncRate;
for (;;) {
{ // scope lock
MonitorAutoLock lock(mVsyncEnabledLock);
if (!mVsyncEnabled) return;
}
// Large parts of gecko assume that the refresh driver timestamp
// must be <= Now() and cannot be in the future.
MOZ_ASSERT(vsync <= TimeStamp::Now());
Display::NotifyVsync(vsync);
// DwmComposition can be dynamically enabled/disabled
// so we have to check every time that it's available.
// When it is unavailable, we fallback to software but will try
// to get back to dwm rendering once it's re-enabled
if (!DwmCompositionEnabled()) {
ScheduleSoftwareVsync(vsync);
return;
}
// Use a combination of DwmFlush + DwmGetCompositionTimingInfoPtr
// Using WaitForVBlank, the whole system dies :/
WinUtils::dwmFlushProcPtr();
HRESULT hr = WinUtils::dwmGetCompositionTimingInfoPtr(0, &vblankTime);
vsync = TimeStamp::Now();
if (SUCCEEDED(hr)) {
vsync = GetAdjustedVsyncTimeStamp(frequency, vblankTime.qpcVBlank);
}
} // end for
}
private:
virtual ~D3DVsyncDisplay()
{
MOZ_ASSERT(NS_IsMainThread());
DisableVsync();
mVsyncThread->Stop();
delete mVsyncThread;
}
bool IsInVsyncThread()
{
return mVsyncThread->thread_id() == PlatformThread::CurrentId();
}
TimeDuration mSoftwareVsyncRate;
TimeStamp mPrevVsync; // Only used on Windows 10
Monitor mVsyncEnabledLock;
base::Thread* mVsyncThread;
TimeDuration mVsyncRate;
bool mVsyncEnabled;
}; // end d3dvsyncdisplay
D3DVsyncSource()
{
mPrimaryDisplay = new D3DVsyncDisplay();
}
virtual Display& GetGlobalDisplay() override
{
return *mPrimaryDisplay;
}
private:
virtual ~D3DVsyncSource()
{
}
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<D3DVsyncDisplay> mPrimaryDisplay;
}; // end D3DVsyncSource
already_AddRefed<mozilla::gfx::VsyncSource>
gfxWindowsPlatform::CreateHardwareVsyncSource()
{
MOZ_RELEASE_ASSERT(NS_IsMainThread());
if (!WinUtils::dwmIsCompositionEnabledPtr) {
NS_WARNING("Dwm composition not available, falling back to software vsync");
return gfxPlatform::CreateHardwareVsyncSource();
}
BOOL dwmEnabled = false;
WinUtils::dwmIsCompositionEnabledPtr(&dwmEnabled);
if (!dwmEnabled) {
NS_WARNING("DWM not enabled, falling back to software vsync");
return gfxPlatform::CreateHardwareVsyncSource();
}
Bug 1207245 - part 6 - rename nsRefPtr<T> to RefPtr<T>; r=ehsan; a=Tomcat The bulk of this commit was generated with a script, executed at the top level of a typical source code checkout. The only non-machine-generated part was modifying MFBT's moz.build to reflect the new naming. CLOSED TREE makes big refactorings like this a piece of cake. # The main substitution. find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \ xargs perl -p -i -e ' s/nsRefPtr\.h/RefPtr\.h/g; # handle includes s/nsRefPtr ?</RefPtr</g; # handle declarations and variables ' # Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h. perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h # Handle nsRefPtr.h itself, a couple places that define constructors # from nsRefPtr, and code generators specially. We do this here, rather # than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename # things like nsRefPtrHashtable. perl -p -i -e 's/nsRefPtr/RefPtr/g' \ mfbt/nsRefPtr.h \ xpcom/glue/nsCOMPtr.h \ xpcom/base/OwningNonNull.h \ ipc/ipdl/ipdl/lower.py \ ipc/ipdl/ipdl/builtin.py \ dom/bindings/Codegen.py \ python/lldbutils/lldbutils/utils.py # In our indiscriminate substitution above, we renamed # nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up. find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \ xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g' if [ -d .git ]; then git mv mfbt/nsRefPtr.h mfbt/RefPtr.h else hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h fi --HG-- rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
2015-10-18 08:24:48 +03:00
RefPtr<VsyncSource> d3dVsyncSource = new D3DVsyncSource();
return d3dVsyncSource.forget();
}
bool
gfxWindowsPlatform::SupportsApzTouchInput() const
{
int value = gfxPrefs::TouchEventsEnabled();
return value == 1 || value == 2;
}
void
gfxWindowsPlatform::GetAcceleratedCompositorBackends(nsTArray<LayersBackend>& aBackends)
{
if (gfxPrefs::LayersPreferOpenGL()) {
aBackends.AppendElement(LayersBackend::LAYERS_OPENGL);
}
if (!gfxPrefs::LayersPreferD3D9()) {
if (gfxPlatform::CanUseDirect3D11() && GetD3D11Device()) {
aBackends.AppendElement(LayersBackend::LAYERS_D3D11);
} else {
NS_WARNING("Direct3D 11-accelerated layers are not supported on this system.");
}
}
if (gfxPrefs::LayersPreferD3D9() || !IsVistaOrLater()) {
// We don't want D3D9 except on Windows XP
if (gfxPlatform::CanUseDirect3D9()) {
aBackends.AppendElement(LayersBackend::LAYERS_D3D9);
} else {
NS_WARNING("Direct3D 9-accelerated layers are not supported on this system.");
}
}
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
// Some features are dependent on other features. If this is the case, we
// try to propagate the status of the parent feature if it wasn't available.
FeatureStatus
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
gfxWindowsPlatform::GetD3D11Status() const
{
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
if (mAcceleration != FeatureStatus::Available) {
return mAcceleration;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
return mD3D11Status;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
FeatureStatus
gfxWindowsPlatform::GetD2D1Status() const
{
if (GetD3D11Status() != FeatureStatus::Available) {
return FeatureStatus::Unavailable;
}
Refactor graphics device initialization on Windows. (bug 1183910 part 1, r=mattwoodrow,bas) This patch addresses a number of inconsistencies in the device initialization process, as well as simplifying it for future use. All device decisions are now explicitly made up-front during startup, rather than implicitly or on-demand. In addition a number of restrictions have been placed on when we can construct devices. Full change list: (1) We no longer attempt to use D3D11 if acceleration is disabled or D3D9 is preferred. This is a departure from our previous behavior, where we would construct these devices but then not use them as a compositor backend. (2) D3D11 startup no longer creates a content device (this is reserved for D2D initialization). (3) D2D is only attempted if we managed to create a D3D11 compositor device. This is a departure from previous behavior where if D3D11 was not used for compositing, we could still create its machinery to use D2D as a content backend. (4) D2D 1.1 initialization is now directly responsible for creating a D3D11 content device. (5) D2D 1.0 and 1.1 logic have been disentangled for clarity. (6) UpdateRenderMode() has been split up, so we can update backend prefs out of band with device resets. (7) mUseGDIFonts and mUseDirectWrite have been removed as their state was confusing. Instead, D2D now depends on DWrite initialization succeeding. If later we fail to get a DWrite font list, we revert our decision to use Direct2D. (8) Device resets now clear a little more state, including the devices set in Moz2D Factory. (9) We no longer create a DWrite text analyzer as it was unused.
2015-07-29 02:52:54 +03:00
return mD2D1Status;
}
unsigned
gfxWindowsPlatform::GetD3D11Version()
{
ID3D11Device* device = GetD3D11Device();
if (!device) {
return 0;
}
return device->GetFeatureLevel();
}
void
gfxWindowsPlatform::GetDeviceInitData(DeviceInitData* aOut)
{
// Check for device resets before giving back new graphics information.
UpdateRenderMode();
gfxPlatform::GetDeviceInitData(aOut);
// IPDL initializes each field to false for us so we can early return.
if (GetD3D11Status() != FeatureStatus::Available) {
return;
}
aOut->useD3D11() = true;
aOut->useD3D11ImageBridge() = !!mD3D11ImageBridgeDevice;
aOut->d3d11TextureSharingWorks() = mCompositorD3D11TextureSharingWorks;
aOut->useD3D11WARP() = mIsWARP;
aOut->useD2D1() = (GetD2D1Status() == FeatureStatus::Available);
if (mD3D11Device) {
DXGI_ADAPTER_DESC desc;
if (!GetDxgiDesc(mD3D11Device, &desc)) {
return;
}
aOut->adapter() = DxgiAdapterDesc::From(desc);
}
}
bool
gfxWindowsPlatform::SupportsPluginDirectDXGIDrawing()
{
if (!GetD3D11ContentDevice() || !CompositorD3D11TextureSharingWorks()) {
return false;
}
return true;
}