March 2022 - Wine-devel - list.winehq.org

[PATCH 1/2] qedit/tests: Check the Sample Grabber's buffer during and after flushing.
by Gabriel Ivăncescu 04 Oct '22

04 Oct '22

Signed-off-by: Gabriel Ivăncescu <gabrielopcode(a)gmail.com> --- dlls/qedit/tests/samplegrabber.c | 234 +++++++++++++++++++++++++++++++ 1 file changed, 234 insertions(+) diff --git a/dlls/qedit/tests/samplegrabber.c b/dlls/qedit/tests/samplegrabber.c index 8e68e3e..e66180f 100644 --- a/dlls/qedit/tests/samplegrabber.c +++ b/dlls/qedit/tests/samplegrabber.c @@ -1036,6 +1036,239 @@ static void test_connect_pin(void) ok(!ref, "Got outstanding refcount %d.\n", ref); } +struct frame_thread_params +{ + IMemInputPin *sink; + IMediaSample *sample; + DWORD delay; +}; + +static DWORD WINAPI frame_thread(void *arg) +{ + struct frame_thread_params *params = arg; + HRESULT hr; + + if (params->delay) + { + if (winetest_debug > 1) trace("%04x: Sleeping %u ms.\n", GetCurrentThreadId(), params->delay); + Sleep(params->delay); + } + if (winetest_debug > 1) trace("%04x: Sending frame.\n", GetCurrentThreadId()); + hr = IMemInputPin_Receive(params->sink, params->sample); + if (winetest_debug > 1) trace("%04x: Returned %#x.\n", GetCurrentThreadId(), hr); + IMediaSample_Release(params->sample); + free(params); + return hr; +} + +static HANDLE send_frame_time(IMemInputPin *sink, REFERENCE_TIME start_time, unsigned char color, DWORD delay) +{ + struct frame_thread_params *params = malloc(sizeof(*params)); + IMemAllocator *allocator; + REFERENCE_TIME end_time; + IMediaSample *sample; + HANDLE thread; + HRESULT hr; + BYTE *data; + + hr = IMemInputPin_GetAllocator(sink, &allocator); + ok(hr == S_OK, "Got hr %#x.\n", hr); + + hr = IMemAllocator_GetBuffer(allocator, &sample, NULL, NULL, 0); + if (hr == VFW_E_NOT_COMMITTED) + { + IMemAllocator_Commit(allocator); + hr = IMemAllocator_GetBuffer(allocator, &sample, NULL, NULL, 0); + } + ok(hr == S_OK, "Got hr %#x.\n", hr); + + hr = IMediaSample_GetPointer(sample, &data); + ok(hr == S_OK, "Got hr %#x.\n", hr); + memset(data, 0x55, 32 * 16 * 2); + + hr = IMediaSample_SetActualDataLength(sample, 32 * 16 * 2); + ok(hr == S_OK, "Got hr %#x.\n", hr); + + start_time *= 10000000; + end_time = start_time + 10000000; + hr = IMediaSample_SetTime(sample, &start_time, &end_time); + ok(hr == S_OK, "Got hr %#x.\n", hr); + + params->sink = sink; + params->sample = sample; + params->delay = delay; + thread = CreateThread(NULL, 0, frame_thread, params, 0, NULL); + + IMemAllocator_Release(allocator); + return thread; +} + +static HANDLE send_frame(IMemInputPin *sink, DWORD delay) +{ + return send_frame_time(sink, 0, 0x55, delay); /* purple */ +} + +static HRESULT join_thread_(int line, HANDLE thread) +{ + DWORD ret; + ok_(__FILE__, line)(!WaitForSingleObject(thread, 1000), "Wait failed.\n"); + GetExitCodeThread(thread, &ret); + CloseHandle(thread); + return ret; +} +#define join_thread(a) join_thread_(__LINE__, a) + +static void test_buffer_flush(void) +{ + AM_MEDIA_TYPE req_mt = + { + .majortype = MEDIATYPE_Video, + .subtype = MEDIASUBTYPE_RGB565, + .formattype = FORMAT_VideoInfo, + .bTemporalCompression = TRUE, + }; + LONG buf[(32 * 16 * 2) / sizeof(LONG)], size = sizeof(buf); + IPin *sink, *source, *renderer_pin; + struct testfilter testsource; + ISampleGrabber *grabber; + IMediaControl *control; + IFilterGraph2 *graph; + IMemInputPin *input; + IBaseFilter *filter; + OAFilterState state; + IMediaFilter *mf; + HANDLE thread; + DWORD ticks; + unsigned i; + HRESULT hr; + ULONG ref; + + testfilter_init(&testsource); + CoCreateInstance(&CLSID_FilterGraph, NULL, CLSCTX_INPROC_SERVER, + &IID_IFilterGraph2, (void **)&graph); + CoCreateInstance(&CLSID_NullRenderer, NULL, CLSCTX_INPROC_SERVER, + &IID_IBaseFilter, (void **)&filter); + IFilterGraph2_AddFilter(graph, filter, L"sink"); + IBaseFilter_FindPin(filter, L"In", &renderer_pin); + IBaseFilter_Release(filter); + + filter = create_sample_grabber(); + IFilterGraph2_AddFilter(graph, &testsource.filter.IBaseFilter_iface, L"source"); + IFilterGraph2_AddFilter(graph, filter, L"sample grabber"); + IBaseFilter_FindPin(filter, L"In", &sink); + IBaseFilter_FindPin(filter, L"Out", &source); + IBaseFilter_QueryInterface(filter, &IID_ISampleGrabber, (void **)&grabber); + IFilterGraph2_QueryInterface(graph, &IID_IMediaControl, (void **)&control); + IPin_QueryInterface(sink, &IID_IMemInputPin, (void **)&input); + + ISampleGrabber_SetMediaType(grabber, &req_mt); + ISampleGrabber_SetBufferSamples(grabber, TRUE); + + testsource.sink_mt = &req_mt; + hr = IFilterGraph2_ConnectDirect(graph, &testsource.source.pin.IPin_iface, sink, &req_mt); + ok(hr == S_OK, "Got hr %#x.\n", hr); + hr = IFilterGraph2_ConnectDirect(graph, source, renderer_pin, &req_mt); + ok(hr == S_OK, "Got hr %#x.\n", hr); + IPin_Release(renderer_pin); + + IFilterGraph2_QueryInterface(graph, &IID_IMediaFilter, (void **)&mf); + IMediaFilter_SetSyncSource(mf, NULL); + IMediaFilter_Release(mf); + + hr = IMemInputPin_ReceiveCanBlock(input); + ok(hr == S_OK, "Got hr %#x.\n", hr); + + hr = ISampleGrabber_GetCurrentBuffer(grabber, &size, buf); + ok(hr == VFW_E_WRONG_STATE, "Got hr %#x.\n", hr); + + hr = IMediaControl_Pause(control); + ok(hr == S_FALSE, "Got hr %#x.\n", hr); + hr = IMediaControl_GetState(control, 0, &state); + ok(hr == VFW_S_STATE_INTERMEDIATE, "Got hr %#x.\n", hr); + + thread = send_frame(input, 0); + hr = IMediaControl_GetState(control, 1000, &state); + ok(hr == S_OK, "Got hr %#x.\n", hr); + ok(WaitForSingleObject(thread, 100) == WAIT_TIMEOUT, "Thread should block in Receive().\n"); + + for (i = 0; i < 3; i++) + { + ticks = GetTickCount(); + hr = ISampleGrabber_GetCurrentBuffer(grabber, &size, buf); + ticks = GetTickCount() - ticks; + ok(hr == S_OK, "Got hr %#x.\n", hr); + ok(ticks < 100, "Got %u ticks.\n", ticks); + ok(size == sizeof(buf), "Got size %d.\n", size); + } + + hr = IMediaControl_Stop(control); + ok(hr == S_OK, "Got hr %#x.\n", hr); + hr = join_thread(thread); + ok(hr == S_OK, "Got hr %#x.\n", hr); + hr = IMediaControl_GetState(control, 1000, &state); + ok(hr == S_OK, "Got hr %#x.\n", hr); + ok(state == State_Stopped, "Got state %d.\n", state); + hr = ISampleGrabber_GetCurrentBuffer(grabber, &size, buf); + ok(hr == S_OK, "Got hr %#x.\n", hr); + + hr = IMediaControl_Pause(control); + ok(hr == S_FALSE, "Got hr %#x.\n", hr); + hr = IMediaControl_GetState(control, 0, &state); + ok(hr == VFW_S_STATE_INTERMEDIATE, "Got hr %#x.\n", hr); + + thread = send_frame(input, 0); + hr = IMediaControl_GetState(control, 1000, &state); + ok(hr == S_OK, "Got hr %#x.\n", hr); + ok(WaitForSingleObject(thread, 100) == WAIT_TIMEOUT, "Thread should block in Receive().\n"); + + for (i = 0; i < 3; i++) + { + ticks = GetTickCount(); + hr = ISampleGrabber_GetCurrentBuffer(grabber, &size, buf); + ticks = GetTickCount() - ticks; + ok(hr == S_OK, "Got hr %#x.\n", hr); + ok(ticks < 100, "Got %u ticks.\n", ticks); + ok(size == sizeof(buf), "Got size %d.\n", size); + } + + IPin_BeginFlush(sink); + hr = join_thread(thread); + ok(hr == S_OK, "Got hr %#x.\n", hr); + hr = ISampleGrabber_GetCurrentBuffer(grabber, &size, buf); + ok(hr == S_OK, "Got hr %#x.\n", hr); + IPin_EndFlush(sink); + + hr = IMediaControl_GetState(control, 0, &state); + todo_wine ok(hr == VFW_S_STATE_INTERMEDIATE, "Got hr %#x.\n", hr); + + ticks = GetTickCount(); + thread = send_frame(input, 150); + hr = ISampleGrabber_GetCurrentBuffer(grabber, &size, buf); + ticks = GetTickCount() - ticks; + ok(hr == S_OK, "Got hr %#x.\n", hr); + ok(ticks < 100, "Got %u ticks.\n", ticks); + ok(size == sizeof(buf), "Got size %d.\n", size); + + hr = IMediaControl_GetState(control, 1000, &state); + ok(hr == S_OK, "Got hr %#x.\n", hr); + hr = IMediaControl_Stop(control); + ok(hr == S_OK, "Got hr %#x.\n", hr); + hr = join_thread(thread); + todo_wine ok(hr == S_OK, "Got hr %#x.\n", hr); + + IPin_Release(sink); + IPin_Release(source); + IMemInputPin_Release(input); + IMediaControl_Release(control); + ref = IFilterGraph2_Release(graph); + ok(!ref, "Got outstanding refcount %d.\n", ref); + ISampleGrabber_Release(grabber); + ref = IBaseFilter_Release(filter); + ok(!ref, "Got outstanding refcount %d.\n", ref); + ref = IBaseFilter_Release(&testsource.filter.IBaseFilter_iface); + ok(!ref, "Got outstanding refcount %d.\n", ref); +} + START_TEST(samplegrabber) { IBaseFilter *filter; @@ -1059,6 +1292,7 @@ START_TEST(samplegrabber) test_aggregation(); test_media_types(); test_connect_pin(); + test_buffer_flush(); CoUninitialize(); } -- 2.21.0

5 7

[PATCH v4 1/8] server: Actually set initial status in set_async_direct_result handler.
by Jinoh Kang 19 Jun '22

19 Jun '22

Commit 15483b1a126 (server: Allow calling async_handoff() with status code STATUS_ALERTED., 2022-02-10) introduced the set_async_direct_result handler which calls async_set_initial_status(). However, the async_set_initial_status() call does nothing since async->terminated is set, leaving the async in a confusing state (unknown_status = 1 but pending/completed). So far, this issue is unlikely to have been a problem in practice for the following reasons: 1. async_set_initial_status() would have unset unknown_status, but it remains set instead. This is usually not a problem, since unknown_status is usually ever read by code paths effectively unreachable for non-device (e.g. socket) asyncs. It would still potentially allow set_async_direct_result to be called multiple times, but it wouldn't actually happen in practice unless something goes wrong. 2. async_set_initial_status() would have set initial_status; however, it is left with the default value STATUS_PENDING. If the actual status is something other than that, the handler closes the wait handle and async_satisfied (the only realconsumer of initial_status) would never be called anyway. For reasons above, this issue is not effectively observable or testable. Nonetheless, the current code does leave the async object in an inconsistent state. Fix this by removing the !async->terminated check in async_set_initial_status(). Signed-off-by: Jinoh Kang <jinoh.kang.kr(a)gmail.com> --- Notes: v1 -> v2: no changes v2 -> v3: no changes v3 -> v4: reuse async_set_initial_status() per feedback server/async.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) diff --git a/server/async.c b/server/async.c index 7d0ff10f7a4..e92d4583c1e 100644 --- a/server/async.c +++ b/server/async.c @@ -303,11 +303,8 @@ struct async *create_async( struct fd *fd, struct thread *thread, const async_da void async_set_initial_status( struct async *async, unsigned int status ) { assert( async->unknown_status ); - if (!async->terminated) - { - async->initial_status = status; - async->unknown_status = 0; - } + async->initial_status = status; + async->unknown_status = 0; } void set_async_pending( struct async *async ) -- 2.34.1

3 10

[PATCH 1/2] shell32: Only call HCR_GetFolderAttributes with simple pidls
by Fabian Maurer 27 Apr '22

27 Apr '22

The function returns false with not-simple pidls anyways. This fixes an "ERR" message when opening a control panel applet from explorer Signed-off-by: Fabian Maurer <dark.shadow4(a)web.de> --- dlls/shell32/shlfolder.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dlls/shell32/shlfolder.c b/dlls/shell32/shlfolder.c index 10aafc160c0..8e3b561cfd1 100644 --- a/dlls/shell32/shlfolder.c +++ b/dlls/shell32/shlfolder.c @@ -412,7 +412,7 @@ HRESULT SHELL32_GetItemAttributes (IShellFolder2 *psf, LPCITEMIDLIST pidl, LPDWO if (_ILIsDrive (pidl)) { *pdwAttributes &= SFGAO_HASSUBFOLDER|SFGAO_FILESYSTEM|SFGAO_FOLDER|SFGAO_FILESYSANCESTOR| SFGAO_DROPTARGET|SFGAO_HASPROPSHEET|SFGAO_CANLINK; - } else if (has_guid && HCR_GetFolderAttributes(pidl, &dwAttributes)) { + } else if (has_guid && _ILIsPidlSimple(pidl) && HCR_GetFolderAttributes(pidl, &dwAttributes)) { *pdwAttributes = dwAttributes; } else if (_ILGetDataPointer (pidl)) { DWORD file_attr = _ILGetFileAttributes (pidl, NULL, 0); -- 2.34.1

2 5

[PATCH] dxgi: Fake success in dxgi_adapter_RegisterVideoMemoryBudgetChangeNotificationEvent
by Fabian Maurer 24 Apr '22

24 Apr '22

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=51665 Signed-off-by: Fabian Maurer <dark.shadow4(a)web.de> --- dlls/dxgi/adapter.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dlls/dxgi/adapter.c b/dlls/dxgi/adapter.c index e38fef5fa4f..987b89c4a88 100644 --- a/dlls/dxgi/adapter.c +++ b/dlls/dxgi/adapter.c @@ -350,7 +350,7 @@ static HRESULT STDMETHODCALLTYPE dxgi_adapter_RegisterVideoMemoryBudgetChangeNot { FIXME("iface %p, event %p, cookie %p stub!\n", iface, event, cookie); - return E_NOTIMPL; + return S_OK; } static void STDMETHODCALLTYPE dxgi_adapter_UnregisterVideoMemoryBudgetChangeNotification( -- 2.35.1

3 3

[PATCH 1/5] include: Add definition for IShellItemImageFactory.
by Jinoh Kang 23 Apr '22

23 Apr '22

Signed-off-by: Jinoh Kang <jinoh.kang.kr(a)gmail.com> --- Notes: v1 -> v2: no changes include/shobjidl.idl | 30 ++++++++++++++++++++++++++++++ 1 file changed, 30 insertions(+) diff --git a/include/shobjidl.idl b/include/shobjidl.idl index c1a36ab5e93..11b84c0d4f7 100644 --- a/include/shobjidl.idl +++ b/include/shobjidl.idl @@ -471,6 +471,36 @@ interface IShellItem2 : IShellItem [out] BOOL *pf); } +/***************************************************************************** + * IShellItemImageFactory interface + */ +[ + object, + uuid(BCC18B79-BA16-442F-80C4-8A59C30C463B), + pointer_default(unique) +] +interface IShellItemImageFactory : IUnknown +{ + [v1_enum] enum _SIIGBF { + SIIGBF_RESIZETOFIT = 0x00000000, + SIIGBF_BIGGERSIZEOK = 0x00000001, + SIIGBF_MEMORYONLY = 0x00000002, + SIIGBF_ICONONLY = 0x00000004, + SIIGBF_THUMBNAILONLY = 0x00000008, + SIIGBF_INCACHEONLY = 0x00000010, + SIIGBF_CROPTOSQUARE = 0x00000020, + SIIGBF_WIDETHUMBNAILS = 0x00000040, + SIIGBF_ICONBACKGROUND = 0x00000080, + SIIGBF_SCALEUP = 0x00000100, + }; + typedef DWORD SIIGBF; + + HRESULT GetImage( + [in] SIZE size, + [in] SIIGBF flags, + [out] HBITMAP *phbm); +} + typedef [v1_enum] enum tagNWMF { NWMF_UNLOADING = 0x00000001, NWMF_USERINITED = 0x00000002, -- 2.34.1

3 9

[PATCH v2] winex11.drv: Call ImeSetCompositionString after IME_SetResultString
by Alex Kwak 19 Apr '22

19 Apr '22

IME_SetResultString function overwrite composition strings. so, next composition is not display on GUI. but, still have an xim. Therefore, there was a string that seemed to have disappeared in the middle of the input. Therefore, change it as follows. When calling IME_SetResultString, if in a Composition state, no longer generate WM_IME_ENDCOMPOSITION and call back the previous Composition to recover the overwritten Composition. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=52700 Signed-off-by: Alex Kwak <take-me-home(a)kakao.com> --- v2: use updateCompStr instead IME_SetResultString. --- dlls/winex11.drv/ime.c | 21 +++++++++++++++++---- dlls/winex11.drv/x11drv.h | 3 ++- dlls/winex11.drv/xim.c | 3 ++- 3 files changed, 21 insertions(+), 6 deletions(-) diff --git a/dlls/winex11.drv/ime.c b/dlls/winex11.drv/ime.c index c1584930861..0212a127205 100644 --- a/dlls/winex11.drv/ime.c +++ b/dlls/winex11.drv/ime.c @@ -1015,7 +1015,8 @@ BOOL IME_SetCompositionString(DWORD dwIndex, LPCVOID lpComp, DWORD dwCompLen, lpRead, dwReadLen); } -void IME_SetResultString(LPWSTR lpResult, DWORD dwResultLen) +void IME_SetResultString(LPWSTR lpResult, DWORD dwResultLen, LPCVOID compStr, + DWORD compStrLen) { HIMC imc; LPINPUTCONTEXT lpIMC; @@ -1028,7 +1029,8 @@ void IME_SetResultString(LPWSTR lpResult, DWORD dwResultLen) if (lpIMC == NULL) return; - newCompStr = updateCompStr(lpIMC->hCompStr, NULL, 0); + newCompStr = updateCompStr(lpIMC->hCompStr, (LPCWSTR) compStr, + compStrLen / sizeof(WCHAR)); ImmDestroyIMCC(lpIMC->hCompStr); lpIMC->hCompStr = newCompStr; @@ -1046,12 +1048,23 @@ void IME_SetResultString(LPWSTR lpResult, DWORD dwResultLen) GenerateIMEMessage(imc, WM_IME_STARTCOMPOSITION, 0, 0); } - GenerateIMEMessage(imc, WM_IME_COMPOSITION, 0, GCS_COMPSTR); GenerateIMEMessage(imc, WM_IME_COMPOSITION, lpResult[0], GCS_RESULTSTR|GCS_RESULTCLAUSE); - GenerateIMEMessage(imc, WM_IME_ENDCOMPOSITION, 0, 0); + + if (compStrLen / sizeof(WCHAR) > 0) + { + GenerateIMEMessage(imc, WM_IME_COMPOSITION, ((LPCWSTR) compStr)[0], + GCS_COMPSTR); + } + else + { + GenerateIMEMessage(imc, WM_IME_COMPOSITION, 0, GCS_COMPSTR); + } if (!inComp) + { + GenerateIMEMessage(imc, WM_IME_ENDCOMPOSITION, 0, 0); ImmSetOpenStatus(imc, FALSE); + } ImmUnlockIMC(imc); } diff --git a/dlls/winex11.drv/x11drv.h b/dlls/winex11.drv/x11drv.h index f389f3e0836..ca077caa45c 100644 --- a/dlls/winex11.drv/x11drv.h +++ b/dlls/winex11.drv/x11drv.h @@ -290,7 +290,8 @@ extern void IME_UpdateAssociation(HWND focus) DECLSPEC_HIDDEN; extern BOOL IME_SetCompositionString(DWORD dwIndex, LPCVOID lpComp, DWORD dwCompLen, LPCVOID lpRead, DWORD dwReadLen) DECLSPEC_HIDDEN; -extern void IME_SetResultString(LPWSTR lpResult, DWORD dwResultlen) DECLSPEC_HIDDEN; +extern void IME_SetResultString(LPWSTR lpResult, DWORD dwResultLen, + LPCVOID compStr, DWORD compStrLen) DECLSPEC_HIDDEN; extern void X11DRV_XDND_EnterEvent( HWND hWnd, XClientMessageEvent *event ) DECLSPEC_HIDDEN; extern void X11DRV_XDND_PositionEvent( HWND hWnd, XClientMessageEvent *event ) DECLSPEC_HIDDEN; diff --git a/dlls/winex11.drv/xim.c b/dlls/winex11.drv/xim.c index 3994c2106cc..3b762d45b3c 100644 --- a/dlls/winex11.drv/xim.c +++ b/dlls/winex11.drv/xim.c @@ -115,7 +115,8 @@ void X11DRV_XIMLookupChars( const char *str, DWORD count ) if ((focus = GetFocus())) IME_UpdateAssociation(focus); - IME_SetResultString(wcOutput, dwOutput); + IME_SetResultString(wcOutput, dwOutput, CompositionString, + dwCompStringLength); HeapFree(GetProcessHeap(), 0, wcOutput); } -- 2.32.0

4 4

[PATCH] winex11.drv: Fix IME_SetResultString for Korean(hangul) IME.
by Alex Kwak 19 Apr '22

19 Apr '22

In a special case for Hangul(Korean), when Composition is completed and this function is called, the new Composition disappears. so, calling IME_SetCompositionString again for makes composition will be starting. Signed-off-by: Alex Kwak <take-me-home(a)kakao.com> --- dlls/winex11.drv/ime.c | 4 ++-- dlls/winex11.drv/x11drv.h | 27 +++++++++++++++++++++++++++ dlls/winex11.drv/xim.c | 13 +++++++++++++ 3 files changed, 42 insertions(+), 2 deletions(-) diff --git a/dlls/winex11.drv/ime.c b/dlls/winex11.drv/ime.c index c1584930861..f7827b31635 100644 --- a/dlls/winex11.drv/ime.c +++ b/dlls/winex11.drv/ime.c @@ -1050,8 +1050,8 @@ void IME_SetResultString(LPWSTR lpResult, DWORD dwResultLen) GenerateIMEMessage(imc, WM_IME_COMPOSITION, lpResult[0], GCS_RESULTSTR|GCS_RESULTCLAUSE); GenerateIMEMessage(imc, WM_IME_ENDCOMPOSITION, 0, 0); - if (!inComp) - ImmSetOpenStatus(imc, FALSE); + myPrivate->bInComposition = FALSE; + ImmSetOpenStatus(imc, FALSE); ImmUnlockIMC(imc); } diff --git a/dlls/winex11.drv/x11drv.h b/dlls/winex11.drv/x11drv.h index f389f3e0836..8ce3a7e8528 100644 --- a/dlls/winex11.drv/x11drv.h +++ b/dlls/winex11.drv/x11drv.h @@ -841,4 +841,31 @@ static inline BOOL is_window_rect_mapped( const RECT *rect ) max( rect->bottom, rect->top + 1 ) > virtual_rect.top); } +#define HANGUL_BASE 0xac00 +#define HANGUL_LAST 0xd7a3 +#define HANGUL_JAMO_BASE 0x1100 +#define HANGUL_JAMO_LAST 0x11F9 +#define HANGUL_COMPAT_JAMO_BASE 0x3130 +#define HANGUL_COMPAT_JAMO_LAST 0x318E +#define HANGUL_LCOUNT 19 +#define HANGUL_VCOUNT 21 +#define HANGUL_TCOUNT 28 +#define HANGUL_NCOUNT (HANGUL_VCOUNT * HANGUL_TCOUNT) +#define HANGUL_SCOUNT (HANGUL_LCOUNT * HANGUL_NCOUNT) + +/* + * This method exists to handle special cases related to Korean(Hangul). + */ +static inline BOOL IsHangul(WPARAM wParam) +{ + if (!wParam) + { + return FALSE; + } + + return (wParam >= HANGUL_BASE && wParam < HANGUL_LAST) || + (wParam >= HANGUL_COMPAT_JAMO_BASE && wParam < HANGUL_COMPAT_JAMO_LAST) || + (wParam >= HANGUL_JAMO_BASE && wParam < HANGUL_JAMO_LAST); +} + #endif /* __WINE_X11DRV_H */ diff --git a/dlls/winex11.drv/xim.c b/dlls/winex11.drv/xim.c index 3994c2106cc..884db67c59c 100644 --- a/dlls/winex11.drv/xim.c +++ b/dlls/winex11.drv/xim.c @@ -117,6 +117,19 @@ void X11DRV_XIMLookupChars( const char *str, DWORD count ) IME_SetResultString(wcOutput, dwOutput); HeapFree(GetProcessHeap(), 0, wcOutput); + + /* + * In a special case for Hangul(Korean), when Composition is completed and + * this function is called, the new Composition disappears. so, calling + * IME_SetCompositionString again for makes composition will be + * starting. + */ + if (CompositionString && dwCompStringLength >= sizeof(WCHAR) && + IsHangul(((const WCHAR*) CompositionString)[0])) + { + IME_SetCompositionString(SCS_SETSTR, CompositionString, + dwCompStringLength, NULL, 0); + } } static BOOL XIMPreEditStateNotifyCallback(XIC xic, XPointer p, XPointer data) -- 2.32.0

4 5

[PATCH v3 1/7] jscript: Implement Object.prototype.__defineGetter__.
by Gabriel Ivăncescu 11 Apr '22

11 Apr '22

Signed-off-by: Gabriel Ivăncescu <gabrielopcode(a)gmail.com> --- dlls/jscript/object.c | 44 ++++++++++++++++ dlls/mshtml/tests/documentmode.js | 88 +++++++++++++++++++++++++++++++ 2 files changed, 132 insertions(+) diff --git a/dlls/jscript/object.c b/dlls/jscript/object.c index 9d4a747..d7650ff 100644 --- a/dlls/jscript/object.c +++ b/dlls/jscript/object.c @@ -294,6 +294,49 @@ done: return hres; } +static HRESULT Object_defineGetter(script_ctx_t *ctx, jsval_t vthis, WORD flags, unsigned argc, jsval_t *argv, jsval_t *r) +{ + property_desc_t desc; + const WCHAR *name; + jsstr_t *name_str; + jsdisp_t *jsthis; + HRESULT hres; + + TRACE("\n"); + + if(!is_object_instance(vthis) || !(jsthis = to_jsdisp(get_object(vthis)))) + goto done; + + if(argc < 2 || !is_object_instance(argv[1])) + return JS_E_FUNCTION_EXPECTED; + + desc.getter = to_jsdisp(get_object(argv[1])); + if(!desc.getter) { + FIXME("getter is not JS object\n"); + return E_NOTIMPL; + } + /* FIXME: Check IsCallable */ + + hres = to_flat_string(ctx, argv[0], &name_str, &name); + if(FAILED(hres)) + return hres; + + desc.flags = desc.mask = PROPF_CONFIGURABLE | PROPF_ENUMERABLE; + desc.explicit_getter = TRUE; + desc.explicit_setter = FALSE; + desc.explicit_value = FALSE; + desc.setter = NULL; + hres = jsdisp_define_property(jsthis, name, &desc); + + jsstr_release(name_str); + if(FAILED(hres)) + return hres; +done: + if(r) + *r = jsval_undefined(); + return S_OK; +} + HRESULT Object_get_proto_(script_ctx_t *ctx, jsval_t vthis, WORD flags, unsigned argc, jsval_t *argv, jsval_t *r) { jsdisp_t *jsthis; @@ -365,6 +408,7 @@ static void Object_destructor(jsdisp_t *dispex) } static const builtin_prop_t Object_props[] = { + {L"__defineGetter__", Object_defineGetter, PROPF_METHOD|PROPF_ES6|2}, {L"hasOwnProperty", Object_hasOwnProperty, PROPF_METHOD|1}, {L"isPrototypeOf", Object_isPrototypeOf, PROPF_METHOD|1}, {L"propertyIsEnumerable", Object_propertyIsEnumerable, PROPF_METHOD|1}, diff --git a/dlls/mshtml/tests/documentmode.js b/dlls/mshtml/tests/documentmode.js index 2902902..18dd65e 100644 --- a/dlls/mshtml/tests/documentmode.js +++ b/dlls/mshtml/tests/documentmode.js @@ -1415,6 +1415,94 @@ sync_test("__proto__", function() { } }); +sync_test("__defineGetter__", function() { + var v = document.documentMode; + var r, x = 42; + + if(v < 11) { + ok(x.__defineGetter__ === undefined, "x.__defineGetter__ = " + x.__defineGetter__); + ok(!("__defineGetter__" in Object), "Object.__defineGetter__ = " + Object.__defineGetter__); + return; + } + ok(Object.prototype.hasOwnProperty("__defineGetter__"), "__defineGetter__ is not a property of Object.prototype"); + ok(Object.prototype.__defineGetter__.length === 2, "__defineGetter__.length = " + Object.prototype.__defineGetter__.length); + + function getter() { return "wine"; } + function setter(val) { } + + r = x.__defineGetter__("foo", getter); + ok(r === undefined, "__defineGetter__ on 42 returned " + r); + ok(x.foo === undefined, "42.foo = " + x.foo); + + x = {}; + r = x.__defineGetter__("foo", getter); + ok(r === undefined, "__defineGetter__ returned " + r); + ok(x.foo === "wine", "x.foo = " + x.foo); + r = Object.getOwnPropertyDescriptor(x, "foo"); + ok(r.value === undefined, "x.foo value = " + r.value); + ok(r.get === getter, "x.foo get = " + r.get); + ok(r.set === undefined, "x.foo set = " + r.set); + ok(r.writable === undefined, "x.foo writable = " + r.writable); + ok(r.enumerable === true, "x.foo enumerable = " + r.enumerable); + ok(r.configurable === true, "x.foo configurable = " + r.configurable); + + Object.defineProperty(x, "foo", { get: undefined, set: setter, configurable: false }); + r = Object.getOwnPropertyDescriptor(x, "foo"); + ok(r.value === undefined, "x.foo setter value = " + r.value); + ok(r.get === undefined, "x.foo setter get = " + r.get); + ok(r.set === setter, "x.foo setter set = " + r.set); + ok(r.writable === undefined, "x.foo setter writable = " + r.writable); + ok(r.enumerable === true, "x.foo setter enumerable = " + r.enumerable); + ok(r.configurable === false, "x.foo setter configurable = " + r.configurable); + try { + x.__defineGetter__("foo", getter); + ok(false, "expected exception calling __defineGetter__ on non-configurable property"); + }catch(e) { + ok(e.number === 0xa13d6 - 0x80000000, "__defineGetter__ on non-configurable property threw exception " + e.number); + } + + r = Object.prototype.__defineGetter__.call(undefined, "bar", getter); + ok(r === undefined, "__defineGetter__ on undefined returned " + r); + r = Object.prototype.__defineGetter__.call(null, "bar", getter); + ok(r === undefined, "__defineGetter__ on null returned " + r); + r = x.__defineGetter__(undefined, getter); + ok(r === undefined, "__defineGetter__ undefined prop returned " + r); + ok(x["undefined"] === "wine", "x.undefined = " + x["undefined"]); + r = x.__defineGetter__(false, getter); + ok(r === undefined, "__defineGetter__ undefined prop returned " + r); + ok(x["false"] === "wine", "x.false = " + x["false"]); + + try { + x.__defineGetter__("bar", "string"); + ok(false, "expected exception calling __defineGetter__ with string"); + }catch(e) { + ok(e.number === 0xa138a - 0x80000000, "__defineGetter__ with string threw exception " + e.number); + } + try { + x.__defineGetter__("bar", undefined); + ok(false, "expected exception calling __defineGetter__ with undefined"); + }catch(e) { + ok(e.number === 0xa138a - 0x80000000, "__defineGetter__ with undefined threw exception " + e.number); + } + try { + x.__defineGetter__("bar", null); + ok(false, "expected exception calling __defineGetter__ with null"); + }catch(e) { + ok(e.number === 0xa138a - 0x80000000, "__defineGetter__ with null threw exception " + e.number); + } + try { + Object.prototype.__defineGetter__.call(x, "bar"); + ok(false, "expected exception calling __defineGetter__ with only one arg"); + }catch(e) { + ok(e.number === 0xa138a - 0x80000000, "__defineGetter__ with only one arg threw exception " + e.number); + } + + x.bar = "test"; + ok(x.bar === "test", "x.bar = " + x.bar); + x.__defineGetter__("bar", getter); + ok(x.bar === "wine", "x.bar with getter = " + x.bar); +}); + async_test("postMessage", function() { var v = document.documentMode; var onmessage_called = false; -- 2.34.1

3 20

[PATCH vkd3d 1/2] vkd3d: Use Vulkan timeline semaphores for D3D12 fences.
by Henri Verbeet 06 Apr '22

06 Apr '22

From: Conor McCarthy <cmccarthy(a)codeweavers.com> D3D12 supports signalling a fence to a lower value, while Vulkan timeline semaphores do not. On the GPU side this is handled by simply submitting the signal anyway, if a test for this passes on device creation, because working around this is impractical. For CPU signals the Vulkan semaphore is replaced with a new one at the lower value only if no waits and/or signals are pending on the GPU. Otherwise, a fixme is emitted. Partly based on a vkd3d-proton patch by Hans-Kristian Arntzen (not including the handling of lower fence values). The old implementation is used if KHR_timeline_semaphore is not available or GPU signals do not work for a lower value. Signed-off-by: Conor McCarthy <cmccarthy(a)codeweavers.com> Signed-off-by: Henri Verbeet <hverbeet(a)codeweavers.com> --- This supersedes patch 230633. v4: - Fix the (trivial) merge conflict with patch 230641. - Simplify setting "device->use_timeline_semaphores". libs/vkd3d/command.c | 571 ++++++++++++++++++++++++++++++++++--- libs/vkd3d/device.c | 87 ++++++ libs/vkd3d/vkd3d_private.h | 30 ++ libs/vkd3d/vulkan_procs.h | 5 + tests/d3d12.c | 11 +- 5 files changed, 663 insertions(+), 41 deletions(-) diff --git a/libs/vkd3d/command.c b/libs/vkd3d/command.c index 952675cf..09171fe4 100644 --- a/libs/vkd3d/command.c +++ b/libs/vkd3d/command.c @@ -46,6 +46,9 @@ HRESULT vkd3d_queue_create(struct d3d12_device *device, object->vk_queue_flags = properties->queueFlags; object->timestamp_bits = properties->timestampValidBits; + object->wait_completion_semaphore = VK_NULL_HANDLE; + object->pending_wait_completion_value = 0; + object->semaphores = NULL; object->semaphores_size = 0; object->semaphore_count = 0; @@ -61,6 +64,20 @@ HRESULT vkd3d_queue_create(struct d3d12_device *device, return S_OK; } +bool vkd3d_queue_init_timeline_semaphore(struct vkd3d_queue *queue, struct d3d12_device *device) +{ + VkResult vr; + + if (!queue->wait_completion_semaphore + && (vr = vkd3d_create_timeline_semaphore(device, 0, &queue->wait_completion_semaphore)) < 0) + { + WARN("Failed to create timeline semaphore, vr %d.\n", vr); + return false; + } + + return true; +} + void vkd3d_queue_destroy(struct vkd3d_queue *queue, struct d3d12_device *device) { const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; @@ -75,6 +92,8 @@ void vkd3d_queue_destroy(struct vkd3d_queue *queue, struct d3d12_device *device) vkd3d_free(queue->semaphores); + VK_CALL(vkDestroySemaphore(device->vk_device, queue->wait_completion_semaphore, NULL)); + for (i = 0; i < ARRAY_SIZE(queue->old_vk_semaphores); ++i) { if (queue->old_vk_semaphores[i]) @@ -268,6 +287,7 @@ static HRESULT vkd3d_enqueue_gpu_fence(struct vkd3d_fence_worker *worker, } worker->enqueued_fences[worker->enqueued_fence_count].vk_fence = vk_fence; + worker->enqueued_fences[worker->enqueued_fence_count].vk_semaphore = VK_NULL_HANDLE; waiting_fence = &worker->enqueued_fences[worker->enqueued_fence_count].waiting_fence; waiting_fence->fence = fence; waiting_fence->value = value; @@ -317,6 +337,7 @@ static void vkd3d_fence_worker_remove_fence(struct vkd3d_fence_worker *worker, s static void vkd3d_fence_worker_move_enqueued_fences_locked(struct vkd3d_fence_worker *worker) { unsigned int i; + bool timeline; size_t count; bool ret; @@ -325,8 +346,18 @@ static void vkd3d_fence_worker_move_enqueued_fences_locked(struct vkd3d_fence_wo count = worker->fence_count + worker->enqueued_fence_count; - ret = vkd3d_array_reserve((void **)&worker->vk_fences, &worker->vk_fences_size, - count, sizeof(*worker->vk_fences)); + if ((timeline = worker->device->use_timeline_semaphores)) + { + ret = vkd3d_array_reserve((void **) &worker->vk_semaphores, &worker->vk_semaphores_size, + count, sizeof(*worker->vk_semaphores)); + ret &= vkd3d_array_reserve((void **) &worker->semaphore_wait_values, &worker->semaphore_wait_values_size, + count, sizeof(*worker->semaphore_wait_values)); + } + else + { + ret = vkd3d_array_reserve((void **)&worker->vk_fences, &worker->vk_fences_size, + count, sizeof(*worker->vk_fences)); + } ret &= vkd3d_array_reserve((void **)&worker->fences, &worker->fences_size, count, sizeof(*worker->fences)); if (!ret) @@ -339,7 +370,16 @@ static void vkd3d_fence_worker_move_enqueued_fences_locked(struct vkd3d_fence_wo { struct vkd3d_enqueued_fence *current = &worker->enqueued_fences[i]; - worker->vk_fences[worker->fence_count] = current->vk_fence; + if (timeline) + { + worker->vk_semaphores[worker->fence_count] = current->vk_semaphore; + worker->semaphore_wait_values[worker->fence_count] = current->waiting_fence.value; + } + else + { + worker->vk_fences[worker->fence_count] = current->vk_fence; + } + worker->fences[worker->fence_count] = current->waiting_fence; ++worker->fence_count; } @@ -347,6 +387,66 @@ static void vkd3d_fence_worker_move_enqueued_fences_locked(struct vkd3d_fence_wo worker->enqueued_fence_count = 0; } +static void vkd3d_wait_for_gpu_timeline_semaphores(struct vkd3d_fence_worker *worker) +{ + const struct d3d12_device *device = worker->device; + const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; + VkSemaphoreWaitInfoKHR wait_info; + VkSemaphore vk_semaphore; + uint64_t counter_value; + unsigned int i, j; + HRESULT hr; + int vr; + + if (!worker->fence_count) + return; + + wait_info.sType = VK_STRUCTURE_TYPE_SEMAPHORE_WAIT_INFO_KHR; + wait_info.pNext = NULL; + wait_info.flags = VK_SEMAPHORE_WAIT_ANY_BIT_KHR; + wait_info.pSemaphores = worker->vk_semaphores; + wait_info.semaphoreCount = worker->fence_count; + wait_info.pValues = worker->semaphore_wait_values; + + vr = VK_CALL(vkWaitSemaphoresKHR(device->vk_device, &wait_info, ~(uint64_t)0)); + if (vr == VK_TIMEOUT) + return; + if (vr != VK_SUCCESS) + { + ERR("Failed to wait for Vulkan timeline semaphores, vr %d.\n", vr); + return; + } + + for (i = 0, j = 0; i < worker->fence_count; ++i) + { + struct vkd3d_waiting_fence *current = &worker->fences[i]; + + vk_semaphore = worker->vk_semaphores[i]; + if ((vr = VK_CALL(vkGetSemaphoreCounterValueKHR(device->vk_device, vk_semaphore, &counter_value))) < 0) + { + ERR("Failed to get Vulkan semaphore value, vr %d.\n", vr); + } + else if (counter_value >= current->value) + { + TRACE("Signaling fence %p value %#"PRIx64".\n", current->fence, current->value); + if (FAILED(hr = d3d12_fence_signal(current->fence, counter_value, VK_NULL_HANDLE))) + ERR("Failed to signal D3D12 fence, hr %#x.\n", hr); + + InterlockedDecrement(&current->fence->pending_worker_operation_count); + continue; + } + + if (i != j) + { + worker->vk_semaphores[j] = worker->vk_semaphores[i]; + worker->semaphore_wait_values[j] = worker->semaphore_wait_values[i]; + worker->fences[j] = worker->fences[i]; + } + ++j; + } + worker->fence_count = j; +} + static void vkd3d_wait_for_gpu_fences(struct vkd3d_fence_worker *worker) { struct d3d12_device *device = worker->device; @@ -408,7 +508,7 @@ static void *vkd3d_fence_worker_main(void *arg) for (;;) { - vkd3d_wait_for_gpu_fences(worker); + worker->wait_for_gpu_fences(worker); if (!worker->fence_count || InterlockedAdd(&worker->enqueued_fence_count, 0)) { @@ -473,6 +573,13 @@ HRESULT vkd3d_fence_worker_start(struct vkd3d_fence_worker *worker, worker->vk_fences_size = 0; worker->fences = NULL; worker->fences_size = 0; + worker->vk_semaphores = NULL; + worker->vk_semaphores_size = 0; + worker->semaphore_wait_values = NULL; + worker->semaphore_wait_values_size = 0; + + worker->wait_for_gpu_fences = device->use_timeline_semaphores + ? vkd3d_wait_for_gpu_timeline_semaphores : vkd3d_wait_for_gpu_fences; if ((rc = vkd3d_mutex_init(&worker->mutex))) { @@ -535,6 +642,8 @@ HRESULT vkd3d_fence_worker_stop(struct vkd3d_fence_worker *worker, vkd3d_free(worker->enqueued_fences); vkd3d_free(worker->vk_fences); vkd3d_free(worker->fences); + vkd3d_free(worker->vk_semaphores); + vkd3d_free(worker->semaphore_wait_values); return S_OK; } @@ -684,6 +793,7 @@ static void d3d12_fence_destroy_vk_objects(struct d3d12_fence *fence) } d3d12_fence_garbage_collect_vk_semaphores_locked(fence, true); + VK_CALL(vkDestroySemaphore(device->vk_device, fence->timeline_semaphore, NULL)); vkd3d_mutex_unlock(&fence->mutex); } @@ -802,31 +912,21 @@ static HRESULT d3d12_fence_add_vk_semaphore(struct d3d12_fence *fence, return hr; } -static HRESULT d3d12_fence_signal(struct d3d12_fence *fence, uint64_t value, VkFence vk_fence) +static bool d3d12_fence_signal_external_events_locked(struct d3d12_fence *fence) { struct d3d12_device *device = fence->device; - struct vkd3d_signaled_semaphore *current; bool signal_null_event_cond = false; unsigned int i, j; - int rc; - - if ((rc = vkd3d_mutex_lock(&fence->mutex))) - { - ERR("Failed to lock mutex, error %d.\n", rc); - return hresult_from_errno(rc); - } - - fence->value = value; for (i = 0, j = 0; i < fence->event_count; ++i) { struct vkd3d_waiting_event *current = &fence->events[i]; - if (current->value <= value) + if (current->value <= fence->value) { if (current->event) { - fence->device->signal_event(current->event); + device->signal_event(current->event); } else { @@ -841,9 +941,28 @@ static HRESULT d3d12_fence_signal(struct d3d12_fence *fence, uint64_t value, VkF ++j; } } + fence->event_count = j; - if (signal_null_event_cond) + return signal_null_event_cond; +} + +static HRESULT d3d12_fence_signal(struct d3d12_fence *fence, uint64_t value, VkFence vk_fence) +{ + struct d3d12_device *device = fence->device; + struct vkd3d_signaled_semaphore *current; + unsigned int i; + int rc; + + if ((rc = vkd3d_mutex_lock(&fence->mutex))) + { + ERR("Failed to lock mutex, error %d.\n", rc); + return hresult_from_errno(rc); + } + + fence->value = value; + + if (d3d12_fence_signal_external_events_locked(fence)) vkd3d_cond_broadcast(&fence->null_event_cond); if (vk_fence) @@ -1069,12 +1188,160 @@ static HRESULT STDMETHODCALLTYPE d3d12_fence_SetEventOnCompletion(ID3D12Fence *i return S_OK; } +static inline bool d3d12_fence_gpu_wait_is_completed(const struct d3d12_fence *fence, unsigned int i) +{ + const struct d3d12_device *device = fence->device; + const struct vkd3d_vk_device_procs *vk_procs; + uint64_t value; + VkResult vr; + + vk_procs = &device->vk_procs; + + if ((vr = VK_CALL(vkGetSemaphoreCounterValueKHR(device->vk_device, + fence->gpu_waits[i].queue->wait_completion_semaphore, &value))) >= 0) + { + return value >= fence->gpu_waits[i].pending_value; + } + + ERR("Failed to get Vulkan semaphore status, vr %d.\n", vr); + return true; +} + +static inline bool d3d12_fence_has_pending_gpu_ops_locked(struct d3d12_fence *fence) +{ + const struct d3d12_device *device = fence->device; + const struct vkd3d_vk_device_procs *vk_procs; + uint64_t value; + unsigned int i; + VkResult vr; + + for (i = 0; i < fence->gpu_wait_count; ++i) + { + if (d3d12_fence_gpu_wait_is_completed(fence, i) && i < --fence->gpu_wait_count) + fence->gpu_waits[i] = fence->gpu_waits[fence->gpu_wait_count]; + } + if (fence->gpu_wait_count) + return true; + + /* Check for pending signals too. */ + if (fence->value >= fence->pending_timeline_value) + return false; + + vk_procs = &device->vk_procs; + + /* Check the actual semaphore value in case fence->value update is lagging. */ + if ((vr = VK_CALL(vkGetSemaphoreCounterValueKHR(device->vk_device, fence->timeline_semaphore, &value))) < 0) + { + ERR("Failed to get Vulkan semaphore status, vr %d.\n", vr); + return false; + } + + return value < fence->pending_timeline_value; +} + +/* Replace the VkSemaphore with a new one to allow a lower value to be set. Ideally apps will + * only use this to reset the fence when no operations are pending on the queue. */ +static HRESULT d3d12_fence_reinit_timeline_semaphore_locked(struct d3d12_fence *fence, uint64_t value) +{ + const struct d3d12_device *device = fence->device; + const struct vkd3d_vk_device_procs *vk_procs; + VkSemaphore timeline_semaphore; + VkResult vr; + + if (d3d12_fence_has_pending_gpu_ops_locked(fence)) + { + /* This situation is not very likely because it means a fence with pending waits and/or signals was + * signalled on the CPU to a lower value. For now, emit a fixme so it can be patched if necessary. + * A patch already exists for this but it's not pretty. */ + FIXME("Unable to re-initialise timeline semaphore to a lower value due to pending GPU ops.\n"); + return E_FAIL; + } + + if ((vr = vkd3d_create_timeline_semaphore(device, value, &timeline_semaphore)) < 0) + { + WARN("Failed to create timeline semaphore, vr %d.\n", vr); + return hresult_from_vk_result(vr); + } + + fence->value = value; + fence->pending_timeline_value = value; + + WARN("Replacing timeline semaphore with a new object.\n"); + + vk_procs = &device->vk_procs; + + VK_CALL(vkDestroySemaphore(device->vk_device, fence->timeline_semaphore, NULL)); + fence->timeline_semaphore = timeline_semaphore; + + return S_OK; +} + +static HRESULT d3d12_fence_signal_cpu_timeline_semaphore(struct d3d12_fence *fence, uint64_t value) +{ + const struct d3d12_device *device = fence->device; + VkSemaphoreSignalInfoKHR info; + HRESULT hr = S_OK; + VkResult vr; + int rc; + + if ((rc = vkd3d_mutex_lock(&fence->mutex))) + { + ERR("Failed to lock mutex, error %d.\n", rc); + return hresult_from_errno(rc); + } + + /* We must only signal a value which is greater than the current value. + * That value can be in the range of current known value (fence->value), or as large as pending_timeline_value. + * Pending timeline value signal might be blocked by another synchronization primitive, and thus statically + * cannot be that value, so the safest thing to do is to check the current value which is updated by the fence + * wait thread continuously. This check is technically racy since the value might be immediately out of date, + * but there is no way to avoid this. */ + if (value > fence->value) + { + const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; + + /* Sanity check against the delta limit. */ + if (value - fence->value > device->vk_info.timeline_semaphore_properties.maxTimelineSemaphoreValueDifference) + { + FIXME("Timeline semaphore delta is %"PRIu64", but implementation only supports a delta of %"PRIu64".\n", + value - fence->value, device->vk_info.timeline_semaphore_properties.maxTimelineSemaphoreValueDifference); + } + + info.sType = VK_STRUCTURE_TYPE_SEMAPHORE_SIGNAL_INFO_KHR; + info.pNext = NULL; + info.semaphore = fence->timeline_semaphore; + info.value = value; + if ((vr = VK_CALL(vkSignalSemaphoreKHR(device->vk_device, &info))) >= 0) + { + fence->value = value; + if (value > fence->pending_timeline_value) + fence->pending_timeline_value = value; + } + else + { + ERR("Failed to signal timeline semaphore, vr %d.\n", vr); + hr = hresult_from_vk_result(vr); + } + } + else if (value < fence->value) + { + hr = d3d12_fence_reinit_timeline_semaphore_locked(fence, value); + } + + d3d12_fence_signal_external_events_locked(fence); + + vkd3d_mutex_unlock(&fence->mutex); + return hr; +} + static HRESULT STDMETHODCALLTYPE d3d12_fence_Signal(ID3D12Fence *iface, UINT64 value) { struct d3d12_fence *fence = impl_from_ID3D12Fence(iface); TRACE("iface %p, value %#"PRIx64".\n", iface, value); + if (fence->timeline_semaphore) + return d3d12_fence_signal_cpu_timeline_semaphore(fence, value); return d3d12_fence_signal(fence, value, VK_NULL_HANDLE); } @@ -1108,6 +1375,7 @@ static struct d3d12_fence *unsafe_impl_from_ID3D12Fence(ID3D12Fence *iface) static HRESULT d3d12_fence_init(struct d3d12_fence *fence, struct d3d12_device *device, UINT64 initial_value, D3D12_FENCE_FLAGS flags) { + VkResult vr; HRESULT hr; int rc; @@ -1136,6 +1404,16 @@ static HRESULT d3d12_fence_init(struct d3d12_fence *fence, struct d3d12_device * fence->events_size = 0; fence->event_count = 0; + fence->timeline_semaphore = VK_NULL_HANDLE; + if (device->use_timeline_semaphores && (vr = vkd3d_create_timeline_semaphore(device, initial_value, + &fence->timeline_semaphore)) < 0) + { + WARN("Failed to create timeline semaphore, vr %d.\n", vr); + return hresult_from_vk_result(vr); + } + fence->pending_timeline_value = initial_value; + fence->gpu_wait_count = 0; + list_init(&fence->semaphores); fence->semaphore_count = 0; @@ -1172,6 +1450,25 @@ HRESULT d3d12_fence_create(struct d3d12_device *device, return S_OK; } +VkResult vkd3d_create_timeline_semaphore(const struct d3d12_device *device, uint64_t initial_value, + VkSemaphore *timeline_semaphore) +{ + const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; + VkSemaphoreTypeCreateInfoKHR type_info; + VkSemaphoreCreateInfo info; + + info.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO; + info.pNext = &type_info; + info.flags = 0; + + type_info.sType = VK_STRUCTURE_TYPE_SEMAPHORE_TYPE_CREATE_INFO_KHR; + type_info.pNext = NULL; + type_info.semaphoreType = VK_SEMAPHORE_TYPE_TIMELINE_KHR; + type_info.initialValue = initial_value; + + return VK_CALL(vkCreateSemaphore(device->vk_device, &info, NULL, timeline_semaphore)); +} + /* Command buffers */ static void d3d12_command_list_mark_as_invalid(struct d3d12_command_list *list, const char *message, ...) @@ -6138,18 +6435,88 @@ static void STDMETHODCALLTYPE d3d12_command_queue_EndEvent(ID3D12CommandQueue *i FIXME("iface %p stub!\n", iface); } +static HRESULT d3d12_fence_update_gpu_signal_timeline_semaphore(struct d3d12_fence *fence, uint64_t value) +{ + const struct d3d12_device *device = fence->device; + int rc; + + if ((rc = vkd3d_mutex_lock(&fence->mutex))) + { + ERR("Failed to lock mutex, error %d.\n", rc); + return hresult_from_errno(rc); + } + + /* If we're attempting to async signal a fence with a value which is not strictly increasing the payload value, + * warn about this case. Do not treat this as an error since it works at least with RADV and Nvidia drivers and + * there's no workaround on the GPU side. */ + if (value <= fence->pending_timeline_value) + { + WARN("Fence %p values are not strictly increasing. Pending values: old %"PRIu64", new %"PRIu64".\n", + fence, fence->pending_timeline_value, value); + } + /* Sanity check against the delta limit. Use the current fence value. */ + else if (value - fence->value > device->vk_info.timeline_semaphore_properties.maxTimelineSemaphoreValueDifference) + { + FIXME("Timeline semaphore delta is %"PRIu64", but implementation only supports a delta of %"PRIu64".\n", + value - fence->value, device->vk_info.timeline_semaphore_properties.maxTimelineSemaphoreValueDifference); + } + fence->pending_timeline_value = value; + + vkd3d_mutex_unlock(&fence->mutex); + + return S_OK; +} + +static HRESULT vkd3d_enqueue_timeline_semaphore(struct vkd3d_fence_worker *worker, VkSemaphore vk_semaphore, + struct d3d12_fence *fence, uint64_t value, struct vkd3d_queue *queue) +{ + struct vkd3d_waiting_fence *waiting_fence; + int rc; + + TRACE("worker %p, fence %p, value %#"PRIx64".\n", worker, fence, value); + + if ((rc = vkd3d_mutex_lock(&worker->mutex))) + { + ERR("Failed to lock mutex, error %d.\n", rc); + return hresult_from_errno(rc); + } + + if (!vkd3d_array_reserve((void **)&worker->enqueued_fences, &worker->enqueued_fences_size, + worker->enqueued_fence_count + 1, sizeof(*worker->enqueued_fences))) + { + ERR("Failed to add GPU timeline semaphore.\n"); + vkd3d_mutex_unlock(&worker->mutex); + return E_OUTOFMEMORY; + } + + worker->enqueued_fences[worker->enqueued_fence_count].vk_semaphore = vk_semaphore; + waiting_fence = &worker->enqueued_fences[worker->enqueued_fence_count].waiting_fence; + waiting_fence->fence = fence; + waiting_fence->value = value; + waiting_fence->queue = queue; + ++worker->enqueued_fence_count; + + InterlockedIncrement(&fence->pending_worker_operation_count); + + vkd3d_cond_signal(&worker->cond); + vkd3d_mutex_unlock(&worker->mutex); + + return S_OK; +} + static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Signal(ID3D12CommandQueue *iface, ID3D12Fence *fence_iface, UINT64 value) { struct d3d12_command_queue *command_queue = impl_from_ID3D12CommandQueue(iface); + VkTimelineSemaphoreSubmitInfoKHR timeline_submit_info; const struct vkd3d_vk_device_procs *vk_procs; VkSemaphore vk_semaphore = VK_NULL_HANDLE; VkFence vk_fence = VK_NULL_HANDLE; struct vkd3d_queue *vkd3d_queue; + uint64_t sequence_number = 0; struct d3d12_device *device; struct d3d12_fence *fence; VkSubmitInfo submit_info; - uint64_t sequence_number; VkQueue vk_queue; VkResult vr; HRESULT hr; @@ -6162,10 +6529,21 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Signal(ID3D12CommandQueue * fence = unsafe_impl_from_ID3D12Fence(fence_iface); - if ((vr = d3d12_fence_create_vk_fence(fence, &vk_fence)) < 0) + if (device->use_timeline_semaphores) { - WARN("Failed to create Vulkan fence, vr %d.\n", vr); - goto fail_vkresult; + if (FAILED(hr = d3d12_fence_update_gpu_signal_timeline_semaphore(fence, value))) + return hr; + + vk_semaphore = fence->timeline_semaphore; + assert(vk_semaphore); + } + else + { + if ((vr = d3d12_fence_create_vk_fence(fence, &vk_fence)) < 0) + { + WARN("Failed to create Vulkan fence, vr %d.\n", vr); + goto fail_vkresult; + } } if (!(vk_queue = vkd3d_queue_acquire(vkd3d_queue))) @@ -6175,7 +6553,8 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Signal(ID3D12CommandQueue * goto fail; } - if ((vr = vkd3d_queue_create_vk_semaphore_locked(vkd3d_queue, device, &vk_semaphore)) < 0) + if (!device->use_timeline_semaphores && (vr = vkd3d_queue_create_vk_semaphore_locked(vkd3d_queue, + device, &vk_semaphore)) < 0) { ERR("Failed to create Vulkan semaphore, vr %d.\n", vr); vk_semaphore = VK_NULL_HANDLE; @@ -6191,7 +6570,19 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Signal(ID3D12CommandQueue * submit_info.signalSemaphoreCount = vk_semaphore ? 1 : 0; submit_info.pSignalSemaphores = &vk_semaphore; - if ((vr = VK_CALL(vkQueueSubmit(vk_queue, 1, &submit_info, vk_fence))) >= 0) + if (device->use_timeline_semaphores) + { + timeline_submit_info.sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO_KHR; + timeline_submit_info.pNext = NULL; + timeline_submit_info.pSignalSemaphoreValues = &value; + timeline_submit_info.signalSemaphoreValueCount = submit_info.signalSemaphoreCount; + timeline_submit_info.waitSemaphoreValueCount = 0; + timeline_submit_info.pWaitSemaphoreValues = NULL; + submit_info.pNext = &timeline_submit_info; + } + + vr = VK_CALL(vkQueueSubmit(vk_queue, 1, &submit_info, vk_fence)); + if (!device->use_timeline_semaphores && vr >= 0) { sequence_number = ++vkd3d_queue->submitted_sequence_number; @@ -6208,6 +6599,9 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Signal(ID3D12CommandQueue * goto fail_vkresult; } + if (device->use_timeline_semaphores) + return vkd3d_enqueue_timeline_semaphore(&device->fence_worker, vk_semaphore, fence, value, vkd3d_queue); + if (vk_semaphore && SUCCEEDED(hr = d3d12_fence_add_vk_semaphore(fence, vk_semaphore, vk_fence, value))) vk_semaphore = VK_NULL_HANDLE; @@ -6243,32 +6637,27 @@ fail_vkresult: hr = hresult_from_vk_result(vr); fail: VK_CALL(vkDestroyFence(device->vk_device, vk_fence, NULL)); - VK_CALL(vkDestroySemaphore(device->vk_device, vk_semaphore, NULL)); + if (!device->use_timeline_semaphores) + VK_CALL(vkDestroySemaphore(device->vk_device, vk_semaphore, NULL)); return hr; } -static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Wait(ID3D12CommandQueue *iface, - ID3D12Fence *fence_iface, UINT64 value) +static HRESULT d3d12_command_queue_wait_binary_semaphore(struct d3d12_command_queue *command_queue, + struct d3d12_fence *fence, uint64_t value) { static const VkPipelineStageFlagBits wait_stage_mask = VK_PIPELINE_STAGE_ALL_COMMANDS_BIT; - struct d3d12_command_queue *command_queue = impl_from_ID3D12CommandQueue(iface); const struct vkd3d_vk_device_procs *vk_procs; struct vkd3d_signaled_semaphore *semaphore; uint64_t completed_value = 0; struct vkd3d_queue *queue; - struct d3d12_fence *fence; VkSubmitInfo submit_info; VkQueue vk_queue; VkResult vr; HRESULT hr; - TRACE("iface %p, fence %p, value %#"PRIx64".\n", iface, fence_iface, value); - vk_procs = &command_queue->device->vk_procs; queue = command_queue->vkd3d_queue; - fence = unsafe_impl_from_ID3D12Fence(fence_iface); - semaphore = d3d12_fence_acquire_vk_semaphore(fence, value, &completed_value); if (!semaphore && completed_value >= value) { @@ -6346,6 +6735,122 @@ fail: return hr; } +static inline void d3d12_fence_update_gpu_wait(struct d3d12_fence *fence, const struct vkd3d_queue *queue) +{ + unsigned int i; + bool found; + int rc; + + if ((rc = vkd3d_mutex_lock(&fence->mutex))) + { + ERR("Failed to lock mutex, error %d.\n", rc); + return; + } + + for (i = 0, found = false; i < fence->gpu_wait_count; ++i) + { + if (fence->gpu_waits[i].queue == queue) + { + fence->gpu_waits[i].pending_value = queue->pending_wait_completion_value; + found = true; + } + else if (d3d12_fence_gpu_wait_is_completed(fence, i) && i < --fence->gpu_wait_count) + { + fence->gpu_waits[i] = fence->gpu_waits[fence->gpu_wait_count]; + } + } + + if (!found) + { + if (fence->gpu_wait_count < ARRAY_SIZE(fence->gpu_waits)) + { + fence->gpu_waits[fence->gpu_wait_count].queue = queue; + fence->gpu_waits[fence->gpu_wait_count++].pending_value = queue->pending_wait_completion_value; + } + else + { + FIXME("Unable to track GPU fence wait.\n"); + } + } + + vkd3d_mutex_unlock(&fence->mutex); +} + +static HRESULT d3d12_command_queue_wait_timeline_semaphore(struct d3d12_command_queue *command_queue, + struct d3d12_fence *fence, uint64_t value) +{ + static const VkPipelineStageFlagBits wait_stage_mask = VK_PIPELINE_STAGE_ALL_COMMANDS_BIT; + VkTimelineSemaphoreSubmitInfoKHR timeline_submit_info; + const struct vkd3d_vk_device_procs *vk_procs; + struct vkd3d_queue *queue; + VkSubmitInfo submit_info; + VkQueue vk_queue; + VkResult vr; + + vk_procs = &command_queue->device->vk_procs; + queue = command_queue->vkd3d_queue; + + assert(fence->timeline_semaphore); + timeline_submit_info.sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO_KHR; + timeline_submit_info.pNext = NULL; + timeline_submit_info.signalSemaphoreValueCount = 0; + timeline_submit_info.pSignalSemaphoreValues = NULL; + timeline_submit_info.waitSemaphoreValueCount = 1; + timeline_submit_info.pWaitSemaphoreValues = &value; + + if (!(vk_queue = vkd3d_queue_acquire(queue))) + { + ERR("Failed to acquire queue %p.\n", queue); + return E_FAIL; + } + + submit_info.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO; + submit_info.pNext = &timeline_submit_info; + submit_info.waitSemaphoreCount = 1; + submit_info.pWaitSemaphores = &fence->timeline_semaphore; + submit_info.pWaitDstStageMask = &wait_stage_mask; + submit_info.commandBufferCount = 0; + submit_info.pCommandBuffers = NULL; + submit_info.signalSemaphoreCount = 0; + submit_info.pSignalSemaphores = NULL; + + ++queue->pending_wait_completion_value; + + submit_info.signalSemaphoreCount = 1; + submit_info.pSignalSemaphores = &queue->wait_completion_semaphore; + timeline_submit_info.signalSemaphoreValueCount = 1; + timeline_submit_info.pSignalSemaphoreValues = &queue->pending_wait_completion_value; + + d3d12_fence_update_gpu_wait(fence, queue); + + vr = VK_CALL(vkQueueSubmit(vk_queue, 1, &submit_info, VK_NULL_HANDLE)); + + vkd3d_queue_release(queue); + + if (vr < 0) + { + WARN("Failed to submit wait operation, vr %d.\n", vr); + return hresult_from_vk_result(vr); + } + + return S_OK; +} + +static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Wait(ID3D12CommandQueue *iface, + ID3D12Fence *fence_iface, UINT64 value) +{ + struct d3d12_command_queue *command_queue = impl_from_ID3D12CommandQueue(iface); + struct d3d12_fence *fence = unsafe_impl_from_ID3D12Fence(fence_iface); + + TRACE("iface %p, fence %p, value %#"PRIx64".\n", iface, fence_iface, value); + + if (command_queue->device->use_timeline_semaphores) + return d3d12_command_queue_wait_timeline_semaphore(command_queue, fence, value); + + FIXME_ONCE("KHR_timeline_semaphore is not available or incompatible. Some wait commands may be unsupported.\n"); + return d3d12_command_queue_wait_binary_semaphore(command_queue, fence, value); +} + static HRESULT STDMETHODCALLTYPE d3d12_command_queue_GetTimestampFrequency(ID3D12CommandQueue *iface, UINT64 *frequency) { diff --git a/libs/vkd3d/device.c b/libs/vkd3d/device.c index eb470c12..ab087d60 100644 --- a/libs/vkd3d/device.c +++ b/libs/vkd3d/device.c @@ -129,6 +129,7 @@ static const struct vkd3d_optional_extension_info optional_device_extensions[] = VK_EXTENSION(KHR_MAINTENANCE3, KHR_maintenance3), VK_EXTENSION(KHR_PUSH_DESCRIPTOR, KHR_push_descriptor), VK_EXTENSION(KHR_SAMPLER_MIRROR_CLAMP_TO_EDGE, KHR_sampler_mirror_clamp_to_edge), + VK_EXTENSION(KHR_TIMELINE_SEMAPHORE, KHR_timeline_semaphore), /* EXT extensions */ VK_EXTENSION(EXT_CALIBRATED_TIMESTAMPS, EXT_calibrated_timestamps), VK_EXTENSION(EXT_CONDITIONAL_RENDERING, EXT_conditional_rendering), @@ -791,6 +792,7 @@ struct vkd3d_physical_device_info VkPhysicalDeviceTexelBufferAlignmentPropertiesEXT texel_buffer_alignment_properties; VkPhysicalDeviceTransformFeedbackPropertiesEXT xfb_properties; VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT vertex_divisor_properties; + VkPhysicalDeviceTimelineSemaphorePropertiesKHR timeline_semaphore_properties; VkPhysicalDeviceProperties2KHR properties2; @@ -803,6 +805,7 @@ struct vkd3d_physical_device_info VkPhysicalDeviceTexelBufferAlignmentFeaturesEXT texel_buffer_alignment_features; VkPhysicalDeviceTransformFeedbackFeaturesEXT xfb_features; VkPhysicalDeviceVertexAttributeDivisorFeaturesEXT vertex_divisor_features; + VkPhysicalDeviceTimelineSemaphoreFeaturesKHR timeline_semaphore_features; VkPhysicalDeviceFeatures2 features2; }; @@ -814,11 +817,13 @@ static void vkd3d_physical_device_info_init(struct vkd3d_physical_device_info *i VkPhysicalDeviceDescriptorIndexingPropertiesEXT *descriptor_indexing_properties; VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT *vertex_divisor_properties; VkPhysicalDeviceTexelBufferAlignmentPropertiesEXT *buffer_alignment_properties; + VkPhysicalDeviceTimelineSemaphorePropertiesKHR *timeline_semaphore_properties; VkPhysicalDeviceDescriptorIndexingFeaturesEXT *descriptor_indexing_features; VkPhysicalDeviceRobustness2FeaturesEXT *robustness2_features; VkPhysicalDeviceVertexAttributeDivisorFeaturesEXT *vertex_divisor_features; VkPhysicalDeviceTexelBufferAlignmentFeaturesEXT *buffer_alignment_features; VkPhysicalDeviceShaderDemoteToHelperInvocationFeaturesEXT *demote_features; + VkPhysicalDeviceTimelineSemaphoreFeaturesKHR *timeline_semaphore_features; VkPhysicalDeviceDepthClipEnableFeaturesEXT *depth_clip_features; VkPhysicalDeviceMaintenance3Properties *maintenance3_properties; VkPhysicalDeviceTransformFeedbackPropertiesEXT *xfb_properties; @@ -838,6 +843,8 @@ static void vkd3d_physical_device_info_init(struct vkd3d_physical_device_info *i buffer_alignment_properties = &info->texel_buffer_alignment_properties; vertex_divisor_features = &info->vertex_divisor_features; vertex_divisor_properties = &info->vertex_divisor_properties; + timeline_semaphore_features = &info->timeline_semaphore_features; + timeline_semaphore_properties = &info->timeline_semaphore_properties; xfb_features = &info->xfb_features; xfb_properties = &info->xfb_properties; @@ -859,6 +866,8 @@ static void vkd3d_physical_device_info_init(struct vkd3d_physical_device_info *i vk_prepend_struct(&info->features2, xfb_features); vertex_divisor_features->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VERTEX_ATTRIBUTE_DIVISOR_FEATURES_EXT; vk_prepend_struct(&info->features2, vertex_divisor_features); + timeline_semaphore_features->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TIMELINE_SEMAPHORE_FEATURES_KHR; + vk_prepend_struct(&info->features2, timeline_semaphore_features); if (vulkan_info->KHR_get_physical_device_properties2) VK_CALL(vkGetPhysicalDeviceFeatures2KHR(physical_device, &info->features2)); @@ -877,6 +886,8 @@ static void vkd3d_physical_device_info_init(struct vkd3d_physical_device_info *i vk_prepend_struct(&info->properties2, xfb_properties); vertex_divisor_properties->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VERTEX_ATTRIBUTE_DIVISOR_PROPERTIES_EXT; vk_prepend_struct(&info->properties2, vertex_divisor_properties); + timeline_semaphore_properties->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TIMELINE_SEMAPHORE_PROPERTIES_KHR; + vk_prepend_struct(&info->properties2, timeline_semaphore_properties); if (vulkan_info->KHR_get_physical_device_properties2) VK_CALL(vkGetPhysicalDeviceProperties2KHR(physical_device, &info->properties2)); @@ -1465,6 +1476,7 @@ static HRESULT vkd3d_init_device_caps(struct d3d12_device *device, vulkan_info->rasterization_stream = physical_device_info->xfb_properties.transformFeedbackRasterizationStreamSelect; vulkan_info->transform_feedback_queries = physical_device_info->xfb_properties.transformFeedbackQueries; vulkan_info->max_vertex_attrib_divisor = max(physical_device_info->vertex_divisor_properties.maxVertexAttribDivisor, 1); + vulkan_info->timeline_semaphore_properties = physical_device_info->timeline_semaphore_properties; device->feature_options.DoublePrecisionFloatShaderOps = features->shaderFloat64; device->feature_options.OutputMergerLogicOp = features->logicOp; @@ -1589,6 +1601,8 @@ static HRESULT vkd3d_init_device_caps(struct d3d12_device *device, vulkan_info->EXT_shader_demote_to_helper_invocation = false; if (!physical_device_info->texel_buffer_alignment_features.texelBufferAlignment) vulkan_info->EXT_texel_buffer_alignment = false; + if (!physical_device_info->timeline_semaphore_features.timelineSemaphore) + vulkan_info->KHR_timeline_semaphore = false; vulkan_info->texel_buffer_alignment_properties = physical_device_info->texel_buffer_alignment_properties; @@ -1939,6 +1953,75 @@ static bool d3d12_is_64k_msaa_supported(struct d3d12_device *device) && info.Alignment <= 0x10000; } +/* A lower value can be signalled on a D3D12 fence. Vulkan timeline semaphores + * do not support this, but test if it works anyway. */ +static bool d3d12_is_timeline_semaphore_supported(const struct d3d12_device *device) +{ + const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; + VkTimelineSemaphoreSubmitInfoKHR timeline_submit_info; + VkSemaphore timeline_semaphore; + VkSubmitInfo submit_info; + bool result = false; + uint64_t value = 0; + VkQueue vk_queue; + VkResult vr; + + if (!device->vk_info.KHR_timeline_semaphore) + return false; + + if ((vr = vkd3d_create_timeline_semaphore(device, 1, &timeline_semaphore)) < 0) + { + WARN("Failed to create timeline semaphore, vr %d.\n", vr); + return false; + } + + if (!(vk_queue = vkd3d_queue_acquire(device->direct_queue))) + { + ERR("Failed to acquire queue %p.\n", device->direct_queue); + VK_CALL(vkDestroySemaphore(device->vk_device, timeline_semaphore, NULL)); + return false; + } + + submit_info.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO; + submit_info.pNext = &timeline_submit_info; + submit_info.waitSemaphoreCount = 0; + submit_info.pWaitSemaphores = NULL; + submit_info.pWaitDstStageMask = NULL; + submit_info.commandBufferCount = 0; + submit_info.pCommandBuffers = NULL; + submit_info.signalSemaphoreCount = 1; + submit_info.pSignalSemaphores = &timeline_semaphore; + + timeline_submit_info.sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO_KHR; + timeline_submit_info.pNext = NULL; + timeline_submit_info.pSignalSemaphoreValues = &value; + timeline_submit_info.signalSemaphoreValueCount = 1; + timeline_submit_info.waitSemaphoreValueCount = 0; + timeline_submit_info.pWaitSemaphoreValues = NULL; + + vr = VK_CALL(vkQueueSubmit(vk_queue, 1, &submit_info, VK_NULL_HANDLE)); + + if (vr >= 0) + { + if ((vr = VK_CALL(vkQueueWaitIdle(vk_queue))) < 0) + WARN("Failed to wait for queue, vr %d.\n", vr); + + if ((vr = VK_CALL(vkGetSemaphoreCounterValueKHR(device->vk_device, timeline_semaphore, &value))) < 0) + ERR("Failed to get Vulkan semaphore status, vr %d.\n", vr); + else if (!(result = !value)) + WARN("Disabling timeline semaphore use due to incompatible behaviour.\n"); + } + else + { + WARN("Failed to submit signal operation, vr %d.\n", vr); + } + + vkd3d_queue_release(device->direct_queue); + VK_CALL(vkDestroySemaphore(device->vk_device, timeline_semaphore, NULL)); + + return result; +} + static HRESULT vkd3d_create_vk_device(struct d3d12_device *device, const struct vkd3d_device_create_info *create_info) { @@ -2037,6 +2120,10 @@ static HRESULT vkd3d_create_vk_device(struct d3d12_device *device, } device->feature_options4.MSAA64KBAlignedTextureSupported = d3d12_is_64k_msaa_supported(device); + device->use_timeline_semaphores = d3d12_is_timeline_semaphore_supported(device) + && vkd3d_queue_init_timeline_semaphore(device->direct_queue, device) + && vkd3d_queue_init_timeline_semaphore(device->compute_queue, device) + && vkd3d_queue_init_timeline_semaphore(device->copy_queue, device); TRACE("Created Vulkan device %p.\n", vk_device); diff --git a/libs/vkd3d/vkd3d_private.h b/libs/vkd3d/vkd3d_private.h index 796dfefd..56060b6d 100644 --- a/libs/vkd3d/vkd3d_private.h +++ b/libs/vkd3d/vkd3d_private.h @@ -59,6 +59,7 @@ #define VKD3D_MAX_SHADER_EXTENSIONS 3u #define VKD3D_MAX_SHADER_STAGES 5u #define VKD3D_MAX_VK_SYNC_OBJECTS 4u +#define VKD3D_MAX_FENCE_WAITING_QUEUES 4u #define VKD3D_MAX_DESCRIPTOR_SETS 64u /* D3D12 binding tier 3 has a limit of 2048 samplers. */ #define VKD3D_MAX_DESCRIPTOR_SET_SAMPLERS 2048u @@ -125,6 +126,7 @@ struct vkd3d_vulkan_info bool KHR_maintenance3; bool KHR_push_descriptor; bool KHR_sampler_mirror_clamp_to_edge; + bool KHR_timeline_semaphore; /* EXT device extensions */ bool EXT_calibrated_timestamps; bool EXT_conditional_rendering; @@ -150,6 +152,8 @@ struct vkd3d_vulkan_info VkPhysicalDeviceTexelBufferAlignmentPropertiesEXT texel_buffer_alignment_properties; + VkPhysicalDeviceTimelineSemaphorePropertiesKHR timeline_semaphore_properties; + unsigned int shader_extension_count; enum vkd3d_shader_spirv_extension shader_extensions[VKD3D_MAX_SHADER_EXTENSIONS]; @@ -348,6 +352,7 @@ struct vkd3d_fence_worker struct vkd3d_enqueued_fence { VkFence vk_fence; + VkSemaphore vk_semaphore; struct vkd3d_waiting_fence waiting_fence; } *enqueued_fences; size_t enqueued_fences_size; @@ -357,6 +362,12 @@ struct vkd3d_fence_worker size_t vk_fences_size; struct vkd3d_waiting_fence *fences; size_t fences_size; + VkSemaphore *vk_semaphores; + size_t vk_semaphores_size; + uint64_t *semaphore_wait_values; + size_t semaphore_wait_values_size; + + void (*wait_for_gpu_fences)(struct vkd3d_fence_worker *worker); struct d3d12_device *device; }; @@ -511,6 +522,12 @@ struct vkd3d_signaled_semaphore bool is_acquired; }; +struct vkd3d_pending_fence_wait +{ + const struct vkd3d_queue *queue; + uint64_t pending_value; +}; + /* ID3D12Fence */ struct d3d12_fence { @@ -530,6 +547,11 @@ struct d3d12_fence size_t events_size; size_t event_count; + VkSemaphore timeline_semaphore; + uint64_t pending_timeline_value; + struct vkd3d_pending_fence_wait gpu_waits[VKD3D_MAX_FENCE_WAITING_QUEUES]; + unsigned int gpu_wait_count; + struct list semaphores; unsigned int semaphore_count; @@ -545,6 +567,9 @@ struct d3d12_fence HRESULT d3d12_fence_create(struct d3d12_device *device, uint64_t initial_value, D3D12_FENCE_FLAGS flags, struct d3d12_fence **fence); +VkResult vkd3d_create_timeline_semaphore(const struct d3d12_device *device, uint64_t initial_value, + VkSemaphore *timeline_semaphore); + /* ID3D12Heap */ struct d3d12_heap { @@ -1284,6 +1309,9 @@ struct vkd3d_queue VkQueueFlags vk_queue_flags; uint32_t timestamp_bits; + VkSemaphore wait_completion_semaphore; + uint64_t pending_wait_completion_value; + struct { VkSemaphore vk_semaphore; @@ -1298,6 +1326,7 @@ struct vkd3d_queue VkQueue vkd3d_queue_acquire(struct vkd3d_queue *queue); HRESULT vkd3d_queue_create(struct d3d12_device *device, uint32_t family_index, const VkQueueFamilyProperties *properties, struct vkd3d_queue **queue); +bool vkd3d_queue_init_timeline_semaphore(struct vkd3d_queue *queue, struct d3d12_device *device); void vkd3d_queue_destroy(struct vkd3d_queue *queue, struct d3d12_device *device); void vkd3d_queue_release(struct vkd3d_queue *queue); @@ -1456,6 +1485,7 @@ struct d3d12_device VkDescriptorPoolSize vk_pool_sizes[VKD3D_DESCRIPTOR_POOL_COUNT]; struct vkd3d_vk_descriptor_heap_layout vk_descriptor_heap_layouts[VKD3D_SET_INDEX_COUNT]; bool use_vk_heaps; + bool use_timeline_semaphores; }; HRESULT d3d12_device_create(struct vkd3d_instance *instance, diff --git a/libs/vkd3d/vulkan_procs.h b/libs/vkd3d/vulkan_procs.h index 60556735..34e0ab4b 100644 --- a/libs/vkd3d/vulkan_procs.h +++ b/libs/vkd3d/vulkan_procs.h @@ -195,6 +195,11 @@ VK_DEVICE_EXT_PFN(vkGetDescriptorSetLayoutSupportKHR) /* VK_KHR_push_descriptor */ VK_DEVICE_EXT_PFN(vkCmdPushDescriptorSetKHR) +/* VK_KHR_timeline_semaphore */ +VK_DEVICE_EXT_PFN(vkGetSemaphoreCounterValueKHR) +VK_DEVICE_EXT_PFN(vkWaitSemaphoresKHR) +VK_DEVICE_EXT_PFN(vkSignalSemaphoreKHR) + /* VK_EXT_calibrated_timestamps */ VK_DEVICE_EXT_PFN(vkGetCalibratedTimestampsEXT) diff --git a/tests/d3d12.c b/tests/d3d12.c index 29bd40c8..2059b182 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -33240,7 +33240,9 @@ static void test_queue_wait(void) command_list = context.list; queue = context.queue; - queue2 = create_command_queue(device, D3D12_COMMAND_LIST_TYPE_DIRECT, D3D12_COMMAND_QUEUE_PRIORITY_NORMAL); + /* 'queue2' must not map to the same command queue as 'queue', or Wait() before GPU signal will fail. + * Using a compute queue fixes this on most hardware, but it may still fail on low spec hardware. */ + queue2 = create_command_queue(device, D3D12_COMMAND_LIST_TYPE_COMPUTE, D3D12_COMMAND_QUEUE_PRIORITY_NORMAL); event = create_event(); ok(event, "Failed to create event.\n"); @@ -33305,12 +33307,6 @@ static void test_queue_wait(void) check_readback_data_uint(&rb, NULL, 0xff00ff00, 0); release_resource_readback(&rb); - if (!vkd3d_test_platform_is_windows()) - { - skip("Wait() is not implemented yet.\n"); /* FIXME */ - goto skip_tests; - } - /* Wait() before CPU signal */ update_buffer_data(cb, 0, sizeof(blue), &blue); queue_wait(queue, fence, 2); @@ -33386,7 +33382,6 @@ static void test_queue_wait(void) check_readback_data_uint(&rb, NULL, 0xff00ff00, 0); release_resource_readback(&rb); -skip_tests: /* Signal() and Wait() in the same command queue */ update_buffer_data(cb, 0, sizeof(blue), &blue); queue_signal(queue, fence, 7); -- 2.20.1

2 1

[PATCH vkd3d v2 1/9] vkd3d-shader/hlsl: Use an unsigned variable for iteration.
by Giovanni Mascellani 06 Apr '22

06 Apr '22

Signed-off-by: Giovanni Mascellani <gmascellani(a)codeweavers.com> --- v2: * New --- libs/vkd3d-shader/hlsl_constant_ops.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/libs/vkd3d-shader/hlsl_constant_ops.c b/libs/vkd3d-shader/hlsl_constant_ops.c index 51cee179..109fc2ee 100644 --- a/libs/vkd3d-shader/hlsl_constant_ops.c +++ b/libs/vkd3d-shader/hlsl_constant_ops.c @@ -194,11 +194,12 @@ static bool fold_mul(struct hlsl_ctx *ctx, struct hlsl_ir_constant *dst, struct hlsl_ir_constant *src1, struct hlsl_ir_constant *src2) { enum hlsl_base_type type = dst->node.data_type->base_type; + unsigned int k; assert(type == src1->node.data_type->base_type); assert(type == src2->node.data_type->base_type); - for (int k = 0; k < 4; ++k) + for (k = 0; k < 4; ++k) { switch (type) { -- 2.35.1

4 35