Stenzek
296d48a67c
ImGuiManager: Ensure DeltaTime can never be negative
2026-10-25 21:59:36 +10:00
Stenzek
0b5b8774cf
Common: Add PackedOptionalArray
2026-10-25 21:59:17 +10:00
Stenzek
cde17ace5a
CDImageCueSheet: Fix single-file pregap and LBA count
...
The implicit pregap read the first data sectors instead of silence, and
the LBA count excluded it.
2026-10-25 14:38:36 +10:00
Stenzek
b2b68c3ced
CDImageCCD: Fix LBA count including pregaps twice
...
Track start is index 1 but length includes the pregap, so the lead-out
position reported by GetTD was too late.
2026-10-25 14:38:36 +10:00
Stenzek
92a7758b53
CDImageCHD: Use generated subchannel Q for virtual pregaps
...
They aren't in the file, so reading the hunk returned the subcode of the
first data frames instead.
2026-10-25 14:38:36 +10:00
Stenzek
9bbf341beb
FullscreenUI: Copy/move checkable choice dialog callback before invoking
...
If the callback reopened the dialog, it would destroy the running
callback and the option it was passed.
2026-10-25 14:29:03 +10:00
Stenzek
b73f8b2731
FullscreenUI: Keep horizontal menu/sidebar state balanced when hidden
...
Their End functions are always called, so returning early left the draw
split unstarted and the style stack unbalanced.
2026-10-25 14:25:51 +10:00
Stenzek
ae4658a7b5
FullscreenUI: Pass p_open through in BeginBlurWindow()
2026-10-25 14:25:51 +10:00
Stenzek
f4fe9de23f
FullscreenUI: Pass button flags through in MenuButtonFrame() overload
...
The height-based overload ignored its flags argument.
2026-10-25 14:25:51 +10:00
Stenzek
aa9916b882
FullscreenUI: Avoid inverted std::clamp() bounds in dropdown width
...
A value box wider than half the screen made the lower bound exceed the
upper bound.
2026-10-25 14:25:51 +10:00
Stenzek
49a2ff938e
FullscreenUI: Fix indeterminate progress bar sweep range
...
The bar was interpolated to the absolute end of the box rather than its
width, so it moved too fast and was hidden for part of each cycle.
2026-10-25 14:25:51 +10:00
Stenzek
414d8f2049
FullscreenUI: Clear all dialogs on shutdown
...
The directory selector, fixed popup and progress dialogs were left open
across a shutdown/reinitialize.
2026-10-25 14:25:51 +10:00
Stenzek
7d64f8223e
FullscreenUI: Include directory selector in open dialog checks
...
Otherwise the menu could be hidden with the directory selector's
callback still pending, and window switches would steal its focus.
2026-10-25 14:25:51 +10:00
Stenzek
fd8bcdcf18
FullscreenUI: Fix ForceKeyNavEnabled() clobbering gamepad nav source
...
The nav input source was assigned from the active ID source, which could
switch a gamepad source to keyboard, e.g. after closing the OSK.
2026-10-25 14:25:51 +10:00
Stenzek
968d20a418
dep/imgui: Don't smooth scroll requests made while window is appearing
...
They're applied on the following frame, when Appearing is no longer set.
Fixes jump scroll when later options are selected in a choice dialog.
2026-10-25 14:25:51 +10:00
Stenzek
c3b58f2563
FullscreenUI: Support drag/swipe scrolling with inertia
...
Dragging past a threshold cancels the pending button press.
2026-10-25 14:25:51 +10:00
Stenzek
20916bd898
FullscreenUI: Activate toggle buttons on release instead of press
...
Allows the press to be cancelled, e.g. by a drag/swipe.
2026-10-25 14:25:50 +10:00
Stenzek
8962d84367
GPU/HW: Add missing logging of depth clears
2026-10-25 13:31:29 +10:00
Stenzek
dd4a0eb779
GPU/HW: Track dirty depth rect to shrink depth merges/clears
...
Merges and drawing area clears now only cover what was drawn with depth
since the last clear.
v0.1-12442
2026-10-25 12:48:59 +10:00
Stenzek
8d9f6e8bf7
Data: Update SDL game controller database
2026-10-25 01:37:45 +10:00
Stenzek
7763167550
VideoPresenter: Probe for presentation backlog when gap baseline rises
...
Otherwise lag left by a long hitch, e.g. a system overlay, resizing window,
etc. looks normal and is never dropped.
2026-10-25 01:37:18 +10:00
Stenzek
3551eef7f8
VideoPresenter: Split late present check out of ShouldPresentFrame()
2026-10-25 01:37:09 +10:00
Stenzek
85fe47af8e
VulkanDevice: Wait for presents with VK_KHR_present_wait
...
Acquiring the next image doesn't always block until the frame is displayed,
so late frames weren't detected.
Note: On NVIDIA drivers at least, this gives you a whacky frame time
graph with vsync on. But it's actually better input latency, comparable
to D3D.
2026-10-25 01:36:32 +10:00
Stenzek
6caf8b2ede
Android: Update translation strings
2026-10-25 01:06:24 +10:00
Stenzek
c5ffc4a978
Data: Add MC_SSAO.fx
2026-10-24 23:01:50 +10:00
Stenzek
c71899031e
GPU/HW: Accumulate PGXP depth across depth buffer clears
...
Gives postprocessing depth for the whole frame, not just the last segment.
2026-10-24 19:28:08 +10:00
Stenzek
d691f5ae21
Regtest: Split -setting key at the last slash
...
Allows nested sections such as InternalPostProcessing/Stage1.
2026-10-24 19:14:27 +10:00
Stenzek
113f297a05
PostProcessing: Fix crash loading ReShade shaders without a swap chain
2026-10-24 19:14:27 +10:00
Stenzek
796c4751b7
GameDatabase: Use PackedOptional for overrides
2026-10-24 18:56:37 +10:00
Stenzek
507e249522
Common: Add PackedOptional class
...
Move the valid bit outside of the value, so we don't waste struct
padding bytes.
2026-10-24 18:56:05 +10:00
Stenzek
d512707236
GPU/HW: Bias PGXP depth towards the camera by draw order
...
Fixes Z-fighting with near-coplanar polygons such as road markings,
building decals, etc.
2026-10-24 18:40:58 +10:00
Stenzek
7200c8ce2f
CPU/PGXP: Add option to normalize depth by GTE matrix scale
...
Some games shrink the view transform of distant objects to stay within 16-bit range,
making raw Z incomparable between objects. Divide it by the rotation matrix's Z row scale.
2026-10-24 17:37:41 +10:00
Stenzek
1d508acef7
GPU/HW: Store reciprocal Z in PGXP depth buffer
...
Screen-space interpolation of W gave incorrect depth inside sloped polygons.
1/z interpolates correctly, and reversing it keeps float precision at distance.
2026-10-24 17:36:08 +10:00
Stenzek
db0b5535d2
Qt: Fix stale mouse grab after recreating display widget
...
QWidget::releaseMouse() clears the grabber even when called on a non-grabber,
so the old widget's deferred release orphaned the new widget's window grab.
Fixes stuck mouse input with Mouse controller when toggling pause in fullscreen.
2026-10-24 16:24:00 +10:00
Stenzek
8db5d62076
Qt: Tweak bold group box headings
2026-10-24 15:12:11 +10:00
Stenzek
450bc59447
Qt: Add icon next to device indicator in bindings
2026-10-24 15:05:22 +10:00
Stenzek
167c6e3323
Misc: Fix gitattributes typo for vcxproj filters
2026-10-24 14:50:59 +10:00
Stenzek
efc6ff524b
Controller: Produce center for equal opposing half-axis inputs
2026-10-24 14:50:46 +10:00
Stenzek
c6747677b4
Controller: Initialize analog axes to AXIS_CENTER
...
Fixes Persona 2 up/down not working after resuming state until X is moved.
2026-10-24 14:50:31 +10:00
Stenzek
43268aa6e9
Controller: Restore 127 as the analog axis center
...
Persona 2 breaks with pure up/down input when the X axis reads exactly 0x80.
2026-10-24 14:50:31 +10:00
Stenzek
de1cb7225d
Cheats: Add override for ForcePGXPPreserveProjectionPrecision
2026-10-24 14:12:16 +10:00
Stenzek
d9880a923a
Tests: Add PGXP CPU shift, bitwise and ADDI tests
2026-10-24 14:12:16 +10:00
Stenzek
7c143e4329
CPU/PGXP: Use double precision and only valid components in ADDI
2026-10-24 14:12:16 +10:00
Stenzek
1648a292cb
CPU/PGXP: Keep precision through ANDI/ORI that don't change the lower half
2026-10-24 14:12:16 +10:00
Stenzek
e67b8528bc
CPU/PGXP: Keep precision through OR/XOR with zero, prefer precise sources
...
OR/XOR with zero is now a move like in the recompiler, instead of dropping
fractions in natively zero halves.
2026-10-24 14:12:16 +10:00
Stenzek
8404d325a6
CPU/PGXP: Extract the upper half directly in right shifts of 16 or more
...
Precision was lost when the native upper half was 0 or -1, the validity of the
upper half wasn't moved to the lower half, and a negative lower half was off by one.
2026-10-24 14:12:15 +10:00
Stenzek
f86e410937
CPU/PGXP: Don't use invalid components in shifts
2026-10-24 14:12:15 +10:00
Stenzek
fa591af76c
CPU/PGXP: Take the upper half sign from the precise value in right shifts
...
When the native upper half is only sign bits, it could disagree with the sign
of the precise value, giving a value 65536 away. Fixes the Rage Racer sky.
2026-10-24 14:12:15 +10:00
Stenzek
c90e55081e
CPU/PGXP: Don't leave shifted-out bits in the upper half in right shifts
...
The bits shifted out of the upper half were left there as a fraction as well as
being moved into the lower half, breaking the sky in Rage Racer.
2026-10-24 14:12:15 +10:00
Stenzek
6eef7c84c0
CPU/PGXP: Mirror GTE halfword register extension in MTC2/LWC2
...
Garbage in the upper halfword caused the shadow to mismatch the register and
lose its precision on the next read.
2026-10-24 14:12:15 +10:00