ecbfeb7dc1
Fix linux compilation
2020-08-27 00:02:17 +01:00
50089207c1
Fix 32-bit compile error
2020-08-26 20:56:28 +01:00
8973ac0241
Don't serialise present chunk for glFrameTerminatorGREMEDY
2020-08-26 19:52:34 +01:00
f63a948392
Cache FBO attachments to avoid repeated query & lookup
2020-08-26 19:27:42 +01:00
3a1b4500ee
Avoid some unnecessary potentially expensive work on frequent calls
2020-08-26 19:27:42 +01:00
022ba9ce67
Combine GL maps for looking up ID/Resource Record from resource name
2020-08-26 19:27:42 +01:00
5e79465ab5
Avoid use of virtual function call for GetID within GL driver
2020-08-26 19:27:42 +01:00
c5887eb4e4
Reduce wasted iteration in GLRenderState::MarkDirty
...
* Only iterate up to the highest known bind for indexed targets, which will
generally be low or 0.
2020-08-26 19:27:42 +01:00
61d15e00ef
Don't create empty GLRenderState for use with MarkDirty
...
* MarkDirty re-queries the state and doesn't use anything so it's wasted time
initialising a massive structure
2020-08-26 19:27:42 +01:00
55c62dd307
Serialise timestamps and durations as tick counts, and convert on replay
...
* When we're capturing programs with high-frequency enough API calls, the
overhead of the math & extra function calls over Timing::GetTick() is
measurable. Removing it and serialising pure tick based durations/timestamps
and then converting on replay gives us identical results and saves time while
background capturing.
2020-08-26 19:27:42 +01:00
9fd7f447d2
Switch thumbnail to bytebuf instead of byte pointer and length
2020-08-26 19:27:42 +01:00
51b228b042
Disable resource record chunk locking for command buffers
...
* The API user is supposed to lock this for us, so don't waste time locking
ourselves.
2020-08-26 19:27:42 +01:00
1b844d1498
Add resettable chunk allocator for recording command buffers
2020-08-26 19:27:41 +01:00
d922db6dff
Fix broken re-use of resource shadow deferred ctx pointers. Closes #2023
2020-08-26 19:14:42 +01:00
62f55a374c
Add proper fence around overlay rendering command buffers
2020-08-26 18:46:56 +01:00
a457aa5a8f
Fix handling of CONCURRENT swapchain images
2020-08-26 18:46:56 +01:00
f664d13bfc
Clear up queue data when device is destroyed
2020-08-26 18:46:56 +01:00
753806d526
Add a specialisation for buffer reference add/remove
...
* This skips some potentially expensive (and redundant) checks if the buffer is
a view.
2020-08-25 17:24:23 +01:00
f013824f97
Cache D3D11 buffers being read-only to speed up bound-for-write checks
2020-08-25 16:16:39 +01:00
a32bf777be
Add extra case to assert
2020-08-24 10:52:30 +01:00
155d455f01
Fix typo in glClearFramebufferfv variants
2020-08-24 10:04:23 +01:00
a6871327aa
GL renderbuffers do not have mips, return early from function
...
* Some of the other sub-functions don't handle non-texture targets
2020-08-21 11:49:44 +01:00
409f256411
Fix event ID indexing for multidraw commands on GL. Closes #2019
2020-08-21 11:49:44 +01:00
07ee28c97f
Allow Python 3.8 to be detected
...
Adds python 3.8 to the list of python 3 versions.
Thank you so much for renderdoc!!
2020-08-21 11:07:11 +01:00
5c0dadae84
Test that RTV descriptor copies in D3D12 are properly recorded mid-frame
2020-08-20 17:09:51 +01:00
3e100109da
Test destroying/freeing NULL handles, fix validation errors
2020-08-20 17:02:03 +01:00
0b0beb3a20
Call virtual functions by explicit class in constructor
2020-08-20 16:11:05 +01:00
81fbf4bcc9
Fix some dodgy memcpy calls
2020-08-20 16:11:05 +01:00
112741a74c
Remove some parameter name shadowing
2020-08-20 16:11:05 +01:00
fa24dc8872
Silence some PVS warnings
2020-08-20 16:11:05 +01:00
b5de095f53
Use explicit enum comparisons to check for empty flags
2020-08-20 15:26:02 +01:00
52632ccc1c
Fix incorrect array index used in disassembly
2020-08-20 15:26:02 +01:00
274185f808
Fix wrong size for memset
2020-08-20 15:26:02 +01:00
c01dd73bf6
Remove some dead code/variables
2020-08-20 15:26:02 +01:00
64f98f286e
Add test of heavy descriptor reallocation and reuse
2020-08-20 13:48:17 +01:00
077aebc9e1
Fix sanitisation of aspect mask in subresource range
2020-08-20 10:49:23 +01:00
83714b825c
Remove note about android support being new
2020-08-20 10:47:16 +01:00
a48ab556f3
Fix compile errors
2020-08-19 19:18:37 +01:00
ffac439cce
Record which queries are reset in each frame & restore them to validity
...
* It's possible that a frame capture could copy from a query and then reset it,
without then recording something valid into the query. Then the next replay
the copy would be unavailable and if WAIT_BIT is set that would lead to a
device lost.
* We already fill out queries with dummy valid data on create time, so now we
record any queries that are reset and re-fill them with valid data again.
2020-08-19 18:02:03 +01:00
4afc97e6dd
Fix backwards assert
2020-08-19 17:57:50 +01:00
213aded173
Fix compilation of single flush validate
2020-08-19 16:35:58 +01:00
15f35931df
Fix wrong string format specifier
2020-08-19 15:21:00 +01:00
73cc1f5476
Add specialised rdcarray which implements key/value lookup
2020-08-19 15:21:00 +01:00
d2e0b7ceb1
Implement move semantics support for rdcarray/pair
2020-08-19 15:21:00 +01:00
babbdadc36
Cache a single value inline in ImageSubresourceMap
...
* Having a single subresource (range) is a common case, so avoid allocating
storage in an array for that, only switch to the array when we have more than
one range to store.
2020-08-19 14:24:51 +01:00
8a350f28c8
Batch updates to background ref cache
...
* If the background cache is empty also (for a descriptor set that has never
been written before) we can duplicate it straight from bindFrameRefs).
2020-08-19 14:24:51 +01:00
0859aaa163
Change descriptor set lock to spinlock
...
* We don't expect contention on this, the only time it will contend is when
actively capturing a frame between updates and submits reading the descriptor
contents, so we penalise that case while making the background case faster -
since a spinlock is 'free' to take when there's no contention.
2020-08-19 14:24:51 +01:00
6eca4a4cd4
Cache call to FormatImageAspects
2020-08-19 14:24:51 +01:00
d0f59de405
Re-use descriptors in descriptor pool
...
* If an application allocates from and resets descriptor pools at very high
frequency the overhead of freeing and reallocating those descriptor sets can
be high. Instead use the descriptor pool as a pool for children and look up
the freelist first for an existing descriptor set before trying to allocate a
new one.
2020-08-19 14:24:51 +01:00
488a9e3faf
Block allocate storage for current descriptor set bindings
2020-08-19 14:24:51 +01:00
baldurk