Commit Graph

5158 Commits

Author SHA1 Message Date
Bartosz Taudul
2df1eaaa7e Pack color indices using SSE. 2019-06-28 21:58:10 +02:00
Bartosz Taudul
d593e5dfa9 DXT1 SIMD color index table generator. 2019-06-28 21:57:38 +02:00
Bartosz Taudul
3208d6c803 Add ARM NEON DXT1 compression timings to manual. 2019-06-28 14:26:00 +02:00
Bartosz Taudul
fcb5b4b888 NEON DXT1 compression. 2019-06-28 14:24:16 +02:00
Bartosz Taudul
e8d4ba492b Unify shifts. 2019-06-28 13:05:32 +02:00
Bartosz Taudul
be4900c822 NEON CheckSolid. 2019-06-28 01:47:04 +02:00
Bartosz Taudul
33486fa3cf Update ARM timings. 2019-06-27 22:47:26 +02:00
Bartosz Taudul
3c066f1527 Simplify code. 2019-06-27 22:40:03 +02:00
Bartosz Taudul
77c6acbc48 Update manual. 2019-06-27 22:30:05 +02:00
Bartosz Taudul
72a0d4c2ab Rest of SSE DXTC compression. 2019-06-27 22:29:44 +02:00
Bartosz Taudul
137b28e110 SSE CheckSolid. 2019-06-27 22:29:44 +02:00
Bartosz Taudul
3d590b6b8c Initialize rpmalloc in compression thread. 2019-06-27 19:14:51 +02:00
Bartosz Taudul
10bcc8c770 Switch to DXT1 textures in profiler utility. 2019-06-27 19:14:51 +02:00
Bartosz Taudul
1939c31165 Experimental DXT1 compressor. 2019-06-27 19:14:51 +02:00
Bartosz Taudul
79eb1b9029 Swap queue and dequeue only if queue has contents. 2019-06-27 13:37:09 +02:00
Bartosz Taudul
aa4ce30dff Update manual. 2019-06-27 13:32:57 +02:00
Bartosz Taudul
7dc7ece2bd Add staging area for frame images.
Compressing frame images on a separate thread may cause frame image
arrival before frames are sent. Fix this issue by creating a staging
area in which frame images will wait for frames to arrive.

This probably breaks playback functionality, as non-existent frames may
be queried, but this problem seems to be very hard to find, so let's
ignore it for now.
2019-06-27 13:24:35 +02:00
Bartosz Taudul
bb35f9a897 Compress frame images in a separate thread. 2019-06-27 13:24:35 +02:00
Bartosz Taudul
7ebd2162c6 Add ETC1 compression thread. 2019-06-26 22:57:24 +02:00
Bartosz Taudul
f565e11976 Store frame images in queue. 2019-06-26 22:52:24 +02:00
Bartosz Taudul
fc106079c5 Remove CPU migration highlight for zones. 2019-06-26 21:35:09 +02:00
Bartosz Taudul
3bf23e15bb Update manual. 2019-06-26 21:07:12 +02:00
Bartosz Taudul
e3aa0a5c88 Update NEWS. 2019-06-26 21:03:34 +02:00
Bartosz Taudul
bc3c375b02 Display crash icon in notification area. 2019-06-26 21:02:04 +02:00
Bartosz Taudul
b8794f64be Extract crash tooltip to a separate function. 2019-06-26 21:01:54 +02:00
Bartosz Taudul
281dcf7c1f Cast to proper types. 2019-06-26 19:33:37 +02:00
Bartosz Taudul
8ce41b3543 Proper init order of thread local thread handle. 2019-06-26 19:32:52 +02:00
Bartosz Taudul
64980a1e6f Use async resolv service. 2019-06-26 18:49:21 +02:00
Bartosz Taudul
5e97e83401 Address can't change. 2019-06-26 18:46:51 +02:00
Bartosz Taudul
913c1e57a6 Add threaded resolv service. 2019-06-26 18:46:51 +02:00
Bartosz Taudul
a8cb257474 Revert "Resolve client host name using DNS."
This reverts commit 48df667a37.
2019-06-26 17:58:23 +02:00
Bartosz Taudul
0aa0b4ac8a Try lower query counts in out-of-memory situations. 2019-06-26 16:43:56 +02:00
Bartosz Taudul
659631fc06 Make vulkan query count variable. 2019-06-26 16:42:51 +02:00
Bartosz Taudul
bc7f2c49c8 GetThreadHandle() might be used by application's code. 2019-06-25 15:44:49 +02:00
Bartosz Taudul
0b656c3469 Update manual. 2019-06-24 21:15:33 +02:00
Bartosz Taudul
9ca254307a Add callstack versions of C API macros. 2019-06-24 21:10:41 +02:00
Bartosz Taudul
c749a2e3fe Add C API for plots and messages. 2019-06-24 21:03:39 +02:00
Bartosz Taudul
48e08acb62 Add C API for frame markup. 2019-06-24 21:03:39 +02:00
Bartosz Taudul
ee99ce833c Implement memory allocation tracking for C API. 2019-06-24 21:03:39 +02:00
Bartosz Taudul
281477f7f9 Tokens must be retrieved for each enqueue. 2019-06-24 20:12:14 +02:00
Bartosz Taudul
06a41708a7 Move TLS accesses close together. 2019-06-24 19:38:44 +02:00
Bartosz Taudul
c4f0965851 Don't use cached thread id to retrieve main thread id. 2019-06-24 19:38:07 +02:00
Bartosz Taudul
a56c47a6a0 Store thread handle in a thread local variable.
This saves us a non-inlineable function call. Thread local block is
accessed anyway, since we need to get the token, so we already have the
pointer and don't need to get it a second time (which is done inside
Windows' GetCurrentThreadId()). We also don't need to store the thread
id in ScopedZone anymore, as it was a micro-optimization to save us the
second GetThreadHandle() call.

This change has a measurable effect of reducing enqueue time from ~10 to
~8 ns.

A further optimization would be to completely skip thread handle
retrieval during zone capture and do it instead on retrieval of data
from the queue. Since each thread has its own producer ("token"), the
thread handle should be accessible during the dequeue operation. This is
a much more invasive change, that would require a) modification of the
queue, b) additional processing of dequeued data to inject the thread
handle.
2019-06-24 19:19:47 +02:00
Bartosz Taudul
46b75c5a19 Only enable tracy-internal GetThreadHandle if tracy is enabled. 2019-06-24 19:18:52 +02:00
Bartosz Taudul
79bfac9ca9 Use proper popcnt for gcc/clang (including cygwin). 2019-06-24 18:56:04 +02:00
Bartosz Taudul
9375afdbed All variables must be defined before goto. 2019-06-23 00:36:25 +02:00
Bartosz Taudul
6bdfedead2 Update nfd to ceb75f7abf3. 2019-06-23 00:35:19 +02:00
Bartosz Taudul
815ad7df28 Update manual. 2019-06-23 00:21:56 +02:00
Bartosz Taudul
a8dcd5d153 Ctrl-click on frame in frame overview to show playback window. 2019-06-23 00:11:46 +02:00
Bartosz Taudul
f125254d14 Cosmetics. 2019-06-23 00:00:16 +02:00