José Fonseca [Thu, 30 May 2013 13:00:44 +0000 (14:00 +0100)]
dxgitrace: Handle simultanous D3D11 maps.
On D3D11 Map/Unmap are context methods, not resource methods, but in order
to handle simultanous maps we need to put the book-keeping in the
resource objects.
Vasil Dimov [Fri, 24 May 2013 11:50:51 +0000 (14:50 +0300)]
In Android.mk: explicitly specify the toolchain
If NDK is cloned from git://codeaurora.org/platform/prebuilt instead of
from git://android.git.linaro.org/platform/prebuilt then
prebuilt/linux-x86/toolchain/ contains some extra entries that confuse
the toolchain pickup logic in cmake/toolchain/android.toolchain.cmake,
thus we explicitly specify the toolchain via ANDROID_TOOLCHAIN_NAME.
libbacktrace #include's two header files from GCC. Provide minimal
replacements for those files. For filenames.h, declare missing macros
assuming POSIX-style paths. For dwarf2.h, #include system dwarf.h and
declare missing enum tags and define missing enum values as macros.
There was a bool field that was caching results of 'is_backtrace_needed'
lookups. Now that field is gone, and 'is_backtrace_needed' is called
repeatedly. I guess it's not that bad (it boils down to a lookup in an
std::set), but that function was also printing notices for functions
with enabled stack trace recording, and now those messages are printed
repeatedly as well. The following patch removes the notices completely
and slightly cleans up the surrounding code.
Ben Kelly [Mon, 6 May 2013 17:25:15 +0000 (13:25 -0400)]
Use ndk-android-r7 for FirefoxOS builds.
Commit 7445733 added support for FirefoxOS, but appears to have been using
a newer version of the NDK than we typically build with. The codeaurora.org
prebuilts repo we are tracking has not pulled in r8 yet and only provides r7.
For the time being it would be helpful to drop back to r7. Admittedly, we
should probably have a better mechanism for selecting the latest version
available.
Changes from v3: Instead of writing the backtrace as Array, the backtrace is
now recorded as a list of stack frame nodes with optional stack frame details
(the scheme is below).
This patch implements backtrace recording during tracing, and adds support in
'apitrace dump' and QApitrace. Backtrace is obtained via platform-specific
functions (and, internally, in platform-specific format). Then it is parsed to
produce an std::vector of stack frame structs: { char *module, *function,
*filename, *linenumber, *offset } (some fields may be NULL) and is written
into the trace file in the Enter call section as a call detail:
BACKTRACE
FRAME
MODULE "module"
FUNCTION "Foo"
FILENAME "foo.cpp"
LINENUMBER "1234"
OFFSET "0xSDF"
FRAME
FUNCTION "Boo"
// no filename line info available for this frame
END_BACKTRACE
A platform-dependent mechanism is provided to specify a set of traced
calls for which backtraces will be recorded. It is possible to specify
either function names, or prefixes of names by appending a '*' (e.g.
"glUniform*").
On Android the backtrace is retrieved from Dalvik via libdvm functions
imported at runtime.
Function set is specified in /data/apitrace.fnames, one per line.
On Linux the backtrace is retrieved via glibc backtrace(), and will not always
yield filename:linenumber information.
Function set is specified via APITRACE_BT_FUNCTIONS environment variable.
On other platforms, obtaining a backtrace is not implemented by this patch.
Ben Kelly [Sat, 4 May 2013 04:35:37 +0000 (21:35 -0700)]
Fix Android target build on Apple hosts.
Currently CMakeLists.txt forces CMAKE_C_COMPILER to clang whenever
compiling on an APPLE host. Unfortunately, this prevents the
android.toolchain.cmake script from selecting the correct cross
compiler. See [line 1136][].
To avoid this problem, only force clang if ANDROID_NDK is not set. This
is one of the variables used by Android.mk.
Shuang He [Wed, 24 Apr 2013 02:01:15 +0000 (10:01 +0800)]
Add option '--snapshot-format=' to allow write raw RGB directly to stdout
This allows to create h264 video with gstreamer-vaapi very fast, and save much more disk space.
Using PNM format couldn't reach same performance, since PNM decoding seems inefficient in gstreamer.
Following are some experiment data with smokinguns demo 1920x1080 resolution on Ivybridge:
1. With png saving, it runs at around 4.5 FPS, which has compression rate 5.15x against raw RGB data
Reference command:
glretrace -s snapshot/ smokinguns.trace
2. Using PNM with gstreamer-vaapi, which could run at around 9 FPS, which has compression rate
485x (QP dependent) against raw RGB data.
Reference command:
glretrace -s - smokinguns.trace | gst-launch-0.10 fdsrc blocksize=409600 ! queue \
! pnmdec ! videoparse format=rgb width=1920 height=1080 ! ffmpegcolorspace ! queue \
! vaapiupload direct-rendering=0 ! queue ! vaapiencodeh264 ! filesink location=xxx.264
3. With following command that directly write raw RGB stream and encoded into H264 video, it
runs at around 18.5 FPS, which has compression rate 485x (QP dependent) against raw RGB data,
Reference command:
glretrace --snapshot-format=RGB -s - smokinguns.trace | gst-launch-0.10 fdsrc blocksize=409600 ! queue \
! videoparse format=rgb width=1920 height=1080 ! queue ! ffmpegcolorspace ! queue \
! vaapiupload direct-rendering=0 ! queue ! vaapiencodeh264 ! filesink location=xxx.264
v2: Use --snapshot-format= option to specify which format is used to write to stdout output
v3: Use enum for snapshotFormat and add example in README.markdown
This Android.mk is part of "Bug 831147 - Integrate apitrace into build"
https://bugzilla.mozilla.org/831147 With this addition, when apitrace
sources are cloned into B2GROOT/external/apitrace, then the FirefoxOS
build system picks up automatically apitrace, compiles and installs it
to the right place.
Signed-off-by: José Fonseca <jose.r.fonseca@gmail.com>