| 31805e60533f21575d584aa2a8c1d738b7fda56e |
|
21-Jul-2015 |
Pyry Haulos <phaulos@google.com> |
Fix warning in modf() precision test code
Bug: 22512069
Change-Id: I2a0ad0af5b9d4da3c55ca0925d375c4b5c3df307
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 18243b3e2d961ecbd5f00139a2ed1a5e777da562 |
|
15-Jul-2015 |
Pyry Haulos <phaulos@google.com> |
Relax modf(Inf) fractional part verification
See Khronos bug 13907
Bug: 22512069
Change-Id: Ic608a8a520d34fe84c3a4c5b5d96d0623e6aebd0
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| e8efb729845f35160e5ab692226a1394154ccb24 |
|
07-Jul-2015 |
Pyry Haulos <phaulos@google.com> |
Relax asin() precision requirements
Use absolute error of 2^-11 for highp and 2^-8 for mediump.
Bug: 22302334
Change-Id: Id3dea981bfc6c1db1e0b4098ab844a77739a930c
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| b5d3366020ce9abfdbd6d10686d8c2fea7787ce9 |
|
07-Jul-2015 |
Pyry Haulos <phaulos@google.com> |
Fix scaling of precision for mediump sin/cos/tan
Code should have been scaling expected precision linearly based on input
value, but was actually using a fixed precision of ~3 ULP.
Bug: 21668605
Change-Id: I00640889319bc7673def3e1541fdd574e8e13306
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 71438518d5d9fc0d4d1b987aab13bd5caefbcf88 |
|
30-May-2015 |
Pyry Haulos <phaulos@google.com> |
Merge "Allow alternative mix() implementation" into mnc-dev
|
| 6c1b3db5fb439806b9b67442f1042615ffa32816 |
|
29-May-2015 |
Pyry Haulos <phaulos@google.com> |
Allow alternative mix() implementation
Some compilers translate mix(a, b, x) to a + (b - a)*x which produces
slightly different results than the form a*(1 - x) + b*x suggested by
the GLSL specification. This change relaxes tests to allow both forms.
Bug: 21110253
Change-Id: I3af056145ef8e64c0fc998302770fd38d105481a
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 123c801f106dd1157fc889eaec934272ac6be937 |
|
29-May-2015 |
Pyry Haulos <phaulos@google.com> |
Allow different multiplication order in reflect()
GLSL spec allows compiler to assume that multiplication is commutative.
reflect() contains three multiplications which compiler can re-order
based on the spec language. This change relaxes reflect() tests to
allow both evaluation orders.
Bug: 19408572
Change-Id: I78e5b6e6f7bc761b4c49f5e1d4a283fcc60bf512
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 0557a707716b8a8722fa509455badf81633ad83b |
|
29-May-2015 |
Pyry Haulos <phaulos@google.com> |
Add support for alternative forms to precision tests
This adds alternative(a, b) that can be used to allow two alternative
forms of sub-expression. Alternative forms will be printed to the test
log output as "{a | b}".
Bug: 19408572
Bug: 21110253
Change-Id: Id6a7508e041be2bc3a42ec44dd318b49ad7fb841
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 4fdcdd50d72a1317dbab9f75738e5bb9ccd30634 |
|
05-May-2015 |
Pyry Haulos <phaulos@google.com> |
Relax builtin precision tests, fix issue in div rounding
* sin(), cos() and derived: Use |x| * 2^-10 for mediump inputs outside
of range [-pi,pi].
* asin(), acos(), and atan(): Use 4096 ULP for highp (from OpenCL's
fast relaxed math).
* modf(): allow any value, including NaN, as fractional if Inf support
is not mandated for the precision (lowp and mediump).
* atan(y, x): allow any value if y or x is Inf and no Inf support is
mandated for the precision.
* div (and derived): allow rounding of -/+inf to -/+maxval when both
inputs are ordinary and denominator is not 0 (legal in RTZ mode).
Bug: 20453754
Bug: 18323265
Change-Id: I8e2fad8909e29993e961d66a175963071ea2e5a8
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| e641b2dde1339ceeabb54ece7f063e255992a198 |
|
03-Apr-2015 |
Mika Isojärvi <misojarvi@google.com> |
Merge "Fix / Clean up de::SharedPtr."
|
| 6801c0680107ff001b065db07b125d622926f311 |
|
31-Mar-2015 |
Mika Isojärvi <misojarvi@google.com> |
Fix / Clean up de::SharedPtr.
- Fix SharedPtr removing input ptr if memory allocation fails.
- Remove almost unused threadSafe template argument.
- Remove deleter template parameter. Use templated constructor instead.
- Use virtual base class for SharedPtrState to fix issue with invalid
reinterpret_casts and some nasty issues with deleters.
- If assigning WeakPtr to last referencing SharedPtr succeed instead of
releasing last reference and failing to acquire WeakPtr.
- If SharedPtr fails to acquire WeakPtr do not release existing reference.
Change-Id: I4a0d522485e085397ff059dc4db4e3bc7986d629
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 2fe928514327a48b70de83c293881347797da3e2 |
|
19-Mar-2015 |
Pyry Haulos <phaulos@google.com> |
Allow choosing random input count in builtin_functions.precision tests
Builtin function precision tests now use --deqp-test-iteration-count=N
command line option, if given, to choose number of random input vectors.
Otherwise default sample count of 16k is used.
Change-Id: Ib9990d607c0d7bf114d8ae4fc52a19f612205114
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| d6148171f88da1301f053e2e0236afc69416137c |
|
16-Mar-2015 |
Jarkko Pöyry <jpoyry@google.com> |
Move array utils out of deDefs.
Change-Id: I77748493f96d9a25288f903eff1eebcafaa9b2c8
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 5730719a54e92e0816ef2b4217a2277907d67bcc |
|
13-Feb-2015 |
Jarkko Pöyry <jpoyry@google.com> |
Prevent incorrect usage of de::getSizedArrayElement.
- Prevent usage without <..._LAST> template argument.
- Preserve old usage as de::getArrayElement.
Change-Id: I1ab7cf128ebe8b1805f2fc4ac062dc8f6cc7e859
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| ab44fc6da25bb126919615ad2ded101695251161 |
|
13-Feb-2015 |
Jarkko Pöyry <jpoyry@google.com> |
Move tcu::ResultCollector to a separate file.
- Move tcu::ResultCollector to a separate file.
- Fix includes in tests.
Change-Id: I688d84e5de7bfa4c80b0034cdac4ea0e4fd2faa7
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 233ccc95381db673521e4e95dee2d201dc018cd6 |
|
21-Nov-2014 |
Pyry Haulos <phaulos@google.com> |
Remove sin/cos workaround for libhoudini
Bug: 18093947
Change-Id: I20dea41a38245ccc6318275207d7981973714cf9
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 6541c9eb90bec1066d78d566a6de4551683fa5bb |
|
22-Oct-2014 |
Pyry Haulos <phaulos@google.com> |
Workaround for libhoudini.so sin&cos bug
This is a workaround for Intel ARM->x86 translator (houdini) bug.
sin() & cos() return garbage for very large inputs. The outcome is
that when codomain is applied the result interval becomes empty.
Workaround is to bring the input value to the base range via modulo
if sin/cos returns an invalid values.
Change-Id: I2178daf2f9ffec543c1cf9fad2881719dd7736cc
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 2e751e3e77060f699c560104e82379d3ce54f67a |
|
25-Sep-2014 |
Pyry Haulos <phaulos@google.com> |
Import several fixes from main repo:
* Rounding of overflowed operation results in shader precision tests
* GL object leak in glu::ObjectVector
* False negatives due to ambiguous cube corner texel selection
* x11_glx build from this repo (bonus)
Change-Id: Ifeb6570cad5dec961057decc4af85365c3054f1c
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 8852c82a1ffa4760985c17cc6875d5d521daf343 |
|
11-Sep-2014 |
Jarkko Poyry <jpoyry@google.com> |
Update dEQP.
Migrate drawElements Quality Program changes from an internal repository.
Bug: 17388917
Change-Id: I21e3f7bc75813f0510025d697d91a2554dc995d4
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|
| 3c827367444ee418f129b2c238299f49d3264554 |
|
02-Sep-2014 |
Jarkko Poyry <jpoyry@google.com> |
Import dEQP.
Import drawElements Quality Program from an internal repository.
Bug: 17388917
Change-Id: Ic109fe4a57e31b2a816113d90fbdf51a43e7abeb
/external/deqp/modules/glshared/glsBuiltinPrecisionTests.cpp
|