Benoit Steiner
a498ff7df6
Fixed incorrect comment
2016-10-06 15:27:27 -07:00
Benoit Steiner
8ba3c41fcf
Revergted unecessary change
2016-10-06 15:12:15 -07:00
Benoit Steiner
a7473d6d5a
Fixed compilation error with gcc >= 5.3
2016-10-06 14:33:22 -07:00
Benoit Steiner
5e64cea896
Silenced a compilation warning
2016-10-06 14:24:17 -07:00
Benoit Steiner
33fba3f08d
Merged in rryan/eigen/tensorfunctors (pull request PR-233)
...
Fully support complex types in SumReducer and MeanReducer when building for CUDA by using scalar_sum_op and scalar_product_op instead of operator+ and operator*.
2016-10-06 12:29:19 -07:00
RJ Ryan
bfc264abe8
Add a test that GPU complex product reductions match CPU reductions.
2016-10-06 11:10:14 -07:00
RJ Ryan
e2e9cdd169
Fully support complex types in SumReducer and MeanReducer when building for CUDA by using scalar_sum_op and scalar_product_op instead of operator+ and operator*.
2016-10-06 10:49:48 -07:00
Benoit Steiner
d485d12c51
Added missing AVX intrinsics for fp16: in particular, implemented predux which is required by the matrix-vector code.
2016-10-06 10:41:03 -07:00
Rasmus Munk Larsen
48c635e223
Add a simple cost model to prevent Eigen's parallel GEMM from using too many threads when the inner dimension is small.
...
Timing for square matrices is unchanged, but both CPU and Wall time are significantly improved for skinny matrices. The benchmarks below are for multiplying NxK * KxN matrices with test names of the form BM_OuterishProd/N/K.
Improvements in Wall time:
Run on [redacted] (12 X 3501 MHz CPUs); 2016-10-05T17:40:02.462497196-07:00
CPU: Intel Haswell with HyperThreading (6 cores) dL1:32KB dL2:256KB dL3:15MB
Benchmark Base (ns) New (ns) Improvement
------------------------------------------------------------------
BM_OuterishProd/64/1 3088 1610 +47.9%
BM_OuterishProd/64/4 3562 2414 +32.2%
BM_OuterishProd/64/32 8861 7815 +11.8%
BM_OuterishProd/128/1 11363 6504 +42.8%
BM_OuterishProd/128/4 11128 9794 +12.0%
BM_OuterishProd/128/64 27691 27396 +1.1%
BM_OuterishProd/256/1 33214 28123 +15.3%
BM_OuterishProd/256/4 34312 36818 -7.3%
BM_OuterishProd/256/128 174866 176398 -0.9%
BM_OuterishProd/512/1 7963684 104224 +98.7%
BM_OuterishProd/512/4 7987913 112867 +98.6%
BM_OuterishProd/512/256 8198378 1306500 +84.1%
BM_OuterishProd/1k/1 7356256 324432 +95.6%
BM_OuterishProd/1k/4 8129616 331621 +95.9%
BM_OuterishProd/1k/512 27265418 7517538 +72.4%
Improvements in CPU time:
Run on [redacted] (12 X 3501 MHz CPUs); 2016-10-05T17:40:02.462497196-07:00
CPU: Intel Haswell with HyperThreading (6 cores) dL1:32KB dL2:256KB dL3:15MB
Benchmark Base (ns) New (ns) Improvement
------------------------------------------------------------------
BM_OuterishProd/64/1 6169 1608 +73.9%
BM_OuterishProd/64/4 7117 2412 +66.1%
BM_OuterishProd/64/32 17702 15616 +11.8%
BM_OuterishProd/128/1 45415 6498 +85.7%
BM_OuterishProd/128/4 44459 9786 +78.0%
BM_OuterishProd/128/64 110657 109489 +1.1%
BM_OuterishProd/256/1 265158 28101 +89.4%
BM_OuterishProd/256/4 274234 183885 +32.9%
BM_OuterishProd/256/128 1397160 1408776 -0.8%
BM_OuterishProd/512/1 78947048 520703 +99.3%
BM_OuterishProd/512/4 86955578 1349742 +98.4%
BM_OuterishProd/512/256 74701613 15584661 +79.1%
BM_OuterishProd/1k/1 78352601 3877911 +95.1%
BM_OuterishProd/1k/4 78521643 3966221 +94.9%
BM_OuterishProd/1k/512 258104736 89480530 +65.3%
2016-10-06 10:33:10 -07:00
Benoit Steiner
9f3276981c
Enabling AVX512 should also enable AVX2.
2016-10-06 10:29:48 -07:00
Gael Guennebaud
80b5133789
Fix compilation of qr.inverse() for column and full pivoting variants.
2016-10-06 09:55:50 +02:00
Benoit Steiner
4131074818
Deleted unecessary CMakeLists.txt file
2016-10-05 18:54:35 -07:00
Benoit Steiner
cb5cd69872
Silenced a compilation warning.
2016-10-05 18:50:53 -07:00
Benoit Steiner
78b569f685
Merged latest updates from trunk
2016-10-05 18:48:55 -07:00
Benoit Steiner
9c2b6c049b
Silenced a few compilation warnings
2016-10-05 18:37:31 -07:00
Benoit Steiner
6f3cd529af
Pulled latest updates from trunk
2016-10-05 18:31:43 -07:00
Benoit Steiner
d7f9679a34
Fixed a couple of compilation warnings
2016-10-05 15:00:32 -07:00
Benoit Steiner
ae1385c7e4
Pull the latest updates from trunk
2016-10-05 14:54:36 -07:00
Benoit Steiner
73b0012945
Fixed compilation warnings
2016-10-05 14:24:24 -07:00
Benoit Steiner
c84084c0c0
Fixed compilation warning
2016-10-05 14:15:41 -07:00
Benoit Steiner
4387433acf
Increased the robustness of the reduction tests on fp16
2016-10-05 10:42:41 -07:00
Benoit Steiner
aad20d700d
Increase the tolerance to numerical noise.
2016-10-05 10:39:24 -07:00
Benoit Steiner
8b69d5d730
::rand() returns a signed integer on win32
2016-10-05 08:55:02 -07:00
Benoit Steiner
ed7a220b04
Fixed a typo that impacts windows builds
2016-10-05 08:51:31 -07:00
Benoit Steiner
ceee1c008b
Silenced compilation warning
2016-10-04 18:47:53 -07:00
Benoit Steiner
698ff69450
Properly characterize the CUDA packet primitives for fp16 as device only
2016-10-04 16:53:30 -07:00
Rasmus Munk Larsen
7f67e6dfdb
Update comment for fast sqrt.
2016-10-04 15:09:11 -07:00
Rasmus Munk Larsen
765615609d
Update comment for fast sqrt.
2016-10-04 15:08:41 -07:00
Rasmus Munk Larsen
3ed67cb0bb
Fix a bug in the implementation of Carmack's fast sqrt algorithm in Eigen (enabled by EIGEN_FAST_MATH), which causes the vectorized parts of the computation to return -0.0 instead of NaN for negative arguments.
...
Benchmark speed in Giga-sqrts/s
Intel(R) Xeon(R) CPU E5-1650 v3 @ 3.50GHz
-----------------------------------------
SSE AVX
Fast=1 2.529G 4.380G
Fast=0 1.944G 1.898G
Fast=1 fixed 2.214G 3.739G
This table illustrates the worst case in terms speed impact: It was measured by repeatedly computing the sqrt of an n=4096 float vector that fits in L1 cache. For large vectors the operation becomes memory bound and the differences between the different versions almost negligible.
2016-10-04 14:22:56 -07:00
Benoit Steiner
6af5ac7e27
Cleanup the cuda executor code.
2016-10-04 08:52:13 -07:00
Benoit Steiner
2f6d1607c8
Cleaned up the random number generation code.
2016-10-04 08:38:23 -07:00
Benoit Steiner
881b90e984
Use explicit type casting to generate packets of zeros.
2016-10-04 08:23:38 -07:00
Benoit Steiner
616a7a1912
Improved support for compiling CUDA code with clang as the host compiler
2016-10-03 17:09:33 -07:00
Benoit Steiner
409e887d78
Added support for constand std::complex numbers on GPU
2016-10-03 11:06:24 -07:00
Gael Guennebaud
9d6d0dff8f
bug #1317 : fix performance regression with some Block expressions and clang by helping it to remove dead code.
...
The trick is to get rid of the nested expression in the evaluator by copying only the required information (here, the strides).
2016-10-01 15:37:00 +02:00
Gael Guennebaud
8b84801f7f
bug #1310 : workaround a compilation regression from 3.2 regarding triangular * homogeneous
2016-09-30 22:49:59 +02:00
Benoit Steiner
422530946f
Renamed the SYCL tests to follow the standard naming convention.
2016-09-30 08:22:10 -07:00
Gael Guennebaud
67b4f45836
Fix angle range
2016-09-30 12:46:33 +02:00
Gael Guennebaud
27f3970453
Remove std:: prefix
2016-09-30 12:40:41 +02:00
Gael Guennebaud
3860a0bc8f
bug #1312 : Quaternion to AxisAngle conversion now ensures the angle will be in the range [-pi,pi]. This also increases accuracy when q.w is negative.
2016-09-29 23:23:35 +02:00
Gael Guennebaud
33500050c3
bug #1308 : fix compilation of some small products involving nullary-expressions.
2016-09-29 09:40:44 +02:00
Benoit Steiner
27d7628f16
Updated the list of warnings to reflect the new message ids introduced in cuda 8.0
2016-09-28 17:42:59 -07:00
Benoit Steiner
2bda1b0d93
Updated the tensor sum and mean reducer to enable them to process complex numbers on cuda gpus.
2016-09-28 17:08:41 -07:00
Mehdi Goli
dd602e62c8
Converting alias template to nested struct in order to be compatible with CXX-03
2016-09-27 16:21:19 +01:00
Gael Guennebaud
f3a00dd2b5
Merged in sergiu/eigen (pull request PR-229)
...
Disabled MSVC level 4 warning C4714
2016-09-27 09:28:08 +02:00
Gael Guennebaud
892afb9416
Add debug info.
2016-09-26 23:53:57 +02:00
Gael Guennebaud
779774f98c
bug #1311 : fix alignment logic in some cases of (scalar*small).lazyProduct(small)
2016-09-26 23:53:40 +02:00
Benoit Steiner
6565f8d60f
Made the initialization of a CUDA device thread safe.
2016-09-26 11:00:32 -07:00
Gael Guennebaud
48dfe98abd
bug #1308 : fix compilation of vector * rowvector::nullary.
2016-09-25 14:54:35 +02:00
Sergiu Deitsch
fe29157d02
disabled MSVC level 4 warning C4714
...
The level 4 warning (/W4) warns about functions marked as __forceinline not
inlined, and generates a lot of noise.
2016-09-25 14:25:47 +02:00
