Commit message (Collapse) | Author | Age | |
---|---|---|---|
* | Fix computeProductBlockingSizes with m==0, and add respective unit test. | 2015-03-31 | |
| | |||
* | merge | 2015-03-27 | |
|\ | |||
* | | Fix hypot(0,0). | 2015-03-27 | |
| | | |||
| * | Fixed the CUDA packet primitives | 2015-03-24 | |
| | | |||
* | | Make MatrixBase::is* methods aware of nested_eval. | 2015-03-24 | |
|/ | |||
* | Fix MSVC compilation: aligned type must be passed by reference | 2015-03-19 | |
| | |||
* | Fix comparison warning | 2015-03-19 | |
| | |||
* | Improve random number generation for integer and add unit test | 2015-03-19 | |
| | |||
* | use unsigned short instead of uint16_t which doesn't exist in c++98 | 2015-03-17 | |
| | |||
* | Similar to cset 3589a9c115a892ea3ca5dac74d71a1526764cb38 | 2015-03-16 | |
| | | | | , also in 2px4 kernel: actual_panel_rows computation should always be resilient to parameters not consistent with the known L1 cache size, see comment | ||
* | fix bug in maxsize calculation, which would cause products of size > 2048 to ↵ | 2015-03-16 | |
| | | | | address the lookup table out of bounds | ||
* | Update Nexus 5 lookup table from combining now 2 runs of the benchmark, ↵ | 2015-03-16 | |
| | | | | using the analyze-blocking-sizes partition tool. Gives better worst-case performance. | ||
* | fix compilation with GCC 4.8 | 2015-03-16 | |
| | |||
* | Fix bug in case where EIGEN_TEST_SPECIFIC_BLOCKING_SIZE is defined but false | 2015-03-15 | |
| | |||
* | Provide a empirical lookup table for blocking sizes measured on a Nexus 5. ↵ | 2015-03-15 | |
| | | | | Only for float, only for Android on ARM 32bit for now. | ||
* | actual_panel_rows computation should always be resilient to parameters not ↵ | 2015-03-15 | |
| | | | | consistent with the known L1 cache size, see comment | ||
* | Fix a unused-var warning | 2015-03-15 | |
| | |||
* | Refactor computeProductBlockingSizes to make room for the possibility of ↵ | 2015-03-15 | |
| | | | | using lookup tables | ||
* | organize a little our default cache sizes, and use a saner default L1 ↵ | 2015-03-13 | |
| | | | | outside of x86 (10% faster on Nexus 5) | ||
* | bug #973, improve AVX support by enabling vectorization of Vector4i-like ↵ | 2015-03-13 | |
| | | | | types, and enforcing alignement of Vector4f/Vector2d-like types to preserve compatibility with SSE and future Eigen versions that will vectorize them with AVX enabled. | ||
* | Fix internal::random(x,y) for integer types. The previous implementation ↵ | 2015-03-13 | |
| | | | | could return y+1. The new implementation uses rejection sampling to get an unbiased behabior. | ||
* | Avoid undeflow when blocking size are tuned manually. | 2015-03-06 | |
| | |||
* | bug #969: workaround abiguous calls to Ref using enable_if. | 2015-03-06 | |
| | |||
* | bug #978: early return for vanishing products | 2015-03-06 | |
| | |||
* | Improve blocking heuristic: if the lhs fit within L1, then block on the rhs ↵ | 2015-03-06 | |
| | | | | in L1 (allows to keep packed rhs in L1) | ||
* | Improve product kernel: replace the previous dynamic loop swaping strategy ↵ | 2015-03-06 | |
| | | | | | | by a more general one: It consists in increasing the actual number of rows of lhs's micro horizontal panel for small depth such that L1 cache is fully exploited. | ||
* | Product optimization: implement a dynamic loop-swapping startegy to improve ↵ | 2015-03-05 | |
| | | | | memory accesses to the destination matrix in the case of K-rank-update like products, i.e., for products of the kind: "large x small" * "small x large" | ||
* | Fix asm comments in 1px1 kernel | 2015-03-03 | |
| | |||
* | Add a benchmark-default-sizes action to benchmark-blocking-sizes.cpp | 2015-03-03 | |
| | |||
* | New scoring functor to select the pivot. | 2015-03-03 | |
| | | | | This is can be useful for non-floating point scalars, where choosing the biggest element is generally not the best choice. | ||
* | must also disable complex<double> when disabling double vectorization | 2015-03-03 | |
| | |||
* | Work around an ICE in Clang 3.5 in the iOS toolchain with double NEON ↵ | 2015-03-03 | |
| | | | | intrinsics. | ||
* | HalfPacket also needed to be disabled for double, on ARMv8. | 2015-03-02 | |
| | |||
* | Increase unit-test L1 cache size to ensure we are doing at least 2 peeled ↵ | 2015-02-27 | |
| | | | | loop within product kernel. | ||
* | Re-enbale detection of min/max parentheses protection, and re-enable ↵ | 2015-02-27 | |
| | | | | mpreal_support unit test. | ||
* | Reimplement the selection between rotating and non-rotating kernels | 2015-02-27 | |
| | | | | | | using templates instead of macros and if()'s. That was needed to fix the build of unit tests on ARM, which I had broken. My bad for not testing earlier. | ||
* | remove trailing comma | 2015-02-27 | |
| | |||
* | Disable Packet2f/2i halfpacket support in NEON. | 2015-02-27 | |
| | | | | | | I believe that it was erroneously turned on, since Packet2f/2i intrinsics are unimplemented, and code trying to use halfpackets just fails to compile on NEON, as it tries to use the default implementation of pload/pstore and the types don't match. | ||
* | Replace a static assert by a runtime one, fixes the build of unit tests on ARM | 2015-02-27 | |
| | | | | | Also safely assert in the non-implemented path that should never be taken in practice, and would return wrong results. | ||
* | Avoid packing rhs multiple-times when blocking on the lhs only. | 2015-02-26 | |
| | |||
* | Make sure that the block size computation is tested by our unit test. | 2015-02-26 | |
| | |||
* | Implement a more generic blocking-size selection algorithm. See explanations ↵ | 2015-02-26 | |
| | | | | | | | inlines. It performs extremely well on Haswell. The main issue is to reliably and quickly find the actual cache size to be used for our 2nd level of blocking, that is: max(l2,l3/nb_core_sharing_l3) | ||
* | Fix typos in block-size testing code, and set peeling on k to 8. | 2015-02-26 | |
| | |||
* | So I extensively measured the impact of the offset in this prefetch. I tried ↵ | 2015-02-25 | |
| | | | | | | | | | | | | | | offset values from 0 to 128 (on this float* pointer, so implicitly times 4 bytes). On x86, I tested a Sandy Bridge with AVX with 12M cache and a Haswell with AVX+FMA with 6M cache on MatrixXf sizes up to 2400. I could not see any significant impact of this offset. On Nexus 5, the offset has a slight effect: values around 32 (times sizeof float) are worst. Anything else is the same: the current 64 (8*pk), or... 0. So let's just go with 0! Note that we needed a fix anyway for not accounting for the value of RhsProgress. 0 nicely avoids the issue altogether! | ||
* | bug #970: Add EIGEN_DEVICE_FUNC to RValue functions, in case Cuda supports ↵ | 2015-02-24 | |
| | | | | RValue-references. | ||
* | Fix my recent prefetch changes: | 2015-02-23 | |
| | | | | | | | | | | | - the first prefetch is actually harmful on Haswell with FMA, but it is the most beneficial on ARM. - the second prefetch... I was very stupid and multiplied by sizeof(scalar) and offset of a scalar* pointer. The old offset was 64; pk = 8, so 64=pk*8. So this effectively restores the older offset. Actually, there were two prefetches here, one with offset 48 and one with offset 64. I could not confirm any benefit from this strange 48 offset on either the haswell or my ARM device. | ||
* | log1p is defined only for real Scalars in C++11 | 2015-02-21 | |
| | |||
* | Fix compilation of unit tests disabling assertion cheking | 2015-02-21 | |
| | |||
* | Fix doc of Ref<> | 2015-02-20 | |
| | |||
* | In C++11 destructors do not throw by default (fix CommaInitializer unit test) | 2015-02-20 | |
| |