Commits · eb65e67eeab84d8a5dcefe13720e2f1d4d99be01 · Roger Ferrer / llvm-epi · GitLab

Jun 17, 2013
- Fix for a regression caused by the LoopVectorizer when · eb90fd1c
  Pekka Jaaskelainen authored Jun 17, 2013
  eb90fd1c
May 31, 2013
- LoopVectorize: PHIs with only outside users should prevent vectorization · 70a9be52
  Arnold Schwaighofer authored May 31, 2013
  70a9be52
May 28, 2013
- Add support for llvm.vectorizer metadata · 5fdf836b
  Paul Redmond authored May 28, 2013
  5fdf836b
May 24, 2013
- LoopVectorize: LoopSimplify can't canonicalize loops with an indirectbr in it,... · 6ac1e623
  Benjamin Kramer authored May 24, 2013
  6ac1e623
May 22, 2013
- LoopVectorize: Make Value pointers that could be RAUW'ed a VH · 12b0d1cd
  Arnold Schwaighofer authored May 22, 2013
  12b0d1cd
May 18, 2013
- LoopVectorize: Handle single edge PHIs · 693a1ca6
  Arnold Schwaighofer authored May 18, 2013
  693a1ca6
May 15, 2013
- LoopVectorize: Hoist conditional loads if possible · 2d920477
  Arnold Schwaighofer authored May 15, 2013
  2d920477
May 14, 2013
- LoopVectorize: Handle loops with multiple forward inductions · 2e7a922a
  Arnold Schwaighofer authored May 14, 2013
  2e7a922a
May 12, 2013
- LoopVectorize: Use the widest induction variable type · f2305e44
  Arnold Schwaighofer authored May 11, 2013
  f2305e44
May 11, 2013
- Add an additional testcase for PR15882. · 5ff6db11
  Nadav Rotem authored May 10, 2013
  5ff6db11
May 09, 2013
- LoopVectorizer: Don't assert on the absence of induction variables · 2e8c69cf
  Arnold Schwaighofer authored May 09, 2013
  2e8c69cf
May 07, 2013
- LoopVectorizer: Improve reduction variable identification · 3610139a
  Arnold Schwaighofer authored May 07, 2013
  3610139a
- LoopVectorize: getConsecutiveVector must respect signed arithmetic · e78b76fb
  Arnold Schwaighofer authored May 07, 2013
  e78b76fb
May 05, 2013
- LoopVectorize: Add support for floating point min/max reductions · d96e427e
  Arnold Schwaighofer authored May 05, 2013
  d96e427e
- LoopVectorize: We don't need an identity element for min/max reductions · a670a0a3
  Arnold Schwaighofer authored May 05, 2013
  a670a0a3
May 03, 2013
- LoopVectorizer: Add support for if-conversion of PHINodes with 3+ incoming values. · 4ce060b3
  Nadav Rotem authored May 03, 2013
  4ce060b3
May 02, 2013
- TBAA: remove !tbaa from testing cases if not used. · 16649b01
  Manman Ren authored May 02, 2013
  16649b01
Apr 30, 2013
- TBAA: remove !tbaa from testing cases if not used. · 1a5ff287
  Manman Ren authored Apr 30, 2013
  1a5ff287
Apr 26, 2013
- LoopVectorizer: Calculate the number of pointers to disambiguate at runtime... · 13306816
  Nadav Rotem authored Apr 26, 2013
  13306816
Apr 25, 2013
- LoopVectorizer: No need to generate pointer disambiguation checks between readonly pointers. · f43cbeee
  Nadav Rotem authored Apr 25, 2013
  f43cbeee
Apr 24, 2013
- LoopVectorize: Scalarize padded types · a6578f70
  Arnold Schwaighofer authored Apr 24, 2013
  a6578f70
- LoopVectorizer: Bail out if we don't have datalayout we need it · 23a0589b
  Arnold Schwaighofer authored Apr 24, 2013
  23a0589b
Apr 23, 2013
- LoopVectorizer: Fix 15830. When scalarizing and unrolling stores make sure... · 71c9d6d3
  Nadav Rotem authored Apr 23, 2013
  71c9d6d3
- Call the potentially costly isAnnotatedParallel() only once. · d3c90e13
  Pekka Jaaskelainen authored Apr 23, 2013
  d3c90e13
- Refuse to (even try to) vectorize loops which have uniform writes, · 6f2f66b6
  Pekka Jaaskelainen authored Apr 23, 2013
  6f2f66b6
Apr 18, 2013
- LoopVectorizer: Recognize min/max reductions · 4cd6aa11
  Arnold Schwaighofer authored Apr 18, 2013
  4cd6aa11
- LoopVectorize: Use a set to avoid longer cycles in the reduction chain too. · 8df2cfb8
  Benjamin Kramer authored Apr 18, 2013
  8df2cfb8
Apr 12, 2013
- LoopVectorizer: integer division is not a reduction operation · f9cea17f
  Arnold Schwaighofer authored Apr 12, 2013
  f9cea17f
Apr 05, 2013
- LoopVectorizer: Pass OperandValueKind information to the cost model · df6f67ed
  Arnold Schwaighofer authored Apr 04, 2013
  df6f67ed
Apr 01, 2013
- X86TTI: Add accurate costs for itofp operations, based on the actual instruction counts. · 52ceb443
  Benjamin Kramer authored Apr 01, 2013
  52ceb443
Mar 14, 2013
- LoopVectorizer: Insert some white space to make test case more readable · 9b55e31b
  Arnold Schwaighofer authored Mar 14, 2013
  9b55e31b
- Add missing asserts flag to test - it uses debug flags · 4991ce9d
  Arnold Schwaighofer authored Mar 14, 2013
  4991ce9d
- LoopVectorize: Invert case when we use a vector cmp value to query select cost · c63cf3a0
  Arnold Schwaighofer authored Mar 14, 2013
  c63cf3a0
Mar 09, 2013
- Test case hygiene. · 01b75cc0
  Benjamin Kramer authored Mar 09, 2013
  01b75cc0
- LoopVectorizer: Ignore dbg.value instructions · 4090b61a
  Arnold Schwaighofer authored Mar 09, 2013
  4090b61a
Mar 08, 2013
- Force cpu in test. · 10a74ed4
  Benjamin Kramer authored Mar 08, 2013
  10a74ed4
- Insert the reduction start value into the first bypass block to preserve domination. · 37c2d65c
  Benjamin Kramer authored Mar 08, 2013
  37c2d65c
Mar 02, 2013
- X86 cost model: Adjust cost for custom lowered vector multiplies · 20ef54f4
  Arnold Schwaighofer authored Mar 02, 2013
  20ef54f4
- PR14448 - prevent the loop vectorizer from vectorizing the same loop twice. · 739e37a0
  Nadav Rotem authored Mar 02, 2013
  739e37a0
Mar 01, 2013
- LoopVectorize: Don't hang forever if a PHI only has skipped PHI uses. · 12f98fae
  Benjamin Kramer authored Mar 01, 2013
  12f98fae