Changes in reduction-small-size.ll

be25f52f - [LV] Autogen several tests for ease of update in upcoming change

Wed, 20 Jul 2022 14:17:44 +0000

[LV] Autogen several tests for ease of update in upcoming change List of files: /llvm-project-15.0.7/llvm/test/Transforms/LoopVectorize/reduction-small-size.ll

2e14900d - [test][NewPM] Use -passes=loop-vectorize instead of -loop-vectorize

Thu, 28 Apr 2022 14:46:00 +0000

[test][NewPM] Use -passes=loop-vectorize instead of -loop-vectorizeUpdate a bunch of loop-vectorize regression tests to use the new PMsyntax (opt -passes=loop-vectorize) instead of the deprecated legacyPM syntax (opt -loop-vectorize). List of files: /llvm-project-15.0.7/llvm/test/Transforms/LoopVectorize/reduction-small-size.ll

112c1c34 - [IVDescriptor] Make sure the sign is included for negative extension.

Mon, 15 Nov 2021 10:31:07 +0000

[IVDescriptor] Make sure the sign is included for negative extension.At the moment, computeRecurrenceType does not include any sign bits inthe maximum bit width. If the value can be negative, this means the signbit will be missing and the sext won't properly extend the value.If the value can be negative, increment the bitwidth by one to make surethere is at least one sign bit in the result value.Note that the increment is also needed *if* the value is *known* to benegative, as a sign bit needs to be preserved for the sext to work.Note that this at the moment prevents vectorization, because theanalysis computes i1 as type for the recurrence when looking through theAND in lookThroughAnd.Fixes PR51794, PR52485.Reviewed By: spatelDifferential Revision: https://reviews.llvm.org/D113056 List of files: /llvm-project-15.0.7/llvm/test/Transforms/LoopVectorize/reduction-small-size.ll

30ebdf8a - [LV] Precommit test case from PR52485.

Fri, 12 Nov 2021 16:09:19 +0000

[LV] Precommit test case from PR52485. List of files: /llvm-project-15.0.7/llvm/test/Transforms/LoopVectorize/reduction-small-size.ll

e515d3a4 - [LV] Add test case from PR51794 for over-eager truncation.

Tue, 02 Nov 2021 20:17:55 +0000

[LV] Add test case from PR51794 for over-eager truncation.This patch adds a test case for PR51794 where reductions are performedon types that are too small. List of files: /llvm-project-15.0.7/llvm/test/Transforms/LoopVectorize/reduction-small-size.ll

23c2f2e6 - [LV] Mark increment of main vector loop induction variable as NUW.

Mon, 07 Jun 2021 08:24:27 +0000

[LV] Mark increment of main vector loop induction variable as NUW.This patch marks the induction increment of the main induction variableof the vector loop as NUW when not folding the tail.If the tail is not folded, we know that End - Start >= Step (eitherstatically or through the minimum iteration checks). We also know that bothStart % Step == 0 and End % Step == 0. We exit the vector loop if %IV +%Step == %End. Hence we must exit the loop before %IV + %Step unsignedoverflows and we can mark the induction increment as NUW.This should make SCEV return more precise bounds for the created vectorloops, used by later optimizations, like late unrolling.At the moment quite a few tests still need to be updated, but beforedoing so I'd like to get initial feedback to make sure I am not missinganything.Note that this could probably be further improved by using informationfrom the original IV.Attempt of modeling of the assumption in Alive2:https://alive2.llvm.org/ce/z/H_DL_gPart of a set of fixes required for PR50412.Reviewed By: mkazantsevDifferential Revision: https://reviews.llvm.org/D103255 List of files: /llvm-project-15.0.7/llvm/test/Transforms/LoopVectorize/reduction-small-size.ll

e498be57 - [LV] Strip wrap flags from vectorized reductions

Thu, 19 Dec 2019 22:04:49 +0000

[LV] Strip wrap flags from vectorized reductionsA sequence of additions or multiplications that is known not to wrap, may wrapif it's order is changed (i.e., reassociated). Therefore when vectorizinginteger sum or product reductions, their no-wrap flags need to be removed.Fixes PR43828Patch by Denis AntrushinDifferential Revision: https://reviews.llvm.org/D69563 List of files: /llvm-project-15.0.7/llvm/test/Transforms/LoopVectorize/reduction-small-size.ll

cee313d2 - Revert "Temporarily Revert "Add basic loop fusion pass.""

Wed, 17 Apr 2019 04:52:47 +0000

Revert "Temporarily Revert "Add basic loop fusion pass.""The reversion apparently deleted the test/Transforms directory.Will be re-reverting again.llvm-svn: 358552 List of files: /llvm-project-15.0.7/llvm/test/Transforms/LoopVectorize/reduction-small-size.ll

a097bc69 - [LV] Use Demanded Bits and ValueTracking for reduction type-shrinking

Sun, 04 Feb 2018 15:42:24 +0000

[LV] Use Demanded Bits and ValueTracking for reduction type-shrinkingThe type-shrinking logic in reduction detection, although narrow in scope, isalso rather ad-hoc, which has led to bugs (e.g., PR35734). This patch modifiesthe approach to rely on the demanded bits and value tracking analyses, ifavailable. We currently perform type-shrinking separately for reductions andother instructions in the loop. Long-term, we should probably think aboutcomputing minimal bit widths in a more complete way for the loops we want tovectorize.PR35734Differential Revision: https://reviews.llvm.org/D42309llvm-svn: 324195 List of files: /llvm-project-15.0.7/llvm/test/Transforms/LoopVectorize/reduction-small-size.ll

f4bb480b - [LV] Use correct insertion point when type shrinking reductions

Fri, 29 Sep 2017 18:07:39 +0000

[LV] Use correct insertion point when type shrinking reductionsWhen type shrinking reductions, we should insert the truncations and extends atthe end of the loop latch block. Previously, these instructions were insertedat the end of the loop header block. The difference is only a problem for loopswith predicated instructions (e.g., conditional stores and instructions thatmay divide by zero). For these instructions, we create new basic blocks insidethe vectorized loop, which cause the loop header and latch to no longer be thesame block. This should fix PR34687.Reference: https://bugs.llvm.org/show_bug.cgi?id=34687llvm-svn: 314542 List of files: /llvm-project-15.0.7/llvm/test/Transforms/LoopVectorize/reduction-small-size.ll

c94f8e29 - [LoopVectorize] Add Support for Small Size Reductions.

Thu, 27 Aug 2015 14:12:17 +0000

[LoopVectorize] Add Support for Small Size Reductions.Unlike scalar operations, we can perform vector operations on element types thatare smaller than the native integer types. We type-promote scalar operations ifthey are smaller than a native type (e.g., i8 arithmetic is promoted to i32arithmetic on Arm targets). This patch detects and removes type-promotionswithin the reduction detection framework, enabling the vectorization of smallsize reductions.In the legality phase, we look through the ANDs and extensions that InstCombinecreates during promotion, keeping track of the smaller type. In theprofitability phase, we use the smaller type and ignore the ANDs and extensionsin the cost model. Finally, in the code generation phase, we truncate the resultof the reduction to allow InstCombine to rewrite the entire expression in thesmaller type.This fixes PR21369.http://reviews.llvm.org/D12202Patch by Matt Simpson !llvm-svn: 246149 List of files: /llvm-project-15.0.7/llvm/test/Transforms/LoopVectorize/reduction-small-size.ll