search for: vectorutils

I was trying to extend "VectorUtils.h", found all functions are defined as static. The common problem when we include this and only use few function, compiler throw warning for unused functions. i.e. > llvm/include/llvm/Transforms/Utils/VectorUtils.h:102:1: warning: > 'llvm::Intrinsic::ID llvm::getIntrinsicIDForCall...

...9/llvm/lib/Target/Hexagon/HexagonFrameLowering.cpp#L1533 4 - https://github.com/llvm/llvm-project/blob/b725d27350ffdda9e8e9144668ad54c922297f59/llvm/lib/CodeGen/CodeGenPrepare.cpp#L6751 5 - https://github.com/llvm/llvm-project/blob/d0307f93a7658e6d0eef1ffd0b0ed4f1506bfc13/llvm/include/llvm/Analysis/VectorUtils.h#L278 6 - https://github.com/llvm/llvm-project/blob/fafec5155e39f5dad098376c1beb4a56604aa655/llvm/lib/Target/ARM/ARMCallingConv.cpp#L207 7 - https://github.com/llvm/llvm-project/blob/d0307f93a7658e6d0eef1ffd0b0ed4f1506bfc13/llvm/lib/Transforms/Instrumentation/ThreadSanitizer.cpp#L563 8 - https://g...

A rotate intrinsic should be relatively close in cost/complexity to the existing bswap. A grep of intrinsic::bswap says we'd probably add code in: InstCombine InstructionSimplify ConstantFolding DemandedBits ValueTracking VectorUtils SelectionDAGBuilder But I don't think it's fair to view those additions as pure added cost. As an example, consider that we have to add hacks to EarlyCSE to recognize multi-IR-instruction min/max/abs patterns. Intrinsics just work as-is there. So if you search for 'matchSelectPattern...

Hi, We have a function in IRBuilder.h Value *CreateVectorSplat(unsigned NumElts, Value *V, const Twine &Name = "") { .. } This function creates 2 instructions - "insertelement" and "shuffle" with all-zero mask. Now I want to add Value *getSplatValue(Value *Val). This function will try to recognize the pattern - insertelement+shuffle and return the splat value

Am Mi., 24. Juni 2020 um 10:12 Uhr schrieb David Blaikie <dblaikie at gmail.com>: > > As mentioned in the Differential, generating the tests automatically > > will lose information about what actually is intended to be tested, > > Agreed - and I didn't mean to suggest tests should be automatically > generated. I work pretty hard in code reviews to encourage tests to

..., Sanjay Patel wrote: > A rotate intrinsic should be relatively close in cost/complexity to the > existing bswap. > > A grep of intrinsic::bswap says we'd probably add code in: > InstCombine > InstructionSimplify > ConstantFolding > DemandedBits > ValueTracking > VectorUtils > SelectionDAGBuilder > > But I don't think it's fair to view those additions as pure added cost. > As an example, consider that we have to add hacks to EarlyCSE to > recognize multi-IR-instruction min/max/abs patterns. Intrinsics just > work as-is there. So if you sea...

On 5/16/18 1:58 PM, Sanjay Patel via llvm-dev wrote: > An informal metric might be: if the operation is supported as a > primitive op or built-in in source languages and it is supported as a > single target instruction, can we guarantee that 1-to-1 translation > through optimization? It seems perfectly reasonable for LLVM users to expect this to happen reliably. I'd like to

...xagon/HexagonFrameLowering.cpp#L1533 > 4 - > https://github.com/llvm/llvm-project/blob/b725d27350ffdda9e8e9144668ad54c922297f59/llvm/lib/CodeGen/CodeGenPrepare.cpp#L6751 > 5 - > https://github.com/llvm/llvm-project/blob/d0307f93a7658e6d0eef1ffd0b0ed4f1506bfc13/llvm/include/llvm/Analysis/VectorUtils.h#L278 > 6 - > https://github.com/llvm/llvm-project/blob/fafec5155e39f5dad098376c1beb4a56604aa655/llvm/lib/Target/ARM/ARMCallingConv.cpp#L207 > 7 - > https://github.com/llvm/llvm-project/blob/d0307f93a7658e6d0eef1ffd0b0ed4f1506bfc13/llvm/lib/Transforms/Instrumentation/ThreadSanitizer.cp...

On 7/1/20 12:40 AM, Michael Kruse via llvm-dev wrote: > To illustrate some difficulties with FileCheck, lets make a > non-semantic change in LLVM: > >     --- a/llvm/lib/Analysis/VectorUtils.cpp >     +++ b/llvm/lib/Analysis/VectorUtils.cpp >     @@ -642,8 +642,8 @@ MDNode *llvm::uniteAccessGroups(MDNode > *AccGroups1, MDNode *AccGroups2) { >          return AccGroups1; > >        SmallSetVector<Metadata *, 4> Union; >     -  addToAccessGroupList(Union, AccG...

> On Nov 15, 2019, at 1:52 AM, Alex Denisov <1101.debian at gmail.com> wrote: > >> I think I can just get the patch and remove the `llvm` on top of the paths, but that’s not a scalable approach. > > IIRC, the -p option of 'patch' is exactly for doing this. Would that simplify your use-case? > Yes, for a single patch that would work. If there is a way to do

...vely close in > cost/complexity to the existing bswap. > > A grep of intrinsic::bswap says we'd probably add code in: > InstCombine > InstructionSimplify > ConstantFolding > DemandedBits > ValueTracking > VectorUtils > SelectionDAGBuilder > > But I don't think it's fair to view those additions as pure > added cost. As an example, consider that we have to add hacks > to EarlyCSE to recognize multi-IR-instruction min/max/abs > patterns. Intrinsics...

Hello All, I am trying to understand the demanded-bits pass. The example in the source code (below) seems quite explicit. In the example, only 16 bits from a 32 bit variable in IR are demanded and therefore the variable is truncated to a 16 bit variable. %1 = add i32 %x, %y %2 = trunc i32 %1 to i16 However I was wondering if for example, the addition demanded, say 18 bits, would this pass

...ay, July 1, 2020 1:40 AM To: Michael Kruse <llvmdev at meinersbur.de> Cc: llvm-dev <llvm-dev at lists.llvm.org> Subject: Re: [llvm-dev] [RFC] Compiled regression tests. To illustrate some difficulties with FileCheck, lets make a non-semantic change in LLVM: --- a/llvm/lib/Analysis/VectorUtils.cpp +++ b/llvm/lib/Analysis/VectorUtils.cpp @@ -642,8 +642,8 @@ MDNode *llvm::uniteAccessGroups(MDNode *AccGroups1, MDNode *AccGroups2) { return AccGroups1; SmallSetVector<Metadata *, 4> Union; - addToAccessGroupList(Union, AccGroups1); addToAccessGroupLis...

...patch cleans up a lot of the old code which was written specifically to match various forms of the log2 shuffle pattern, and therefore the tests are significantly reduced in size. The target independent changes introduce a new transitional API documented below, by moving some duplicated code into VectorUtils (to generate shuffle patterns for targets which don't opt-in). Some basic legalization is implemented, but only enough to support the cases required by AArch64, there may be some things I've missed that are exposed by other targets. Transitional API ================ To support both the ol...

...asAnalysis.cpp for Release+Asserts build llvm[3]: Compiling AsmPrinter.cpp for Release+Asserts build llvm[3]: Compiling JumpThreading.cpp for Release+Asserts build llvm[2]: Compiling ValueTracking.cpp for Release+Asserts build llvm[3]: Compiling LICM.cpp for Release+Asserts build llvm[2]: Compiling VectorUtils.cpp for Release+Asserts build llvm[3]: Compiling AsmPrinterDwarf.cpp for Release+Asserts build llvm[3]: Compiling X86InstrInfo.cpp for Release+Asserts build llvm[2]: Building Release+Asserts Archive Library libLLVMAnalysis.a llvm[3]: Compiling LoadCombine.cpp for Release+Asserts build make[3]: Ente...