similar to: RFC: Exposing TargetTransformInfo factories from TargetMachine

On Fri, Dec 15, 2017 at 5:30 AM, Hal Finkel <hfinkel at anl.gov> wrote: > Are there reasons why we might not want to do this? Other options we should > consider? It does make the TargetMachine -> TargetIRAnalysis path less abstract, but given that all targets have the same pattern of instantiating a TargetIRAnalysis with a Function->TargetTransformInfo hook, the abstraction does

Dear R users, I have used the following function (in blue) aiming to find the linear regression between MOE and XLA and nesting my data by Species. I have obtained the following results (in green). model4<-lme(MOE~XLA, random = ~ XLA|Species, method="ML")summary(model4) Linear mixed-effects model fit by maximum likelihood Data: NULL         AIC     BIC   logLik  -1.040187 8.78533

Hi, I’m writing this email behalf of my manager in Alibaba. Our team is working on a deep learning engine called MNN. We open sourced it this May. (Github link here: https://github.com/alibaba/MNN ). It has industry-leading blazing fast inference speed on mobile and it is tiny in size. In the internal experiments we ran, MNN outperforms other inference engines under almost all settings by

Some optimizations depend on whether alloca instruction always has non-zero value. Currently, this checking is done by isKnownNonZero() in ValueTracking, and it assumes alloca in address space 0 always has non-zero value but alloca in non-zero address spaces does not always have non-zero value. However, this assumption is incorrect for certain targets. For example, amdgcn---amdgiz target has

Very cool! The first thing that jumps out to me is how tidy and modular the code structure is. The code feels very familiar (stylistically, organizationally, etc.) to me as an LLVM developer. One thing that wasn't at all clear to me is how this is different/similar to TensorFlow XLA (previously mentioned on this list). Can you briefly compare and contrast this with TensorFlow XLA? -- Sean

Hi, We are having connection timeout issues in Excel and Word. Was this an issue that was resolved in post 3.0 versions? We are running 2.2.8a. Below is the samba log file and I have attached a netmon output. Thanks, Domenic [2005/01/24 09:13:48, 0] smbd/oplock.c:oplock_break(797) oplock_break: receive_smb timed out after 30 seconds. oplock_break failed for file

Hal, Thanks for your suggestion. I think that makes sense. Currently, non-zero alloca address space is already represented by data layout, e.g., the last component of the data layout of amdgcn---amdgiz target is -A5, which means alloca is in address space 5. How about adding a letter z to -A5 to indicate alloca may have zero value? i.e. -A5 means alloca is in address space 5 and always has

I know, I said a bad word -- analysis group. But it works pretty much the way I think we want here. We *always* want a TargetTransformInfo, and we have reasonable (conservative) stubs in place. We would just like the option of providing one from the target that has very clever implementations. I would propose that we make TargetTransformInfo be an analysis group, and provide

Hi, I ran into this issue recently and wanted to know if it was a bug or expected behavior. In the R600 backend's TargetTransformInfo implementation, we were setting UnrollingPreferences::Count = UINT_MAX. This was a mistake as we should have been setting UnrollingPreferences::MaxCount instead. However, as a result of setting Count to UINT_MAX, this loop would be unrolled 15 times: if (b

<font face="Verdana,Arial,Helvetica,sans-serif" size="2"><div><div><div style="box-sizing: inherit;"><font face="Verdana, Arial, Helvetica, sans-serif" size="2" style="box-sizing: inherit;">Hi,</font></div><div style="box-sizing: inherit;"><font face="Verdana, Arial,

On 7/16/2013 11:38 PM, Andrew Trick wrote: > Since introducing the new TargetTransformInfo analysis, there has been some confusion over the role of target heuristics in IR passes. A few patches have led to interesting discussions. > > To centralize the discussion, until we get some documentation and better APIs in place, let me throw out an oversimplified Straw Man for a new pass pipline.

Since introducing the new TargetTransformInfo analysis, there has been some confusion over the role of target heuristics in IR passes. A few patches have led to interesting discussions. To centralize the discussion, until we get some documentation and better APIs in place, let me throw out an oversimplified Straw Man for a new pass pipline. It serves two purposes: (1) an overdue reorganization of

Hi, I'm trying to implement the TTI hooks for AMDGPU to avoid unrolling loops for operations with huge expansions (i.e. integer division). The values that are ultimately reported by opt -cost-model -analyze (the actual cost model tests) seem to not matter for this. The huge cost I've assigned division doesn't prevent the loop from being unrolled, because it isn't actually

Hi, I can't seem to get any Excel Add-Ins to work. In particular, I need the 'Solver' for an assignment ... I've tried with two office versions (2007/2003), which I've managed to install successfully (you have to enable the Solver add-in during installation), but it's the same every time. When I try to start it up I get a message saying can't access SOLVER.XLA, and

Hello, I got this problem: I want to mount a directory without providing a login/password [indicaf$]. I thought that "guest ok = yes" would be sufficient. But it not works. What option is the right one to open a share to everyone without a login box in windows ? Thanks. This is my smb.conf (samab v3.0.7): tccwebcom# more smb.conf # Samba config file created using SWAT # from

Hi, I need to add a flush-denormals-to-zero (FTZ) flag to FastMathFlags, but we've already used up the 7 bits available in Value::SubclassOptionalData (the "backing storage" for FPMathOperator::getFastMathFlags()). These are the possibilities I can think of: 1. Increase the size of FPMathOperator. This gives us some additional bits for FTZ and other fastmath flags we'd want

Hello LLVM community, We have been working hard on a new domain specific optimizing compiler, and we are pleased to announce that we have recently open sourced the project! We would like to introduce you to Glow, an optimizing compiler for neural networks! This new compiler is built on the hard work of this community and we would like to thank all of the contributors to the LLVM project. We

On Sun, Mar 17, 2019 at 1:47 PM Craig Topper <craig.topper at gmail.com> wrote: > Can we move HasValueHandle out of the byte used for SubClassOptionalData and move it to the flags at the bottom of value by shrinking NumUserOperands to 27? I like this approach because it is less work for me. :) But I agree with Sanjay below that this only kicks the can slightly further down the road

Hello, OS: Ubuntu 10.04.2 LTS (lucid) wine: wine-1.3.21 (AFAIK, that's the lastest) After following the instruction from http://www.winehq.org/download/ubuntu, I cannot get Office2000 to run... I get the following errors when running Word and Excel, then they hangs repeating the same message. Code: $ wine WINWORD.EXE fixme:x11drv:X11DRV_GetDeviceCaps (0x530): CAPS1 is unimplemented, will

> On Dec 14, 2017, at 20:28, Hal Finkel via llvm-dev <llvm-dev at lists.llvm.org> wrote: > > Would that address your use case? Or can you have null dereferenceable pointers in that address space, just not ones from alloca? I would like to clarify what “null” means exactly. One related thing I would like in the future is for the DataLayout to specify what numeric value is the