similar to: [LLVMdev] "Anti" scheduling with OoO cores?

Hello, My out-of-tree target features some high latency instructions (let's call them FXLV). When an FXLV issues, it reserves its destination register and execution continues; if a subsequent instruction attempts to read or write that register, the pipline will stall until the FXLV completes. I have attempted to encode this constraint in the machine scheduler (excerpt at bottom of email).

On Dec 16, 2013, at 4:26 PM, Hal Finkel <hfinkel at anl.gov> wrote: >> At the end of each iteration, quality of generated code is estimated >> by >> executing newly introduced target dependent pass. Based on results >> path for >> the following iteration is calculated. At the moment, this has been >> proved >> for MIPS only and it is based on code

On Friday 18 December 2009 13:56, David Greene wrote: > On Friday 18 December 2009 13:53, David Greene wrote: > > > > + void releaseStream() { > > > > + // Delete the stream if needed. Otherwise, transfer the buffer > > > > + // settings from this raw_ostream back to the underlying > > > > stream. + if (!TheStream) > > >

On Wednesday 16 December 2009 13:35, David Greene wrote: > > Please make BufferSize an 'unsigned' and default it to 8192. Please use > > PRESERVE_STREAM instead of 'false'. Here's an updated version of the circular buffer. Ok to check in? -Dave Index: include/llvm/Support/circular_raw_ostream.h

On Friday 18 December 2009 17:27, David Greene wrote: > > Here's the updated patch. > > Well, that didn't go through right. Here it is again. Argh! Stupid bug fixed. :) Index: include/llvm/Support/circular_raw_ostream.h =================================================================== --- include/llvm/Support/circular_raw_ostream.h (revision 0) +++

On Friday 18 December 2009 12:28, Chris Lattner wrote: > This is looking a lot better, here are some more comments: > > + /// current_pos - Return the current position within the stream, > > + /// not counting the bytes currently in the buffer. > > + virtual uint64_t current_pos() { > > I didn't notice this earlier, but is there any reason for current_pos

On Friday 18 December 2009 19:47, Chris Lattner wrote: > On Dec 18, 2009, at 3:46 PM, David Greene wrote: > > + /// circular_raw_ostream - A raw_ostream that saves its output in a > > + /// circular buffer. > > A better description would be "which *can* save its data to a circular > buffer, or can pass it through directly to an underlying stream if > specified

On Friday 18 December 2009 13:53, David Greene wrote: > > > + void releaseStream() { > > > + // Delete the stream if needed. Otherwise, transfer the buffer > > > + // settings from this raw_ostream back to the underlying stream. > > > + if (!TheStream) > > > + return; > > > + if (DeleteStream) > > > +

On Dec 16, 2009, at 8:12 AM, David Greene wrote: >>> >>> What, specifically, do you want reviewed before I start checking in? >>> I'll be happy to prepare a patch that shows just that. >> >> I want a patch that does one thing (e.g. implements dbgs(), > > Ok, let's start with that. Here's the patch to add the circular raw_ostream. This

On Wednesday 16 December 2009 13:19, Chris Lattner wrote: > + int BufferSize; > + std::vector<char> BufferArray; > + bool DelayOutput; > + std::vector<char>::iterator Cur; > > Please doxygenify these. Ok. > Instead of using a std::vector for BufferArray, please just new[] an array > since it is fixed size. Ok. > Why is BufferSize needed with

On Friday 18 December 2009 17:22, David Greene wrote: > On Friday 18 December 2009 13:56, David Greene wrote: > > On Friday 18 December 2009 13:53, David Greene wrote: > > > > > + void releaseStream() { > > > > > + // Delete the stream if needed. Otherwise, transfer the > > > > > buffer + // settings from this raw_ostream back to

On Dec 17, 2009, at 3:41 PM, David Greene wrote: > On Wednesday 16 December 2009 13:35, David Greene wrote: > >>> Please make BufferSize an 'unsigned' and default it to 8192. Please use >>> PRESERVE_STREAM instead of 'false'. > > Here's an updated version of the circular buffer. Ok to check in? This is looking a lot better, here are some more

On Dec 18, 2009, at 3:46 PM, David Greene wrote: > > + /// circular_raw_ostream - A raw_ostream that saves its output in a > + /// circular buffer. A better description would be "which *can* save its data to a circular buffer, or can pass it through directly to an underlying stream if specified with a buffer of zero." When it is buffering, what causes it to flush? Your

Hi, I would like to ask what IssueWidth and NumMicroOps refer to in MachineScheduler, just to be 100% sure what the intent is. Are we modeling the decoder phase or the execution stage? Background: First of all, there seems to be different meanings of "issue" depending on which platform you're on:

Hi, I'm working on defining a SchedMachineModel for the Southern Islands family of GPUs, and I have two questions related to the MachineScheduler. 1. I have a resource that can process 15 instructions at the same time. In the TableGen definitions, should I do: def HWVMEM : ProcResource<15>; or let BufferSize = 15 in { def HWVMEM : ProcResource<1>; } 2. Southern Islands has

On Dec 21, 2009, at 8:05 AM, David Greene wrote: > On Friday 18 December 2009 20:20, David Greene wrote: > >> I wanted something to signify in the output where the debug dump starts. >> That's for that case where some stuff might still get output through >> errs(). I anticipate that while I'll change most uses of errs() to dbgs(), >> we probably don't

On Friday 18 December 2009 20:20, David Greene wrote: > I wanted something to signify in the output where the debug dump starts. > That's for that case where some stuff might still get output through > errs(). I anticipate that while I'll change most uses of errs() to dbgs(), > we probably don't want to change all of them. > > I agree the naming here is poor.

From: Andrew Trick [mailto:atrick at apple.com] Sent: 24 January 2014 21:52 To: Daniel Sanders Cc: LLVM Developers Mailing List (llvmdev at cs.uiuc.edu) Subject: Re: New machine model questions On Jan 24, 2014, at 2:21 AM, Daniel Sanders <Daniel.Sanders at imgtec.com<mailto:Daniel.Sanders at imgtec.com>> wrote: Hi Andrew, I seem to be making good progress on the P5600 scheduler

On Tuesday 15 December 2009 23:02, Chris Lattner wrote: > On Dec 15, 2009, at 6:33 PM, David Greene wrote: > > On Tuesday 15 December 2009 20:11, Chris Lattner wrote: > >> Please send complete and minimal patches to avoid wasting reviewer's > >> time, > > > > What do you want, complete or minimal? I don't want to pre-send 50 > > or so patches

On Jul 22, 2013, at 11:50 AM, Tom Stellard <tom at stellard.net> wrote: > Hi, > > I'm working on defining a SchedMachineModel for the Southern Islands > family of GPUs, and I have two questions related to the > MachineScheduler. > > 1. I have a resource that can process 15 instructions at the same time. > In the TableGen definitions, should I do: > > def