[+CC Matthias, Quentin]
Inter-procedural register allocation can be a big win, but my estimate
is that it will be challenging to complete within one summer unless
you're already familiar with LLVM's register allocator.
I've CC'ed some people who can give you some more detailed information.
-- Sanjoy
> _______________________________________________
> LLVM Developers mailing list
> llvm...@lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
--
Sanjoy Das
http://playingwithpointers.com
_______________________________________________
LLVM Developers mailing list
llvm...@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
From my perspective this project sounds doable. I would expect the register allocation parts to be not too hard: I imagine this being just distilling a new clobber regmask after allocating a function. I would expect the challenging (or annoying) part to get a machine module pass (or a similar mechanism to influence the order in which functions are processed) and a callgraph in the backend. So this might end up being more pass manager / infrastructure work than register allocation.
I'd be happy to answer detail questions or give guidance on the register allocation aspects.
- Matthias
I have a very tiny patch that wrap the backend in a CGSCC pass manager, which will achieve what is needed here I believe: i.e. running CodeGen for every callee before any caller.
I can rebase it if anyone is interested.
--
Mehdi
Hi Vivek,
[+CC Matthias, Quentin]
Inter-procedural register allocation can be a big win, but my estimate
is that it will be challenging to complete within one summer unless
you're already familiar with LLVM's register allocator.
Apologies: didn't notice how old this thread is before replying.
No need to apologize this thread surely deserved some answers :)
From my perspective this project sounds doable. I would expect the register allocation parts to be not too hard: I imagine this being just distilling a new clobber regmask after allocating a function. I would expect the challenging (or annoying) part to get a machine module pass (or a similar mechanism to influence the order in which functions are processed) and a callgraph in the backend. So this might end up being more pass manager / infrastructure work than register allocation.
I'd be happy to answer detail questions or give guidance on the register allocation aspects.
> On Mar 22, 2016, at 6:04 PM, Matthias Braun via llvm-dev <llvm...@lists.llvm.org> wrote:
>
> No need to apologize this thread surely deserved some answers :)
>
> From my perspective this project sounds doable. I would expect the register allocation parts to be not too hard: I imagine this being just distilling a new clobber regmask after allocating a function. I would expect the challenging (or annoying) part to get a machine module pass (or a similar mechanism to influence the order in which functions are processed) and a callgraph in the backend.
I have a very tiny patch that wrap the backend in a CGSCC pass manager, which will achieve what is needed here I believe: i.e. running CodeGen for every callee before any caller.
I can rebase it if anyone is interested.
From the research and code I've seen - Doesn't this break regalloc
down into a global and location allocation strategy? (maybe I'm
remembering incorrectly)
I think the main challenge of a real inter-procedural register allocator is to change all of the calling convention dynamically and more importantly convey the right information to other tools (via CFA, CFI, etc.).
Cheers,
Q.
The pass manager already has support for calligraph connected region IIRC.
As for the regmask part, we probably could hack something up in a week or so, but I believe this is not what Vivek had in mind.
I think the main challenge of a real inter-procedural register allocator is to change all of the calling convention dynamically and more importantly convey the right information to other tools (via CFA, CFI, etc.).
On Mar 23, 2016, at 2:44 PM, vivek pandya <vivekv...@gmail.com> wrote:Vivek PandyaOn Wed, Mar 23, 2016 at 10:18 PM, Quentin Colombet <qcol...@apple.com> wrote:The pass manager already has support for calligraph connected region IIRC.If I am not wrong Quentin and Mehdi Amini refers to CallGraphSCCPass.cpp
As for the regmask part, we probably could hack something up in a week or so, but I believe this is not what Vivek had in mind.
Which operands should be kept in registers between function call should be justifying and for that we can take help from some research work ( some of I mentioned previously I have to read it again. Please suggest some more relevant papers ) once that is implemented we can update the regmask for a call instruction to indicate which registers are free to be used. Am I going in correct direction ?
I think the main challenge of a real inter-procedural register allocator is to change all of the calling convention dynamically and more importantly convey the right information to other tools (via CFA, CFI, etc.).
Here for calling convention do you mean that has to be handle for different kind of backends differently or you are referring some thing I don't know. I don't understand what do you mean by 'convey the right information to other tool' if we have updated regmask for a call instruction then MachineFunction should be able to reflect that fact in MachineFunction pass which is used for intra-procedural register allocation, all we have done is allocated some registers that should live across the function call.
On Mar 23, 2016, at 2:59 PM, Quentin Colombet via llvm-dev <llvm...@lists.llvm.org> wrote:On Mar 23, 2016, at 2:44 PM, vivek pandya <vivekv...@gmail.com> wrote:Vivek PandyaOn Wed, Mar 23, 2016 at 10:18 PM, Quentin Colombet <qcol...@apple.com> wrote:The pass manager already has support for calligraph connected region IIRC.If I am not wrong Quentin and Mehdi Amini refers to CallGraphSCCPass.cppYes.As for the regmask part, we probably could hack something up in a week or so, but I believe this is not what Vivek had in mind.
Which operands should be kept in registers between function call should be justifying and for that we can take help from some research work ( some of I mentioned previously I have to read it again. Please suggest some more relevant papers ) once that is implemented we can update the regmask for a call instruction to indicate which registers are free to be used. Am I going in correct direction ?I do not know if there is a paper on this as this is quite trivial, but IIRC Open64 register allocator does that.Anyhow, the algo is:Given a call graph SCC- Allocate the function with no calls or where each callee has been allocated- Propagate the clobbered registers to the callers of that function by updating the related regmasks on the callsites.Repeat until no more candidate.
On Mar 23, 2016, at 6:38 PM, Gerolf Hoflehner via llvm-dev <llvm...@lists.llvm.org> wrote:Right direction overall. The simplest approach to this is feasible within a summer and should definitely give you good results when you have cases of hot calls with many spill/fills around it that could be eliminated.
On Mar 23, 2016, at 2:59 PM, Quentin Colombet via llvm-dev <llvm...@lists.llvm.org> wrote:On Mar 23, 2016, at 2:44 PM, vivek pandya <vivekv...@gmail.com> wrote:Vivek PandyaOn Wed, Mar 23, 2016 at 10:18 PM, Quentin Colombet <qcol...@apple.com> wrote:The pass manager already has support for calligraph connected region IIRC.If I am not wrong Quentin and Mehdi Amini refers to CallGraphSCCPass.cppYes.As for the regmask part, we probably could hack something up in a week or so, but I believe this is not what Vivek had in mind.Which operands should be kept in registers between function call should be justifying and for that we can take help from some research work ( some of I mentioned previously I have to read it again. Please suggest some more relevant papers ) once that is implemented we can update the regmask for a call instruction to indicate which registers are free to be used. Am I going in correct direction ?I do not know if there is a paper on this as this is quite trivial, but IIRC Open64 register allocator does that.Anyhow, the algo is:Given a call graph SCC- Allocate the function with no calls or where each callee has been allocated- Propagate the clobbered registers to the callers of that function by updating the related regmasks on the callsites.Repeat until no more candidate.One does not necessarily need the call graph. The compiler can do this as an opportunistic optimization. The callee collects a resource mask and the caller consumes it when it is “there”. Within a module when the callee”leaf” is compiled before the caller the information is “there”. When the call graph is available you want a bottom up walk for this optimization.A few things to keep an eye on:- The twist here could be that the bottom up order conflicts with the layout order, so the two optimizations would have to run independently. ( I have not looked into the layout algorithm so this might not be an actual issue here).
foo:r0 = 1000…retbar:…call foovreg1 = vreg2 + 1000
int A::foo() {return this->value;}
foo:r0 = load r0, #offset_of_valueret
On Mar 23, 2016, at 2:44 PM, vivek pandya <vivekv...@gmail.com> wrote:Vivek PandyaOn Wed, Mar 23, 2016 at 10:18 PM, Quentin Colombet <qcol...@apple.com> wrote:The pass manager already has support for calligraph connected region IIRC.If I am not wrong Quentin and Mehdi Amini refers to CallGraphSCCPass.cppYes.As for the regmask part, we probably could hack something up in a week or so, but I believe this is not what Vivek had in mind.
Which operands should be kept in registers between function call should be justifying and for that we can take help from some research work ( some of I mentioned previously I have to read it again. Please suggest some more relevant papers ) once that is implemented we can update the regmask for a call instruction to indicate which registers are free to be used. Am I going in correct direction ?I do not know if there is a paper on this as this is quite trivial, but IIRC Open64 register allocator does that.Anyhow, the algo is:Given a call graph SCC- Allocate the function with no calls or where each callee has been allocated- Propagate the clobbered registers to the callers of that function by updating the related regmasks on the callsites.Repeat until no more candidate.Allocate remaining functions “normally”.I think the main challenge of a real inter-procedural register allocator is to change all of the calling convention dynamically and more importantly convey the right information to other tools (via CFA, CFI, etc.).
Here for calling convention do you mean that has to be handle for different kind of backends differently or you are referring some thing I don't know. I don't understand what do you mean by 'convey the right information to other tool' if we have updated regmask for a call instruction then MachineFunction should be able to reflect that fact in MachineFunction pass which is used for intra-procedural register allocation, all we have done is allocated some registers that should live across the function call.My mistake, I though you had in mind what I call a “true” inter procedural registers allocator: one that changes the allocation at function boundaries as well. I.e., it may choose that it is more efficient to put the first argument of function foo is register FP0 even if the ABI says R0.With this kind of scheme, you break the ABI (and you need LTO to be allowed to do that), you need to “dynamically” adjust the calling convention to what the register allocator chooses, and moreover you need to be able to communicate to the other tools (dynamic linker, debugger, etc.) where are the things that are usually defined by the ABI, like the frame pointer, the return value, etc.
On Mar 24, 2016, at 11:09 AM, Pete Cooper <peter_...@apple.com> wrote:Layout is just the order functions reach the AsmPrinter, so you’re right that this is going to make the function output different. If we care about the order, which we may do, then we’d need to cache the data in the AsmPrinter and reorder it there somehow.
On Mar 23, 2016, at 6:38 PM, Gerolf Hoflehner via llvm-dev <llvm...@lists.llvm.org> wrote:Right direction overall. The simplest approach to this is feasible within a summer and should definitely give you good results when you have cases of hot calls with many spill/fills around it that could be eliminated.
On Mar 23, 2016, at 2:59 PM, Quentin Colombet via llvm-dev <llvm...@lists.llvm.org> wrote:On Mar 23, 2016, at 2:44 PM, vivek pandya <vivekv...@gmail.com> wrote:Vivek PandyaOn Wed, Mar 23, 2016 at 10:18 PM, Quentin Colombet <qcol...@apple.com> wrote:The pass manager already has support for calligraph connected region IIRC.If I am not wrong Quentin and Mehdi Amini refers to CallGraphSCCPass.cppYes.As for the regmask part, we probably could hack something up in a week or so, but I believe this is not what Vivek had in mind.Which operands should be kept in registers between function call should be justifying and for that we can take help from some research work ( some of I mentioned previously I have to read it again. Please suggest some more relevant papers ) once that is implemented we can update the regmask for a call instruction to indicate which registers are free to be used. Am I going in correct direction ?I do not know if there is a paper on this as this is quite trivial, but IIRC Open64 register allocator does that.Anyhow, the algo is:Given a call graph SCC- Allocate the function with no calls or where each callee has been allocated- Propagate the clobbered registers to the callers of that function by updating the related regmasks on the callsites.Repeat until no more candidate.One does not necessarily need the call graph. The compiler can do this as an opportunistic optimization. The callee collects a resource mask and the caller consumes it when it is “there”. Within a module when the callee”leaf” is compiled before the caller the information is “there”. When the call graph is available you want a bottom up walk for this optimization.A few things to keep an eye on:- The twist here could be that the bottom up order conflicts with the layout order, so the two optimizations would have to run independently. ( I have not looked into the layout algorithm so this might not be an actual issue here).
Some bonus features that come from codegen on the calligraphy, and specifically having accurate regmasks and similar information:- The X86 VZeroUpper pass should insert fewer VZeroUpper instructions before calls, and could possibly even learn that after the call the state of vzeroupper is known.- Values in registers can be used by the callee instead of loading them.The second one here is fun. Imagine this pseudo code:foo:r0 = 1000…retbar:…call foovreg1 = vreg2 + 1000You know which registers contain which values after the call to foo. In this case you know that the value of 1000 is available in a register already so you can avoid loading it for use in the add. You could have other values in registers too, even those which are passed in to foo. The ‘this’ pointer is the best example as its probably incredibly likely that r0 contains the this pointer after a function call which didn’t override r0 for the return.
One more, just for fun: Inter-procedural stack allocation. That is of calls bar, bar needs 4 bytes of stack space. Instead of bar allocating 4 bytes, it adds an attribute to itself, then foo allocates 4 bytes of space at the bottom of the stack for bar to use.The this pointer example is actually related to what Quentin mentioned as a future direction here: rewriting calling conventions. If you haveint A::foo() {return this->value;}then you are going to have code something likefoo:r0 = load r0, #offset_of_valueretIf the this pointer is live after the call, and it almost certainly is, then it would be better to rewrite this call to avoid clobbering r0. That is, return the this pointer in r0 and the value in r1. That could actually be done as an IR level pass too though if its deemed profitable.Anyway, didn’t mean to distract from the immediate goals of this project. I’m excited to see the SCC code make it in tree and see what else it enables.
Some bonus features that come from codegen on the calligraphy, and specifically having accurate regmasks and similar information:- The X86 VZeroUpper pass should insert fewer VZeroUpper instructions before calls, and could possibly even learn that after the call the state of vzeroupper is known.- Values in registers can be used by the callee instead of loading them.The second one here is fun. Imagine this pseudo code:foo:r0 = 1000…retbar:…call foovreg1 = vreg2 + 1000You know which registers contain which values after the call to foo. In this case you know that the value of 1000 is available in a register already so you can avoid loading it for use in the add. You could have other values in registers too, even those which are passed in to foo. The ‘this’ pointer is the best example as its probably incredibly likely that r0 contains the this pointer after a function call which didn’t override r0 for the return.The above mentioned case is interesting and useful, perhaps and simple analysis pass which can return a map from value to register will help.
One more, just for fun: Inter-procedural stack allocation. That is of calls bar, bar needs 4 bytes of stack space. Instead of bar allocating 4 bytes, it adds an attribute to itself, then foo allocates 4 bytes of space at the bottom of the stack for bar to use.The this pointer example is actually related to what Quentin mentioned as a future direction here: rewriting calling conventions. If you haveint A::foo() {return this->value;}then you are going to have code something likefoo:r0 = load r0, #offset_of_valueretIf the this pointer is live after the call, and it almost certainly is, then it would be better to rewrite this call to avoid clobbering r0. That is, return the this pointer in r0 and the value in r1. That could actually be done as an IR level pass too though if its deemed profitable.Anyway, didn’t mean to distract from the immediate goals of this project. I’m excited to see the SCC code make it in tree and see what else it enables.Can you please provide some links to understand benefits of IP stack allocation ?
I have also write the draft proposal, I will share it through the summer of code site.Here is the link https://docs.google.com/document/d/1DrsaFJdtxV73Zpns2bEgjATLFcWuaYMPHuvt5THLeLk/edit?usp=sharingThis is not much effective but still I would like to give it a try. Please review it quickly I have 23 hours to submit the final PDF.
On Mar 23, 2016, at 6:38 PM, Gerolf Hoflehner via llvm-dev <llvm...@lists.llvm.org> wrote:Right direction overall. The simplest approach to this is feasible within a summer and should definitely give you good results when you have cases of hot calls with many spill/fills around it that could be eliminated.
On Mar 23, 2016, at 2:59 PM, Quentin Colombet via llvm-dev <llvm...@lists.llvm.org> wrote:On Mar 23, 2016, at 2:44 PM, vivek pandya <vivekv...@gmail.com> wrote:Vivek PandyaOn Wed, Mar 23, 2016 at 10:18 PM, Quentin Colombet <qcol...@apple.com> wrote:The pass manager already has support for calligraph connected region IIRC.If I am not wrong Quentin and Mehdi Amini refers to CallGraphSCCPass.cppYes.As for the regmask part, we probably could hack something up in a week or so, but I believe this is not what Vivek had in mind.Which operands should be kept in registers between function call should be justifying and for that we can take help from some research work ( some of I mentioned previously I have to read it again. Please suggest some more relevant papers ) once that is implemented we can update the regmask for a call instruction to indicate which registers are free to be used. Am I going in correct direction ?I do not know if there is a paper on this as this is quite trivial, but IIRC Open64 register allocator does that.Anyhow, the algo is:Given a call graph SCC- Allocate the function with no calls or where each callee has been allocated- Propagate the clobbered registers to the callers of that function by updating the related regmasks on the callsites.Repeat until no more candidate.One does not necessarily need the call graph. The compiler can do this as an opportunistic optimization. The callee collects a resource mask and the caller consumes it when it is “there”. Within a module when the callee”leaf” is compiled before the caller the information is “there”. When the call graph is available you want a bottom up walk for this optimization.A few things to keep an eye on:- The twist here could be that the bottom up order conflicts with the layout order, so the two optimizations would have to run independently. ( I have not looked into the layout algorithm so this might not be an actual issue here).
- You also need to consider the supported preemption model. When a function can be preempted dynamically the statically collected information for a callee cannot be used and the optimization may not kick in.
- Most of the work I would expect to be tuning the assignment heuristics in the allocator (a live range that spans two calls sites, should it go into a scratch register that is not used in one call but in the other? How could profile change that? etc). But again, perhaps the cheapest approach is not to go into the heuristics and only remove a scratch register fill/spill around a call sit when that register is not destroyed anywhere down in the call tree.
On Mar 24, 2016, at 1:55 PM, Mehdi Amini <mehdi...@apple.com> wrote:One more, just for fun: Inter-procedural stack allocation. That is of calls bar, bar needs 4 bytes of stack space. Instead of bar allocating 4 bytes, it adds an attribute to itself, then foo allocates 4 bytes of space at the bottom of the stack for bar to use.This is something that can be performed with a module pass at the IR level right? The codegen would just have to be teached to recognize the attribute.
--Mehdi
On Mar 23, 2016, at 2:59 PM, Quentin Colombet via llvm-dev <llvm...@lists.llvm.org> wrote:
On Mar 23, 2016, at 2:44 PM, vivek pandya <vivekv...@gmail.com> wrote:Vivek PandyaOn Wed, Mar 23, 2016 at 10:18 PM, Quentin Colombet <qcol...@apple.com> wrote:The pass manager already has support for calligraph connected region IIRC.If I am not wrong Quentin and Mehdi Amini refers to CallGraphSCCPass.cppYes.As for the regmask part, we probably could hack something up in a week or so, but I believe this is not what Vivek had in mind.
Which operands should be kept in registers between function call should be justifying and for that we can take help from some research work ( some of I mentioned previously I have to read it again. Please suggest some more relevant papers ) once that is implemented we can update the regmask for a call instruction to indicate which registers are free to be used. Am I going in correct direction ?I do not know if there is a paper on this as this is quite trivial, but IIRC Open64 register allocator does that.Anyhow, the algo is:Given a call graph SCC- Allocate the function with no calls or where each callee has been allocated- Propagate the clobbered registers to the callers of that function by updating the related regmasks on the callsites.Repeat until no more candidate.
On Mar 24, 2016, at 1:59 PM, Pete Cooper <peter_...@apple.com> wrote:I thought you would need to run codegen to get a specific number of bytes to allocate. You would compile bar, note down how many bytes of stack it would have required, then add that as an attribute. The IR level could only make a good guess as to how many bytes we need.On Mar 24, 2016, at 1:55 PM, Mehdi Amini <mehdi...@apple.com> wrote:One more, just for fun: Inter-procedural stack allocation. That is of calls bar, bar needs 4 bytes of stack space. Instead of bar allocating 4 bytes, it adds an attribute to itself, then foo allocates 4 bytes of space at the bottom of the stack for bar to use.This is something that can be performed with a module pass at the IR level right? The codegen would just have to be teached to recognize the attribute.
could I ask you if you could share your patch?
We are testing whether at least some leaf call register allocation optimization could help and you patch or some pointers on what to do will be very useful.
Thank you
Adam
On Wed, 23 Mar 2016 05:04:41 +0100, Mehdi Amini via llvm-dev <llvm...@lists.llvm.org> wrote:
>
>> On Mar 22, 2016, at 6:04 PM, Matthias Braun via llvm-dev <llvm...@lists.llvm.org> wrote:
>>
>> No need to apologize this thread surely deserved some answers :)
>>
>> From my perspective this project sounds doable. I would expect the register allocation parts to be not too hard: I imagine this being just distilling a new clobber regmask after allocating a function. I would expect the challenging (or annoying) part to get a machine module pass (or a similar mechanism to influence the order in which functions are processed) and a callgraph in the backend.
>
> I have a very tiny patch that wrap the backend in a CGSCC pass manager, which will achieve what is needed here I believe: i.e. running CodeGen for every callee before any caller.
> I can rebase it if anyone is interested.