similar to: [LLVMdev] question on constant folding

Is there a pass I can use after reg2mem to get rid of occurances like this?: store i32 %cond, i32* %cond.reg2mem %cond.reload = load i32* %cond.reg2mem store i32 %cond.reload, i32* %cond4.reg2mem Essentially, in this case, reg2mem creates an extra memory space to store and load a value from just here, and never uses the value again; since this isn't efficient code, I'm wondering if

Hello, I have two for loops (one inside the other), that after indvars, looprotate, etc. (the important here is the loop rotate), is similar to this (I've stripped the real operations): define i32 @f() nounwind { entry: br label %bb1 bb1: ; preds = %bb3, %bb1, %entry %i.0.reg2mem.0.ph = phi i32 [ 0, %entry ], [ %i.0.reg2mem.0.ph, %bb1 ], [ %indvar.next9, %bb3 ] ;

Hi, when I delete some instruction, I got some error prompt message. - %i.0.reg2mem.0 = phi i32 [ 0, %bb5 ], [ %indvar.next, %bb12 ] ; <i32> [#uses=2] - %s.0.reg2mem.0 = phi i32 [ 0, %bb5 ], [ %tmp16, %bb12 ] ; <i32> [#uses=1] - %tmp14 = tail call i32 @foobar(i32 %i.0.reg2mem.0) nounwind ; <i32> [#uses=1] - %tmp16 = add i32 %tmp14, %s.0.reg2mem.0 ; <i32>

On 08/03/2011 08:35 AM, Jimborean Alexandra wrote: > Hello Tobi, > > You are right, we need to run some other passes before running the > scalar evolution pass. The sequence that I run for this example is -O3 > -loop-simplify -reg2mem. This is why I did not obtain the expressions > depending on the loop indices. So I removed the reg2mem pass and scalar > evolution computes the

Hello Tobi, You are right, we need to run some other passes before running the scalar evolution pass. The sequence that I run for this example is -O3 -loop-simplify -reg2mem.  This is why I did not obtain the expressions depending on the loop indices. So I removed the reg2mem pass and scalar evolution computes the correct functions. However, I need to run the reg2mem pass (or any other that

Hi, I work on a pass that requires as input LLVM code without any phi nodes. For this, I use the reg2mem pass which produces suitable code, the pass runs correctly, but I obtain a significant performance decrease. I expect that this is because there are more reads / writes to memory after running the reg2mem pass. How can I optimize the code, without inserting any phi nodes? Or is there a

I am currently building a BasicBlock pass which requires to run -reg2mem before it, and need to run -mem2reg after it to clean up. So, I want to specify -reg2mem as one of the pre-requisite passes to it, as: class MyPass: public BasicBlockPass{ virtual void getAnalysisUsage(AnalysisUsage &AU){ ... AU.addRequired<RegToMem>(); ... } }; I searched all passes under

Codeprepare and independent blocks are introducing these loads and stores. These are prepasses that polly runs prior to building the dependence graph to transform scalar dependences into data dependences. Ether was working on eliminating the rewrite of scalar dependences. On Thu, Aug 15, 2013 at 5:32 AM, Star Tan <tanmx_star at yeah.net> wrote: > Hi all, > > I have investigated the

Somewhere during optimization PHI nodes on pointer types (including alloca instructions) are being introduced, and they persist through the scalar replacement of aggregates pass and others. I can't seem to find a combination of passes or transformations to get rid of them. Has anyone had this problem before, and know a transformation to eliminate it? My optimization passes rely on

Hi Sebpop, Thanks for your explanation. I noticed that Polly would finally run the SROA pass to transform these load/store instructions into scalar operations. Is it possible to run such a pass before polly-dependence analysis? Star Tan At 2013-08-15 21:12:53,"Sebastian Pop" <sebpop at gmail.com> wrote: >Codeprepare and independent blocks are introducing these loads and

I do not think that running SROA before polly is a good idea: it would defeat the purpose of the code preparation passes that polly intentionally schedules for the data dependence analysis. If you remove the data references before polly runs, you would miss them in the dependence graph: that could lead to incorrect transforms. On Thu, Aug 15, 2013 at 7:28 PM, Star Tan <tanmx_star at

So this change did indeed have an effect! :) I’m seeing regressions in a number of benchmarks mainly due to a host of extra bitcasts that get introduced. Here’s the problem I’m seeing in a nutshell: 1) There is a Phi with input type double 2) Polly demotes the phi into a load/store of type double 3) InstCombine canonicalizes the load/store to use i64 instead of double 4)

Hi, Seems pass scalar-evolution+indvars fail to get the loop trip count of the following case: int foo(int x, int y, int lam[256], int alp[256]) { int i; int z = y; for (i = 255; i >= 0; i--) { z += x; lam[i] = alp[i]; } return z; } The final optimized ll code is : define i32 @foo(i32 %x, i32 %y, i32* %lam, i32* %alp) nounwind { entry: br label %bb bb:

I guess AU.addRequiredID(DemoteRegisterToMemoryID) would do the trick. - xi On Jun 19, 2011, at 11:03 PM, Chuck Zhao wrote: > I am currently building a BasicBlock pass which requires to run -reg2mem before it, and need to run -mem2reg after it to clean up. > > So, I want to specify -reg2mem as one of the pre-requisite passes to it, as: > > > class MyPass: public

On 31.8.2011, at 11.41, Eli Friedman wrote: > Mmm... reg2mem will transform IR with PHI's into IR without them, but > it generates a bunch of alloca's, which I would assume are not cheap > to lower to VHDL. You might have to write your own pass to get the > precise transformation you're looking for. Right. Thanks. I need to see the reg2mem source code. Teemu

Hello, guys. I just tested -reg2mem pass to see how it changes my bitcode. E.g., for the following simple C code: ------------------------------------------------------------- int foo() { int i,j; int sum = 0; for (i=0; i<10; i++) { sum += i; for (j=0; j<3; j++) sum += 2; } return sum; } ------------------------------------------------------------- I could get the

On Wed, Aug 31, 2011 at 2:35 AM, Teemu Rinta-aho <teemu.rinta-aho at nomadiclab.com> wrote: > On 31.8.2011, at 11.41, Eli Friedman wrote: > >> Mmm... reg2mem will transform IR with PHI's into IR without them, but >> it generates a bunch of alloca's, which I would assume are not cheap >> to lower to VHDL.  You might have to write your own pass to get the

On 10/20/10 6:05 AM, Luke Dalessandro wrote: > I have a transformation where I'd like to use both DominatorTree (for ExtractCodeRegion), and DemoteRegisterToMemory (i.e., reg2mem). The transformation is phased, so all occurrences of getAnalysis<DominatorTree>(Function) happen before any occurrence of getAnalysisID<FunctionPass>(&DemoteRegisterToMemoryID, Function). I

Thanks for the reply Pete. Unfortunately, I don’t think it is going to be as simple as ignoring those loads which only store. In findCommonType(), only one alloca is passed in at a time. So, while you could find those cases where that alloca was loaded from and stored elsewhere, you can’t find those places that store to that alloca from somewhere else (at least not easily that I can see). So

When using the makefiles in LLVM, the commands that Make executes are hidden from the user. Instead something like: llvm[3]: Compiling Reg2Mem.cpp for Debug build is printed to the terminal, rather than the commands that are used to build Reg2Mem.cpp. Is there a way to get the makefiles to print the commands that they are executing? Regards, Ryan