search for: bval

...E,discrete=FALSE,names=NULL,sparse=FALSE) lik = profile11a.obj$lik proc = profile11a.obj$proc qwts = rep(1/length(knots),length(knots)) qwts = qwts %*% t(lambda) weights = array(1,dim(data11a)) likmore = make.id() likmore$weights = weights lik = make.SSElik() lik$more = likmore lik$bvals = eval.basis(times,bbasis) procmore = make.fhn() procmore$weights = qwts procmore$qpts = qpts procmore$names = varnames procmore$parnames = parnames proc = make.SSEproc() proc$more = procmore proc$bvals = list(bvals = eval.basis(procmore$qpts,bbasis,0), dbvals = eval.basis(procmo...

...on _X X; inline short2 as_short2(int x) { X result; result.i = x; return result.s; } inline int as_int(short2 x) { X result; result.s = x; return result.i; } void short2_int_swap( short2* b, int* c) { int gidx = get_id(0); short2 bval = b[gidx]; int cval = c[gidx]; b[gidx] = as_short2(cval); c[gidx] = as_int(bval); } -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20090904/cd0940bf/attachment.html>

...AM, Sjoerd Meijer via llvm-dev wrote: what we would like to generate is a vector loop with implicit predication, which works by setting up the the number of elements processed by the loop: hwloop 10 [i:4] = b[i:4] + c[i:4] Why couldn't you use VP intrinsics and scalable types for this? %bval = <4 x vscale x double> call @vp.load(..., /* %evl */ 10) %cval = <4 x vscale x double> call @vp.load(..., /* %evl */ 10) %sum = <4 x vscale x double> fadd %bval, %cval store [..] I see three issues with the llvm.set.loop.elements approach: 1) It is conceptually broken:...

...t>(V)) { ConstantInt *CV = dyn_cast<ConstantInt>(V); ConstantInt *CB = dyn_cast<ConstantInt>(B); ConstantInt *CC = dyn_cast<ConstantInt>(C); if (CV && CB && CC) { const APInt &VVal = CV->getValue(); const APInt &BVal = CB->getValue(); const APInt &CVal = CC->getValue(); ****** But isn't APInt considered to represent unsigned integers ? Is there another way of doing this ? Thanks in advance, Pranav

...quot;,list(rownames(Y),varnames,colnames(Y))) Y2<-as.array(Y,dimnames=dimnames(Y)) assign("Y",Y2) then rerun the analysis with the same result Now traceback() gives: 7: as.array.default(Y) 6: as.array(Y) 5: kronecker(X, Y) 4: kronecker(X, Y) 3: diag(rep(1, nrep)) %x% basisvals$bvals.obs 2: LS.setup(pars, coefs, fn, basisvals, lambda, fd.obj, more, data, weights, times, quadrature, eps = 1e-06, posproc, poslik, discrete, names, sparse) 1: Profile.LS(fhn, data = data2, times = times, pars = pars, coefs = coefs, lambda = lambda, out.meth = "nls", c...

...AM, Sjoerd Meijer via llvm-dev wrote: what we would like to generate is a vector loop with implicit predication, which works by setting up the the number of elements processed by the loop: hwloop 10 [i:4] = b[i:4] + c[i:4] Why couldn't you use VP intrinsics and scalable types for this? %bval = <4 x vscale x double> call @vp.load(..., /* %evl */ 10) %cval = <4 x vscale x double> call @vp.load(..., /* %evl */ 10) %sum = <4 x vscale x double> fadd %bval, %cval store [..] I see three issues with the llvm.set.loop.elements approach: 1) It is conceptually broken:...

Hi Roger, That's a good example, that shows most of the moving parts involved here. In a nutshell, the difference is, and what we would like to make explicit, is the vector trip versus the scalar loop trip count. In your IR example, the loads/stores are predicated on a mask that is calculated from a splat induction variable, which is compared with the vector trip count. Illustrated with your

...literal, i32 0, i32 0 %y141 = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 1 %yVal = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 2 %crVal = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 3 %cbVal = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 4 %rVal = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 5 %gVal = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 6 %bVal = getelementptr inbounds %...

...oundliteral, i32 0, i32 0 %y141 = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 1 %yVal = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 2 %crVal = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 3 %cbVal = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 4 %rVal = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 5 %gVal = getelementptr inbounds %struct.AKEKernelContext* %.compoundliteral, i32 0, i32 6 %bVal = getelementptr inbounds %str...

...AM, Sjoerd Meijer via llvm-dev wrote: what we would like to generate is a vector loop with implicit predication, which works by setting up the the number of elements processed by the loop: hwloop 10 [i:4] = b[i:4] + c[i:4] Why couldn't you use VP intrinsics and scalable types for this? %bval = <4 x vscale x double> call @vp.load(..., /* %evl */ 10) %cval = <4 x vscale x double> call @vp.load(..., /* %evl */ 10) %sum = <4 x vscale x double> fadd %bval, %cval store [..] I see three issues with the llvm.set.loop.elements approach: 1) It is conceptually broken:...

...nstantInt *CV = dyn_cast<ConstantInt>(V); >      ConstantInt *CB = dyn_cast<ConstantInt>(B); >      ConstantInt *CC = dyn_cast<ConstantInt>(C); >      if (CV && CB && CC) { >        const APInt &VVal = CV->getValue(); >        const APInt &BVal = CB->getValue(); >        const APInt &CVal = CC->getValue(); > ****** > > But isn't APInt considered to represent unsigned integers ? Is there another > way of doing this ? APInt's aren't inherently signed or unsigned; see http://llvm.org/doxygen/classllvm_1_...

...AM, Sjoerd Meijer via llvm-dev wrote: what we would like to generate is a vector loop with implicit predication, which works by setting up the the number of elements processed by the loop: hwloop 10 [i:4] = b[i:4] + c[i:4] Why couldn't you use VP intrinsics and scalable types for this? %bval = <4 x vscale x double> call @vp.load(..., /* %evl */ 10) %cval = <4 x vscale x double> call @vp.load(..., /* %evl */ 10) %sum = <4 x vscale x double> fadd %bval, %cval store [..] I see three issues with the llvm.set.loop.elements approach: 1) It is conceptually broken:...