search for: tmp7

...%entry.header.loop.end ], [ 0, %entry ] > %global_id_0 = add i32 %7, %6 > %8 = sext i32 %global_id_0 to i64 > %arrayidx.i = getelementptr float addrspace(1)* %1, i64 %8 > %tmp3.i = load float addrspace(1)* %arrayidx.i, align 4 > %tmp5.i = fmul float %tmp3.i, 1.000000e+01 > %tmp7.i = fsub float 1.000000e+00, %tmp3.i > %tmp8.i = fmul float %tmp7.i, 1.000000e+02 > %tmp9.i = fadd float %tmp5.i, %tmp8.i > %tmp20.i = fmul float %tmp7.i, 1.000000e+01 > %tmp21.i = fadd float %tmp3.i, %tmp20.i > %tmp23.i = fmul float %tmp3.i, 0x3F847AE140000000 > %tmp26.i =...

Hello everybody, I have stumbled upon a test case (the attached module is a slightly reduced version) that shows extremely reduced performance on linux compared to windows when executed using LLVM's JIT. We narrowed the problem down to the actual code being generated, the source IR on both systems is the same. Try compiling the attached module: llc -O3 -filetype=asm -o BAD.s BAD.ll Under

...out shuffle simplification. This is a testcase from test/Transforms/InstCombine/vec_shuffle.ll: define <4 x i32> @test10(<4 x i32> %tmp5) nounwind { %tmp6 = shufflevector <4 x i32> %tmp5, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %tmp7 = shufflevector <4 x i32> %tmp6, <4 x i32> undef, <4 x i32> zeroinitializer ret <4 x i32> %tmp7 } opt –instcombine will combine to: define <4 x i32> @test10(<4 x i32> %tmp5) nounwind { %tmp7 = shufflevector <4 x i32> %tmp5, <4 x i32> undef, &lt...

...=c(timelineMSEC,X,Y)) userSubset x<-as.numeric(userSubset$X) y<-as.numeric(userSubset$Y) scatter3d(x,y,userSubset$timeline,xlim = c(0,1280), ylim = c(0,1024), zlim=c(0,1800000),type="h",main=sessionU01[i,],sub=sessionU01[i,]) tmp6=rep(".ps") tmp7=paste(sessionU01[i,],tmp6,sep="") tmp7 rgl.postscript(tmp7,"ps",drawText=FALSE) #pdf(file=tmp7) #dev.print(file=tmp7, device=pdf, width=600) #dev.off(2) }

...]* %x, i32 0, i32 1 ; <i32**> [#uses=1] %tmp23 = ptrtoint i32** %tmp1 to i32 ; <i32> [#uses=1] %x45 = ptrtoint [2 x i32*]* %x to i32 ; <i32> [#uses=1] %tmp6 = sub i32 %tmp23, %x45 ; <i32> [#uses=1] %tmp7 = ashr i32 %tmp6, 2 ; <i32> [#uses=1] ret i32 %tmp7 } The return value is 1. The ashr exposes the pointer size by shifting the 4 byte distance over by 2. For the analysis that I am doing, it would be nice to have an instruction that explicitly performs this distance c...

...ch=20,col='blue',cex=0.7,xlim=xl,ylim=yl,...) panel.abline(v=0, col='gray30',lty=2,lwd=1.5,...) panel.loess(x,y, span=2/3,family='gaussian',normalize=T,col='red',lwd=1.5,...) # panel.lmline(x,y, col='red',lwd=1.5,...) } nms <- names(tmp7)[c(3:7,9)] for (i in nms) { tmp1 <- xyplot(i~pk, data=tmp7, # different y's in the loop; x is same for all as.table=T,panel=myPanel,main=paste('PK vs',i,'- overall')) pdf(paste("pkvs",i,"a.pdf",sep=""),w=10,h=8) print(tmp1) dev...

...[2 x i32*]* %x to i32 ; <i32> [#uses=1] %tmp6 = sub i32 %tmp23, %x45 ; <i32> [#uses=1] %size = getelementptr i32** null, i32 1 ; <i32**> [#uses=1] %sizeI = ptrtoint i32** %size to i32 ; <i32> [#uses=1] %tmp7 = ashr i32 %tmp6, %sizeI ; <i32> [#uses=1] ret i32 %tmp7 } There, pointer size independent. The problem you see is you are using a frontend targeting a specific platform, so pointersize is known (see the target datalayout line). Andrew

...2] %tmp3 = and i32 %tmp, -65536 ; <i32> [#uses=2] %tmp4 = icmp ne i32 %tmp3, 168296448 ; <i1> [#uses=1] %tmp5 = and i1 %tmp2, %tmp4 ; <i1> [#uses=1] %tmp6 = and i32 %tmp, -256 ; <i32> [#uses=2] %tmp7 = icmp eq i32 %tmp6, 168296704 ; <i1> [#uses=1] %tmp8 = icmp eq i32 %tmp3, 168296448 ; <i1> [#uses=2] %tmp9 = icmp ne i32 %tmp6, 168296704 ; <i1> [#uses=1] %tmp10 = and i1 %tmp8, %tmp9 ; <i1> [#uses=1] %tmp11 = load...

...declare i32 @printf(i8*, ...) define i32 @main (){ entry: %n = alloca i32 %f = alloca i32 %i = alloca i32 %zero = alloca i32 %inc = alloca i32 store i32 5, i32* %n store i32 1, i32* %f store i32 1, i32* %i store i32 0 , i32* %zero store i32 1 , i32* %inc %tmp7 = load i32* %n %tmp8 = load i32* %zero %tmp6 = icmp sgt i32 %tmp7, %tmp8 br i1 %tmp6, label %then4, label %else12 then4: %tmp14 = load i32* %i %tmp16 = load i32* %n %tmp18 = load i32* %inc %tmp15 = add i32 %tmp16, %tmp18 %tmp13 = icmp slt i32 %tmp14, %tmp15 br i1 %t...

...ull ; <i1> [#uses=1] br i1 %tmp2, label %UnifiedReturnBlock, label %bb bb: ; preds = %entry %tmp5 = tail call void (i8*)* (void (i8*)*)* asm "", "=r,0"( void (i8*)* @_Jv_RegisterClasses ) ; <void (i8*)*> [#uses=2] %tmp7 = icmp eq void (i8*)* %tmp5, null ; <i1> [#uses=1] br i1 %tmp7, label %UnifiedReturnBlock, label %bb10 bb10: ; preds = %bb tail call void %tmp5( i8* bitcast ([0 x i8*]* @__JCR_LIST__ to i8*) ) ret void UnifiedReturnBlock: ; preds = %bb,...

...tion Attrs: nounwind readonly define i32 @get32Bits(i8* inreg nocapture readonly %x_arg) #0 { %tmp1 = getelementptr inbounds i8* %x_arg, i64 3 %tmp2 = load i8* %tmp1, align 1 %tmp3 = zext i8 %tmp2 to i32 %tmp4 = shl nuw nsw i32 %tmp3, 24 %tmp6 = getelementptr inbounds i8* %x_arg, i64 2 %tmp7 = load i8* %tmp6, align 1 %tmp8 = zext i8 %tmp7 to i32 %tmp9 = shl nuw nsw i32 %tmp8, 16 %tmp10 = or i32 %tmp9, %tmp4 %tmp12 = getelementptr inbounds i8* %x_arg, i64 1 %tmp13 = load i8* %tmp12, align 1 %tmp14 = zext i8 %tmp13 to i32 %tmp15 = shl nuw nsw i32 %tmp14, 8 %tmp16 = or i32...

...ts as: %str1 = alloca [8 x i8], align 16 ; <[8 x i8]*> [#uses=9] %tmp1 = getelementptr [8 x i8]* %str1, i32 0, i32 0 ; <i8*> [#uses=2] store i8 109, i8* %tmp1, align 16 %tmp4 = getelementptr [8 x i8]* %str1, i32 0, i32 1 ; <i8*> [#uses=1] store i8 111, i8* %tmp4, align 1 %tmp7 = getelementptr [8 x i8]* %str1, i32 0, i32 2 ; <i8*> [#uses=1] store i8 116, i8* %tmp7, align 1 %tmp10 = getelementptr [8 x i8]* %str1, i32 0, i32 3 ; <i8*> [#uses=1] store i8 104, i8* %tmp10, align 1 %tmp13 = getelementptr [8 x i8]* %str1, i32 0, i32 4 ; <i8*> [#uses=1]...

As Sanjay noted in D31426<https://reviews.llvm.org/D31426#712701>, InstructionSimplify is missing the following simplification: This function: define <4 x i32> @splat_operand(<4 x i32> %x) { %splat = shufflevector <4 x i32> %x, <4 x i32> undef, <4 x i32> zeroinitializer %shuf = shufflevector <4 x i32> %splat, <4 x i32> undef, <4 x i32>

...- dbConnect(MySQL(), user='root',dbname='test',host='localhost',password='') listUser <- dbSendQuery(mycon, "select distinct userID from ulogdata") userList <- fetch(listUser, n= -1) par(mfrow=c(5,3)) tmp6=rep("UserScatterActionRegion.pdf") tmp7=paste(tmp6,sep="") tmp7 #pdf(tmp7) jpeg(file="UserScatterActionRegion.jpeg") for (i in 1:20){ x<-subset(user01 ,userID == userList[i,],select=c(X,Y)) x bin<-hexbin(x$X, x$Y, xbins=100) part1=rep("Action Regions") nameFull=paste(userList[...

...I422"))) float2 shuffle(float2, uint2); In LLVM 2.9 and LLVM 3.0, our front-end generates: @__shuffle_2f32_2u32 = alias weak <2 x i32> (<2 x i32>, <2 x i32>)* @4 And the calls, before linking, look like: %call9 = call <2 x float> @__shuffle_2f32_2u32(<2 x float> %tmp7, <2 x i32> %tmp8) nounwind After linking with LLVM 3.0, the call looks like: %call9 = call <2 x float> bitcast (<2 x i32> (<2 x i32>, <2 x i32>)* @__shuffle_2f32_2u32 to <2 x float> (<2 x float>, <2 x i32>)*)(<2 x float> %tmp7, <2 x i32> %...

...%f = alloca i32 > > %i = alloca i32 > > %zero = alloca i32 > > %inc = alloca i32 > > store i32 5, i32* %n > > store i32 1, i32* %f > > store i32 1, i32* %i > > store i32 0 , i32* %zero > > store i32 1 , i32* %inc > > > > %tmp7 = load i32* %n > > %tmp8 = load i32* %zero > > %tmp6 = icmp sgt i32 %tmp7, %tmp8 > > br i1 %tmp6, label %then4, label %else12 > > > > then4: > > %tmp14 = load i32* %i > > %tmp16 = load i32* %n > > %tmp18 = load i32* %inc > > %tmp15...

...created for different basic block are independent from each other, for example in the following IR the temp variables used in the basic block bb: i.e. %tmp, %tmp1, %tmp2 are only dependent on the variables defined in the entry: basic block, and same holds for basic block bb5: i.e. variables %tmp6, %tmp7, %tmp8 etc are only dependent on the variables in the entry block (%limit_addr, %i.. etc) and the temp variables defined in the same basic block. Is this true ?, as far as i have seen, by generating the bitcode for various test codes, and it holds true for all. So, now the proposition is, that one...

...32 %x, i32 %y, i32* %lam, i32* %alp) nounwind { entry: br label %bb bb: ; preds = %bb, %entry %indvar = phi i32 [ 0, %entry ], [ %indvar.next, %bb ] ; <i32> [#uses=4] %i.0.reg2mem.0 = sub i32 255, %indvar ; <i32> [#uses=2] %tmp7 = getelementptr i32* %alp, i32 %i.0.reg2mem.0 ; <i32*> [#uses=1] %tmp8 = load i32* %tmp7, align 4 ; <i32> [#uses=1] %tmp11 = getelementptr i32* %lam, i32 %i.0.reg2mem.0 ; <i32*> [#uses=1] store i32 %tmp8, i32* %tmp11, a...

...} return ret; } generates: define i32 @loop(i32 %x) nounwind readnone { %1 = icmp sgt i32 %x, 0 br i1 %1, label %bb.nph, label %3 bb.nph: ; preds = %0 %tmp4 = add i32 %x, -1 %tmp6 = add i32 %x, -2 %tmp16 = add i32 %x, -3 %tmp7 = zext i32 %tmp6 to i33 %tmp5 = zext i32 %tmp4 to i33 %tmp17 = zext i32 %tmp16 to i33 %tmp15 = mul i33 %tmp5, %tmp7 %tmp18 = mul i33 %tmp15, %tmp17 %tmp8 = mul i32 %tmp4, %tmp6 %tmp19 = lshr i33 %tmp18, 1 %2 = shl i32 %tmp8, 2 %tmp20 = trunc i33 %tmp19 to i32 %tm...

...n A[2*i + j - 10]= .. always accesses memory using a function       f(i,j)   =   2*i + j - 10 + base_address_of_A If I run the scalar evolution pass on this code I obtain: %arrayidx = getelementptr inbounds [200 x i32]* @main.A, i32 0, i64 %idxprom   -->  ((4 * (sext i32 (-10 + (2 * %tmp6) + %tmp7) to i64)) + @main.A)         Could you please offer an insight on how can I obtain the function from the internals of the scalar evolution pass?  Thank you. Alexandra -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/atta...