similar to: [hexagon][PowerPC] code regression (sub-optimal code) on LLVM 9 when generating hardware loops, and the "llvm.uadd" intrinsic.

The Hexagon part is fixed in r364790. -- Krzysztof Parzyszek kparzysz at quicinc.com<mailto:kparzysz at quicinc.com> LLVM compiler development From: llvm-dev <llvm-dev-bounces at lists.llvm.org> On Behalf Of Joan Lluch via llvm-dev Sent: Sunday, June 30, 2019 2:04 PM To: llvm-dev <llvm-dev at lists.llvm.org> Subject: [EXT] [llvm-dev] [hexagon][PowerPC] code regression

I've looked into that and it's not going to be simple, unfortunately. Here's the original example again: --- name: fred tracksRegLiveness: true body: | bb.0: successors: %bb.1, %bb.2 J2_jumpt undef %p0, %bb.2, implicit-def %pc J2_jump %bb.1, implicit-def %pc bb.1: successors: %bb.3 %r0 = L2_loadruh_io undef %r0, 0 PS_storerhabs 0, killed

Yes, immediately after branch folding the code would still behave the same as the original. At the same time, any subsequent optimization may "exploit" the incorrect liveness information to do something bad. If you add -run-pass if-converter, you'll get: # After If Converter # Machine code for function fred: IsSSA, NoPHIs, TracksLiveness, NoVRegs BB#0: %R0<def>

On 7/28/2017 1:59 PM, Quentin Colombet wrote: > Hi Krzysztof, > > Thanks for digging into this. > >> On Jul 28, 2017, at 11:08 AM, Krzysztof Parzyszek <kparzysz at codeaurora.org> wrote: >> >> I've looked into that and it's not going to be simple, unfortunately. >> >> Here's the original example again: >> >> --- >>

The comment in test/CodeGen/X86/branchfolding-undef.mir states that such merging is legal, however doing so can actually generate wrong code: Consider this (valid code): --- name: fred tracksRegLiveness: true body: | bb.0: successors: %bb.1, %bb.2 J2_jumpt undef %p0, %bb.2, implicit-def %pc J2_jump %bb.1, implicit-def %pc bb.1: successors: %bb.3 %r0 =

Rust pattern matching code can sometimes generate very redundant LLVM IR, in which several branches contain exactly the same code. LLVM successfully unifies that, but the dispatching mechanism does not simplify accordingly. I created a gist here: https://gist.github.com/ranma42/d2e6d50999e801ffd4ed (based on two examples available in Rust issues:

Hi @ll, while clang/LLVM recognizes common bit-twiddling idioms/expressions like unsigned int rotate(unsigned int x, unsigned int n) { return (x << n) | (x >> (32 - n)); } and typically generates "rotate" machine instructions for this expression, it fails to recognize other also common bit-twiddling idioms/expressions. The standard IEEE CRC-32 for "big

On Jun 13, 2012, at 1:15 PM, Sergei Larin <slarin at codeaurora.org> wrote: > Andy, > > You are probably right here – look at this – before phi elimination this code looks much more sane: > > # *** IR Dump After Live Variable Analysis ***: > # Machine code for function push: SSA > Function Live Outs: %R0 > > BB#0: derived from LLVM BB %entry >

On Jun 13, 2012, at 10:49 AM, Sergei Larin <slarin at codeaurora.org> wrote: > So if this early exit is taken: > > // SSA defs do not have output/anti dependencies. > // The current operand is a def, so we have at least one. > if (llvm::next(MRI.def_begin(Reg)) == MRI.def_end()) > return; > > we do not ever get to this point: > >

Ok, after a long detour I am back to where I have started. I think there is a problem at dep DAG construction. Let me try to convince you. Here is the C code we are dealing with: push () { struct xx_stack *stack, *top; for (stack = xx_stack; stack; stack = stack->next) top = stack; yy_instr = top->first; } If the loop never iterates, "top" will have

On Sep 18, 2012, at 1:45 PM, Bjorn De Sutter <bjorn.desutter at elis.ugent.be> wrote: > I am working on a backend for a CGRA architecture with advanced predicate support (as on EPIC machines and as first used in the OpenIMPACT compiler). Until last month, the backend was working fine, but since the r161643 commit by stoklund, my backend doesn't work anymore. I think I noticed some

Sergei, Absolutely right, the copy/ldriw should not be reordered. I was attempting to explain that I consider it a phi-elimination bug, not a DAG builder bug. Liveness will also have problems with this code in the long run. To avoid confusion, I filed PR13112: Phi elimination generates uninitialized vreg uses, and disabled the SSA check until its fixes in r158461. However, your C code is also

Hi I have some inline function C code, that llvm could be optimizing better. Since I am new to this, I wonder if someone could give me a few pointers, how to approach this in LLVM. Should I try to change the IR code -somehow- to get the code generator to generate better code, or should I rather go to the code generator and try to add an optimization pass ? Thanks for any feedback. Ciao

The Problem? Nowaday, SCEV called "Scalar Evolution" does only evolate instructions that has predictable operand, Constant-Based operand. such as that can evolute as a constant. otherwise we couldn't evolate it as SCEV node, evolated as SCEVUnknown. important thing that we remember is, we do not use SCEV only for Loop Deletion, which that doesn't really needed on nature loops

Andy, You are probably right here - look at this - before phi elimination this code looks much more sane: # *** IR Dump After Live Variable Analysis ***: # Machine code for function push: SSA Function Live Outs: %R0 BB#0: derived from LLVM BB %entry %vreg5<def> = IMPLICIT_DEF; IntRegs:%vreg5 %vreg4<def> = TFRI_V4 <ga:@xx_stack>; IntRegs:%vreg4

Hello LLVM, This is a proposal for a new pass that improves performance and code size in some nested loop situations. The pass is target independent. >From the description in the file header: This optimization finds loop exit values reevaluated after the loop execution and replaces them by the corresponding exit values if they are available. Such sequences can arise after the

Hi, I am working on a backend for a CGRA architecture with advanced predicate support (as on EPIC machines and as first used in the OpenIMPACT compiler). Until last month, the backend was working fine, but since the r161643 commit by stoklund, my backend doesn't work anymore. I think I noticed some related commits later on, and the assertion I get on the latest trunk (r164162) differs from

Hi, I was wondering if there is a way to specify macros to help shorten rewriting patterns like these: def : Pat <(v4i8 (mul (v4i8 IntRegs:$a), (v4i8 IntRegs:$b))), (v4i8 (VTRUNEHB (v4i16 (VTRUNEWH (v2i32 (VMPYH (v2i16 (EXTRACT_SUBREG (v4i16 (VSXTBH (v4i8 IntRegs:$a))), subreg_hireg)), (v2i16 (EXTRACT_SUBREG (v4i16 (VSXTBH (v4i8

I want to generate the instruction like addi 40, r3 ! i.e. r3 = r3 + 40 The format i wrote is def ADDI : F1<opcode, (outs IntRegs:$dst), (ins IntRegs:$dst, i32imm:$imm) "addi $imm, $dst", [(set $IntRegs:$dst, (add $IntRegs:$dst, i32imm:$c))] but it is not compiling. what should be the format. vikram -- View this message

Andy, Lang, Thanks for the suggestion. I have spent more time with it today, and I do see some strange things in liveness update. I am not at the actual cause yet, but here is what I got so far: I have the following live ranges when I start scheduling a region: R2 = [0B,48r:0)[352r,416r:5)... R3 = [0B,48r:0)[368r,416r:5)... R4 = [0B,32r:0)[384r,416r:4)... R5 = [0B,32r:0)[400r,416r:4)...