[llvm-dev] [Hexagon] Type Legalization

Haidl, Michael via llvm-dev llvm-dev at lists.llvm.org
Fri Sep 22 00:17:27 PDT 2017


Ok I double checked it and TLI.isTypeLegal does work. My setup was 
messed up and an earlier version of the *.so was loaded.

Cheers,
Michael

On 22.09.2017 07:46, Haidl, Michael via llvm-dev wrote:
> Hi Craig,
> 
> protecting the transformation with:
>     if (TLI.isTypeLegal(VT)
>         && TLI.isOperationLegal(ISD::SUB, VT)
>         && TLI.isOperationLegal(ISD::ADD, VT)
>         && TLI.isOperationLegal(ISD::SHL, VT)
>         && TLI.isOperationLegal(ISD::SRA, VT)) {
> 
> shows the same result.
> 
> Michael
> 
> On 22.09.2017 07:19, Craig Topper wrote:
>> Is VT a legal type on Hexagon? It looks like Hexagon may be setting SHL
>> as Custom for every defined vector type. Try adding TLI.isTypeLegal(VT) too.
>>
>> ~Craig
>>
>> On Thu, Sep 21, 2017 at 10:06 PM, Haidl, Michael
>> <michael.haidl at uni-muenster.de <mailto:michael.haidl at uni-muenster.de>>
>> wrote:
>>
>>      Hi Sanjay,
>>
>>      thanks for this information. I did get a little bit further with the
>>      patch. However, Hexagon gives me headaches.
>>
>>      I tried to limit the scope of the patch to the BeforeLegalizeTypes phase
>>      and Hexagon still reaches the unreachable. Hexagon tries to split or
>>      widen a vector type for a node with custom lowering where the
>>      unreachable arises from inside TargetLowering::ReplaceNodeResults which
>>      is not overwritten in the Hexagon backend.
>>
>>      To avoid generating illegal code after all I queried TLI if all
>>      generated operations are legal operations for the given vector type.
>>
>>          if (   TLI.isOperationLegal(ISD::SUB, VT)
>>              && TLI.isOperationLegal(ISD::ADD, VT)
>>              && TLI.isOperationLegal(ISD::SHL, VT)
>>              && TLI.isOperationLegal(ISD::SRA, VT)) {
>>
>>      The Hexagon backend says happily: yes! Go ahead!
>>      When it comes to HexagonDAGToDAGISel and the type should be legalized no
>>      custom lowering is found and TargetLowering::ReplaceNodeResults is
>>      called which falls back to the default implementation and the
>>      unreachable is trigged.
>>
>>      Is this really working as intended or am I missing something here?
>>
>>      Cheers,
>>      Michael
>>
>>      On 20.09.2017 16:10, Sanjay Patel wrote:
>>       > There are multiple problems/questions here:
>>       >
>>       > 1. Make sure you've updated trunk to the latest rev before running
>>       > update_llc_test_checks.py on lea-3.ll. Ie, I would only expect the
>>       > output you're seeing if you're running the script on a version of
>>      that
>>       > test file before r313631. After that commit, each RUN has its own
>>      check
>>       > prefix, so there should be no conflict opportunity.
>>       >
>>       > 2. I didn't realize the scope of the patch covered all targets
>>      and both
>>       > scalars and vectors. That isn't going to work as-is. We can't assume
>>       > that every target and data type will prefer to replace that multiply.
>>       > Even the x86 diffs in lea-3.ll are regressions:
>>       >
>>       > -; LNX1-NEXT:    leal (%rdi,%rdi,2), %eax
>>       > +; LNX1-NEXT:    leal (,%rdi,4), %eax
>>       > +; LNX1-NEXT:    subl %edi, %eax
>>       >
>>       > I suggest taking a smaller first step by limiting the patch to cases
>>       > where the multiply is not a legal op for a target
>>      (TLI.isOperationLegal()).
>>       >
>>       > 3. Since the patch can cause a crash, that needs to be investigated
>>       > first. I didn't look into it, but my guess for the Hexagon crash (and
>>       > probably other targets) is that you're trying to create illegal
>>       > ops after the DAG has been legalized. So that's another potential
>>      way to
>>       > limit the scope of the patch - don't try the transform unless we're
>>       > pre-legalization:
>>       >    if (Level < AfterLegalizeDAG) { // do something }
>>       >
>>       >
>>       >
>>       > On Wed, Sep 20, 2017 at 4:17 AM, Haidl, Michael
>>       > <michael.haidl at uni-muenster.de
>>      <mailto:michael.haidl at uni-muenster.de>
>>      <mailto:michael.haidl at uni-muenster.de
>>      <mailto:michael.haidl at uni-muenster.de>>>
>>       > wrote:
>>       >
>>       >     Hi,
>>       >
>>       >     I am currently working on a more or less intrusive patch
>>      (D37896), which
>>       >     pulls optimizations on multiplications from some back-ends,
>>      e.g., (mul
>>       >     x, 2^N + 1) => (add (shl x, N), x) in AArch64, into the
>>      DAGCombiner to
>>       >     have this optimization generic on all targets.
>>       >     However, running the LLVM Tests leads to 67 unexpected results.
>>       >
>>       >     Am 19.09.2017 um 15:58 schrieb Sanjay Patel:
>>       >      > For the tests that are changing, you should see if those
>>      changes are
>>       >      > improvements, regressions, or neutral. This is
>>      unfortunately not
>>       >     always
>>       >      > obvious for x86 asm, so feel free to just post those diffs
>>      in an
>>       >     updated
>>       >      > version of the patch at D37896.
>>       >      >
>>       >      > If the test files have auto-generated assertions (look for
>>      this
>>       >     string
>>       >      > on the first line of the test file: "NOTE: Assertions have
>>      been
>>       >      > autogenerated by utils/update_llc_test_checks.py"...
>>       >      > and both of these do as of:
>>      https://reviews.llvm.org/rL313631 <https://reviews.llvm.org/rL313631>
>>       >     <https://reviews.llvm.org/rL313631
>>      <https://reviews.llvm.org/rL313631>>
>>       >      > <https://reviews.llvm.org/rL313631
>>      <https://reviews.llvm.org/rL313631>
>>       >     <https://reviews.llvm.org/rL313631
>>      <https://reviews.llvm.org/rL313631>>> ), then it's easy to observe the
>>       >      > diffs by re-running that script after your code patch is
>>      applied:
>>       >      > $ /path/to/update_llc_test_checks.py
>>      --llc=/path/to/local/and/new/llc
>>       >      > lea-3.ll
>>       >      > $ svn diff lea-3.ll
>>       >
>>       >     As described by Sanjay in the original thread, I updated the
>>      llc tests
>>       >     but some of the tests were not updated.
>>       >
>>       >     For example: CodeGen/X86/lea-3.ll produces the following output:
>>       >     llvm/utils/update_llc_test_checks.py
>>      --llc=/home/dev/local/bin/llc
>>       >     CodeGen/X86/lea-3.ll
>>       >     WARNING: Found conflicting asm under the same prefix: 'CHECK'!
>>       >     WARNING: Found conflicting asm under the same prefix: 'CHECK'!
>>       >     WARNING: Found conflicting asm under the same prefix: 'CHECK'!
>>       >
>>       >     leaving parts of the test unchanged and the tests still fail.
>>       >
>>       >     Other tests like CodeGen/AArch64/aarch64-gep-opt.ll results in:
>>       >     WARNING: Skipping non-FileChecked RUN line: llc -O3
>>       >     -aarch64-enable-gep-opt=true -mattr=-use-aa
>>      -print-after=codegenprepare
>>       >     < %s >%t 2>&1 && FileCheck --check-prefix=CHECK-NoAA <%t %s
>>       >     WARNING: Skipping non-FileChecked RUN line: llc -O3
>>       >     -aarch64-enable-gep-opt=true -mattr=+use-aa
>>      -print-after=codegenprepare
>>       >     < %s >%t 2>&1 && FileCheck --check-prefix=CHECK-UseAA <%t %s
>>       >     WARNING: Skipping non-FileChecked RUN line: llc -O3
>>       >     -aarch64-enable-gep-opt=true -print-after=codegenprepare
>>      -mcpu=cyclone <
>>       >     %s >%t 2>&1 && FileCheck --check-prefix=CHECK-NoAA <%t %s
>>       >     WARNING: Skipping non-FileChecked RUN line: llc -O3
>>       >     -aarch64-enable-gep-opt=true -print-after=codegenprepare
>>       >     -mcpu=cortex-a53 < %s >%t 2>&1 && FileCheck
>>      --check-prefix=CHECK-UseAA
>>       >     <%t %s
>>       >
>>       >     And the worst thing I witnessed was with the Hexagon back-end
>>      were my
>>       >     changes in DAGCombiner trigger an Unreachable:
>>       >
>>       >     home/dev/llvm/build/./bin/llc -march=hexagon -mcpu=hexagonv5
>>       >     -disable-hsdr <
>>       >     /home/dev/llvm/llvm/test/CodeGen/Hexagon/vect/vect-cst-v4i32.ll |
>>       >     /home/dev/llvm/build/./bin/FileCheck
>>       >     /home/dev/llvm/llvm/test/CodeGen/Hexagon/vect/vect-cst-v4i32.ll
>>       >     --
>>       >     Exit Code: 2
>>       >
>>       >     Command Output (stderr):
>>       >     --
>>       >     ReplaceNodeResults not implemented for this target!
>>       >     UNREACHABLE executed at
>>       >     /home/dev/llvm/llvm/include/llvm/Target/TargetLowering.h:3164!
>>       >     #0 0x00007f1b2d330cfa
>>      llvm::sys::PrintStackTrace(llvm::raw_ostream&)
>>       >     (/home/dev/llvm/build/bin/../lib/libLLVMSupport.so.6+0x116cfa)
>>       >     #1 0x00007f1b2d32ea9e llvm::sys::RunSignalHandlers()
>>       >     (/home/dev/llvm/build/bin/../lib/libLLVMSupport.so.6+0x114a9e)
>>       >     #2 0x00007f1b2d32ec04 SignalHandler(int)
>>       >     (/home/dev/llvm/build/bin/../lib/libLLVMSupport.so.6+0x114c04)
>>       >     #3 0x00007f1b2c3ed7f0 (/lib/x86_64-linux-gnu/libc.so.6+0x357f0)
>>       >     #4 0x00007f1b2c3ed77f gsignal
>>      (/lib/x86_64-linux-gnu/libc.so.6+0x3577f)
>>       >     #5 0x00007f1b2c3ef37a abort
>>      (/lib/x86_64-linux-gnu/libc.so.6+0x3737a)
>>       >     #6 0x00007f1b2d2b012a
>>       >     (/home/dev/llvm/build/bin/../lib/libLLVMSupport.so.6+0x9612a)
>>       >     #7 0x00007f1b359326ec
>>       >
>>        (/home/dev/llvm/build/bin/../lib/libLLVMHexagonCodeGen.so.6+0x1176ec)
>>       >     #8 0x00007f1b2d6a37a6
>>       >     llvm::DAGTypeLegalizer::CustomLowerNode(llvm::SDNode*,
>>      llvm::EVT, bool)
>>       >
>>        (/home/dev/llvm/build/bin/../lib/libLLVMSelectionDAG.so.6+0x1247a6)
>>       >     #9 0x00007f1b2d6be0d9
>>       >     llvm::DAGTypeLegalizer::SplitVectorResult(llvm::SDNode*,
>>      unsigned int)
>>       >
>>        (/home/dev/llvm/build/bin/../lib/libLLVMSelectionDAG.so.6+0x13f0d9)
>>       >     #10 0x00007f1b2d69f156 llvm::DAGTypeLegalizer::run()
>>       >
>>        (/home/dev/llvm/build/bin/../lib/libLLVMSelectionDAG.so.6+0x120156)
>>       >     #11 0x00007f1b2d6a01ff llvm::SelectionDAG::LegalizeTypes()
>>       >
>>        (/home/dev/llvm/build/bin/../lib/libLLVMSelectionDAG.so.6+0x1211ff)
>>       >     #12 0x00007f1b2d785d42
>>      llvm::SelectionDAGISel::CodeGenAndEmitDAG()
>>       >
>>        (/home/dev/llvm/build/bin/../lib/libLLVMSelectionDAG.so.6+0x206d42)
>>       >     #13 0x00007f1b2d790fd8
>>       >     llvm::SelectionDAGISel::SelectAllBasicBlocks(llvm::Function
>>      const&)
>>       >
>>        (/home/dev/llvm/build/bin/../lib/libLLVMSelectionDAG.so.6+0x211fd8)
>>       >     #14 0x00007f1b2d7930e2
>>       >
>>        llvm::SelectionDAGISel::runOnMachineFunction(llvm::MachineFunction&)
>>       >     [clone .part.873]
>>       >
>>        (/home/dev/llvm/build/bin/../lib/libLLVMSelectionDAG.so.6+0x2140e2)
>>       >     #15 0x00007f1b35931d6a (anonymous
>>       >
>>        namespace)::HexagonDAGToDAGISel::runOnMachineFunction(llvm::MachineFunction&)
>>       >
>>        (/home/dev/llvm/build/bin/../lib/libLLVMHexagonCodeGen.so.6+0x116d6a)
>>       >     #16 0x00007f1b2ef22585
>>       >     llvm::MachineFunctionPass::runOnFunction(llvm::Function&)
>>       >     (/home/dev/llvm/build/bin/../lib/libLLVMCodeGen.so.6+0x21f585)
>>       >     #17 0x00007f1b2e922dd3
>>       >     llvm::FPPassManager::runOnFunction(llvm::Function&)
>>       >     (/home/dev/llvm/build/bin/../lib/libLLVMCore.so.6+0x18ddd3)
>>       >     #18 0x00007f1b2e922e93
>>      llvm::FPPassManager::runOnModule(llvm::Module&)
>>       >     (/home/dev/llvm/build/bin/../lib/libLLVMCore.so.6+0x18de93)
>>       >     #19 0x00007f1b2e9239e1
>>      llvm::legacy::PassManagerImpl::run(llvm::Module&)
>>       >     (/home/dev/llvm/build/bin/../lib/libLLVMCore.so.6+0x18e9e1)
>>       >     #20 0x000055b29c4ef3e2 compileModule(char**, llvm::LLVMContext&)
>>       >     (/home/dev/llvm/build/./bin/llc+0x213e2)
>>       >     #21 0x000055b29c4e23bc main
>>      (/home/dev/llvm/build/./bin/llc+0x143bc)
>>       >     #22 0x00007f1b2c3d83f1 __libc_start_main
>>       >     (/lib/x86_64-linux-gnu/libc.so.6+0x203f1)
>>       >     #23 0x000055b29c4e256a _start
>>      (/home/dev/llvm/build/./bin/llc+0x1456a)
>>       >     Stack dump:
>>       >     0.      Program arguments: /home/dev/llvm/build/./bin/llc
>>      -march=hexagon
>>       >     -mcpu=hexagonv5 -disable-hsdr
>>       >     1.      Running pass 'Function Pass Manager' on module '<stdin>'.
>>       >     2.      Running pass 'Hexagon DAG->DAG Pattern Instruction
>>      Selection' on
>>       >     function '@run'
>>       >     FileCheck error: '-' is empty.
>>       >     FileCheck command line:  /home/dev/llvm/build/./bin/FileCheck
>>       >     /home/dev/llvm/llvm/test/CodeGen/Hexagon/vect/vect-cst-v4i32.ll
>>       >
>>       >     How do I proceed to get the patch running all tests as
>>      expected? The
>>       >     Tests directly focused on the results of the patch are
>>      passing without
>>       >     problems.
>>       >
>>       >     Cheers,
>>       >     Michael
>>       >
>>       >
>>
>>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
> 


More information about the llvm-dev mailing list