[llvm-dev] Instruction selection problems due to SelectionDAGBuilder

Daniel Sanders via llvm-dev llvm-dev at lists.llvm.org
Thu Aug 4 03:16:25 PDT 2016


Hi Alex,

>     However, when using the mips64 back end (subtarget) we get this correct selection DAG:
>        <snip>
>            t55: v8i64 = BUILD_VECTOR Constant:i64<0>, Constant:i64<-1>, Constant:i64<-2>, Constant:i64<-3>, Constant:i64<-4>, Constant:i64<-5>, Constant:i64<-6>, Constant:i64<-7>
>          t56: v8i64 = add t47, t55
>        <snip>

v8i64 isn't a legal type on MIPS64 with MSA so I think you must be looking at the SelectionDAG before type legalization. This can be very different from the SelectionDAG used for instruction selection which may explain the confusion. You can see the DAG that the instruction selector sees using –view-isel-dags.

Eli Bendersky has a good high-level overview of the code generator at http://eli.thegreenplace.net/2012/11/24/life-of-an-instruction-in-llvm but the relevant bit can be roughly summarized as:

·         LLVM-IR is converted to an equivalent SelectionDAG which will almost certainly contain types and operations the target won't be able to handle

·         The legalizer hacks away at the SelectionDAG until it fits the target

·         The 'legal' SelectionDAG nodes are replaced with instructions for the target.

So for your example on Mips, we start with the LLVM-IR:
                %0 = add <8 x i64> %a, <i64 0, i64 -1, i64 -2, i64 -3, i64 -4, i64 -5, i64 -6, i64 -7>
This is converted to a SelectionDAG that looks something like:
                  t1: v8i64 = BUILD_VECTOR Constant:i64<0>, Constant:i64<-1>, Constant:i64<-2>, Constant:i64<-3>, Constant:i64<-4>, Constant:i64<-5>, Constant:i64<-6>, Constant:i64<-7>
                t2: v8i64 = add t0, t2
This SelectionDAG contains illegal vector types (they have too many elements for our target) so the vectors are split:
                  t11: v4i64 = BUILD_VECTOR Constant:i64<0>, Constant:i64<-1>, Constant:i64<-2>, Constant:i64<-3>
                t20: v4i64 = add t10, t11
                  t12: v4i64 = BUILD_VECTOR Constant:i64<-4>, Constant:i64<-5>, Constant:i64<-6>, Constant:i64<-7>
                t21: v4i64 = add t10, t12
which still has illegal vector types so it splits them again to get something like:
                  t31: v2i64 = BUILD_VECTOR Constant:i64<0>, Constant:i64<-1>
                t53: v2i64 = add t44, t31
                  t32: v2i64 = BUILD_VECTOR Constant:i64<-2>, Constant:i64<-3>
                t54: v2i64 = add t45, t32
                  t33: v2i64 = BUILD_VECTOR Constant:i64<-4>, Constant:i64<-5>
                t55: v2i64 = add t45, t33
                  t34: v2i64 = BUILD_VECTOR Constant:i64<-6>, Constant:i64<-7>
                t56: v2i64 = add t46, t34
At this point we have legal types but some illegal operations so the operation legalizer steps in. The 'add' operations are fine since we'll be able to select the addv.d instruction for these but we can't pick instructions for the 'BUILD_VECTOR' nodes. If the constants were different then these nodes might be legal (see MipsSETargetLowering::lowerBUILD_VECTOR() for the code that decides which nodes are ok and which aren't, and also the 'setOperationAction(ISD::BUILD_VECTOR, Ty, Custom)' call that tells SelectionDAG the rules are non-trivial) but we'll have to replace the BUILD_VECTOR's we have with something we can handle. The operation legalizer therefore changes them to something like:
                  t61: v2i64,ch = load<LD64[ConstantPool]> t65, ConstantPool:i64<<2 x i64> <i64 0, i64 -1>> 0, undef:i64
                t83: v2i64 = add t74, t61
                  t62: v2i64,ch = load<LD64[ConstantPool]> t65, ConstantPool:i64<<2 x i64> <i64 -2, i64 -3>> 0, undef:i64
                t84: v2i64 = add t75, t62
                  t63: v2i64,ch = load<LD64[ConstantPool]> t65, ConstantPool:i64<<2 x i64> <i64 -4, i64 -5>> 0, undef:i64
                t85: v2i64 = add t75, t63
                  t64: v2i64,ch = load<LD64[ConstantPool]> t65, ConstantPool:i64<<2 x i64> <i64 -6, i64 -7>> 0, undef:i64
                t86: v2i64 = add t76, t64
At this point, the SelectionDAG is suitable for Mips64 with MSA so the instruction selector runs all the rules defined in tablegen to convert the DAG to a DAG of target instructions (MachineSDNode's).

Hope this helps.

From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of Nemanja Ivanovic via llvm-dev
Sent: 04 August 2016 07:20
To: Alex Susu
Cc: llvm-dev
Subject: Re: [llvm-dev] Instruction selection problems due to SelectionDAGBuilder

I'm not an expert on this at all and there isn't enough information shown to see how the "Good" back end performs the BUILD_VECTOR operation with a constant vector, but it is clear that your back end does that with a Constant Pool load. Furthermore, your back end probably does not specify a matcher in the target description file for the respective load.
As far as debugging is concerned - you can find exactly where the matching fails by opening $LLVM_BUILD/lib/<Target>/<Tgt>GenDAGISel.inc and finding the indices listed above (268, 277, etc.).
So I think that if you don't want BUILD_VECTOR for MVT::v8i64 with constant elements to be legalized as a constant pool load, you should not have the following line in your TargetLowering instance:
setOperationAction(ISD::BUILD_VECTOR, MVT::v8i64, Expand)

At least I think that is a rough description of some of the issues causing this.
N

On Tue, Aug 2, 2016 at 6:00 PM, Alex Susu via llvm-dev <llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:
  Hello.
    I'm having problems at instruction selection with my back end with the following basic-block due to a vector add with immediate constant vector (obtained by vectorizing a simple C program doing vector sum map):
    vector.ph<http://vector.ph>:                                        ; preds = %vector.memcheck50
      %.splatinsert = insertelement <8 x i64> undef, i64 %i.07.unr, i32 0
      %.splat = shufflevector <8 x i64> %.splatinsert, <8 x i64> undef, <8 x i32> zeroinitializer
      %induction = add <8 x i64> %.splat, <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
      %.splatinsert56 = insertelement <8 x i64> undef, i64 %xtraiter, i32 0
      %.splat57 = shufflevector <8 x i64> %.splatinsert56, <8 x i64> undef, <8 x i32> zeroinitializer
      %induction58 = add <8 x i64> %.splat57, <i64 0, i64 -1, i64 -2, i64 -3, i64 -4, i64 -5, i64 -6, i64 -7>
      br label %vector.body25

    The exact problem reported is:
        Selecting: t51: v8i64,ch = load<LD64[ConstantPool]> t0, ConstantPool:i64<<8 x i64> <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>> 0, undef:i64
        ISEL: Starting pattern match on root node: t51: v8i64,ch = load<LD64[ConstantPool]> t0, ConstantPool:i64<<8 x i64> <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>> 0, undef:i64
          Initial Opcode index to 268
          Match failed at index 277
          Continuing at 396
          Match failed at index 398
          Continuing at 422
        LLVM ERROR: Cannot select: t51: v8i64,ch = load<LD64[ConstantPool]> t0, ConstantPool:i64<<8 x i64> <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>> 0, undef:i64
          t50: i64 = ConstantPool<<8 x i64> <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>> 0
          t48: i64 = undef
        In function: foo

    The reason is that for the basic-block my back end generates the following Selection DAG:
        (From 201_LoopVectorize/25_GOOD_map/NEW/6/1/NEW/STDerr3_wo_getSetCCResultType)
        Initial selection DAG: BB#15 'foo:vector.ph<http://vector.ph>'
        SelectionDAG has 41 nodes:
          t0: ch = EntryToken
          t4: i32 = Constant<0>
                      t3: i64,ch = CopyFromReg t0, Register:i64 %vreg12
                    t6: v8i64 = insert_vector_elt undef:v8i64, t3, Constant:i64<0>
                  t7: v8i64 = vector_shuffle<0,0,0,0,0,0,0,0> t6, undef:v8i64
                  t15: v8i64 = BUILD_VECTOR Constant:i64<0>, Constant:i64<1>, Constant:i64<2>, Constant:i64<3>, Constant:i64<4>, Constant:i64<5>, Constant:i64<6>, Constant:i64<7>
                t16: v8i64 = add t7, t15
              t18: ch = CopyToReg t0, Register:v8i64 %vreg16, t16
                        t20: i64,ch = CopyFromReg t0, Register:i64 %vreg5
                      t22: i64 = AssertSext t20, ValueType:ch:i8
                    t23: v8i64 = insert_vector_elt undef:v8i64, t22, Constant:i64<0>
                  t24: v8i64 = vector_shuffle<0,0,0,0,0,0,0,0> t23, undef:v8i64
                  t32: v8i64 = BUILD_VECTOR Constant:i64<0>, Constant:i64<-1>, Constant:i64<-2>, Constant:i64<-3>, Constant:i64<-4>, Constant:i64<-5>, Constant:i64<-6>, Constant:i64<-7>
                t33: v8i64 = add t24, t32
              t35: ch = CopyToReg t0, Register:v8i64 %vreg17, t33
              t37: ch = CopyToReg t0, Register:i64 %vreg117, Constant:i64<0>
            t39: ch = TokenFactor t18, t35, t37
          t40: ch = br t39, BasicBlock:ch<vector.body25 0x1d07660>

    However, when using the mips64 back end (subtarget) we get this correct selection DAG:
        (From 201_LoopVectorize/25_GOOD_map/NEW/6/1/NEW/Mips64/STDerr_llc_mips64)
        Initial selection DAG: BB#15 'foo:vector.ph<http://vector.ph>'
        SelectionDAG has 87 nodes:
          t0: ch = EntryToken
          t4: i32 = Constant<0>
                t3: i64,ch = CopyFromReg t0, Register:i64 %vreg12
              t6: v8i64 = insert_vector_elt undef:v8i64, t3, Constant:i64<0>
            t7: v8i64 = vector_shuffle<0,0,0,0,0,0,0,0> t6, undef:v8i64
            t15: v8i64 = BUILD_VECTOR Constant:i64<0>, Constant:i64<1>, Constant:i64<2>, Constant:i64<3>, Constant:i64<4>, Constant:i64<5>, Constant:i64<6>, Constant:i64<7>
          t16: v8i64 = add t7, t15
                  t43: i64,ch = CopyFromReg t0, Register:i64 %vreg5
                t45: i64 = AssertSext t43, ValueType:ch:i8
              t46: v8i64 = insert_vector_elt undef:v8i64, t45, Constant:i64<0>
            t47: v8i64 = vector_shuffle<0,0,0,0,0,0,0,0> t46, undef:v8i64
            t55: v8i64 = BUILD_VECTOR Constant:i64<0>, Constant:i64<-1>, Constant:i64<-2>, Constant:i64<-3>, Constant:i64<-4>, Constant:i64<-5>, Constant:i64<-6>, Constant:i64<-7>
          t56: v8i64 = add t47, t55
                  t17: i64 = extract_vector_elt t16, Constant:i64<0>
                t26: ch = CopyToReg t0, Register:i64 %vreg16, t17
                  t18: i64 = extract_vector_elt t16, Constant:i64<1>
                t28: ch = CopyToReg t0, Register:i64 %vreg17, t18
                  t19: i64 = extract_vector_elt t16, Constant:i64<2>
                t30: ch = CopyToReg t0, Register:i64 %vreg18, t19
                  t20: i64 = extract_vector_elt t16, Constant:i64<3>
                t32: ch = CopyToReg t0, Register:i64 %vreg19, t20
                  t21: i64 = extract_vector_elt t16, Constant:i64<4>
                t34: ch = CopyToReg t0, Register:i64 %vreg20, t21
                  t22: i64 = extract_vector_elt t16, Constant:i64<5>
                t36: ch = CopyToReg t0, Register:i64 %vreg21, t22
                  t23: i64 = extract_vector_elt t16, Constant:i64<6>
                t38: ch = CopyToReg t0, Register:i64 %vreg22, t23
                  t24: i64 = extract_vector_elt t16, Constant:i64<7>
                t40: ch = CopyToReg t0, Register:i64 %vreg23, t24
              t41: ch = TokenFactor t26, t28, t30, t32, t34, t36, t38, t40
                  t57: i64 = extract_vector_elt t56, Constant:i64<0>
                t66: ch = CopyToReg t0, Register:i64 %vreg24, t57
                  t58: i64 = extract_vector_elt t56, Constant:i64<1>
                t68: ch = CopyToReg t0, Register:i64 %vreg25, t58
                  t59: i64 = extract_vector_elt t56, Constant:i64<2>
                t70: ch = CopyToReg t0, Register:i64 %vreg26, t59
                  t60: i64 = extract_vector_elt t56, Constant:i64<3>
                t72: ch = CopyToReg t0, Register:i64 %vreg27, t60
                  t61: i64 = extract_vector_elt t56, Constant:i64<4>
                t74: ch = CopyToReg t0, Register:i64 %vreg28, t61
                  t62: i64 = extract_vector_elt t56, Constant:i64<5>
                t76: ch = CopyToReg t0, Register:i64 %vreg29, t62
                  t63: i64 = extract_vector_elt t56, Constant:i64<6>
                t78: ch = CopyToReg t0, Register:i64 %vreg30, t63
                  t64: i64 = extract_vector_elt t56, Constant:i64<7>
                t80: ch = CopyToReg t0, Register:i64 %vreg31, t64
              t81: ch = TokenFactor t66, t68, t70, t72, t74, t76, t78, t80
              t83: ch = CopyToReg t0, Register:i64 %vreg209, Constant:i64<0>
            t85: ch = TokenFactor t41, t81, t83
          t86: ch = br t85, BasicBlock:ch<vector.body25 0x1bd35f0>

    I am curious what is wrong - I've tried to match the Mips' back end: I have put most of the vector splat instructions and the vextract and INSERT_D_DESC instruction, etc .
    I also don't get enough DEBUG information to understand where exactly the problem comes from (probably I missed some TableGen record).

    Please let me know if you have any idea.

  Thank you very much,
    Alex
_______________________________________________
LLVM Developers mailing list
llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160804/c28d3ccc/attachment.html>


More information about the llvm-dev mailing list