[LLVMdev] Error w/ Tablegen + Intrinsics

[Villmow, Micah]

I found the problem

 

 

If the intrinsic instructions are not wrapped in:

let Properties = [IntrReadMem] in {

}

 

Then they are off by one.

 

Any reason why this should be required?

 

Thanks,

 

Micah

 

From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu]
On Behalf Of Nate Begeman
Sent: Tuesday, April 14, 2009 6:47 PM
To: LLVM Developers Mailing List
Subject: Re: [LLVMdev] Error w/ Tablegen + Intrinsics

 

Are you using isTarget = 1 in your intrinsics file?

 

On Apr 14, 2009, at 6:34 PM, Villmow, Micah wrote:





It seems that Tablegen is generating intrinsic ID's off by  in
DAGISel.inc

 

In DAGISel.inc, I have the following pattern:

int64_t CN1 = Tmp0->getZExtValue();

 

    // Pattern: (intrinsic_w_chain:f32 103:iPTR, GPRF32:f32:$src0,
GPRF32:f32:$src1, GPRF32:f32:$src2)

    // Emits: (MACRO_FMA_f32:f32 GPRF32:f32:$src0, GPRF32:f32:$src1,
GPRF32:f32:$src2)

    // Pattern complexity = 8  cost = 1  size = 0

    if (CN1 == INT64_C(103)) {

      SDValue N2 = N.getOperand(2);

      SDValue N3 = N.getOperand(3);

      SDValue N4 = N.getOperand(4);

      if (N2.getNode()->getValueType(0) == MVT::f32 &&

          N3.getNode()->getValueType(0) == MVT::f32 &&

          N4.getNode()->getValueType(0) == MVT::f32) {

        SDNode *Result = Emit_28(N, AMDIL::MACRO_FMA_f32, MVT::f32);

        return Result;

      }

    }

 

However, when MACRO_FMA is called, Tmp0->getZExtValue() returns 102!

 

The section of code where this is probably wrong is:

CodeGenDAGPatterns.cpp:1238 unsigned IID =
getDAGPatterns().getIntrinsicID(Operator)+1;

Or here:

CodeGenDAGPatterns.h:524

const CodeGenIntrinsic &getIntrinsicInfo(unsigned IID) const {

    assert(IID-1 < Intrinsics.size() && "Bad intrinsic ID!");

    return Intrinsics[IID-1];

  }

 

These are the only locations where I can find that it is returning the
IID off by 1.

 

 

If someone can let me know what I need to do in order to fix this it
would be greatly appreciated.

 

Thanks,

 

 

_______________________________________________
LLVM Developers mailing list
LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20090414/aa728dcd/attachment.html>