FW: Bug in SelectionDAG visitTargetIntrinsic

We ran into a problem where specifying IntrNoMem was causing our instruction selection to fail with target specific intrinsics. After looking into the code and ISel debug it looks like tablegen and SelectionDAG are using different criteria to generate code for intrinsic_w_chain vs intrinsic_wo_chain.

In CodeGenDAGPatterns.cpp, tablegen decides based on whether IntrNoMem is set or not. However with SelectionDAG, whether to use a chain or not is determined by the call site attributes and not by the intrinsic.

So, we can get the situation where the call site has a different attribute than the intrinsic, and this causes selection dag to fail.

I believe that this is wrong and that whether a chain should be generated or not should come from only the intrinsic and not the call site. Since the mapping of call -> intrinsic is by function name only, it should not matter if the readnone attribute is set or not as that is irrelevant to the code generator. Only what is set in the tablegen definition should be determine how the intrinsic is generated.

So, I'm proposing the following patch. What this patch does is instead of relying on the call site to determine if a chain is required, use instead the read/write attributes of the intrinsic from the backend instead. There is not much documentation on target intrinsics and no other backend uses them in this manner.

This patch sound good?

Thanks,
Micah

target_intrinsic_incorrect_chain.txt (1.3 KB)

void SelectionDAGBuilder::visitTargetIntrinsic(const CallInst &I,
- unsigned Intrinsic) {
- bool HasChain = !I.doesNotAccessMemory();
- bool OnlyLoad = HasChain && I.onlyReadsMemory();
+ unsigned Intrinsic) {
+ // Info is set by getTgtMemInstrinsic
+ TargetLowering::IntrinsicInfo Info;
+ bool IsTgtIntrinsic = TLI.getTgtMemIntrinsic(Info, I, Intrinsic);
+ bool HasChain = Info.readMem || Info.writeMem;
+ bool OnlyLoad = HasChain && Info.readMem;

This doesn't seem right. If a call is marked ReadNone, it doesn't seem legal to select it to an intrinsic that read / write memory. By definition a "ReadNone" function cannot touch memory, no?

Evan

At this point, the intrinsic has already been selected, as it was done in its grandparent function(visitCall). This function is only generating the SDNodes that corresponds to the intrinsic and the question is should the call be used to determine whether the chain should be generated, or should the intrinsic be used instead? Since Tablegen uses the intrinsic to do so, I believe normalizing the behavior here is the correct approahc.

Micah

Hi Micah,

*From:*Villmow, Micah
*Sent:* Tuesday, November 06, 2012 1:37 PM
*To:* 'llvm-dev@cs.uiuc.edu'
*Cc:* Guo, Xiaoyi
*Subject:* Bug in SelectionDAG visitTargetIntrinsic

We ran into a problem where specifying IntrNoMem was causing our instruction
selection to fail with target specific intrinsics. After looking into the code
and ISel debug it looks like tablegen and SelectionDAG are using different
criteria to generate code for intrinsic_w_chain vs intrinsic_wo_chain.

In CodeGenDAGPatterns.cpp, tablegen decides based on whether IntrNoMem is set or
not. However with SelectionDAG, whether to use a chain or not is determined by
the call site attributes and not by the intrinsic.

So, we can get the situation where the call site has a different attribute than
the intrinsic, and this causes selection dag to fail.

I'm curious to know how this occurs? I mean, what code is generating a call to
an intrinsic with different attributes on the call site?

I believe that this is wrong and that whether a chain should be generated or not
should come from only the intrinsic and not the call site. Since the mapping of
call -> intrinsic is by function name only, it should not matter if the readnone
attribute is set or not as that is irrelevant to the code generator. Only what
is set in the tablegen definition should be determine how the intrinsic is
generated.

In my opinion whether the node returns a chain result or not should only depend
on the intrinsic, not on the call site. However if the intrinsic has a chain
but the call-site says "read-only" you could, as an optimization, not use the
chain, so getting some benefit from the "read-only" at the codegen level.

So, I'm proposing the following patch. What this patch does is instead of
relying on the call site to determine if a chain is required, use instead the
read/write attributes of the intrinsic from the backend instead. There is not
much documentation on target intrinsics and no other backend uses them in this
manner.

As for your patch, you need a testcase. Also,

@@ -3508,9 +3508,12 @@
/// visitTargetIntrinsic - Lower a call of a target intrinsic to an INTRINSIC
/// node.
void SelectionDAGBuilder::visitTargetIntrinsic(const CallInst &I,
- unsigned Intrinsic) {
- bool HasChain = !I.doesNotAccessMemory();
- bool OnlyLoad = HasChain && I.onlyReadsMemory();
+ unsigned Intrinsic) {
+ // Info is set by getTgtMemInstrinsic
+ TargetLowering::IntrinsicInfo Info;
+ bool IsTgtIntrinsic = TLI.getTgtMemIntrinsic(Info, I, Intrinsic);
+ bool HasChain = Info.readMem || Info.writeMem;
+ bool OnlyLoad = HasChain && Info.readMem;

I don't see what this issue has to do with target intrinsics. Can't you write
some IR by hand that calls some ordinary intrinsic with an unexpected "readnone"
flag and hit the same issue?

Ciao, Duncan.

Hi Evan,

  void SelectionDAGBuilder::visitTargetIntrinsic(const CallInst &I,
- unsigned Intrinsic) {
- bool HasChain = !I.doesNotAccessMemory();
- bool OnlyLoad = HasChain && I.onlyReadsMemory();
+ unsigned Intrinsic) {
+ // Info is set by getTgtMemInstrinsic
+ TargetLowering::IntrinsicInfo Info;
+ bool IsTgtIntrinsic = TLI.getTgtMemIntrinsic(Info, I, Intrinsic);
+ bool HasChain = Info.readMem || Info.writeMem;
+ bool OnlyLoad = HasChain && Info.readMem;

This doesn't seem right. If a call is marked ReadNone, it doesn't seem legal to
select it to an intrinsic that read / write memory. By definition a "ReadNone"
function cannot touch memory, no?

you can have a readnone call to a function that writes memory, it's just that
you somehow know that this particular call doesn't write memory. For example
consider a call to
   int @my_func(int param, bool write_mem);
which only writes memory if write_mem is 'true'. Then it would be correct
to mark calls to my_func "readonly" when the write_mem parameter is 'false',
even though the my_func function itself could't have the readonly parameter,
since it can write memory sometimes, i.e. when write_mem is 'true'.

Presumably Micah has an intrinsic that works this way: it is capable of writing
memory, but in some special identifiable circumstances it doesn't.

Ciao, Duncan.