[LLVMdev] [RFC] Stripping unusable intrinsics

[Philip Reames]

On 12/23/2014 10:41 AM, Chris Lattner wrote:
> On Dec 23, 2014, at 10:28 AM, Chris Bieneman <beanz at apple.com 
> <mailto:beanz at apple.com>> wrote:
>>>> It should be straight-forward to have something like 
>>>> LLVMInitializeX86Target/RegisterTargetMachine install the 
>>>> intrinsics into a registry.
>>>
>>> I tried doing that a few years ago. It’s not nearly as easy as it 
>>> sounds because we’ve got hardcoded references to various target 
>>> intrinsics scattered throughout the code.
>>
>> I was just writing to say exactly this. There are a number of ways we 
>> could work toward something like this. I’m completely in favor of a 
>> world where Intrinsics are properties of the targets and don’t leach 
>> out, however today they do in a lot of places.
>
> What are the specific problems here?  Anything that does an equality 
> comparison with the IntrinsicID can be changed to do strcmp with the 
> name.  That would handle the one-off cases like 
> InstCombiner::SimplifyDemandedUseBits in InstCombine.
>
> The other cases in InstCombine could be handled similarly, but may be 
> better handled by adding a intrinsic behavior query APIs to the 
> intrinsic registry, or would be better handled (eventually) by adding 
> new attributes to the intrinsics themselves.
I find the idea of replacing a bunch of case statements in a switch with 
string comparisons to be highly objectionable.  This would be a marked 
decrease both in readability and maintainability.  I also don't see how 
it helps reduce code bloat.

Let me make an alternate proposal.  Let's separate out the target 
specific cases into their own switch.  We can introduce a function along 
the lines of isTargetArchitectureEnabled(X86) which internally is 
implemented via macros, but externally is just a function call.

We then have code that looks like:
switch(intrinsic_id) {
case Intrinsic::llvm_X: {}
case ....
};
if (isTargetArchitectureEnabled(X86)) {
   // We could even break this into a helper function...
   switch(intrinsic_id) {
   case Intrinsic::x86_X: {}
   case ....
   };
}

If we've done our jobs right in the optimizer, we should be able to 
constant fold that condition at compile time and merge the active 
switches back into a single switch case.  If the architecture isn't 
enabled, the code will trivially fold away.  All without (visible) 
macros or changing the use of intrinsic ids.

Philip






-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20141229/601cd091/attachment.html>