[LLVMdev] Proposal: Loads/stores with deterministic trap/unwind behavior

[Andrew Trick]

On Mar 31, 2014, at 6:58 PM, Peter Collingbourne <peter at pcc.me.uk> wrote:

> Hi,
> 
> I wanted to propose an IR extension that would allow us to support zero-cost
> exception handling for non-call operations that may trap. I wanted to start
> with loads and stores through a null pointer, and later we might extend this to
> div/rem/mod zero. This feature is obviously useful for implementing languages
> such as Java and Go which deterministically translate such operations into
> exceptions which may be caught by the user.
> 
> There are a couple of somewhat orthogonal features that this would entail:
> 
> 1) Deterministic handling for loads and stores through a null pointer.
> 2) Ability to unwind a load/store to a specific basic block, like invoke.
> 
> At the moment, we do not exactly have 1), as the optimizer considers
> non-volatile loads/stores through a null pointer to have undefined
> behavior. Volatile loads/stores are closer, but they come with their own
> set of baggage that can inhibit optimization. (For example, if we can prove
> that a load would always succeed, 'volatile' prevents us from reordering
> the load or deleting it if it is dead.) So I propose to add an attribute to
> 'load' and 'store', which we can call, say, 'nullcheck', with the following
> additional semantics:
> 
> - If the memory address is between zero and a target-defined value (i.e. the
>   size of the zero page) the instruction is guaranteed to trap in a
>   target-defined manner.
> 
> - The optimizer may only delete or reorder nullcheck instructions if the
>   program cannot observe such a transformation by installing a signal handler
>   for the trap.  Therefore, the optimizer would be able to drop the attribute
>   if it can prove that the address will always be non-null.
> 
> To support 2), I propose a couple of new instructions. I haven't come up with
> great names for these instructions, but:
> 
> - 'iload' is to 'load' as 'invoke' is to 'call'. That is, the instruction is
>   a terminator and has normal and unwind destinations. e.g.
> 
>   %v = iload i8* %ptr to label %try.cont unwind label %lpad
> 
> - Similarly, 'istore' is to 'store' as 'invoke' is to 'call'.
> 
>   istore i8 %v, i8* %ptr to label %try.cont unwind label %lpad
> 
> These instructions always have 'nullcheck' semantics, plus:
> 
> - If the instruction traps and the program has installed a signal handler
>   for the trap which unwinds, the unwind is guaranteed to land at the
>   landing pad.
> 
> I've been working on an implementation of 'iload' and 'istore' which are
> in the attached patches, if you are interested. (They aren't ready to go
> in yet.) I have asm parsing/printing for both, and code generation for
> 'iload'. Would be interested in getting feedback on code generation as this
> is my first serious foray into the backend -- I haven't tried running the
> generated code yet and the DAG builder is a mashup of the DAG builders for
> 'invoke' and 'load', but I've eyeballed the asm it generates (e.g. llc produces
> iload-exception.s for the attached iload-exception.ll) and it looks reasonable.

Hi Peter. All due respect, I don’t think it’s right to introduce new load/store instructions with equivalent semantics to and+icmp+br+load/store.

‘invoke’ is different. It is needed because there is no way for the caller to explicitly check for an exception.

We do introduce intrinsics that encapsulate things like overflow checks. This is done to eliminate control flow edges that tend to inhibit LLVM’s optimizer and instruction selection. But you’re not removing the control flow, so this technique does not apply. Null checks should actually be exposed in IR so general optimizations can remove redundant checks.

Ideally this would just be a machine code pass that can hoist a load/store above a branch and nuke the compare. However, I can see how it’s easier to relate the compare operand to the address arithmetic at IR level.

To do this at IR level, you could introduce a pre-CodeGen pass that converts cmp+br+load/store into a checked load intrinsic. Since this intrinsic only exists for instruction selection, the optimizer doesn’t need to know about it.

The intrinsic would need to be lowered to an MI pseudo-instruction that feels like a load/store to the backend, but is a terminator. During code emission you could grab the address of that pseudo load/store and its resulting branch target to inform the runtime.

I can envision having a front end generate an intrinsic for explicit null checks so that an optimization pass can do something special with these, like reorder null checks and let the runtime recover. But I would not include the load/store semantics in the check. And this is entirely independent from the codegen issue of whether to use trapping loads and stores.

-Andy

> 
> Thanks,
> -- 
> Peter
> <0001-Create-UnwindPoint-as-a-base-class-of-InvokeInst.patch><0002-Add-ILoadInst-and-IStoreInst-classes.patch><0003-Add-SelectionDAG-support-for-iload.patch><iload-exception.ll><iload-exception.s>_______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev