134 lines · plain
1//===----------------------------------------------------------------------===//2// Representing sign/zero extension of function results3//===----------------------------------------------------------------------===//4 5Mar 25, 2009 - Initial Revision6 7Most ABIs specify that functions which return small integers do so in a8specific integer GPR. This is an efficient way to go, but raises the question:9if the returned value is smaller than the register, what do the high bits hold?10 11There are three (interesting) possible answers: undefined, zero extended, or12sign extended. The number of bits in question depends on the data-type that13the front-end is referencing (typically i1/i8/i16/i32).14 15Knowing the answer to this is important for two reasons: 1) we want to be able16to implement the ABI correctly. If we need to sign extend the result according17to the ABI, we really really do need to do this to preserve correctness. 2)18this information is often useful for optimization purposes, and we want the19mid-level optimizers to be able to process this (e.g. eliminate redundant20extensions).21 22For example, lets pretend that X86 requires the caller to properly extend the23result of a return (I'm not sure this is the case, but the argument doesn't24depend on this). Given this, we should compile this:25 26int a();27short b() { return a(); }28 29into:30 31_b:32 subl $12, %esp33 call L_a$stub34 addl $12, %esp35 cwtl36 ret37 38An optimization example is that we should be able to eliminate the explicit39sign extension in this example:40 41short y();42int z() {43 return ((int)y() << 16) >> 16;44}45 46_z:47 subl $12, %esp48 call _y49 ;; movswl %ax, %eax -> not needed because eax is already sext'd50 addl $12, %esp51 ret52 53//===----------------------------------------------------------------------===//54// What we have right now.55//===----------------------------------------------------------------------===//56 57Currently, these sorts of things are modelled by compiling a function to return58the small type and a signext/zeroext marker is used. For example, we compile59Z into:60 61define i32 @z() nounwind {62entry:63 %0 = tail call signext i16 (...)* @y() nounwind64 %1 = sext i16 %0 to i3265 ret i32 %166}67 68and b into:69 70define signext i16 @b() nounwind {71entry:72 %0 = tail call i32 (...)* @a() nounwind ; <i32> [#uses=1]73 %retval12 = trunc i32 %0 to i16 ; <i16> [#uses=1]74 ret i16 %retval1275}76 77This has some problems: 1) the actual precise semantics are really poorly78defined (see PR3779). 2) some targets might want the caller to extend, some79might want the callee to extend 3) the mid-level optimizer doesn't know the80size of the GPR, so it doesn't know that %0 is sign extended up to 32-bits 81here, and even if it did, it could not eliminate the sext. 4) the code82generator has historically assumed that the result is extended to i32, which is83a problem on PIC16 (and is also probably wrong on alpha and other 64-bit84targets).85 86//===----------------------------------------------------------------------===//87// The proposal88//===----------------------------------------------------------------------===//89 90I suggest that we have the front-end fully lower out the ABI issues here to91LLVM IR. This makes it 100% explicit what is going on and means that there is92no cause for confusion. For example, the cases above should compile into:93 94define i32 @z() nounwind {95entry:96 %0 = tail call i32 (...)* @y() nounwind97 %1 = trunc i32 %0 to i1698 %2 = sext i16 %1 to i3299 ret i32 %2100}101define i32 @b() nounwind {102entry:103 %0 = tail call i32 (...)* @a() nounwind104 %retval12 = trunc i32 %0 to i16105 %tmp = sext i16 %retval12 to i32106 ret i32 %tmp107}108 109In this model, no functions will return an i1/i8/i16 (and on a x86-64 target110that extends results to i64, no i32). This solves the ambiguity issue, allows us 111to fully describe all possible ABIs, and now allows the optimizers to reason112about and eliminate these extensions.113 114The one thing that is missing is the ability for the front-end and optimizer to115specify/infer the guarantees provided by the ABI to allow other optimizations.116For example, in the y/z case, since y is known to return a sign extended value,117the trunc/sext in z should be eliminable.118 119This can be done by introducing new sext/zext attributes which mean "I know120that the result of the function is sign extended at least N bits. Given this,121and given that it is stuck on the y function, the mid-level optimizer could122easily eliminate the extensions etc with existing functionality.123 124The major disadvantage of doing this sort of thing is that it makes the ABI125lowering stuff even more explicit in the front-end, and that we would like to126eventually move to having the code generator do more of this work. However,127the sad truth of the matter is that this is a) unlikely to happen anytime in128the near future, and b) this is no worse than we have now with the existing129attributes.130 131C compilers fundamentally have to reason about the target in many ways. 132This is ugly and horrible, but a fact of life.133 134