1 Calling Conventions Architecture Review: Caller- and Callee-Saved - PowerPoint PPT Presentation

Prelude Register Allocation II The Sears Tower in Chicago was the world’s Last time tallest building from 1974 until 1998. – Register allocation – Global allocation via graph coloring Today – More register allocation Fun facts – Allocation across procedure calls – The Sears Tower contains enough concrete to build an eight-lane, five- mile-long highway, enough steel to build 50,000 automobiles, and enough telephone wiring to wrap around the world 1.75 times. – At the very top of the building, the maximum wind drift is just one foot. – Six roof-mounted robotic window-washing machines clean all 16,1000 windows on the Sears tower. – My son Kevin visited the “Serious” tower no less than 5 times during the two years we lived in Chicago. CS553 Lecture Register Allocation II 1 CS553 Lecture Register Allocation II 2 Interference Graph Allocators Register Allocation and Procedure Calls Chaitin Briggs Problem – Register values may change across procedure calls – The allocator must be sensitive to this Two approaches – Work within a well-defined calling convention – Use interprocedural allocation (not covering this) CS553 Lecture Register Allocation II 3 CS553 Lecture Register Allocation II 4 1

Calling Conventions Architecture Review: Caller- and Callee-Saved Registers Partition registers into two categories Goals – Caller-saved – Fast calls (pass arguments in registers, minimal register saving/restoring) – Callee-saved – Language-independent – Support debugging, profiling, garbage collection, etc. Caller-saved registers – Caller must save/restore these registers when live across call Complicating Issues – Callee is free to use them – Varargs – Passing/returning aggregates caller Example foo() – Exceptions, non-local returns callee – setjmp() / longjmp() { r caller = 4 goo() is free to goo() save r caller modify r caller { goo() r caller = 99 restore r caller } r caller ? } CS553 Lecture Register Allocation II 5 CS553 Lecture Register Allocation II 6 Architecture Review: Caller- and Callee-Saved Registers Architectures with Callee and Caller Registers Callee-saved registers SPARC – Callee must save/restore these registers when it uses them – hardware-saved %i0-%i7, %o0-%o8 – Caller expects callee to not change them Alpha – 7 callee-saved out of 32 registers Example callee MIPS caller foo() goo() promises – caller-saved: $t0-$t9, $a0-$a3, $v0-$v1 { not to modify goo() – callee-saved: $s0-$s7, $ra, $fp r callee = 4 r callee { PPC save r callee goo() r callee = 99 – 18 callee-saved restore r callee – 14 caller-saved r callee ? } } StarCore EABI – 4 callee-saved – 28 caller-saved CS553 Lecture Register Allocation II 7 CS553 Lecture Register Allocation II 8 2

Register Allocation and Calling Conventions Precolored Nodes Add architectural registers to interference graph Insensitive register allocation – Precolored (mutually interfering) – Save all live caller-saved registers before call; restore after – Not simplifiable – Save all used callee-saved registers at procedure entry; restore at return – Not spillable – Suboptimal foo() Express allocation constraints { – Integers usually can’t be stored in floating point registers t = … A variable that is not live across calls should go in – Some instructions can only store result in certain registers … = t caller-saved registers – Caller-saved and callee-saved registers. . . s = … A variable that is live across multiple calls should f() s3 s2 r1 go in callee-saved registers g() … = s } s4 s1 r2 f3 Sensitive register allocation – Encode calling convention constraints in the IR and interference graph – How? Use precolored nodes floating point floating point integer integer CS553 Lecture Register Allocation II 9 CS553 Lecture Register Allocation II 10 Precolored Nodes and Calling Conventions Example Callee-saved registers r1 , r2 caller-saved – Treat entry as def of all callee-saved registers foo(): r3 callee-saved – Treat exit as use of them all def(r3) – Allocator must “spill” callee-saved registers to use them t1 := r3 a := ... b := ... foo() ... a ... { r1 call goo t1 def(r3) ... b ... Live range of callee-saved registers r3 := t1 use(r3) use(r3) r3 r2 b a } return Caller-saved registers – Variables live across call interfere with all caller-saved registers CS553 Lecture Register Allocation II 11 CS553 Lecture Register Allocation II 12 3

MiniJava compiler, Tiger pg. 198, Ch 11, Ch 12 Tradeoffs Callee-saved registers Callee-saved registers – MipsFrame.java, calleeSaves = { RA, S0, S1, S2, S3, S4, S5, S6, S7}, + Decreases code size: one procedure body may have multiple calls – In MipsFrameRegAlloc.java procEntryExit1 – at entry, copy calleeSaves into temporaries + Small procedures tend to need fewer registers than large ones; callee-save – at end, copy from temporaries back into calleeSaves makes sense because procedure sizes are shrinking – In MipsFrameRegAlloc.java procEntryExit2 − May increase execution time: For long-lived variables, may save and – at entry, could put a source statement with defs, but is not necessary restore registers multiple times, once for each procedure, instead of a – at end, already inserting a sink statement that uses all calleeSaves Caller-saved registers single end-to-end save/restore – MipsFrame.java, callerSaves = {T0, T1, T2, T3, T4, T5, T6, T7, T8, T9, V0, V1}; – Put them all in destlist for the function call, see Codegen.java call to frame.callDefs() The larger “problem” – Put use in sink statement so they are live from call to end of procedure Special registers – We’re making local decisions for policies that require global information – Cannot use for register allocation, might want to make FP (which is S8) special – In MipsFrame.java procEntryExit2 put uses of the special registers in sink stmt Argument registers can also be used as caller-saved registers – Codegen.java pushes all arguments onto the stack right now CS553 Lecture Register Allocation II 13 CS553 Lecture Register Allocation II 14 Concepts Next Time Register allocation and procedure calls Reading – Make sure to read the Briggs paper Calling conventions – Caller- vs. callee-saved registers Lecture – Precoloring – Register Allocation III CS553 Lecture Register Allocation II 15 CS553 Lecture Register Allocation II 16 4

Problem with Callee-Saved Registers Run-time systems ( e.g., setjmp() / longjmp() and debuggers) need to know register values in any stack frame – Caller-saved registers are on stack frame at known location – Callee-saved registers? F4: save r3 F3: r1, r2 caller-saved F2: save r1,r2 r3 callee-saved F1: CS553 Lecture Register Allocation II 17 5