qbe/ops.h
Roland Paterson-Jones c2ff93e75e Global Value Numbering / Global Code Motion
More or less as proposed in its ninth iteration with the
addition of a gcmmove() functionality to restore coherent
local schedules.

Changes since RFC 8:

Features:

 - generalization of phi 1/0 detection
 - collapse linear jmp chains before GVN; simplifies if-graph
     detection used in 0/non-0 value inference and if-elim...
 - infer 0/non-0 values from dominating blk jnz; eliminates
     redundant cmp eq/ne 0 and associated jnz/blocks, for example
     redundant null pointer checks (hare codebase likes this)
 - remove (emergent) empty if-then-else graphlets between GVN and
     GCM; improves GCM instruction placement, particularly cmps.
 - merge %addr =l add %addr1, N sequences - reduces tmp count,
     register pressure.
 - squash consecutive associative ops with constant args, e.g.
     t1 = add t, N ... t2 = add t2, M -> t2 = add t, N+M

Bug Fixes:

 - remove "cmp eq/ne of non-identical RCon's " in copyref().
   RCon's are not guaranteed to be dedup'ed, and symbols can
   alias.

Codebase:

  - moved some stuff into cfg.c including blkmerge()
  - some refactoring in gvn.c
  - simplification of reassoc.c - always reassoc all cmp ops
      and Kl add %t, N. Better on coremark, smaller codebase.
  - minor simplification of movins() - use vins

Testing - standard QBE, cproc, hare, harec, coremark
          [still have Rust build issues with latest roland]

Benchmark
- coremark is ~15%+ faster than master
- hare "HARETEST_INCLUDE='slow' make check" ~8% faster
    (crypto::sha1::sha1_1gb is biggest obvious win - ~25% faster)

Changes since RFC 7:

Bug fixes:

- remove isbad4gcm() in GVN/GCM - it is unsound due to different state
    at GVN vs GCM time; replace with "reassociation" pass after GCM
- fix intra-blk use-before-def after GCM
- prevent GVN from deduping trapping instructions cos GCM will not
    move them
- remove cmp eq/ne identical arg copy detection for floating point, it
    is not valid for NaN
- fix cges/cged flagged as commutative in ops.h instead of cnes/cned
    respectively; just a typo

Minor features:

- copy detection handles cmp le/lt/ge/gt with identical args
- treat (integer) div/rem by non-zero constant as non-trapping
- eliminate add N/sub N pairs in copy detection
- maintain accurate tmp use in GVN; not strictly necessary but enables
    interim global state sanity checking
- "reassociation" of trivial constant offset load/store addresses, and
    cmp ops with point-of-use in pass after GCM
- normalise commutative op arg order - e.g. op con, tmp -> op tmp, con
    to simplify copy detection and GVN instruction dedup

Codebase:

- split out core copy detection and constant folding (back) out into
    copy.c, fold.c respectively; gvn.c was getting monolithic
- generic support for instruction moving in ins.c - used by GCM and
    reassoc
- new reassociation pass in reassoc.c
- other minor clean-up/refactor

Changes since RFC 6:

- More ext elimination in GVN by examination of def and use bit width
- elimination of redundant and mask by bit width examination
- Incorporation of Song's patch

Changes since RFC 5:

- avoidance of "bad" candidates for GVN/GCM - trivial address offset
    calculations, and comparisons
- more copy detection mostly around boolean values
- allow elimination of unused load, alloc, trapping instructions
- detection of trivial boolean v ? 1 : 0 phi patterns
- bug fix for (removal of) "chg" optimisation in ins recreation - it
    was missing removal of unused instructions in some cases

ifelim() between GVN and GCM; deeper nopunused()
2025-03-14 09:58:37 +01:00

197 lines
10 KiB
C

#ifndef X /* amd64 */
#define X(NMemArgs, SetsZeroFlag, LeavesFlags)
#endif
#ifndef V /* riscv64 */
#define V(Imm)
#endif
#ifndef F
#define F(CanFold, HasId, IdVal, Commutes, Associates, Idemp, IsCmpEq, IsCmpLgte, CmpEqVal, IsPinned)
#endif
#define T(a,b,c,d,e,f,g,h) { \
{[Kw]=K##a, [Kl]=K##b, [Ks]=K##c, [Kd]=K##d}, \
{[Kw]=K##e, [Kl]=K##f, [Ks]=K##g, [Kd]=K##h} \
}
/*********************/
/* PUBLIC OPERATIONS */
/*********************/
/* Arithmetic and Bits */
O(add, T(w,l,s,d, w,l,s,d), F(1,1,0,1,1,0,0,0,0,0)) X(2,1,0) V(1)
O(sub, T(w,l,s,d, w,l,s,d), F(1,1,0,0,0,0,0,0,0,0)) X(2,1,0) V(0)
O(neg, T(w,l,s,d, x,x,x,x), F(1,0,0,0,0,0,0,0,0,0)) X(1,1,0) V(0)
O(div, T(w,l,s,d, w,l,s,d), F(1,1,1,0,0,0,0,0,0,0)) X(0,0,0) V(0)
O(rem, T(w,l,e,e, w,l,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,0) V(0)
O(udiv, T(w,l,e,e, w,l,e,e), F(1,1,1,0,0,0,0,0,0,0)) X(0,0,0) V(0)
O(urem, T(w,l,e,e, w,l,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,0) V(0)
O(mul, T(w,l,s,d, w,l,s,d), F(1,1,1,1,0,0,0,0,0,0)) X(2,0,0) V(0)
O(and, T(w,l,e,e, w,l,e,e), F(1,0,0,1,1,1,0,0,0,0)) X(2,1,0) V(1)
O(or, T(w,l,e,e, w,l,e,e), F(1,1,0,1,1,1,0,0,0,0)) X(2,1,0) V(1)
O(xor, T(w,l,e,e, w,l,e,e), F(1,1,0,1,1,0,0,0,0,0)) X(2,1,0) V(1)
O(sar, T(w,l,e,e, w,w,e,e), F(1,1,0,0,0,0,0,0,0,0)) X(1,1,0) V(1)
O(shr, T(w,l,e,e, w,w,e,e), F(1,1,0,0,0,0,0,0,0,0)) X(1,1,0) V(1)
O(shl, T(w,l,e,e, w,w,e,e), F(1,1,0,0,0,0,0,0,0,0)) X(1,1,0) V(1)
/* Comparisons */
O(ceqw, T(w,w,e,e, w,w,e,e), F(1,1,1,1,0,0,1,0,1,0)) X(0,1,0) V(0)
O(cnew, T(w,w,e,e, w,w,e,e), F(1,1,0,1,0,0,1,0,0,0)) X(0,1,0) V(0)
O(csgew, T(w,w,e,e, w,w,e,e), F(1,0,0,0,0,0,0,1,1,0)) X(0,1,0) V(0)
O(csgtw, T(w,w,e,e, w,w,e,e), F(1,0,0,0,0,0,0,1,0,0)) X(0,1,0) V(0)
O(cslew, T(w,w,e,e, w,w,e,e), F(1,0,0,0,0,0,0,1,1,0)) X(0,1,0) V(0)
O(csltw, T(w,w,e,e, w,w,e,e), F(1,0,0,0,0,0,0,1,0,0)) X(0,1,0) V(1)
O(cugew, T(w,w,e,e, w,w,e,e), F(1,0,0,0,0,0,0,1,1,0)) X(0,1,0) V(0)
O(cugtw, T(w,w,e,e, w,w,e,e), F(1,0,0,0,0,0,0,1,0,0)) X(0,1,0) V(0)
O(culew, T(w,w,e,e, w,w,e,e), F(1,0,0,0,0,0,0,1,1,0)) X(0,1,0) V(0)
O(cultw, T(w,w,e,e, w,w,e,e), F(1,0,0,0,0,0,0,1,0,0)) X(0,1,0) V(1)
O(ceql, T(l,l,e,e, l,l,e,e), F(1,0,0,1,0,0,1,0,1,0)) X(0,1,0) V(0)
O(cnel, T(l,l,e,e, l,l,e,e), F(1,0,0,1,0,0,1,0,0,0)) X(0,1,0) V(0)
O(csgel, T(l,l,e,e, l,l,e,e), F(1,0,0,0,0,0,0,1,1,0)) X(0,1,0) V(0)
O(csgtl, T(l,l,e,e, l,l,e,e), F(1,0,0,0,0,0,0,1,0,0)) X(0,1,0) V(0)
O(cslel, T(l,l,e,e, l,l,e,e), F(1,0,0,0,0,0,0,1,1,0)) X(0,1,0) V(0)
O(csltl, T(l,l,e,e, l,l,e,e), F(1,0,0,0,0,0,0,1,0,0)) X(0,1,0) V(1)
O(cugel, T(l,l,e,e, l,l,e,e), F(1,0,0,0,0,0,0,1,1,0)) X(0,1,0) V(0)
O(cugtl, T(l,l,e,e, l,l,e,e), F(1,0,0,0,0,0,0,1,0,0)) X(0,1,0) V(0)
O(culel, T(l,l,e,e, l,l,e,e), F(1,0,0,0,0,0,0,1,1,0)) X(0,1,0) V(0)
O(cultl, T(l,l,e,e, l,l,e,e), F(1,0,0,0,0,0,0,1,0,0)) X(0,1,0) V(1)
O(ceqs, T(s,s,e,e, s,s,e,e), F(1,0,0,1,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cges, T(s,s,e,e, s,s,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cgts, T(s,s,e,e, s,s,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cles, T(s,s,e,e, s,s,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,1,0) V(0)
O(clts, T(s,s,e,e, s,s,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cnes, T(s,s,e,e, s,s,e,e), F(1,0,0,1,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cos, T(s,s,e,e, s,s,e,e), F(1,0,0,1,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cuos, T(s,s,e,e, s,s,e,e), F(1,0,0,1,0,0,0,0,0,0)) X(0,1,0) V(0)
O(ceqd, T(d,d,e,e, d,d,e,e), F(1,0,0,1,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cged, T(d,d,e,e, d,d,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cgtd, T(d,d,e,e, d,d,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cled, T(d,d,e,e, d,d,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cltd, T(d,d,e,e, d,d,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cned, T(d,d,e,e, d,d,e,e), F(1,0,0,1,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cod, T(d,d,e,e, d,d,e,e), F(1,0,0,1,0,0,0,0,0,0)) X(0,1,0) V(0)
O(cuod, T(d,d,e,e, d,d,e,e), F(1,0,0,1,0,0,0,0,0,0)) X(0,1,0) V(0)
/* Memory */
O(storeb, T(w,e,e,e, m,e,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(storeh, T(w,e,e,e, m,e,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(storew, T(w,e,e,e, m,e,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(storel, T(l,e,e,e, m,e,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(stores, T(s,e,e,e, m,e,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(stored, T(d,e,e,e, m,e,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(loadsb, T(m,m,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(loadub, T(m,m,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(loadsh, T(m,m,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(loaduh, T(m,m,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(loadsw, T(m,m,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(loaduw, T(m,m,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
O(load, T(m,m,m,m, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
/* Extensions and Truncations */
O(extsb, T(w,w,e,e, x,x,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(extub, T(w,w,e,e, x,x,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(extsh, T(w,w,e,e, x,x,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(extuh, T(w,w,e,e, x,x,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(extsw, T(e,w,e,e, e,x,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(extuw, T(e,w,e,e, e,x,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(exts, T(e,e,e,s, e,e,e,x), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(truncd, T(e,e,d,e, e,e,x,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(stosi, T(s,s,e,e, x,x,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(stoui, T(s,s,e,e, x,x,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(dtosi, T(d,d,e,e, x,x,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(dtoui, T(d,d,e,e, x,x,e,e), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(swtof, T(e,e,w,w, e,e,x,x), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(uwtof, T(e,e,w,w, e,e,x,x), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(sltof, T(e,e,l,l, e,e,x,x), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(ultof, T(e,e,l,l, e,e,x,x), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(cast, T(s,d,w,l, x,x,x,x), F(1,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
/* Stack Allocation */
O(alloc4, T(e,l,e,e, e,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(alloc8, T(e,l,e,e, e,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(alloc16, T(e,l,e,e, e,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
/* Variadic Function Helpers */
O(vaarg, T(m,m,m,m, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(vastart, T(m,e,e,e, x,e,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(copy, T(w,l,s,d, x,x,x,x), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
/* Debug */
O(dbgloc, T(w,e,e,e, w,e,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,1) V(0)
/****************************************/
/* INTERNAL OPERATIONS (keep nop first) */
/****************************************/
/* Miscellaneous and Architecture-Specific Operations */
O(nop, T(x,x,x,x, x,x,x,x), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(addr, T(m,m,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(blit0, T(m,e,e,e, m,e,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,1,0) V(0)
O(blit1, T(w,e,e,e, x,e,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,1,0) V(0)
O(swap, T(w,l,s,d, w,l,s,d), F(0,0,0,0,0,0,0,0,0,0)) X(1,0,0) V(0)
O(sign, T(w,l,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,0) V(0)
O(salloc, T(e,l,e,e, e,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,0) V(0)
O(xidiv, T(w,l,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(1,0,0) V(0)
O(xdiv, T(w,l,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(1,0,0) V(0)
O(xcmp, T(w,l,s,d, w,l,s,d), F(0,0,0,0,0,0,0,0,0,0)) X(1,1,0) V(0)
O(xtest, T(w,l,e,e, w,l,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(1,1,0) V(0)
O(acmp, T(w,l,e,e, w,l,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,0) V(0)
O(acmn, T(w,l,e,e, w,l,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,0) V(0)
O(afcmp, T(e,e,s,d, e,e,s,d), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,0) V(0)
O(reqz, T(w,l,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,0) V(0)
O(rnez, T(w,l,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,0) V(0)
/* Arguments, Parameters, and Calls */
O(par, T(x,x,x,x, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(parsb, T(x,x,x,x, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(parub, T(x,x,x,x, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(parsh, T(x,x,x,x, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(paruh, T(x,x,x,x, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(parc, T(e,x,e,e, e,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(pare, T(e,x,e,e, e,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(arg, T(w,l,s,d, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(argsb, T(w,e,e,e, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(argub, T(w,e,e,e, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(argsh, T(w,e,e,e, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(arguh, T(w,e,e,e, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(argc, T(e,x,e,e, e,l,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(arge, T(e,l,e,e, e,x,e,e), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(argv, T(x,x,x,x, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
O(call, T(m,m,m,m, x,x,x,x), F(0,0,0,0,0,0,0,0,0,1)) X(0,0,0) V(0)
/* Flags Setting */
O(flagieq, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagine, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagisge, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagisgt, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagisle, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagislt, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagiuge, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagiugt, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagiule, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagiult, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagfeq, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagfge, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagfgt, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagfle, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagflt, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagfne, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagfo, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
O(flagfuo, T(x,x,e,e, x,x,e,e), F(0,0,0,0,0,0,0,0,0,0)) X(0,0,1) V(0)
#undef T
#undef X
#undef V
#undef O
/*
| column -t -o ' '
*/