On Tue, Jun 16, 2026 at 5:22 PM <[email protected]> wrote:
>
> From: Kyrylo Tkachov <[email protected]>
>
> The backward jump threader tail-duplicates an if-convertible "diamond" chain
> when the same predicate controls several diamonds.  The astcenc "none" backend
> lowers a per-lane "r = mask ? hi : lo" select, where one comparison mask 
> drives
> two selects, to such a chain:
>
>     m = (a > b) ? -1 : 0;     // a comparison mask, kept as PHI<0, -1>
>     ...
>     if (m == -1) ...          // select 1 under the mask
>     ...
>     if (m == -1) ...          // select 2 under the same mask
>
> The threader sees the later tests are redundant on the path through the first
> (m, hence m == -1, is known) and tail-duplicates the intervening diamonds to
> fold the branch away, turning an N-diamond chain into a 2^N decision tree that
> reconverges at a single multi-predecessor merge.  phiopt and RTL if-conversion
> can no longer recognise the per-select conditional moves, so they come out as
> compare+branch.  On 731.astcenc_r (SPEC CPU2026, aarch64) this is the dominant
> gap in compute_quantized_weights_for_decimation, where -O3 emits the hot 
> select
> kernel with ~34 branches / 12 fcsel instead of ~5 / 24.  This pass gives ~19%
> improvement on that benchmark overall.
>
> Rather than restrict the threader (which is doing a legitimate transform), 
> this
> removes the underlying redundancy with two value/control simplifications that
> improve the GIMPLE in their own right; a beneficial side effect is that the
> threader no longer finds correlated conditions to tail-duplicate.  They run 
> in a
> small pass, pass_merge_diamonds, placed before pass_thread_jumps_full and 
> gated
> on flag_thread_jumps, so it runs exactly when the backward threader does 
> (under
> -fno-thread-jumps the diamonds stay clean and if-conversion handles them):
>
>   1. De-indirection: a test "m == C" / "m != C" whose operand m is a PHI of 
> two
>      integer constants (a mask "cmp ? cst : cst" from a clean diamond) is
>      rewritten to test the mask's own controlling condition cmp.  We already
>      perform this for a single-use mask; here it is exposed across a multi-use
>      mask PHI.  The mask PHI then becomes dead.

So this is (cmp ? CST1 : CST2) ==/!= CST, we should attempt to simplify
that with a match pattern like

(simplify
 (ne (cond^ @0 INTEGER_CST@1 INTEGER_CST@2) INTEGER_CST@3)
 (if (@1 != @3 && @2 == @3)
  @0)))

?  (fix that to actually be correct, of course)


>   2. Merge: two if-convertible diamonds controlled by an identical condition,
>      the first dominating the second, are merged.  The live PHIs of the first
>      are recomputed in the second under its (identical) branch, and the first
>      diamond's now-dead edge is removed with remove_edge_and_dominated_blocks,
>      which keeps the dominator information valid for the rest of the pass.
>
> De-indirection runs as a phase that completes before any merge, so that all
> same-condition diamonds are recognised before one of them is folded away; the
> merge phase then drains each chain of same-condition diamonds, visiting the
> if-convertible diamond heads collected during de-indirection.
>
> The result is a chain of distinct-condition, two-PHI diamonds that the 
> threader
> leaves alone and RTL if-conversion lowers to conditional moves.  Whether each
> select becomes a conditional move stays an RTL if-conversion cost-model
> decision, so code that benefits from branches is unaffected.
>
> Bootstrapped and regtested on aarch64-unknown-linux-gnu.
>
> Signed-off-by: Kyrylo Tkachov <[email protected]>
>
> gcc/ChangeLog:
>
>         PR tree-optimization/125672
>         * tree-ssa-ifcombine.cc (ifcvt_diamond_join): New function.
>         (deindirect_mask_cond): New function.
>         (all_uses_dominated_by): New function.
>         (merge_cond_diamond): New function.
>         (class pass_merge_diamonds): New pass.
>         (make_pass_merge_diamonds): New function.
>         * tree-pass.h (make_pass_merge_diamonds): Declare.
>         * passes.def: Run pass_merge_diamonds before pass_thread_jumps_full.
>
> gcc/testsuite/ChangeLog:
>
>         PR tree-optimization/125672
>         * gcc.dg/tree-ssa/pr125672.c: New test.
>         * gcc.dg/tree-ssa/pr125672-2.c: New test.
>         * gcc.dg/tree-ssa/pr125672-3.c: New test.
>         * gcc.dg/tree-ssa/pr125672-4.c: New test.
>         * gcc.dg/tree-ssa/pr125672-5.c: New test.
>         * gcc.dg/tree-ssa/pr125672-6.c: New test.
> ---
>  gcc/passes.def                             |   1 +
>  gcc/testsuite/gcc.dg/tree-ssa/pr125672-2.c |  32 ++
>  gcc/testsuite/gcc.dg/tree-ssa/pr125672-3.c |  15 +
>  gcc/testsuite/gcc.dg/tree-ssa/pr125672-4.c |  22 ++
>  gcc/testsuite/gcc.dg/tree-ssa/pr125672-5.c |  37 ++
>  gcc/testsuite/gcc.dg/tree-ssa/pr125672-6.c |  26 ++
>  gcc/testsuite/gcc.dg/tree-ssa/pr125672.c   |  26 ++
>  gcc/tree-pass.h                            |   1 +
>  gcc/tree-ssa-ifcombine.cc                  | 400 +++++++++++++++++++++
>  9 files changed, 560 insertions(+)
>  create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr125672-2.c
>  create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr125672-3.c
>  create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr125672-4.c
>  create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr125672-5.c
>  create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr125672-6.c
>  create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr125672.c
>
> diff --git a/gcc/passes.def b/gcc/passes.def
> index 1fc867fae51..3c4f0629e78 100644
> --- a/gcc/passes.def
> +++ b/gcc/passes.def
> @@ -233,6 +233,7 @@ along with GCC; see the file COPYING3.  If not see
>        NEXT_PASS (pass_return_slot);
>        NEXT_PASS (pass_fre, true /* may_iterate */);
>        NEXT_PASS (pass_merge_phi);
> +      NEXT_PASS (pass_merge_diamonds);

I wonder whether we want to do this after VRP to give us a chance
to eliminate redundant compares after inlining first?  There's also likely
a bit of DCE/DSE left on the plate at this point which might inhibit
this transform?  I realize this might require pushing back thread-jumps-full,
but maybe that's not too bad?

>        NEXT_PASS (pass_thread_jumps_full, /*first=*/true);
>        NEXT_PASS (pass_vrp, false /* final_p*/);
>        NEXT_PASS (pass_array_bounds);
> diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr125672-2.c 
> b/gcc/testsuite/gcc.dg/tree-ssa/pr125672-2.c
> new file mode 100644
> index 00000000000..59c7762a3b3
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/tree-ssa/pr125672-2.c
> @@ -0,0 +1,32 @@
> +/* Execution test: merging same-condition if-convertible diamonds must
> +   preserve semantics.  PR tree-optimization/125672.  */
> +/* { dg-do run } */
> +/* { dg-options "-O2" } */
> +
> +__attribute__((noipa)) static void
> +kern (int n, const int *a, const int *b, int *o)
> +{
> +  for (int i = 0; i < n; i++)
> +    {
> +      int m0 = (a[i] > b[i]) ? -1 : 0;        /* a comparison mask */
> +      int m1 = (a[i] < b[i]) ? -1 : 0;        /* a second mask */
> +      o[4*i + 0] = (m0 == -1) ? a[i] : b[i];  /* select under mask 0 */
> +      o[4*i + 1] = (m1 == -1) ? a[i] : b[i];  /* select under mask 1 */
> +      o[4*i + 2] = (m0 == -1) ? b[i] : a[i];  /* mask 0 again */
> +      o[4*i + 3] = (m1 == -1) ? b[i] : a[i];  /* mask 1 again */
> +    }
> +}
> +
> +int
> +main (void)
> +{
> +  int a[3] = { 5, 2, 9 };
> +  int b[3] = { 3, 7, 9 };
> +  int o[12];
> +  int exp[12] = { 5, 3, 3, 5,   7, 2, 2, 7,   9, 9, 9, 9 };
> +  kern (3, a, b, o);
> +  for (int i = 0; i < 12; i++)
> +    if (o[i] != exp[i])
> +      __builtin_abort ();
> +  return 0;
> +}
> diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr125672-3.c 
> b/gcc/testsuite/gcc.dg/tree-ssa/pr125672-3.c
> new file mode 100644
> index 00000000000..958d0d2aaec
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/tree-ssa/pr125672-3.c
> @@ -0,0 +1,15 @@
> +/* A single comparison mask feeding selects is already simplified by phiopt,
> +   so pass_merge_diamonds is not needed and must not fire here; it only acts
> +   when interleaved masks defeat phiopt.  PR tree-optimization/125672.  */
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -fdump-tree-mergediam-details" } */
> +
> +void
> +f (float a, float b, int x, int y, int u, int v, int *o)
> +{
> +  int m = (a > b) ? -1 : 0;
> +  o[0] = (m == -1) ? x : y;
> +  o[1] = (m == -1) ? u : v;
> +}
> +
> +/* { dg-final { scan-tree-dump-not "de-indirecting mask test" "mergediam" } 
> } */
> diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr125672-4.c 
> b/gcc/testsuite/gcc.dg/tree-ssa/pr125672-4.c
> new file mode 100644
> index 00000000000..3c02943ccc8
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/tree-ssa/pr125672-4.c
> @@ -0,0 +1,22 @@
> +/* PR tree-optimization/125672: two comparison masks are interleaved (which
> +   defeats phiopt), and each drives two selects.  pass_merge_diamonds
> +   de-indirects the four mask tests and merges each pair of same-condition
> +   diamonds, so both pairs are merged.  */
> +
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -fdump-tree-mergediam-details" } */
> +
> +void
> +f (float a, float b, float c, float d,
> +   int x0, int y0, int u0, int v0, int x1, int y1, int u1, int v1, int *o)
> +{
> +  int m0 = (a > b) ? -1 : 0;
> +  int m1 = (c > d) ? -1 : 0;
> +  o[0] = (m0 == -1) ? x0 : y0;   /* mask 0 */
> +  o[1] = (m1 == -1) ? x1 : y1;   /* mask 1, interleaved */
> +  o[2] = (m0 == -1) ? u0 : v0;   /* mask 0 again */
> +  o[3] = (m1 == -1) ? u1 : v1;   /* mask 1 again */
> +}
> +
> +/* Both same-condition pairs are merged.  */
> +/* { dg-final { scan-tree-dump-times "merging if-convertible diamond" 2 
> "mergediam" } } */
> diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr125672-5.c 
> b/gcc/testsuite/gcc.dg/tree-ssa/pr125672-5.c
> new file mode 100644
> index 00000000000..15100f54cff
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/tree-ssa/pr125672-5.c
> @@ -0,0 +1,37 @@
> +/* PR tree-optimization/125672: a chain of same-condition diamonds collapses 
> via
> +   successive merges in a single pass.  One mask drives three selects (a 
> second
> +   mask is interleaved so phiopt cannot pre-simplify); after de-indirection 
> the
> +   three mask-0 diamonds form a dominator chain that is drained by repeated
> +   merges.  Each merge removes the dominating diamond's now-dead edge with
> +   remove_edge_and_dominated_blocks, which keeps the dominator tree valid so 
> the
> +   next merge in the chain still sees correct dominator information.  This 
> is also an execution test: a wrong dominator
> +   update would miscompile the recomputed selects.  */
> +
> +/* { dg-do run } */
> +/* { dg-options "-O2 -fdump-tree-mergediam-details" } */
> +
> +__attribute__((noipa)) static void
> +kern (float a, float b, float c, float d,
> +      int x0, int y0, int u0, int v0, int p0, int q0, int x1, int y1, int *o)
> +{
> +  int m0 = (a > b) ? -1 : 0;
> +  int m1 = (c > d) ? -1 : 0;
> +  o[0] = (m0 == -1) ? x0 : y0;   /* mask 0 */
> +  o[1] = (m1 == -1) ? x1 : y1;   /* mask 1, interleaved (defeats phiopt) */
> +  o[2] = (m0 == -1) ? u0 : v0;   /* mask 0 */
> +  o[3] = (m0 == -1) ? p0 : q0;   /* mask 0 (three deep) */
> +}
> +
> +int
> +main (void)
> +{
> +  int o[4];
> +  kern (2, 1, 1, 2, 10, 11, 12, 13, 14, 15, 20, 21, o);
> +  /* a>b true -> m0; c>d false -> !m1.  */
> +  if (o[0] != 10 || o[1] != 21 || o[2] != 12 || o[3] != 14)
> +    __builtin_abort ();
> +  return 0;
> +}
> +
> +/* The three mask-0 diamonds collapse into one via two successive merges.  */
> +/* { dg-final { scan-tree-dump-times "merging if-convertible diamond" 2 
> "mergediam" } } */
> diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr125672-6.c 
> b/gcc/testsuite/gcc.dg/tree-ssa/pr125672-6.c
> new file mode 100644
> index 00000000000..4be14d16b7d
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/tree-ssa/pr125672-6.c
> @@ -0,0 +1,26 @@
> +/* PR tree-optimization/125672: pass_merge_diamonds de-indirects every mask 
> test
> +   before it merges any diamond, so that all same-condition diamonds are
> +   recognised before one of them is folded away.  Two masks drive the selects
> +   (m1 is interleaved to defeat phiopt), and the result R of the first m0 
> select
> +   is consumed only at the end (o[3] = R), placing its use below the later m0
> +   select.  With the phases kept separate the pass performs three merges; a
> +   single walk that merges as it de-indirects commits to a merge before the
> +   remaining mask tests are de-indirected and performs only two.  */
> +
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -fdump-tree-mergediam-details" } */
> +
> +void
> +f (float a, float b, float c, float d,
> +   int p, int q, int x0, int y0, int x1, int y1, int u1, int v1, int *o)
> +{
> +  int m0 = (a > b) ? -1 : 0;
> +  int m1 = (c > d) ? -1 : 0;
> +  int r = (m0 == -1) ? p : q;     /* m0 select; result reused at o[3] */
> +  o[0] = (m1 == -1) ? x1 : y1;    /* m1 select, interleaved (defeats phiopt) 
> */
> +  o[1] = (m0 == -1) ? x0 : y0;    /* m0 select */
> +  o[2] = (m1 == -1) ? u1 : v1;    /* m1 select */
> +  o[3] = r;                       /* late use of the first m0 select */
> +}
> +
> +/* { dg-final { scan-tree-dump-times "merging if-convertible diamond" 3 
> "mergediam" } } */
> diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr125672.c 
> b/gcc/testsuite/gcc.dg/tree-ssa/pr125672.c
> new file mode 100644
> index 00000000000..87acf94a559
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/tree-ssa/pr125672.c
> @@ -0,0 +1,26 @@
> +/* PR tree-optimization/125672: when one comparison mask drives several
> +   per-lane selects, the backward jump threader tail-duplicates the
> +   if-convertible diamond chain.  pass_merge_diamonds removes the redundancy
> +   before threading: it de-indirects the mask comparisons and merges the
> +   diamonds controlled by an identical condition.  */
> +
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -fdump-tree-mergediam-details" } */
> +
> +void
> +f (float a, float b, float c, float d,
> +   int x0, int y0, int x1, int y1,
> +   int u0, int v0, int u1, int v1, int *o)
> +{
> +  int m0 = (a > b) ? -1 : 0;
> +  int m1 = (c > d) ? -1 : 0;
> +  o[0] = (m0 == -1) ? x0 : y0;
> +  o[1] = (m1 == -1) ? x1 : y1;
> +  o[2] = (m0 == -1) ? u0 : v0;
> +  o[3] = (m1 == -1) ? u1 : v1;
> +}
> +
> +/* The mask comparisons are de-indirected to test the comparison directly... 
>  */
> +/* { dg-final { scan-tree-dump "de-indirecting mask test" "mergediam" } } */
> +/* ... and the two diamonds per mask are merged into one.  */
> +/* { dg-final { scan-tree-dump "merging if-convertible diamond" "mergediam" 
> } } */
> diff --git a/gcc/tree-pass.h b/gcc/tree-pass.h
> index b3c97658a8f..1625c1766e0 100644
> --- a/gcc/tree-pass.h
> +++ b/gcc/tree-pass.h
> @@ -464,6 +464,7 @@ extern gimple_opt_pass *make_pass_phiopt (gcc::context 
> *ctxt);
>  extern gimple_opt_pass *make_pass_forwprop (gcc::context *ctxt);
>  extern gimple_opt_pass *make_pass_phiprop (gcc::context *ctxt);
>  extern gimple_opt_pass *make_pass_tree_ifcombine (gcc::context *ctxt);
> +extern gimple_opt_pass *make_pass_merge_diamonds (gcc::context *ctxt);
>  extern gimple_opt_pass *make_pass_dse (gcc::context *ctxt);
>  extern gimple_opt_pass *make_pass_nrv (gcc::context *ctxt);
>  extern gimple_opt_pass *make_pass_rename_ssa_copies (gcc::context *ctxt);
> diff --git a/gcc/tree-ssa-ifcombine.cc b/gcc/tree-ssa-ifcombine.cc
> index 6489abf3a06..7b33862a06a 100644
> --- a/gcc/tree-ssa-ifcombine.cc
> +++ b/gcc/tree-ssa-ifcombine.cc
> @@ -1358,6 +1358,406 @@ tree_ssa_ifcombine_bb (basic_block inner_cond_bb)
>
>  /* Main entry for the tree if-conversion pass.  */
>
> +/* Merge if-convertible "diamonds" controlled by an identical condition, and
> +   de-indirect mask comparisons that feed them, before the backward jump
> +   threaders run.
> +
> +   When a comparison mask drives several selects, as in
> +
> +       m = (a > b) ? -1 : 0; // comparison mask, kept as a PHI of 0 and -1
> +       if (m == -1) ... // a select under the mask
> +       if (m == -1) ... // another select under the same mask
> +
> +   the per-lane selects lower to a chain of if-convertible diamonds in which
> +   the same predicate controls several diamonds.  The backward jump threader
> +   then notices the later tests are redundant on the path through the earlier
> +   one (m, and hence "m == -1", is known there) and tail-duplicates the
> +   intervening diamonds to fold the branch away, turning an N-diamond chain
> +   into a 2^N decision tree that reconverges at a single multi-predecessor
> +   merge.  RTL if-conversion can no longer recover the per-select conditional
> +   moves and emits compare+branch instead.
> +
> +   This pass removes the redundancy at its two sources, so there is nothing 
> for
> +   threading to duplicate.
> +
> +   De-indirection: a test "m == C" / "m != C" whose operand m is a PHI of two
> +   integer constants (a mask "cmp ? cst : cst") is rewritten to test the 
> mask's
> +   own controlling condition cmp; the mask PHI then becomes dead.
> +
> +   Merge: two if-convertible diamonds controlled by an identical condition,
> +   where the first dominates the second, are merged: the PHIs of the first 
> are
> +   recomputed in the second under its identical branch and the first branch 
> is
> +   folded away, so both selects live under one branch.
> +
> +   The result is a chain of distinct-condition diamonds that the threader
> +   leaves alone and RTL if-conversion lowers to conditional moves; whether 
> each
> +   select actually becomes a conditional move stays an RTL if-conversion
> +   cost-model decision.  */
> +
> +/* If COND_BB heads an if-then-else whose two arms are empty (no side 
> effects)
> +   and reconverge at a single join block, return that join block and set
> +   *E_TRUE and *E_FALSE to the edges entering the join on the condition's 
> true
> +   and false side (possibly through one empty forwarder).  Otherwise NULL.  
> */
> +
> +static basic_block
> +ifcvt_diamond_join (basic_block cond_bb, edge *e_true, edge *e_false)
> +{
> +  basic_block tb = NULL, fb = NULL;
> +  if (!recognize_if_then_else (cond_bb, &tb, &fb))
> +    return NULL;
> +  edge te = find_edge (cond_bb, tb);
> +  edge fe = find_edge (cond_bb, fb);
> +
> +  edge tj_e = te, fj_e = fe;
> +  basic_block tj = tb, fj = fb;
> +  if (single_pred_p (tb) && single_succ_p (tb) && empty_block_p (tb))
> +    {
> +      tj = single_succ (tb);
> +      tj_e = single_succ_edge (tb);
> +    }
> +  if (single_pred_p (fb) && single_succ_p (fb) && empty_block_p (fb))
> +    {
> +      fj = single_succ (fb);
> +      fj_e = single_succ_edge (fb);
> +    }
> +
> +  /* One arm may reach the join directly (a triangle), with the other through
> +     an empty forwarder, or both arms may forward to the join.  */
> +  basic_block join;
> +  if (tb == fj && tb != cond_bb)
> +    {
> +      join = tb;
> +      *e_true = te;
> +      *e_false = fj_e;
> +    }
> +  else if (fb == tj && fb != cond_bb)
> +    {
> +      join = fb;
> +      *e_true = tj_e;
> +      *e_false = fe;
> +    }
> +  else if (tj == fj && tj != cond_bb && tb != fb)
> +    {
> +      join = tj;
> +      *e_true = tj_e;
> +      *e_false = fj_e;
> +    }
> +  else
> +    return NULL;
> +
> +  if (EDGE_COUNT (join->preds) != 2
> +      || (*e_true)->src == (*e_false)->src)
> +    return NULL;
> +  if (tb != join && !bb_no_side_effects_p (tb))
> +    return NULL;
> +  if (fb != join && !bb_no_side_effects_p (fb))
> +    return NULL;
> +  return join;
> +}
> +
> +/* Rewrite the GIMPLE_COND ending COND_BB if it tests "m == C" / "m != C"
> +   where m is a PHI of two integer constants controlled by a clean diamond:
> +   replace the test with that diamond's controlling condition.  This removes
> +   the mask indirection so the predicate is exposed directly.  Returns true 
> if
> +   the condition was rewritten, which may swap COND_BB's outgoing true/false
> +   edges.  Requires valid dominator info.  */
> +
> +static bool
> +deindirect_mask_cond (basic_block cond_bb)
> +{
> +  gcond *gc = safe_dyn_cast <gcond *> (*gsi_last_bb (cond_bb));
> +  if (!gc)
> +    return false;
> +  enum tree_code cc = gimple_cond_code (gc);
> +  if (cc != EQ_EXPR && cc != NE_EXPR)
> +    return false;
> +
> +  /* By canonicalization the SSA_NAME is the LHS and the INTEGER_CST the RHS
> +     (tree_swap_operands_p), so no operand swap is needed.  */
> +  tree m = gimple_cond_lhs (gc);
> +  tree c = gimple_cond_rhs (gc);
> +  if (TREE_CODE (m) != SSA_NAME || TREE_CODE (c) != INTEGER_CST)
> +    return false;
> +
> +  gphi *mphi = dyn_cast <gphi *> (SSA_NAME_DEF_STMT (m));
> +  if (!mphi || gimple_phi_num_args (mphi) != 2)
> +    return false;
> +  basic_block jm = gimple_bb (mphi);
> +
> +  basic_block cm = get_immediate_dominator (CDI_DOMINATORS, jm);
> +  if (!cm)
> +    return false;
> +  edge et, ef;
> +  if (ifcvt_diamond_join (cm, &et, &ef) != jm)
> +    return false;
> +
> +  tree vt = PHI_ARG_DEF_FROM_EDGE (mphi, et);
> +  tree vf = PHI_ARG_DEF_FROM_EDGE (mphi, ef);
> +  if (TREE_CODE (vt) != INTEGER_CST || TREE_CODE (vf) != INTEGER_CST)
> +    return false;
> +
> +  bool c_eq_vt = tree_int_cst_equal (vt, c);
> +  bool c_eq_vf = tree_int_cst_equal (vf, c);
> +  /* C must equal exactly one of the two mask values, otherwise the test does
> +     not reduce to the mask's controlling condition.  */
> +  if (c_eq_vt == c_eq_vf)
> +    return false;
> +  /* Whether "m == c" holds when cm's true edge (ET) is taken.  */
> +  bool when_true = c_eq_vt;
> +  if (cc == NE_EXPR)
> +    when_true = !when_true;
> +
> +  /* Copy the mask's controlling condition into GC verbatim.  If "m == c"
> +     instead holds on cm's false edge, swap GC's outgoing true/false edges
> +     rather than inverting the comparison (inversion would fail for an
> +     unordered FP comparison).  */
> +  gcond *ctrl = as_a <gcond *> (*gsi_last_bb (cm));
> +  gimple_cond_set_code (gc, gimple_cond_code (ctrl));
> +  gimple_cond_set_lhs (gc, gimple_cond_lhs (ctrl));
> +  gimple_cond_set_rhs (gc, gimple_cond_rhs (ctrl));
> +  update_stmt (gc);
> +  if (!when_true)
> +    {
> +      EDGE_SUCC (cond_bb, 0)->flags ^= (EDGE_TRUE_VALUE | EDGE_FALSE_VALUE);
> +      EDGE_SUCC (cond_bb, 1)->flags ^= (EDGE_TRUE_VALUE | EDGE_FALSE_VALUE);
> +    }
> +
> +  if (dump_file && (dump_flags & TDF_DETAILS))
> +    fprintf (dump_file,
> +            "de-indirecting mask test in bb%d via diamond bb%d\n",
> +            cond_bb->index, cm->index);
> +  return true;
> +}
> +
> +/* Return true if every use of NAME is dominated by block BY (treating a use
> +   in a PHI as occurring on the corresponding predecessor edge).  */
> +
> +static bool
> +all_uses_dominated_by (tree name, basic_block by)
> +{
> +  imm_use_iterator it;
> +  use_operand_p use_p;
> +  gimple *use_stmt;
> +  FOR_EACH_IMM_USE_FAST (use_p, it, name)
> +    {
> +      use_stmt = USE_STMT (use_p);
> +      if (is_gimple_debug (use_stmt))
> +       continue;
> +      basic_block ub;
> +      if (is_a <gphi *> (use_stmt))
> +       ub = phi_arg_edge_from_use (use_p)->src;
> +      else
> +       ub = gimple_bb (use_stmt);
> +      if (!dominated_by_p (CDI_DOMINATORS, ub, by))
> +       return false;
> +    }
> +  return true;
> +}
> +
> +/* Try to merge the same-condition diamond headed by B2 into a dominating
> +   diamond.  Return true if a merge was performed.  */
> +
> +static bool
> +merge_cond_diamond (basic_block b2)
> +{
> +  edge t2, f2;
> +  basic_block join2 = ifcvt_diamond_join (b2, &t2, &f2);
> +  if (!join2)
> +    return false;
> +  gcond *c2 = as_a <gcond *> (*gsi_last_bb (b2));
> +
> +  for (basic_block b1 = get_immediate_dominator (CDI_DOMINATORS, b2);
> +       b1; b1 = get_immediate_dominator (CDI_DOMINATORS, b1))
> +    {
> +      gcond *c1 = safe_dyn_cast <gcond *> (*gsi_last_bb (b1));
> +      if (!c1)
> +       continue;
> +      if (gimple_cond_code (c1) != gimple_cond_code (c2)
> +         || !operand_equal_p (gimple_cond_lhs (c1), gimple_cond_lhs (c2), 0)
> +         || !operand_equal_p (gimple_cond_rhs (c1), gimple_cond_rhs (c2), 0))
> +       continue;
> +
> +      edge t1, f1;
> +      basic_block join1 = ifcvt_diamond_join (b1, &t1, &f1);
> +      if (!join1 || join1 == join2 || join1 == b2)
> +       continue;
> +
> +      /* Each live data PHI of JOIN1 must be recomputable at JOIN2.  A 
> virtual
> +        PHI is left in place: the diamond arms are side-effect free (checked 
> by
> +        ifcvt_diamond_join), so its two arguments are equal and it 
> degenerates
> +        once B1's branch is folded and cfg cleanup merges the edge.  */
> +      bool ok = true, any_phi = false;
> +      for (gphi_iterator gpi = gsi_start_phis (join1);
> +          !gsi_end_p (gpi); gsi_next (&gpi))
> +       {
> +         gphi *phi = gpi.phi ();
> +         tree res = gimple_phi_result (phi);
> +         if (virtual_operand_p (res))
> +           continue;
> +         any_phi = true;
> +         if (!has_zero_uses (res) && !all_uses_dominated_by (res, join2))
> +           {
> +             ok = false;
> +             break;
> +           }
> +       }
> +      if (!ok || !any_phi)
> +       continue;
> +
> +      if (dump_file && (dump_flags & TDF_DETAILS))
> +       fprintf (dump_file,
> +                "merging if-convertible diamond bb%d into same-condition "
> +                "diamond bb%d\n", b1->index, b2->index);
> +
> +      /* Move each live PHI of JOIN1 to JOIN2 under B2's branch; drop dead
> +        ones.  */
> +      for (gphi_iterator gpi = gsi_start_phis (join1); !gsi_end_p (gpi);)
> +       {
> +         gphi *phi = gpi.phi ();
> +         tree res = gimple_phi_result (phi);
> +         if (virtual_operand_p (res))
> +           {
> +             gsi_next (&gpi);
> +             continue;
> +           }
> +         if (!has_zero_uses (res))
> +           {
> +             tree tv = PHI_ARG_DEF_FROM_EDGE (phi, t1);
> +             tree fv = PHI_ARG_DEF_FROM_EDGE (phi, f1);
> +             location_t tl = gimple_phi_arg_location_from_edge (phi, t1);
> +             location_t fl = gimple_phi_arg_location_from_edge (phi, f1);
> +             tree nres = copy_ssa_name (res);
> +             gphi *nphi = create_phi_node (nres, join2);
> +             add_phi_arg (nphi, tv, t2, tl);
> +             add_phi_arg (nphi, fv, f2, fl);
> +
> +             imm_use_iterator it;
> +             use_operand_p use_p;
> +             gimple *use_stmt;
> +             FOR_EACH_IMM_USE_STMT (use_stmt, it, res)
> +               {
> +                 /* Non-debug uses are all dominated by JOIN2 (checked 
> above).
> +                    A debug use may sit in a block JOIN2 does not dominate; 
> the
> +                    recomputed value is not available there, so reset it 
> rather
> +                    than create invalid SSA.  */
> +                 if (is_gimple_debug (use_stmt)
> +                     && !dominated_by_p (CDI_DOMINATORS,
> +                                         gimple_bb (use_stmt), join2))
> +                   {
> +                     gimple_debug_bind_reset_value (use_stmt);
> +                     update_stmt (use_stmt);
> +                     continue;
> +                   }
> +                 FOR_EACH_IMM_USE_ON_STMT (use_p, it)
> +                   SET_USE (use_p, nres);
> +                 update_stmt (use_stmt);
> +               }
> +           }
> +         remove_phi_node (&gpi, true);
> +       }
> +
> +      /* Remove B1's now-redundant branch.  Both arms reach JOIN1, which no
> +        longer holds a PHI distinguishing them, so the branch is dead.  Keep 
> B1's
> +        true edge as a fallthrough and delete the false edge with any blocks 
> that
> +        become unreachable; remove_edge_and_dominated_blocks keeps the 
> dominator
> +        info valid for the rest of the walk.  cfg cleanup merges the empty 
> arm.  */
> +      edge b1t = EDGE_SUCC (b1, 0);
> +      edge b1f = EDGE_SUCC (b1, 1);
> +      if (b1f->flags & EDGE_TRUE_VALUE)
> +       std::swap (b1t, b1f);
> +      gimple_stmt_iterator gsic1 = gsi_last_bb (b1);
> +      gsi_remove (&gsic1, true);
> +      remove_edge_and_dominated_blocks (b1f);
> +      b1t->flags &= ~(EDGE_TRUE_VALUE | EDGE_FALSE_VALUE);
> +      b1t->flags |= EDGE_FALLTHRU;
> +      b1t->probability = profile_probability::always ();
> +
> +      return true;
> +    }
> +  return false;
> +}
> +
> +namespace {
> +
> +const pass_data pass_data_merge_diamonds =
> +{
> +  GIMPLE_PASS, /* type */
> +  "mergediam", /* name */
> +  OPTGROUP_NONE, /* optinfo_flags */
> +  TV_TREE_IFCOMBINE, /* tv_id */
> +  PROP_cfg | PROP_ssa, /* properties_required */
> +  0, /* properties_provided */
> +  0, /* properties_destroyed */
> +  0, /* todo_flags_start */
> +  0, /* todo_flags_finish */
> +};
> +
> +class pass_merge_diamonds : public gimple_opt_pass
> +{
> +public:
> +  pass_merge_diamonds (gcc::context *ctxt)
> +    : gimple_opt_pass (pass_data_merge_diamonds, ctxt)
> +  {}
> +
> +  /* opt_pass methods: */
> +  /* This pass only removes redundancy that the backward jump threader would
> +     otherwise tail-duplicate, so it is only useful when that threader runs. 
>  */
> +  bool gate (function *) final override
> +  { return flag_thread_jumps && !optimize_debug; }
> +  unsigned int execute (function *) final override;
> +
> +}; // class pass_merge_diamonds
> +
> +unsigned int
> +pass_merge_diamonds::execute (function *fun)
> +{
> +  bool any = false;
> +  calculate_dominance_info (CDI_DOMINATORS);
> +
> +  /* Visit blocks so a block's single predecessor comes first (as in phiopt
> +     and ifcombine).  This is ~dominator order, so the two phases below need 
> no
> +     iteration: in phase 1 an outer mask test is de-indirected before an 
> inner
> +     test that depends on it, and in phase 2 merge_cond_diamond sees a
> +     dominating diamond's controlling condition already de-indirected.  The
> +     phases are kept separate: all conditions must be de-indirected before
> +     any merge so that same-condition diamonds are recognised as such.  */
> +  basic_block *order = single_pred_before_succ_order ();
> +  int n = n_basic_blocks_for_fn (fun) - NUM_FIXED_BLOCKS;
> +
> +  /* Phase 1: de-indirect mask comparisons, collecting the if-convertible
> +     diamond heads to revisit for merging (a worklist for phase 2, so phase 2
> +     does not re-walk every block).  */
> +  auto_vec<basic_block> heads;
> +  for (int i = 0; i < n; i++)
> +    {
> +      if (deindirect_mask_cond (order[i]))
> +       any = true;
> +      edge et, ef;
> +      if (ifcvt_diamond_join (order[i], &et, &ef))
> +       heads.safe_push (order[i]);
> +    }
> +
> +  /* Phase 2: merge diamonds controlled by identical conditions, visiting 
> only
> +     the collected heads; the inner loop drains a whole chain of
> +     same-condition diamonds into B2.  Kept separate from phase 1: every
> +     condition must be de-indirected before any merge folds a diamond away.  
> */
> +  for (basic_block b2 : heads)
> +    while (merge_cond_diamond (b2))
> +      any = true;
> +
> +  free (order);
> +  return any ? TODO_cleanup_cfg : 0;
> +}
> +
> +} // anon namespace
> +
> +gimple_opt_pass *
> +make_pass_merge_diamonds (gcc::context *ctxt)
> +{
> +  return new pass_merge_diamonds (ctxt);
> +}
> +
> +
>  namespace {
>
>  const pass_data pass_data_tree_ifcombine =
> --
> 2.50.1 (Apple Git-155)
>

Reply via email to