[gcc r12-3072] bswap: Recognize (int) __builtin_bswap64 (arg) idioms or __builtin_bswap?? (arg) & mask [PR86723]

[gcc r12-3072] bswap: Recognize (int) __builtin_bswap64 (arg) idioms or __builtin_bswap?? (arg) & mask [PR86723]

[Jakub Jelinek]

https://gcc.gnu.org/g:b320edc0c29c838b0090c3c9be14187d132f73f2

commit r12-3072-gb320edc0c29c838b0090c3c9be14187d132f73f2
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Mon Aug 23 11:52:06 2021 +0200

    bswap: Recognize (int) __builtin_bswap64 (arg) idioms or __builtin_bswap?? (arg) & mask [PR86723]
    
    The following patch recognizes in the bswap pass (only there for now,
    haven't done it for store merging pass yet) code sequences that can
    be handled by (int32) __builtin_bswap64 (arg), i.e. where we have
    0x05060708 n->n with 64-bit non-memory argument (if it is memory, we
    can just load the 32-bit at 4 bytes into the address and n->n would
    be 0x01020304; and only 64 -> 32 bit, because 64 -> 16 bit or 32 -> 16 bit
    would mean only two bytes in the result and probably not worth it),
    and furthermore the case where we have in the 0x0102030405060708 etc.
    numbers some bytes 0 (i.e. known to contain zeros rather than source bytes),
    as long as we have at least two original bytes in the right
    positions (and no unknown bytes).  This can be handled by
    __builtin_bswap64 (arg) & 0xff0000ffffff00ffULL etc.
    The latter change is the reason why counting the bswap messages doesn't work
    too well in optimize-bswap* tests anymore, while the pass iterates from end
    of basic block towards start, it will often match both the bswap at the end
    and some of the earlier bswaps with some masks (not a problem generally,
    we'll just DCE it away whenever possible).  The pass right now doesn't
    handle __builtin_bswap* calls in the pattern matching (which is the reason
    why it operates backwards), but it uses FOR_EACH_BB_FN (bb, fun) order
    of handling blocks and matched sequences can span multiple blocks, so I was
    worried about cases like:
    void bar (unsigned long long);
    unsigned long long
    foo (unsigned long long value, int x)
    {
      unsigned long long tmp = (((value & 0x00000000000000ffull) << 56)
              | ((value & 0x000000000000ff00ull) << 40)
              | ((value & 0x00000000ff000000ull) << 8));
      if (x)
        bar (tmp);
      return (tmp
              | ((value & 0x000000ff00000000ull) >> 8)
              | ((value & 0x0000ff0000000000ull) >> 24)
              | ((value & 0x0000000000ff0000ull) << 24)
              | ((value & 0x00ff000000000000ull) >> 40)
              | ((value & 0xff00000000000000ull) >> 56));
    }
    but it seems we handle even that fine, while bb2 ending in GIMPLE_COND
    is processed first, we recognize there a __builtin_bswap64 (value) & mask1,
    in the last bb we recognize tmp | (__builtin_bswap64 (value) & mask2) and
    PRE optimizes that into t = __builtin_bswap64 (value); tmp = t & mask1;
    in the first bb and return t; in the last one.
    
    2021-08-23  Jakub Jelinek  <jakub@redhat.com>
    
            PR tree-optimization/86723
            * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
            cast64_to_32 argument, set *cast64_to_32 to false, unless n is
            non-memory permutation of 64-bit src which only has bytes of
            0 or [5..8] and n->range is 4.
            (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
            find_bswap_or_nop_finalize caller, support bswap with some bytes
            zeroed, as long as at least two bytes are not zeroed.
            (bswap_replace): Add mask argument and handle masking of bswap
            result.
            (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
            caller, punt if cast64_to_32 or mask is not all ones.
            (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
            caller, for now punt if cast64_to_32.
    
            * gcc.dg/pr86723.c: New test.
            * gcc.target/i386/pr86723.c: New test.
            * gcc.dg/optimize-bswapdi-1.c: Use -fdump-tree-optimized instead of
            -fdump-tree-bswap and scan for number of __builtin_bswap64 calls.
            * gcc.dg/optimize-bswapdi-2.c: Likewise.
            * gcc.dg/optimize-bswapsi-1.c: Use -fdump-tree-optimized instead of
            -fdump-tree-bswap and scan for number of __builtin_bswap32 calls.
            * gcc.dg/optimize-bswapsi-5.c: Likewise.
            * gcc.dg/optimize-bswapsi-3.c: Likewise.  Expect one __builtin_bswap32
            call instead of zero.

Diff:
---
 gcc/gimple-ssa-store-merging.c            | 100 +++++++++++++++++++++++++-----
 gcc/testsuite/gcc.dg/optimize-bswapdi-1.c |   4 +-
 gcc/testsuite/gcc.dg/optimize-bswapdi-2.c |   4 +-
 gcc/testsuite/gcc.dg/optimize-bswapsi-1.c |   4 +-
 gcc/testsuite/gcc.dg/optimize-bswapsi-3.c |   4 +-
 gcc/testsuite/gcc.dg/optimize-bswapsi-5.c |   4 +-
 gcc/testsuite/gcc.dg/pr86723.c            |  63 +++++++++++++++++++
 gcc/testsuite/gcc.target/i386/pr86723.c   |  52 ++++++++++++++++
 8 files changed, 208 insertions(+), 27 deletions(-)

diff --git a/gcc/gimple-ssa-store-merging.c b/gcc/gimple-ssa-store-merging.c
index 5d3094b5f3c..781c02d4ddb 100644
--- a/gcc/gimple-ssa-store-merging.c
+++ b/gcc/gimple-ssa-store-merging.c
@@ -792,7 +792,7 @@ find_bswap_or_nop_1 (gimple *stmt, struct symbolic_number *n, int limit)
 
 void
 find_bswap_or_nop_finalize (struct symbolic_number *n, uint64_t *cmpxchg,
-			    uint64_t *cmpnop)
+			    uint64_t *cmpnop, bool *cast64_to_32)
 {
   unsigned rsize;
   uint64_t tmpn, mask;
@@ -802,6 +802,7 @@ find_bswap_or_nop_finalize (struct symbolic_number *n, uint64_t *cmpxchg,
      according to the size of the symbolic number before using it.  */
   *cmpxchg = CMPXCHG;
   *cmpnop = CMPNOP;
+  *cast64_to_32 = false;
 
   /* Find real size of result (highest non-zero byte).  */
   if (n->base_addr)
@@ -814,7 +815,27 @@ find_bswap_or_nop_finalize (struct symbolic_number *n, uint64_t *cmpxchg,
   if (n->range < (int) sizeof (int64_t))
     {
       mask = ((uint64_t) 1 << (n->range * BITS_PER_MARKER)) - 1;
-      *cmpxchg >>= (64 / BITS_PER_MARKER - n->range) * BITS_PER_MARKER;
+      if (n->base_addr == NULL
+	  && n->range == 4
+	  && int_size_in_bytes (TREE_TYPE (n->src)) == 8)
+	{
+	  /* If all bytes in n->n are either 0 or in [5..8] range, this
+	     might be a candidate for (unsigned) __builtin_bswap64 (src).
+	     It is not worth it for (unsigned short) __builtin_bswap64 (src)
+	     or (unsigned short) __builtin_bswap32 (src).  */
+	  *cast64_to_32 = true;
+	  for (tmpn = n->n; tmpn; tmpn >>= BITS_PER_MARKER)
+	    if ((tmpn & MARKER_MASK)
+		&& ((tmpn & MARKER_MASK) <= 4 || (tmpn & MARKER_MASK) > 8))
+	      {
+		*cast64_to_32 = false;
+		break;
+	      }
+	}
+      if (*cast64_to_32)
+	*cmpxchg &= mask;
+      else
+	*cmpxchg >>= (64 / BITS_PER_MARKER - n->range) * BITS_PER_MARKER;
       *cmpnop &= mask;
     }
 
@@ -837,6 +858,8 @@ find_bswap_or_nop_finalize (struct symbolic_number *n, uint64_t *cmpxchg,
       n->range = rsize;
     }
 
+  if (*cast64_to_32)
+    n->range = 8;
   n->range *= BITS_PER_UNIT;
 }
 
@@ -849,7 +872,8 @@ find_bswap_or_nop_finalize (struct symbolic_number *n, uint64_t *cmpxchg,
    expression.  */
 
 gimple *
-find_bswap_or_nop (gimple *stmt, struct symbolic_number *n, bool *bswap)
+find_bswap_or_nop (gimple *stmt, struct symbolic_number *n, bool *bswap,
+		   bool *cast64_to_32, uint64_t *mask)
 {
   tree type_size = TYPE_SIZE_UNIT (TREE_TYPE (gimple_get_lhs (stmt)));
   if (!tree_fits_uhwi_p (type_size))
@@ -929,17 +953,30 @@ find_bswap_or_nop (gimple *stmt, struct symbolic_number *n, bool *bswap)
     }
 
   uint64_t cmpxchg, cmpnop;
-  find_bswap_or_nop_finalize (n, &cmpxchg, &cmpnop);
+  find_bswap_or_nop_finalize (n, &cmpxchg, &cmpnop, cast64_to_32);
 
   /* A complete byte swap should make the symbolic number to start with
      the largest digit in the highest order byte. Unchanged symbolic
      number indicates a read with same endianness as target architecture.  */
+  *mask = ~(uint64_t) 0;
   if (n->n == cmpnop)
     *bswap = false;
   else if (n->n == cmpxchg)
     *bswap = true;
   else
-    return NULL;
+    {
+      int set = 0;
+      for (uint64_t msk = MARKER_MASK; msk; msk <<= BITS_PER_MARKER)
+	if ((n->n & msk) == 0)
+	  *mask &= ~msk;
+	else if ((n->n & msk) == (cmpxchg & msk))
+	  set++;
+	else
+	  return NULL;
+      if (set < 2)
+	return NULL;
+      *bswap = true;
+    }
 
   /* Useless bit manipulation performed by code.  */
   if (!n->base_addr && n->n == cmpnop && n->n_ops == 1)
@@ -1024,10 +1061,10 @@ bswap_view_convert (gimple_stmt_iterator *gsi, tree type, tree val)
 tree
 bswap_replace (gimple_stmt_iterator gsi, gimple *ins_stmt, tree fndecl,
 	       tree bswap_type, tree load_type, struct symbolic_number *n,
-	       bool bswap)
+	       bool bswap, uint64_t mask)
 {
   tree src, tmp, tgt = NULL_TREE;
-  gimple *bswap_stmt;
+  gimple *bswap_stmt, *mask_stmt = NULL;
   tree_code conv_code = NOP_EXPR;
 
   gimple *cur_stmt = gsi_stmt (gsi);
@@ -1247,6 +1284,15 @@ bswap_replace (gimple_stmt_iterator gsi, gimple *ins_stmt, tree fndecl,
     tgt = make_ssa_name (bswap_type);
   tmp = tgt;
 
+  if (mask != ~(uint64_t) 0)
+    {
+      tree m = build_int_cst (bswap_type, mask);
+      tmp = make_temp_ssa_name (bswap_type, NULL, "bswapdst");
+      gimple_set_lhs (bswap_stmt, tmp);
+      mask_stmt = gimple_build_assign (tgt, BIT_AND_EXPR, tmp, m);
+      tmp = tgt;
+    }
+
   /* Convert the result if necessary.  */
   if (!useless_type_conversion_p (TREE_TYPE (tgt), bswap_type))
     {
@@ -1260,7 +1306,7 @@ bswap_replace (gimple_stmt_iterator gsi, gimple *ins_stmt, tree fndecl,
       gsi_insert_after (&gsi, convert_stmt, GSI_SAME_STMT);
     }
 
-  gimple_set_lhs (bswap_stmt, tmp);
+  gimple_set_lhs (mask_stmt ? mask_stmt : bswap_stmt, tmp);
 
   if (dump_file)
     {
@@ -1277,11 +1323,17 @@ bswap_replace (gimple_stmt_iterator gsi, gimple *ins_stmt, tree fndecl,
 
   if (cur_stmt)
     {
+      if (mask_stmt)
+	gsi_insert_after (&gsi, mask_stmt, GSI_SAME_STMT);
       gsi_insert_after (&gsi, bswap_stmt, GSI_SAME_STMT);
       gsi_remove (&gsi, true);
     }
   else
-    gsi_insert_before (&gsi, bswap_stmt, GSI_SAME_STMT);
+    {
+      gsi_insert_before (&gsi, bswap_stmt, GSI_SAME_STMT);
+      if (mask_stmt)
+	gsi_insert_before (&gsi, mask_stmt, GSI_SAME_STMT);
+    }
   return tgt;
 }
 
@@ -1341,8 +1393,14 @@ maybe_optimize_vector_constructor (gimple *cur_stmt)
       return false;
     }
 
-  gimple *ins_stmt = find_bswap_or_nop (cur_stmt, &n, &bswap);
-  if (!ins_stmt || n.range != (unsigned HOST_WIDE_INT) sz)
+  bool cast64_to_32;
+  uint64_t mask;
+  gimple *ins_stmt = find_bswap_or_nop (cur_stmt, &n, &bswap,
+					&cast64_to_32, &mask);
+  if (!ins_stmt
+      || n.range != (unsigned HOST_WIDE_INT) sz
+      || cast64_to_32
+      || mask != ~(uint64_t) 0)
     return false;
 
   if (bswap && !fndecl && n.range != 16)
@@ -1351,7 +1409,7 @@ maybe_optimize_vector_constructor (gimple *cur_stmt)
   memset (&nop_stats, 0, sizeof (nop_stats));
   memset (&bswap_stats, 0, sizeof (bswap_stats));
   return bswap_replace (gsi_for_stmt (cur_stmt), ins_stmt, fndecl,
-			bswap_type, load_type, &n, bswap) != NULL_TREE;
+			bswap_type, load_type, &n, bswap, mask) != NULL_TREE;
 }
 
 /* Find manual byte swap implementations as well as load in a given
@@ -1405,7 +1463,8 @@ pass_optimize_bswap::execute (function *fun)
 	  tree fndecl = NULL_TREE, bswap_type = NULL_TREE, load_type;
 	  enum tree_code code;
 	  struct symbolic_number n;
-	  bool bswap;
+	  bool bswap, cast64_to_32;
+	  uint64_t mask;
 
 	  /* This gsi_prev (&gsi) is not part of the for loop because cur_stmt
 	     might be moved to a different basic block by bswap_replace and gsi
@@ -1442,7 +1501,8 @@ pass_optimize_bswap::execute (function *fun)
 	      continue;
 	    }
 
-	  ins_stmt = find_bswap_or_nop (cur_stmt, &n, &bswap);
+	  ins_stmt = find_bswap_or_nop (cur_stmt, &n, &bswap,
+					&cast64_to_32, &mask);
 
 	  if (!ins_stmt)
 	    continue;
@@ -1479,7 +1539,7 @@ pass_optimize_bswap::execute (function *fun)
 	    continue;
 
 	  if (bswap_replace (gsi_for_stmt (cur_stmt), ins_stmt, fndecl,
-			     bswap_type, load_type, &n, bswap))
+			     bswap_type, load_type, &n, bswap, mask))
 	    changed = true;
 	}
     }
@@ -2820,7 +2880,8 @@ imm_store_chain_info::try_coalesce_bswap (merged_store_group *merged_store,
     }
 
   uint64_t cmpxchg, cmpnop;
-  find_bswap_or_nop_finalize (&n, &cmpxchg, &cmpnop);
+  bool cast64_to_32;
+  find_bswap_or_nop_finalize (&n, &cmpxchg, &cmpnop, &cast64_to_32);
 
   /* A complete byte swap should make the symbolic number to start with
      the largest digit in the highest order byte.  Unchanged symbolic
@@ -2828,6 +2889,10 @@ imm_store_chain_info::try_coalesce_bswap (merged_store_group *merged_store,
   if (n.n != cmpnop && n.n != cmpxchg)
     return false;
 
+  /* For now.  */
+  if (cast64_to_32)
+    return false;
+
   if (n.base_addr == NULL_TREE && !is_gimple_val (n.src))
     return false;
 
@@ -4161,7 +4226,8 @@ imm_store_chain_info::output_merged_store (merged_store_group *group)
 	    n->vuse = gimple_vuse (ins_stmt);
 	}
       bswap_res = bswap_replace (gsi_start (seq), ins_stmt, fndecl,
-				 bswap_type, load_type, n, bswap);
+				 bswap_type, load_type, n, bswap,
+				 ~(uint64_t) 0);
       gcc_assert (bswap_res);
     }
 
diff --git a/gcc/testsuite/gcc.dg/optimize-bswapdi-1.c b/gcc/testsuite/gcc.dg/optimize-bswapdi-1.c
index a4a3a792375..56a2071b5ec 100644
--- a/gcc/testsuite/gcc.dg/optimize-bswapdi-1.c
+++ b/gcc/testsuite/gcc.dg/optimize-bswapdi-1.c
@@ -1,7 +1,7 @@
 /* { dg-do compile } */
 /* { dg-require-effective-target bswap } */
 /* { dg-require-effective-target stdint_types } */
-/* { dg-options "-O2 -fdump-tree-bswap" } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
 /* { dg-additional-options "-mzarch" { target s390*-*-* } } */
 
 #include <stdint.h>
@@ -58,4 +58,4 @@ swap64_c (uint64_t x)
 }
 
 
-/* { dg-final { scan-tree-dump-times "64 bit bswap implementation found at" 3 "bswap" } } */
+/* { dg-final { scan-tree-dump-times "= __builtin_bswap64 \\\(" 3 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/optimize-bswapdi-2.c b/gcc/testsuite/gcc.dg/optimize-bswapdi-2.c
index 89b251fe461..c6d96049e10 100644
--- a/gcc/testsuite/gcc.dg/optimize-bswapdi-2.c
+++ b/gcc/testsuite/gcc.dg/optimize-bswapdi-2.c
@@ -1,7 +1,7 @@
 /* { dg-do compile } */
 /* { dg-require-effective-target bswap } */
 /* { dg-require-effective-target stdint_types } */
-/* { dg-options "-O2 -fdump-tree-bswap" } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
 /* { dg-additional-options "-mzarch" { target s390*-*-* } } */
 
 #include <stdint.h>
@@ -23,4 +23,4 @@ swap64_c (uint64_t x)
 }
 
 
-/* { dg-final { scan-tree-dump-times "64 bit bswap implementation found at" 1 "bswap" } } */
+/* { dg-final { scan-tree-dump-times "= __builtin_bswap64 \\\(" 1 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/optimize-bswapsi-1.c b/gcc/testsuite/gcc.dg/optimize-bswapsi-1.c
index c403d0498bb..2d24f43e5e6 100644
--- a/gcc/testsuite/gcc.dg/optimize-bswapsi-1.c
+++ b/gcc/testsuite/gcc.dg/optimize-bswapsi-1.c
@@ -1,7 +1,7 @@
 /* { dg-do compile } */
 /* { dg-require-effective-target bswap } */
 /* { dg-require-effective-target stdint_types } */
-/* { dg-options "-O2 -fdump-tree-bswap" } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
 /* { dg-additional-options "-march=z900" { target s390*-*-* } } */
 
 #include <stdint.h>
@@ -89,4 +89,4 @@ swap32_f (unsigned in)
   return in;
 }
 
-/* { dg-final { scan-tree-dump-times "32 bit bswap implementation found at" 6 "bswap" } } */
+/* { dg-final { scan-tree-dump-times "= __builtin_bswap32 \\\(" 6 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/optimize-bswapsi-3.c b/gcc/testsuite/gcc.dg/optimize-bswapsi-3.c
index 9418a83dfe3..7cd1b406a87 100644
--- a/gcc/testsuite/gcc.dg/optimize-bswapsi-3.c
+++ b/gcc/testsuite/gcc.dg/optimize-bswapsi-3.c
@@ -1,7 +1,7 @@
 /* { dg-do compile } */
 /* { dg-require-effective-target bswap } */
 /* { dg-require-effective-target stdint_types } */
-/* { dg-options "-O2 -fdump-tree-bswap" } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
 /* { dg-additional-options "-march=z900" { target s390-*-* } } */
 
 typedef int SItype __attribute__ ((mode (SI)));
@@ -20,4 +20,4 @@ swap32 (SItype in)
 	 | (((in >> 24) & 0xFF) << 0);
 }
 
-/* { dg-final { scan-tree-dump-not "32 bit bswap implementation found at" "bswap" } } */
+/* { dg-final { scan-tree-dump-times "= __builtin_bswap32 \\\(" 1 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/optimize-bswapsi-5.c b/gcc/testsuite/gcc.dg/optimize-bswapsi-5.c
index b4d8b9a8dd9..91a5284fa81 100644
--- a/gcc/testsuite/gcc.dg/optimize-bswapsi-5.c
+++ b/gcc/testsuite/gcc.dg/optimize-bswapsi-5.c
@@ -1,6 +1,6 @@
 /* { dg-do compile } */
 /* { dg-require-effective-target bswap } */
-/* { dg-options "-O2 -fdump-tree-bswap -fno-inline-functions" } */
+/* { dg-options "-O2 -fdump-tree-optimized -fno-inline-functions" } */
 /* { dg-additional-options "-march=z900" { target s390-*-* } } */
 
 struct L { unsigned int l[2]; };
@@ -28,4 +28,4 @@ bar (double a, struct L *p)
   foo (a, p);
 }
 
-/* { dg-final { scan-tree-dump-times "32 bit bswap implementation found at" 2 "bswap" } } */
+/* { dg-final { scan-tree-dump-times "= __builtin_bswap32 \\\(" 2 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/pr86723.c b/gcc/testsuite/gcc.dg/pr86723.c
new file mode 100644
index 00000000000..e3fd6b9d4a1
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr86723.c
@@ -0,0 +1,63 @@
+/* PR tree-optimization/86723 */
+/* { dg-do run { target { ilp32 || lp64 } } } */
+/* { dg-options "-O2" } */
+
+__attribute__((noipa)) int
+foo (unsigned long long value)
+{
+  return (((value & 0x00000000000000ffull) << 56)
+	  | ((value & 0x000000000000ff00ull) << 40)
+	  | ((value & 0x0000000000ff0000ull) << 24)
+	  | ((value & 0x00000000ff000000ull) << 8)
+	  | ((value & 0x000000ff00000000ull) >> 8)
+	  | ((value & 0x0000ff0000000000ull) >> 24)
+	  | ((value & 0x00ff000000000000ull) >> 40)
+	  | ((value & 0xff00000000000000ull) >> 56));
+}
+
+__attribute__((noipa)) int
+bar (unsigned long long value)
+{
+  return (((value & 0x000000ff00000000ull) >> 8)
+	  | ((value & 0x0000ff0000000000ull) >> 24)
+	  | ((value & 0x00ff000000000000ull) >> 40)
+	  | ((value & 0xff00000000000000ull) >> 56));
+}
+
+__attribute__((noipa)) unsigned long long
+baz (unsigned long long value)
+{
+  return (((value & 0x00000000000000ffull) << 56)
+	  | ((value & 0x000000000000ff00ull) << 40)
+	  | ((value & 0x00000000ff000000ull) << 8)
+	  | ((value & 0x000000ff00000000ull) >> 8)
+	  | ((value & 0x0000ff0000000000ull) >> 24)
+	  | ((value & 0xff00000000000000ull) >> 56));
+}
+
+__attribute__((noipa)) unsigned int
+qux (unsigned int value)
+{
+  return (((value & 0x000000ff) << 24)
+	  | ((value & 0x00ff0000) >> 8)
+	  | ((value & 0xff000000) >> 24));
+}
+
+__attribute__((noipa)) unsigned int
+corge (unsigned int value)
+{
+  return (((value & 0x000000ff) << 24)
+	  | ((value & 0xff000000) >> 24));
+}
+
+int
+main ()
+{
+  if (foo (0x0102030405060708ull) != 0x04030201
+      || bar (0x0102030405060708ull) != 0x04030201
+      || baz (0x0102030405060708ull) != 0x0807000504030001ull
+      || qux (0x01020304) != 0x04000201
+      || corge (0x01020304) != 0x04000001)
+    __builtin_abort ();
+  return 0;
+}
diff --git a/gcc/testsuite/gcc.target/i386/pr86723.c b/gcc/testsuite/gcc.target/i386/pr86723.c
new file mode 100644
index 00000000000..b61d750c01a
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr86723.c
@@ -0,0 +1,52 @@
+/* PR tree-optimization/86723 */
+/* { dg-do compile { target lp64 } } */
+/* { dg-options "-O2" } */
+/* { dg-final { scan-assembler-times "\tbswap\t" 5 } } */
+
+int
+foo (unsigned long long value)
+{
+  return (((value & 0x00000000000000ffull) << 56)
+	  | ((value & 0x000000000000ff00ull) << 40)
+	  | ((value & 0x0000000000ff0000ull) << 24)
+	  | ((value & 0x00000000ff000000ull) << 8)
+	  | ((value & 0x000000ff00000000ull) >> 8)
+	  | ((value & 0x0000ff0000000000ull) >> 24)
+	  | ((value & 0x00ff000000000000ull) >> 40)
+	  | ((value & 0xff00000000000000ull) >> 56));
+}
+
+int
+bar (unsigned long long value)
+{
+  return (((value & 0x000000ff00000000ull) >> 8)
+	  | ((value & 0x0000ff0000000000ull) >> 24)
+	  | ((value & 0x00ff000000000000ull) >> 40)
+	  | ((value & 0xff00000000000000ull) >> 56));
+}
+
+unsigned long long
+baz (unsigned long long value)
+{
+  return (((value & 0x00000000000000ffull) << 56)
+	  | ((value & 0x000000000000ff00ull) << 40)
+	  | ((value & 0x00000000ff000000ull) << 8)
+	  | ((value & 0x000000ff00000000ull) >> 8)
+	  | ((value & 0x0000ff0000000000ull) >> 24)
+	  | ((value & 0xff00000000000000ull) >> 56));
+}
+
+unsigned int
+qux (unsigned int value)
+{
+  return (((value & 0x000000ff) << 24)
+	  | ((value & 0x00ff0000) >> 8)
+	  | ((value & 0xff000000) >> 24));
+}
+
+unsigned int
+corge (unsigned int value)
+{
+  return (((value & 0x000000ff) << 24)
+	  | ((value & 0xff000000) >> 24));
+}