store-merging: Disable string_concatenate mode if start or end aren't byte aligned [PR108498]

  Hi!

The first of the following testcases is miscompiled on powerpc64-linux -O2
-m64 at least, the latter at least on x86_64-linux -m32/-m64.
Since GCC 11 store-merging has a separate string_concatenation mode which
turns stores into setting a MEM_REF from a STRING_CST.
This mode is triggered if at least one of the to be merged stores
is a STRING_CST store and either the first store (to earliest address)
is that STRING_CST store or the first store is 8-bit INTEGER_CST store
and then there are some rules when to turn that mode off or not merge
further stores into it.

The problem with these 2 testcases is that the actual implementation
relies on start/width of the store to be at byte boundaries, as it
simply creates a char array, MEM_REF can be only on byte boundaries
and the char array too, plus obviously STRING_CST as well.
But as can be easily seen in the second testcase, nothing verifies this,
while the first store has to be a STRING_CST (which will be aligned)
or 8-bit INTEGER_CST, that 8-bit INTEGER_CST store could be a bitfield
store, nothing verifies any stores in between whether they actually are
8-bit and aligned, the only major requirement is that all the stores
are consecutive.

For GCC 14 I think we should reconsider this, simply treat STRING_CST
stores during the merging like INTEGER_CST stores and deal with it only
during split_group where we can create multiple parts, this part
would be a normal store, this part would be STRING_CST store, this part
another normal store etc.  But that is quite a lot of work, the following
patch just disables the string_concatenate mode if boundaries aren't byte
aligned in the spot where we disable it if it is too short too.
If that happens, we'll just try to do the merging using normal 1/2/4/8 etc.
byte stores as usually with RMW masking for any bits that shouldn't be
touched or punt if we end up with too many stores compared to the original.

Note, an original STRING_CST store will count as one store in that case,
something we might want to reconsider later too (but, after all, CONSTRUCTOR
stores (aka zeroing) already have the same problem, they can be large and
expensive and we still count them as one store).

Bootstrapped/regtested on x86_64-linux and i686-linux plus tested on the
first testcase with cross to powerpc64-linux, ok for trunk?

2023-01-25  Jakub Jelinek  <jakub@redhat.com>

	PR tree-optimization/108498
	* gimple-ssa-store-merging.cc (class store_operand_info):
	End coment with full stop rather than comma.
	(split_group): Likewise.
	(merged_store_group::apply_stores): Clear string_concatenation if
	start or end aren't on a byte boundary.

	* gcc.c-torture/execute/pr108498-1.c: New test.
	* gcc.c-torture/execute/pr108498-2.c: New test.


	Jakub
  

On Wed, 25 Jan 2023, Jakub Jelinek wrote:

> Hi!
> 
> The first of the following testcases is miscompiled on powerpc64-linux -O2
> -m64 at least, the latter at least on x86_64-linux -m32/-m64.
> Since GCC 11 store-merging has a separate string_concatenation mode which
> turns stores into setting a MEM_REF from a STRING_CST.
> This mode is triggered if at least one of the to be merged stores
> is a STRING_CST store and either the first store (to earliest address)
> is that STRING_CST store or the first store is 8-bit INTEGER_CST store
> and then there are some rules when to turn that mode off or not merge
> further stores into it.
> 
> The problem with these 2 testcases is that the actual implementation
> relies on start/width of the store to be at byte boundaries, as it
> simply creates a char array, MEM_REF can be only on byte boundaries
> and the char array too, plus obviously STRING_CST as well.
> But as can be easily seen in the second testcase, nothing verifies this,
> while the first store has to be a STRING_CST (which will be aligned)
> or 8-bit INTEGER_CST, that 8-bit INTEGER_CST store could be a bitfield
> store, nothing verifies any stores in between whether they actually are
> 8-bit and aligned, the only major requirement is that all the stores
> are consecutive.
> 
> For GCC 14 I think we should reconsider this, simply treat STRING_CST
> stores during the merging like INTEGER_CST stores and deal with it only
> during split_group where we can create multiple parts, this part
> would be a normal store, this part would be STRING_CST store, this part
> another normal store etc.  But that is quite a lot of work, the following
> patch just disables the string_concatenate mode if boundaries aren't byte
> aligned in the spot where we disable it if it is too short too.
> If that happens, we'll just try to do the merging using normal 1/2/4/8 etc.
> byte stores as usually with RMW masking for any bits that shouldn't be
> touched or punt if we end up with too many stores compared to the original.
> 
> Note, an original STRING_CST store will count as one store in that case,
> something we might want to reconsider later too (but, after all, CONSTRUCTOR
> stores (aka zeroing) already have the same problem, they can be large and
> expensive and we still count them as one store).
> 
> Bootstrapped/regtested on x86_64-linux and i686-linux plus tested on the
> first testcase with cross to powerpc64-linux, ok for trunk?

LGTM.

> 2023-01-25  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR tree-optimization/108498
> 	* gimple-ssa-store-merging.cc (class store_operand_info):
> 	End coment with full stop rather than comma.
> 	(split_group): Likewise.
> 	(merged_store_group::apply_stores): Clear string_concatenation if
> 	start or end aren't on a byte boundary.
> 
> 	* gcc.c-torture/execute/pr108498-1.c: New test.
> 	* gcc.c-torture/execute/pr108498-2.c: New test.
> 
> --- gcc/gimple-ssa-store-merging.cc.jj	2023-01-02 09:32:33.811120195 +0100
> +++ gcc/gimple-ssa-store-merging.cc	2023-01-24 16:43:59.700075344 +0100
> @@ -1614,7 +1614,7 @@ namespace {
>     then VAL represents the constant and all the other fields are zero, or
>     a memory load, then VAL represents the reference, BASE_ADDR is non-NULL
>     and the other fields also reflect the memory load, or an SSA name, then
> -   VAL represents the SSA name and all the other fields are zero,  */
> +   VAL represents the SSA name and all the other fields are zero.  */
>  
>  class store_operand_info
>  {
> @@ -2309,6 +2309,10 @@ merged_store_group::apply_stores ()
>    if (buf_size <= MOVE_MAX)
>      string_concatenation = false;
>  
> +  /* String concatenation only works for byte aligned start and end.  */
> +  if (start % BITS_PER_UNIT != 0 || width % BITS_PER_UNIT != 0)
> +    string_concatenation = false;
> +
>    /* Create a power-of-2-sized buffer for native_encode_expr.  */
>    if (!string_concatenation)
>      buf_size = 1 << ceil_log2 (buf_size);
> @@ -3631,7 +3635,7 @@ split_group (merged_store_group *group,
>  
>    /* For bswap framework using sets of stores, all the checking has been done
>       earlier in try_coalesce_bswap and the result always needs to be emitted
> -     as a single store.  Likewise for string concatenation,  */
> +     as a single store.  Likewise for string concatenation.  */
>    if (group->stores[0]->rhs_code == LROTATE_EXPR
>        || group->stores[0]->rhs_code == NOP_EXPR
>        || group->string_concatenation)
> --- gcc/testsuite/gcc.c-torture/execute/pr108498-1.c.jj	2023-01-24 16:53:23.920800393 +0100
> +++ gcc/testsuite/gcc.c-torture/execute/pr108498-1.c	2023-01-24 16:52:37.610479583 +0100
> @@ -0,0 +1,82 @@
> +/* PR tree-optimization/108498 */
> +
> +struct A
> +{
> +  signed char a1;
> +  int a2;
> +};
> +
> +struct B
> +{
> +  struct A b1;
> +  unsigned char b2:1, b3:1, b4:2, b5:1, b6:1, b7[4];
> +};
> +
> +struct C
> +{
> +  unsigned char c1;
> +  char c2;
> +  signed char c3;
> +  unsigned char c4, c5[4], c6:1, c7:1, c8:1, c9:3, c10:1;
> +  struct A c11;
> +  struct B c12[3];
> +};
> +
> +static inline struct C
> +foo (unsigned char a, unsigned b, int c, struct A d,
> +     unsigned e, struct B f, struct B g, struct B h)
> +{
> +  struct C x
> +    = { .c1 = b, .c2 = 0, .c3 = c, .c6 = a, .c4 = e, .c7 = 0,
> +        .c8 = 0, .c9 = 7, .c10 = 0, .c5 = {0, 1, 2, 3}, .c11 = d,
> +        .c12 = {f, g, h} };
> +  return x;
> +}
> +
> +static inline struct A
> +bar (int a, int b)
> +{
> +  struct A x = { .a1 = a, .a2 = b };
> +  return x;
> +}
> +
> +static inline struct B
> +baz (struct A b1)
> +{
> +  struct B x = { .b1 = b1, .b6 = 0, .b5 = 0, .b7 = {0, 1, 2, 3}, .b2 = 0 };
> +  return x;
> +}
> +
> +struct C
> +qux (void)
> +{
> +  const struct B a = baz (bar (0, 0));
> +  struct C b;
> +  struct B c[2];
> +  struct A d = { 0, 1 };
> +  c[0].b1.a1 = 0;
> +  c[0].b1.a2 = 2;
> +  c[1].b1.a1 = 4;
> +  c[1].b1.a2 = 8;
> +  return foo (0, 2, -1, d, 3, c[0], c[1], a);
> +}
> +
> +__attribute__((noipa)) void
> +corge (struct C *x)
> +{
> +  char buf[1024];
> +  __builtin_memset (buf, 0xaa, sizeof (buf));
> +  asm volatile ("" : : "r" (buf));
> +  __builtin_memset (x, 0x55, sizeof (struct C));
> +  asm volatile ("" : : "r" (x));
> +}
> +
> +int
> +main ()
> +{
> +  struct C x;
> +  corge (&x);
> +  x = qux ();
> +  if (x.c6 || x.c9 != 7)
> +    __builtin_abort ();
> +}
> --- gcc/testsuite/gcc.c-torture/execute/pr108498-2.c.jj	2023-01-24 16:53:20.871845097 +0100
> +++ gcc/testsuite/gcc.c-torture/execute/pr108498-2.c	2023-01-24 16:54:02.167239462 +0100
> @@ -0,0 +1,91 @@
> +/* PR tree-optimization/108498 */
> +
> +struct U { char c[16]; };
> +struct V { char c[16]; };
> +struct S { unsigned int a : 3, b : 8, c : 21; struct U d; unsigned int e; struct V f; unsigned int g : 5, h : 27; };
> +struct T { unsigned int a : 16, b : 8, c : 8; struct U d; unsigned int e; struct V f; unsigned int g : 5, h : 27; };
> +
> +__attribute__((noipa)) void
> +foo (struct S *p)
> +{
> +  p->b = 231;
> +  p->c = 42;
> +  p->d = (struct U) { "abcdefghijklmno" };
> +  p->e = 0xdeadbeef;
> +  p->f = (struct V) { "ABCDEFGHIJKLMNO" };
> +}
> +
> +__attribute__((noipa)) void
> +bar (struct S *p)
> +{
> +  p->b = 231;
> +  p->c = 42;
> +  p->d = (struct U) { "abcdefghijklmno" };
> +  p->e = 0xdeadbeef;
> +  p->f = (struct V) { "ABCDEFGHIJKLMNO" };
> +  p->g = 12;
> +}
> +
> +__attribute__((noipa)) void
> +baz (struct T *p)
> +{
> +  p->c = 42;
> +  p->d = (struct U) { "abcdefghijklmno" };
> +  p->e = 0xdeadbeef;
> +  p->f = (struct V) { "ABCDEFGHIJKLMNO" };
> +  p->g = 12;
> +}
> +
> +int
> +main ()
> +{
> +  if (__CHAR_BIT__ != 8 || __SIZEOF_INT__ != 4)
> +    return 0;
> +  struct S s = {};
> +  struct T t = {};
> +  foo (&s);
> +  if (s.a != 0 || s.b != 231 || s.c != 42
> +      || __builtin_memcmp (&s.d.c, "abcdefghijklmno", 16) || s.e != 0xdeadbeef
> +      || __builtin_memcmp (&s.f.c, "ABCDEFGHIJKLMNO", 16) || s.g != 0 || s.h != 0)
> +    __builtin_abort ();
> +  __builtin_memset (&s, 0, sizeof (s));
> +  s.a = 7;
> +  s.g = 31;
> +  s.h = (1U << 27) - 1;
> +  foo (&s);
> +  if (s.a != 7 || s.b != 231 || s.c != 42
> +      || __builtin_memcmp (&s.d.c, "abcdefghijklmno", 16) || s.e != 0xdeadbeef
> +      || __builtin_memcmp (&s.f.c, "ABCDEFGHIJKLMNO", 16) || s.g != 31 || s.h != (1U << 27) - 1)
> +    __builtin_abort ();
> +  __builtin_memset (&s, 0, sizeof (s));
> +  bar (&s);
> +  if (s.a != 0 || s.b != 231 || s.c != 42
> +      || __builtin_memcmp (&s.d.c, "abcdefghijklmno", 16) || s.e != 0xdeadbeef
> +      || __builtin_memcmp (&s.f.c, "ABCDEFGHIJKLMNO", 16) || s.g != 12 || s.h != 0)
> +    __builtin_abort ();
> +  __builtin_memset (&s, 0, sizeof (s));
> +  s.a = 7;
> +  s.g = 31;
> +  s.h = (1U << 27) - 1;
> +  bar (&s);
> +  if (s.a != 7 || s.b != 231 || s.c != 42
> +      || __builtin_memcmp (&s.d.c, "abcdefghijklmno", 16) || s.e != 0xdeadbeef
> +      || __builtin_memcmp (&s.f.c, "ABCDEFGHIJKLMNO", 16) || s.g != 12 || s.h != (1U << 27) - 1)
> +    __builtin_abort ();
> +  baz (&t);
> +  if (t.a != 0 || t.b != 0 || t.c != 42
> +      || __builtin_memcmp (&t.d.c, "abcdefghijklmno", 16) || t.e != 0xdeadbeef
> +      || __builtin_memcmp (&t.f.c, "ABCDEFGHIJKLMNO", 16) || t.g != 12 || t.h != 0)
> +    __builtin_abort ();
> +  __builtin_memset (&s, 0, sizeof (s));
> +  t.a = 7;
> +  t.b = 255;
> +  t.g = 31;
> +  t.h = (1U << 27) - 1;
> +  baz (&t);
> +  if (t.a != 7 || t.b != 255 || t.c != 42
> +      || __builtin_memcmp (&t.d.c, "abcdefghijklmno", 16) || t.e != 0xdeadbeef
> +      || __builtin_memcmp (&t.f.c, "ABCDEFGHIJKLMNO", 16) || t.g != 12 || t.h != (1U << 27) - 1)
> +    __builtin_abort ();
> +  return 0;
> +}
> 
> 	Jakub
> 
>