On Thu, 19 Aug 2021, Qing Zhao wrote:
>
>
> > On Aug 19, 2021, at 4:00 AM, Richard Biener <[email protected]> wrote:
> >
> > On Wed, 18 Aug 2021, Qing Zhao wrote:
> >
> >>
> >>
> >>> On Aug 18, 2021, at 2:15 AM, Richard Biener <[email protected]> wrote:
> >>>
> >>> On Tue, 17 Aug 2021, Qing Zhao wrote:
> >>>
> >>>>
> >>>>
> >>>>> On Aug 17, 2021, at 9:50 AM, Qing Zhao via Gcc-patches
> >>>>> <[email protected]> wrote:
> >>>>>
> >>>>>
> >>>>>
> >>>>>> On Aug 17, 2021, at 3:29 AM, Richard Biener <[email protected]> wrote:
> >>>>>>
> >>>>>> On Mon, 16 Aug 2021, Qing Zhao wrote:
> >>>>>>
> >>>>>>> My current code for expand_DEFERRED_INIT is like the following, could
> >>>>>>> you check and see whether there is any issue for it:
> >>>>>>>
> >>>>>>> #define INIT_PATTERN_VALUE 0xFE
> >>>>>>> static void
> >>>>>>> expand_DEFERRED_INIT (internal_fn, gcall *stmt)
> >>>>>>> {
> >>>>>>> tree lhs = gimple_call_lhs (stmt);
> >>>>>>> tree var_size = gimple_call_arg (stmt, 0);
> >>>>>>> enum auto_init_type init_type
> >>>>>>> = (enum auto_init_type) TREE_INT_CST_LOW (gimple_call_arg (stmt, 1));
> >>>>>>> bool is_vla = (bool) TREE_INT_CST_LOW (gimple_call_arg (stmt, 2));
> >>>>>>>
> >>>>>>> tree var_type = TREE_TYPE (lhs);
> >>>>>>> gcc_assert (init_type > AUTO_INIT_UNINITIALIZED);
> >>>>>>>
> >>>>>>> if (is_vla || (!use_register_for_decl (lhs)))
> >>>>>>> {
> >>>>>>> if (TREE_CODE (lhs) == SSA_NAME)
> >>>>>>> lhs = SSA_NAME_VAR (lhs);
> >>>>>>
> >>>>>> this should not be necessary (in fact you shouldn't see a SSA_NAME
> >>>>>> here, if you do then using SSA_NAME_VAR is wrong)
> >>>>> You mean during RTL expansion phase, all SSA_NAMEs are gone already?
> >>>>
> >>>> Actually, the lhs could be SSA_NAME here,
> >>>>
> >>>> Breakpoint 1, expand_DEFERRED_INIT (stmt=0x7fffe96ae348) at
> >>>> ../../latest-gcc/gcc/internal-fn.c:3021
> >>>> 3021 mark_addressable (lhs);
> >>>> (gdb) call debug_tree(lhs)
> >>>> <ssa_name 0x7fffe9584e58
> >>>> type <real_type 0x7fffe959b2a0 float sizes-gimplified SF
> >>>> size <integer_cst 0x7fffe9579f48 constant 32>
> >>>> unit-size <integer_cst 0x7fffe9579f60 constant 4>
> >>>> align:32 warn_if_not_align:0 symtab:0 alias-set 2 canonical-type
> >>>> 0x7fffe959b2a0 precision:32
> >>>> pointer_to_this <pointer_type 0x7fffe959b7e0>>
> >>>> visited var <var_decl 0x7ffff7ff7bd0 temp1>
> >>>> def_stmt temp1_5 = .DEFERRED_INIT (4, 2, 0, &"temp1"[0]);
> >>>> version:5>
> >>>>
> >>>> when I deleted:
> >>>>
> >>>> if (TREE_CODE (lhs) == SSA_NAME
> >>>> lhs = SSA_NAME_VAR (lhs);
> >>>
> >>> but then using SSA_NAME_VAR is broken. I suspect use_register_for_decl
> >>> isn't the correct thing to look at. I think we need to look at what
> >>> the LHS expanded to if it is a SSA_VAR_P (that includes SSA names
> >>> but also plain DECLs but not what we get from VLAs where we'd see
> >>> *ptr). So sth like
> >>>
> >>> bool reg_lhs;
> >>> if (SSA_VAR_P (lhs))
> >>> {
> >>> rtx tem = expand_expr (lhs, NULL_RTX, VOIDmode, EXPAND_WRITE);
> >>> reg_lhs = !MEM_P (tem);
> >>> /* If not MEM_P reg_lhs should be REG_P or SUBREG_P (but maybe
> >>> also CONCAT or lowpart...?) */
> >>> }
> >>> else
> >>> {
> >>> gcc_assert (is_vla);
> >>> reg_lhs = false;
> >>> }
> >>>
> >>> if (!reg_lhs)
> >>> memset path
> >>> else
> >>> expand_assignment path
> >>
> >> After making the following change:
> >>
> >> + bool reg_lhs = true;
> >>
> >> tree var_type = TREE_TYPE (lhs);
> >> gcc_assert (init_type > AUTO_INIT_UNINITIALIZED);
> >>
> >> - if (is_vla || (!use_register_for_decl (lhs)))
> >> + if (SSA_VAR_P (lhs))
> >> + {
> >> + rtx tem = expand_expr (lhs, NULL_RTX, VOIDmode, EXPAND_WRITE);
> >> + reg_lhs = !MEM_P (tem);
> >> + }
> >> + else
> >> + {
> >> + gcc_assert (is_vla);
> >> + reg_lhs = false;
> >> + }
> >> +
> >> + if (!reg_lhs)
> >> {
> >>
> >> I got exactly the same internal error that failed at expr.c:
> >>
> >> 8436 /* We must have made progress. */
> >> 8437 gcc_assert (inner != exp);
> >>
> >>
> >> Looks like for the following code:
> >>
> >> 3026 if (!reg_lhs)
> >> 3027 {
> >> 3028 /* If this is a VLA or the variable is not in register,
> >> 3029 expand to a memset to initialize it. */
> >> 3030 mark_addressable (lhs);
> >> 3031 tree var_addr = build_fold_addr_expr (lhs);
> >> 3032
> >> 3033 tree value = (init_type == AUTO_INIT_PATTERN) ?
> >> 3034 build_int_cst (integer_type_node,
> >> 3035 INIT_PATTERN_VALUE) :
> >> 3036 integer_zero_node;
> >> 3037 tree m_call = build_call_expr (builtin_decl_implicit
> >> (BUILT_IN_MEMSET),
> >> 3038 3, var_addr, value, var_size);
> >> 3039 /* Expand this memset call. */
> >> 3040 expand_builtin_memset (m_call, NULL_RTX, TYPE_MODE (var_type));
> >> 3041 }
> >>
> >> At line 3030, “lhs” could be a SSA_NAME.
> >>
> >> My questions are:
> >>
> >> 1. Could the routine “mark_addressable” and “build_fold_addr_expr” be
> >> applied on SSA_NAME?
> >
> > No.
> >
> >> 2. Could the routine “expand_builtin_memset” be applied on the memset call
> >> whose “DEST” is
> >> an address expression on SSA_NAME?
> >
> > No.
> >
> >> 3. Within “expand_DEFERRED_INIT”, can I call “expand_builtin_memset” to
> >> expand .DEFERRED_INIT?
> >
> > Well, not with "invalid" GENERIC I fear (address of a SSA name).
> >
> >> I suspect that one of the above 3 might be the issue, but not sure which
> >> one?
> >
> > All of the above ;) So while reg_lhs is now precise as to how the
> > variable will end up (the SSA name will end up as a stack variable in this
> > case, for whatever reason), expansion via memcpy only works when
> > working on the RTL representation. The usual "workaround" (ugh)
> > is to use make_tree (), so in the !reg_lhs path you'd do
> >
> > /* Get a new GENERIC representation for the RTL. That's necesary
> > in case LHS is an SSA name. */
> > lhs = make_tree (TREE_TYPE (lhs), tem);
>
> This resolved the issue.
>
> Another question,
>
> Previously, I used
>
> if (TREE_CODE (lhs) == SSA_NAME)
> lhs = SSA_NAME_VAR (lhs);
>
> To resolve this issue. The purpose looks like the same as “make_tree”, just
> get an generic tree for the LHS.
>
> Why you said using SSA_NAME_VAR is broken? Is it because SSA_NAME_VAR will
> not always return a valid TREE?
Because it's simply the wrong entity - I have no idea why that even
worked. Ah, cfgexpand associates it with some DECL_RTL for the
benefit of debug info. But it's still wrong.
> I should use as following
>
>
> If (TREE_CODE (lhs) == SSA_NAME) && SSA_NAME_VAR (lhs))
> Lhs = SSA_NAME_VAR (lhs)
>
> ?
No. A SSA_NAME_VAR can have multiple SSA_NAMEs (obviously) and
they do not necessarily have to be allocated to the same variable
partition - that is, there's no 1:1 relationship between SSA_NAME
and stack slot or (pseudo) register. You want to initialize the
storage associated with the SSA_NAME in the .DEFERRED_INIT call,
not some other storage.
> >
> > alternatively you could maybe do
> >
> > if (DECL_P (lhs))
> > {
> > + rtx tem = expand_expr (lhs, NULL_RTX, VOIDmode, EXPAND_WRITE);
> > + reg_lhs = !MEM_P (tem);
> > }
> > else if (TREE_CODE (lhs) == SSA_NAME)
> > reg_lhs = true;
> > else
> > reg_lhs = false;
> >
> > thus treat SSA names as register storage always (even if it will end
> > up on the stack).
>
> My question here, for a complicate structure SSA_NAME, will expanding through
> memset better than expand_asssignment?
It depends. In the end I'd consider it a missed-optimization bug on
the side that generates worse code - but I do expect cases will exist
for both. Clearly memset will be worse when dealing with register
initialization (thus the !MEM_P check) and I expect memset to be OK
for stack where member-wise init esp. with non-zero might turn up
worse code.
Richard.
> Qing
> >
> > Richard.
> >
> >> Thanks a lot.
> >>
> >> Qing
> >>
> >>
> >>
> >>> bool reg_lhs;
> >>> if (SSA_VAR_P (lhs))
> >>> {
> >>> rtx tem = expand_expr (lhs, NULL_RTX, VOIDmode, EXPAND_WRITE);
> >>> reg_lhs = !MEM_P (tem);
> >>> /* If not MEM_P reg_lhs should be REG_P or SUBREG_P (but maybe
> >>> also CONCAT or lowpart...?) */
> >>> }
> >>> else
> >>> {
> >>> gcc_assert (is_vla);
> >>> reg_lhs = false;
> >>> }
> >>
> >>
> >>>
> >>>> Many testing cases failed with internal compiler error:
> >>>>
> >>>> /home/opc/Work/GCC/latest-gcc/gcc/testsuite/c-c++-common/auto-init-3.c:9:9:
> >>>> internal compiler error: in expand_expr_addr_expr_1, at expr.c:8437
> >>>> 0xe237aa expand_expr_addr_expr_1
> >>>> ../../latest-gcc/gcc/expr.c:8437
> >>>> 0xe24059 expand_expr_addr_expr
> >>>> ../../latest-gcc/gcc/expr.c:8525
> >>>> 0xe32b56 expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
> >>>> expand_modifier, rtx_def**, bool)
> >>>> ../../latest-gcc/gcc/expr.c:11741
> >>>> 0xe2da52 expand_expr_real_1(tree_node*, rtx_def*, machine_mode,
> >>>> expand_modifier, rtx_def**, bool)
> >>>> ../../latest-gcc/gcc/expr.c:10777
> >>>> 0xe24706 expand_expr_real(tree_node*, rtx_def*, machine_mode,
> >>>> expand_modifier, rtx_def**, bool)
> >>>> ../../latest-gcc/gcc/expr.c:8713
> >>>> 0xc13f15 expand_expr
> >>>> ../../latest-gcc/gcc/expr.h:301
> >>>> 0xc17acb get_memory_rtx
> >>>> ../../latest-gcc/gcc/builtins.c:1370
> >>>> 0xc2223d expand_builtin_memset_args
> >>>> ../../latest-gcc/gcc/builtins.c:4102
> >>>> 0xc21a20 expand_builtin_memset(tree_node*, rtx_def*, machine_mode)
> >>>> ../../latest-gcc/gcc/builtins.c:3886
> >>>> 0xfb5c85 expand_DEFERRED_INIT
> >>>> ../../latest-gcc/gcc/internal-fn.c:3031
> >>>>
> >>>>
> >>>> So, did I do anything wrong?
> >>>>
> >>>> Qing
> >>>
> >>> --
> >>> Richard Biener <[email protected]>
> >>> SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg,
> >>> Germany; GF: Felix Imendörffer; HRB 36809 (AG Nuernberg)
> >>
> >>
> >
> > --
> > Richard Biener <[email protected]>
> > SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg,
> > Germany; GF: Felix Imendörffer; HRB 36809 (AG Nuernberg)
>
>
--
Richard Biener <[email protected]>
SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg,
Germany; GF: Felix Imendörffer; HRB 36809 (AG Nuernberg)
