From: Qing Zhao <qing.zhao@oracle.com>
To: Richard Biener <rguenther@suse.de>
Cc: Martin Jambor <mjambor@suse.cz>, Jakub Jelinek <jakub@redhat.com>,
Kees Cook <keescook@chromium.org>,
Nick Alcock via Gcc-patches <gcc-patches@gcc.gnu.org>,
Richard Biener <richard.guenther@gmail.com>
Subject: Re: [patch][version 6] add -ftrivial-auto-var-init and variable attribute "uninitialized" to gcc
Date: Mon, 16 Aug 2021 19:49:13 +0000 [thread overview]
Message-ID: <B86DCCEA-FE60-45AD-848F-2550140BCD93@oracle.com> (raw)
In-Reply-To: <nycvar.YFH.7.76.2108160935370.11781@zhemvz.fhfr.qr>
> On Aug 16, 2021, at 2:40 AM, Richard Biener <rguenther@suse.de> wrote:
>
> On Thu, 12 Aug 2021, Qing Zhao wrote:
>
>> Hi, Richard,
>>
>> For RTL expansion of call to .DEFERRED_INIT, I changed my code per your suggestions like following:
>>
>> ======================
>> #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 || (!can_native_interpret_type_p (var_type)))
>> {
>> /* If this is a VLA or the type of the variable cannot be natively
>> interpreted, expand to a memset to initialize it. */
>> if (TREE_CODE (lhs) == SSA_NAME)
>> lhs = SSA_NAME_VAR (lhs);
>> tree var_addr = NULL_TREE;
>> if (is_vla)
>> var_addr = TREE_OPERAND (lhs, 0);
>> else
>> {
>> TREE_ADDRESSABLE (lhs) = 1;
>> var_addr = build_fold_addr_expr (lhs);
>> }
>> tree value = (init_type == AUTO_INIT_PATTERN) ?
>> build_int_cst (unsigned_char_type_node,
>> INIT_PATTERN_VALUE) :
>> build_zero_cst (unsigned_char_type_node);
>> tree m_call = build_call_expr (builtin_decl_implicit (BUILT_IN_MEMSET),
>> 3, var_addr, value, var_size);
>> /* Expand this memset call. */
>> expand_builtin_memset (m_call, NULL_RTX, TYPE_MODE (var_type));
>> }
>> else
>> {
>> /* If this is not a VLA and the type of the variable can be natively
>> interpreted, expand to assignment to generate better code. */
>> tree pattern = NULL_TREE;
>> unsigned HOST_WIDE_INT total_bytes
>> = tree_to_uhwi (TYPE_SIZE_UNIT (var_type));
>>
>> if (init_type == AUTO_INIT_PATTERN)
>> {
>> unsigned char *buf = (unsigned char *) xmalloc (total_bytes);
>> memset (buf, INIT_PATTERN_VALUE, total_bytes);
>> pattern = native_interpret_expr (var_type, buf, total_bytes);
>> gcc_assert (pattern);
>> }
>>
>> tree init = (init_type == AUTO_INIT_PATTERN) ?
>> pattern :
>> build_zero_cst (var_type);
>> expand_assignment (lhs, init, false);
>> }
>> }
>> ===========================
>>
>> Now, I used “can_native_interpret_type_p (var_type)” instead of “use_register_for_decl (lhs)” to decide
>> whether to use “memset” or use “assign” to expand this function.
>>
>> However, this exposed an bug that is very hard to be addressed:
>>
>> *******For the testing case: test suite/gcc.dg/uninit-I.c:
>>
>> /* { dg-do compile } */
>> /* { dg-options "-O2 -Wuninitialized" } */
>>
>> int sys_msgctl (void)
>> {
>> struct { int mode; } setbuf;
>> return setbuf.mode; /* { dg-warning "'setbuf\.mode' is used" } */
>> ==
>>
>> ******the above auto var “setbuf” has “struct” type, which “can_native_interpret_type_p(var_type)” is false, therefore,
>> Expanding this .DEFERRED_INIT call went down the “memset” expansion route.
>>
>> However, this structure type can be fitted into a register, therefore cannot be taken address anymore at this stage, even though I tried:
>>
>> TREE_ADDRESSABLE (lhs) = 1;
>> var_addr = build_fold_addr_expr (lhs);
>>
>> To create an address variable for it, the expansion still failed at expr.c: line 8412:
>> during RTL pass: expand
>> /home/opc/Work/GCC/latest-gcc/gcc/testsuite/gcc.dg/auto-init-uninit-I.c:6:24: internal compiler error: in expand_expr_addr_expr_1, at expr.c:8412
>> 0xd04104 expand_expr_addr_expr_1
>> ../../latest-gcc/gcc/expr.c:8412
>> 0xd04a95 expand_expr_addr_expr
>> ../../latest-gcc/gcc/expr.c:8525
>> 0xd13592 expand_expr_real_1(tree_node*, rtx_def*, machine_mode, expand_modifier, rtx_def**, bool)
>> ../../latest-gcc/gcc/expr.c:11741
>> 0xd05142 expand_expr_real(tree_node*, rtx_def*, machine_mode, expand_modifier, rtx_def**, bool)
>> ../../latest-gcc/gcc/expr.c:8713
>> 0xaed1d3 expand_expr
>> ../../latest-gcc/gcc/expr.h:301
>> 0xaf0d89 get_memory_rtx
>> ../../latest-gcc/gcc/builtins.c:1370
>> 0xafb4fb expand_builtin_memset_args
>> ../../latest-gcc/gcc/builtins.c:4102
>> 0xafacde expand_builtin_memset(tree_node*, rtx_def*, machine_mode)
>> ../../latest-gcc/gcc/builtins.c:3886
>> 0xe97fb3 expand_DEFERRED_INIT
>>
>> ******That’s the major reason why I chose “use_register_for_decl(lhs)” to decide “memset” expansion or “assign” expansion, “memset” expansion
>> needs to take address of the variable, if the variable has been decided to fit into a register, then its address cannot taken anymore at this stage.
>>
>> ******using “can_native_interpret_type_p” did make the “pattern” generation part much cleaner and simpler, however, looks like it didn’t work correctly.
>>
>> Based on this, I’d like to keep my previous implementation by using “use_register_for_decl” to decide whether to take “memset” expansion or “assign” expansion.
>> Therefore, I might still need to keep the “UGLY” implementation of generatting “pattern” constant for different types?
>>
>> Let me know your opinion on this.
>
> Hmm, I think you can use use_register_for_decl(lhs) to decide to use an
> alternate type to generate the pattern (and feed to
> can_native_interpret_type_p) by using
> lang_hooks.type_for_mode (TYPE_MODE (TREE_TYPE (lhs))).
Do you mean that the TYPE returned by “lang_hooks.type_for_mode(TYPE_MODE (TREE_TYPE (lhs))” will always satisfy “can_native_interpret_type_p”?
Even for big structure types?
i.e,
tree var_type = TREE_TYPE(lhs);
tree pattern = NULL_TREE;
unsigned HOST_WIDE_INT total_bytes
= tree_to_uhwi (TYPE_SIZE_UNIT (var_type));
If (use_register_for_decl(lhs)==false)
{
tree alt_type = lang_hooks.type_for_mode(TYPE_MODE(var_type), TYPE_UNSIGNED(var_type);
If (can_native_interpret_type_p (alt_type))
{
unsigned char *buf = (unsigned char *) xmalloc (total_bytes);
memset (buf, INIT_PATTERN_VALUE, total_bytes);
pattern = native_interpret_expr (alt_type, buf, total_bytes);
gcc_assert (pattern);
}
else
gcc_unreachable ();
}
?
Don’t quite understand here. Please clarify.
> You can then
> build the assignment from the pattern as
>
> VIEW_CONVERT <reg-type> (lhs) = pattern_cst;
What’s the <reg_type> in the above? The type “alt_type” returned by “lang_hooks.type_for_mode(TREE_TYPE(lhs))?
Do I need to build “MODIFY_EXPR” for the above?
lhs = build1 (VIEW_CONVERT_EXPR, alt_type, lhs);
tree final = build2 (MODIFY_EXPR, TREE_TYPE (alt_type), lhs, pattern);
Then how to expand this”final"?
>
> note that more RTL-expand-ish would be to simply expand 'lhs' and
> see whether it's a REG_P or a MEM_P and decide based on that.
You mean that the current RTL expansion will automatically expand LHS to memset route or assignment route based on whether
LHS is a REG_P or MEM_P? I don’t need to explicitly code for “expand_builtin_memset” or “expand_assign”?
> Of course
> that no longer allows you to use the high-level expand_assignment
> or memset but you'd need to work on a lower level then.
Don’t understand here, so, I need more coding on the RTL expansing for what part?
Qing
>
> Richard.
>
next prev parent reply other threads:[~2021-08-16 19:49 UTC|newest]
Thread overview: 60+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-27 3:26 Qing Zhao
2021-07-28 20:21 ` Kees Cook
2021-07-28 21:53 ` Qing Zhao
2021-08-09 14:09 ` Richard Biener
2021-08-09 16:38 ` Qing Zhao
2021-08-09 17:14 ` Richard Biener
2021-08-10 7:36 ` Richard Biener
2021-08-10 13:39 ` Qing Zhao
2021-08-10 14:16 ` Richard Biener
2021-08-10 15:02 ` Qing Zhao
2021-08-10 15:22 ` Richard Biener
2021-08-10 15:55 ` Qing Zhao
2021-08-10 20:16 ` Qing Zhao
2021-08-10 22:26 ` Qing Zhao
2021-08-11 7:02 ` Richard Biener
2021-08-11 13:33 ` Qing Zhao
2021-08-11 13:37 ` Richard Biener
2021-08-11 13:54 ` Qing Zhao
2021-08-11 13:58 ` Richard Biener
2021-08-11 14:00 ` Qing Zhao
2021-08-11 15:30 ` Qing Zhao
2021-08-11 15:53 ` Richard Biener
2021-08-11 16:22 ` Qing Zhao
2021-08-11 16:55 ` Richard Biener
2021-08-11 16:57 ` Qing Zhao
2021-08-11 20:30 ` Qing Zhao
2021-08-11 22:03 ` Qing Zhao
2021-08-16 7:12 ` Richard Biener
2021-08-16 14:48 ` Qing Zhao
2021-08-16 15:08 ` Richard Biener
2021-08-16 15:39 ` Qing Zhao
2021-08-16 7:11 ` Richard Biener
2021-08-16 16:48 ` Qing Zhao
2021-08-17 15:04 ` Qing Zhao
2021-08-17 20:40 ` Qing Zhao
2021-08-18 7:19 ` Richard Biener
2021-08-18 14:39 ` Qing Zhao
2021-08-11 9:02 ` Richard Sandiford
2021-08-11 13:44 ` Qing Zhao
2021-08-11 16:15 ` Richard Sandiford
2021-08-11 16:29 ` Qing Zhao
2021-08-12 19:24 ` Qing Zhao
2021-08-12 22:45 ` Qing Zhao
2021-08-16 7:40 ` Richard Biener
2021-08-16 15:45 ` Qing Zhao
2021-08-17 8:29 ` Richard Biener
2021-08-17 14:50 ` Qing Zhao
2021-08-17 16:08 ` Qing Zhao
2021-08-18 7:15 ` Richard Biener
2021-08-18 16:02 ` Qing Zhao
2021-08-19 9:00 ` Richard Biener
2021-08-19 13:54 ` Qing Zhao
2021-08-20 14:52 ` Qing Zhao
2021-08-23 13:55 ` Richard Biener
2021-09-02 17:24 ` Qing Zhao
2021-08-16 19:49 ` Qing Zhao [this message]
2021-08-17 8:43 ` Richard Biener
2021-08-17 14:03 ` Qing Zhao
2021-08-17 14:45 ` Richard Biener
2021-08-17 14:53 ` Qing Zhao
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=B86DCCEA-FE60-45AD-848F-2550140BCD93@oracle.com \
--to=qing.zhao@oracle.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=jakub@redhat.com \
--cc=keescook@chromium.org \
--cc=mjambor@suse.cz \
--cc=rguenther@suse.de \
--cc=richard.guenther@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).