* [PATCH, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
@ 2008-05-20 13:07 Uros Bizjak
2008-05-20 16:27 ` Richard Guenther
2008-05-20 16:29 ` Jakub Jelinek
0 siblings, 2 replies; 12+ messages in thread
From: Uros Bizjak @ 2008-05-20 13:07 UTC (permalink / raw)
To: GCC Patches
[-- Attachment #1: Type: text/plain, Size: 2216 bytes --]
Hello!
OK, here is the controversial patch that enables cld instruction with
-mcld. It includes --enable-cld configure option, and in this case
-mno-cld disables generation of cld instructions.
2008-05-20 Uros Bizjak <ubizjak@gmail.com>
PR target/36079
* configure.ac: Handle --enable-cld.
* configure.in: Regenerate.
* configure.ac: Ditto.
* config/i386/i386.h (TARGET_CLD): New define.
(struct machine_function): Add needs_cld field.
(ix86_current_function_needs_cld): New define.
* config/i386/i386.md (UNSPEC_CLD): New unspec volatile constant.
("cld"): New isns pattern.
("strmov_singleop"): Set ix86_current_function_needs_cld flag.
("rep_mov"): Ditto.
("strset_singleop"): Ditto.
("rep_stos"): Ditto.
("cmpstrnqi_nz_1"): Ditto.
("cmpstrnqi_1"): Ditto.
("strlenqi_1"): Ditto.
* config/i386/i386.opt (mcld, mno-cld): New options.
* config/i386/i386.c (ix86_expand_prologue): Emit cld insn for
when ix86_current_function_needs_cld is set and when combination
of configure and compile options allow it.
The patch was bootstrapped and regression tested on i686-pc-linux-gnu
a month or so ago. Currently vanila (without patch!) gcc-4_3 branch
doesn't build for me, it dies with:
gcc -c -g -fkeep-inline-functions -DIN_GCC -W -Wall
-Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes
-Wold-style-definition -Wmissing-format-attribute -pedantic
-Wno-long-long -Wno-variadic-macros -Wno-overlength-strings
-fno-common -DHAVE_CONFIG_H -DGENERATOR_FILE -I. -Ibuild
-I../../gcc-svn/branches/gcc-4_3-branch/gcc
-I../../gcc-svn/branches/gcc-4_3-branch/gcc/build
-I../../gcc-svn/branches/gcc-4_3-branch/gcc/../include
-I../../gcc-svn/branches/gcc-4_3-branch/gcc/../libcpp/include
-I/usr/local/include
-I../../gcc-svn/branches/gcc-4_3-branch/gcc/../libdecnumber
-I../../gcc-svn/branches/gcc-4_3-branch/gcc/../libdecnumber/bid
-I../libdecnumber -o build/errors.o
../../gcc-svn/branches/gcc-4_3-branch/gcc/errors.c
build/genmodes -h > tmp-modes.h
/bin/sh: build/genmodes: No such file or directory
gmake[3]: *** [s-modes-h] Error 127
Uros.
[-- Attachment #2: cld.diff.txt --]
[-- Type: text/plain, Size: 8478 bytes --]
Index: gcc/configure
===================================================================
--- gcc/configure (revision 135609)
+++ gcc/configure (working copy)
@@ -898,6 +898,7 @@ Optional Features:
--enable-sjlj-exceptions
arrange to use setjmp/longjmp exception handling
--enable-secureplt enable -msecure-plt by default for PowerPC
+ --enable-cld enable -mcld by default for 32bit x86
--disable-win32-registry
disable lookup of installation paths in the
Registry on Windows hosts
@@ -12951,6 +12952,22 @@ if test "${enable_secureplt+set}" = set;
fi;
+# Check whether --enable-cld or --disable-cld was given.
+if test "${enable_cld+set}" = set; then
+ enableval="$enable_cld"
+
+else
+ enable_cld=no
+fi;
+
+if test "$enable_cld" = yes ; then
+
+cat >>confdefs.h <<\_ACEOF
+#define USE_IX86_CLD 1
+_ACEOF
+
+fi
+
# Windows32 Registry support for specifying GCC installation paths.
# Check whether --enable-win32-registry or --disable-win32-registry was given.
if test "${enable_win32_registry+set}" = set; then
Index: gcc/config.in
===================================================================
--- gcc/config.in (revision 135609)
+++ gcc/config.in (working copy)
@@ -1369,37 +1369,37 @@
#endif
-/* The size of `int', as computed by sizeof. */
+/* The size of a `int', as computed by sizeof. */
#ifndef USED_FOR_TARGET
#undef SIZEOF_INT
#endif
-/* The size of `long', as computed by sizeof. */
+/* The size of a `long', as computed by sizeof. */
#ifndef USED_FOR_TARGET
#undef SIZEOF_LONG
#endif
-/* The size of `long long', as computed by sizeof. */
+/* The size of a `long long', as computed by sizeof. */
#ifndef USED_FOR_TARGET
#undef SIZEOF_LONG_LONG
#endif
-/* The size of `short', as computed by sizeof. */
+/* The size of a `short', as computed by sizeof. */
#ifndef USED_FOR_TARGET
#undef SIZEOF_SHORT
#endif
-/* The size of `void *', as computed by sizeof. */
+/* The size of a `void *', as computed by sizeof. */
#ifndef USED_FOR_TARGET
#undef SIZEOF_VOID_P
#endif
-/* The size of `__int64', as computed by sizeof. */
+/* The size of a `__int64', as computed by sizeof. */
#ifndef USED_FOR_TARGET
#undef SIZEOF___INT64
#endif
@@ -1441,6 +1441,13 @@
#endif
+/* Define if gcc should emit cld at the function entry when stringops are
+ used. */
+#ifndef USED_FOR_TARGET
+#undef USE_IX86_CLD
+#endif
+
+
/* Define to 1 if the 'long long' (or '__int64') is wider than 'long' but
still efficiently supported by the host hardware. */
#ifndef USED_FOR_TARGET
Index: gcc/configure.ac
===================================================================
--- gcc/configure.ac (revision 135609)
+++ gcc/configure.ac (working copy)
@@ -1548,6 +1548,15 @@ AC_ARG_ENABLE(secureplt,
[ --enable-secureplt enable -msecure-plt by default for PowerPC],
[], [])
+AC_ARG_ENABLE(cld,
+[ --enable-cld enable -mcld by default for 32bit x86], [],
+[enable_cld=no])
+
+if test "$enable_cld" = yes ; then
+ AC_DEFINE(USE_IX86_CLD, 1,
+[Define if gcc should emit cld at the function entry when stringops are used.])
+fi
+
# Windows32 Registry support for specifying GCC installation paths.
AC_ARG_ENABLE(win32-registry,
[ --disable-win32-registry
Index: gcc/config/i386/i386.h
===================================================================
--- gcc/config/i386/i386.h (revision 135609)
+++ gcc/config/i386/i386.h (working copy)
@@ -2443,8 +2443,9 @@ struct machine_function GTY(())
int save_varrargs_registers;
int accesses_prev_frame;
int optimize_mode_switching[MAX_386_ENTITIES];
- /* Set by ix86_compute_frame_layout and used by prologue/epilogue expander to
- determine the style used. */
+ int needs_cld;
+ /* Set by ix86_compute_frame_layout and used by prologue/epilogue
+ expander to determine the style used. */
int use_fast_prologue_epilogue;
/* Number of saved registers USE_FAST_PROLOGUE_EPILOGUE has been computed
for. */
@@ -2464,6 +2465,7 @@ struct machine_function GTY(())
#define ix86_stack_locals (cfun->machine->stack_locals)
#define ix86_save_varrargs_registers (cfun->machine->save_varrargs_registers)
#define ix86_optimize_mode_switching (cfun->machine->optimize_mode_switching)
+#define ix86_current_function_needs_cld (cfun->machine->needs_cld)
#define ix86_tls_descriptor_calls_expanded_in_cfun \
(cfun->machine->tls_descriptor_call_expanded_p)
/* Since tls_descriptor_call_expanded is not cleared, even if all TLS
Index: gcc/config/i386/i386.md
===================================================================
--- gcc/config/i386/i386.md (revision 135609)
+++ gcc/config/i386/i386.md (working copy)
@@ -205,6 +205,7 @@
(UNSPECV_XCHG 12)
(UNSPECV_LOCK 13)
(UNSPECV_PROLOGUE_USE 14)
+ (UNSPECV_CLD 15)
])
;; Constants to represent pcomtrue/pcomfalse variants
@@ -18514,6 +18515,14 @@
\f
;; Block operation instructions
+(define_insn "cld"
+ [(unspec_volatile [(const_int 0)] UNSPECV_CLD)]
+ ""
+ "cld"
+ [(set_attr "length" "1")
+ (set_attr "length_immediate" "0")
+ (set_attr "modrm" "0")])
+
(define_expand "movmemsi"
[(use (match_operand:BLK 0 "memory_operand" ""))
(use (match_operand:BLK 1 "memory_operand" ""))
@@ -18586,7 +18595,7 @@
(set (match_operand 2 "register_operand" "")
(match_operand 5 "" ""))])]
"TARGET_SINGLE_STRINGOP || optimize_size"
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strmovdi_rex_1"
[(set (mem:DI (match_operand:DI 2 "register_operand" "0"))
@@ -18703,7 +18712,7 @@
(match_operand 3 "memory_operand" ""))
(use (match_dup 4))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*rep_movdi_rex64"
[(set (match_operand:DI 2 "register_operand" "=c") (const_int 0))
@@ -18863,7 +18872,7 @@
(set (match_operand 0 "register_operand" "")
(match_operand 3 "" ""))])]
"TARGET_SINGLE_STRINGOP || optimize_size"
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strsetdi_rex_1"
[(set (mem:DI (match_operand:DI 1 "register_operand" "0"))
@@ -18957,7 +18966,7 @@
(use (match_operand 3 "register_operand" ""))
(use (match_dup 1))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*rep_stosdi_rex64"
[(set (match_operand:DI 1 "register_operand" "=c") (const_int 0))
@@ -19133,7 +19142,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (match_dup 2))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*cmpstrnqi_nz_1"
[(set (reg:CC FLAGS_REG)
@@ -19180,7 +19189,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (match_dup 2))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*cmpstrnqi_1"
[(set (reg:CC FLAGS_REG)
@@ -19249,7 +19258,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (reg:CC FLAGS_REG))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strlenqi_1"
[(set (match_operand:SI 0 "register_operand" "=&c")
Index: gcc/config/i386/i386.opt
===================================================================
--- gcc/config/i386/i386.opt (revision 135609)
+++ gcc/config/i386/i386.opt (working copy)
@@ -250,6 +250,14 @@ Support SSE5 built-in functions and code
;; Instruction support
+mcld
+Target Report RejectNegative Mask(X86_CLD)
+Generate cld instruction in the function prologue.
+
+mno-cld
+Target Report RejectNegative Mask(NO_X86_CLD)
+Do not generate cld instruction in the function prologue.
+
mabm
Target Report RejectNegative Var(x86_abm)
Support code generation of Advanced Bit Manipulation (ABM) instructions.
Index: gcc/config/i386/i386.c
===================================================================
--- gcc/config/i386/i386.c (revision 135609)
+++ gcc/config/i386/i386.c (working copy)
@@ -6485,6 +6485,16 @@ ix86_expand_prologue (void)
emit_insn (gen_prologue_use (pic_offset_table_rtx));
emit_insn (gen_blockage ());
}
+
+ /* Emit cld instruction if stringops are used in the function. */
+ if (!TARGET_64BIT
+#ifdef USE_IX86_CLD
+ && !TARGET_NO_X86_CLD
+#else
+ && TARGET_X86_CLD
+#endif
+ && ix86_current_function_needs_cld)
+ emit_insn (gen_cld ());
}
/* Emit code to restore saved registers using MOV insns. First register
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
2008-05-20 13:07 [PATCH, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore Uros Bizjak
@ 2008-05-20 16:27 ` Richard Guenther
2008-05-20 16:29 ` Jakub Jelinek
1 sibling, 0 replies; 12+ messages in thread
From: Richard Guenther @ 2008-05-20 16:27 UTC (permalink / raw)
To: Uros Bizjak; +Cc: GCC Patches
On Tue, May 20, 2008 at 2:47 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
> Hello!
>
> OK, here is the controversial patch that enables cld instruction with
> -mcld. It includes --enable-cld configure option, and in this case
> -mno-cld disables generation of cld instructions.
>
> 2008-05-20 Uros Bizjak <ubizjak@gmail.com>
>
> PR target/36079
> * configure.ac: Handle --enable-cld.
> * configure.in: Regenerate.
> * configure.ac: Ditto.
> * config/i386/i386.h (TARGET_CLD): New define.
> (struct machine_function): Add needs_cld field.
> (ix86_current_function_needs_cld): New define.
> * config/i386/i386.md (UNSPEC_CLD): New unspec volatile constant.
> ("cld"): New isns pattern.
> ("strmov_singleop"): Set ix86_current_function_needs_cld flag.
> ("rep_mov"): Ditto.
> ("strset_singleop"): Ditto.
> ("rep_stos"): Ditto.
> ("cmpstrnqi_nz_1"): Ditto.
> ("cmpstrnqi_1"): Ditto.
> ("strlenqi_1"): Ditto.
> * config/i386/i386.opt (mcld, mno-cld): New options.
> * config/i386/i386.c (ix86_expand_prologue): Emit cld insn for
> when ix86_current_function_needs_cld is set and when combination
> of configure and compile options allow it.
>
> The patch was bootstrapped and regression tested on i686-pc-linux-gnu
> a month or so ago. Currently vanila (without patch!) gcc-4_3 branch
> doesn't build for me, it dies with:
>
> gcc -c -g -fkeep-inline-functions -DIN_GCC -W -Wall
> -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes
> -Wold-style-definition -Wmissing-format-attribute -pedantic
> -Wno-long-long -Wno-variadic-macros -Wno-overlength-strings
> -fno-common -DHAVE_CONFIG_H -DGENERATOR_FILE -I. -Ibuild
> -I../../gcc-svn/branches/gcc-4_3-branch/gcc
> -I../../gcc-svn/branches/gcc-4_3-branch/gcc/build
> -I../../gcc-svn/branches/gcc-4_3-branch/gcc/../include
> -I../../gcc-svn/branches/gcc-4_3-branch/gcc/../libcpp/include
> -I/usr/local/include
> -I../../gcc-svn/branches/gcc-4_3-branch/gcc/../libdecnumber
> -I../../gcc-svn/branches/gcc-4_3-branch/gcc/../libdecnumber/bid
> -I../libdecnumber -o build/errors.o
> ../../gcc-svn/branches/gcc-4_3-branch/gcc/errors.c
> build/genmodes -h > tmp-modes.h
> /bin/sh: build/genmodes: No such file or directory
> gmake[3]: *** [s-modes-h] Error 127
I have bootstrapped and tested this patch successfully on i686-pc-linux-gnu
with --enable-cld.
Richard.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
2008-05-20 13:07 [PATCH, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore Uros Bizjak
2008-05-20 16:27 ` Richard Guenther
@ 2008-05-20 16:29 ` Jakub Jelinek
2008-05-20 16:43 ` Jakub Jelinek
` (2 more replies)
1 sibling, 3 replies; 12+ messages in thread
From: Jakub Jelinek @ 2008-05-20 16:29 UTC (permalink / raw)
To: Uros Bizjak; +Cc: GCC Patches
On Tue, May 20, 2008 at 02:47:15PM +0200, Uros Bizjak wrote:
> OK, here is the controversial patch that enables cld instruction with
> -mcld. It includes --enable-cld configure option, and in this case
> -mno-cld disables generation of cld instructions.
The configury bits are IMHO too ugly and the ChangeLog entry doesn't match
the patch either.
I think there is no reason why we need to use two separate target_flags
bits, one for -mcld and one for -mno-cld - we already have
target_flags_explicit, so there are 2 bits for each mask option already.
Also, it is IMHO cleaner to just set target_flags & MASK_CLD bit during
option processing. I think we shouldn't ignore -mcld even for 64-bit code,
remember the original bug has been found in a 64-bit app with weirdo hand
written assembly, but as libc never uses std for 64-bit code, IMHO
--enable-cld shouldn't make -mcld the default for 64-bit code.
2008-05-20 Uros Bizjak <ubizjak@gmail.com>
Jakub Jelinek <jakub@redhat.com>
PR target/36079
* configure.ac: Handle --enable-cld.
* configure.in: Regenerate.
* configure.ac: Ditto.
* config/i386/i386.h (struct machine_function): Add needs_cld field.
(ix86_current_function_needs_cld): New define.
* config/i386/i386.md (UNSPEC_CLD): New unspec volatile constant.
(cld): New isns pattern.
(strmov_singleop, rep_mov, strset_singleop, rep_stos, cmpstrnqi_nz_1,
cmpstrnqi_1, strlenqi_1): Set ix86_current_function_needs_cld flag.
* config/i386/i386.opt (mcld): New option.
* config/i386/i386.c (ix86_expand_prologue): Emit cld insn if
TARGET_CLD and ix86_current_function_needs_cld.
(override_options): Use -mcld by default for 32-bit code if
USE_IX86_CLD.
--- gcc/config/i386/i386.md.jj 2008-04-22 00:19:40.000000000 +0200
+++ gcc/config/i386/i386.md 2008-05-20 16:34:37.000000000 +0200
@@ -205,6 +205,7 @@
(UNSPECV_XCHG 12)
(UNSPECV_LOCK 13)
(UNSPECV_PROLOGUE_USE 14)
+ (UNSPECV_CLD 15)
])
;; Constants to represent pcomtrue/pcomfalse variants
@@ -18514,6 +18515,14 @@
\f
;; Block operation instructions
+(define_insn "cld"
+ [(unspec_volatile [(const_int 0)] UNSPECV_CLD)]
+ ""
+ "cld"
+ [(set_attr "length" "1")
+ (set_attr "length_immediate" "0")
+ (set_attr "modrm" "0")])
+
(define_expand "movmemsi"
[(use (match_operand:BLK 0 "memory_operand" ""))
(use (match_operand:BLK 1 "memory_operand" ""))
@@ -18586,7 +18595,7 @@
(set (match_operand 2 "register_operand" "")
(match_operand 5 "" ""))])]
"TARGET_SINGLE_STRINGOP || optimize_size"
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strmovdi_rex_1"
[(set (mem:DI (match_operand:DI 2 "register_operand" "0"))
@@ -18703,7 +18712,7 @@
(match_operand 3 "memory_operand" ""))
(use (match_dup 4))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*rep_movdi_rex64"
[(set (match_operand:DI 2 "register_operand" "=c") (const_int 0))
@@ -18863,7 +18872,7 @@
(set (match_operand 0 "register_operand" "")
(match_operand 3 "" ""))])]
"TARGET_SINGLE_STRINGOP || optimize_size"
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strsetdi_rex_1"
[(set (mem:DI (match_operand:DI 1 "register_operand" "0"))
@@ -18957,7 +18966,7 @@
(use (match_operand 3 "register_operand" ""))
(use (match_dup 1))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*rep_stosdi_rex64"
[(set (match_operand:DI 1 "register_operand" "=c") (const_int 0))
@@ -19133,7 +19142,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (match_dup 2))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*cmpstrnqi_nz_1"
[(set (reg:CC FLAGS_REG)
@@ -19180,7 +19189,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (match_dup 2))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*cmpstrnqi_1"
[(set (reg:CC FLAGS_REG)
@@ -19249,7 +19258,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (reg:CC FLAGS_REG))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strlenqi_1"
[(set (match_operand:SI 0 "register_operand" "=&c")
--- gcc/config/i386/i386.opt.jj 2008-02-18 23:45:48.000000000 +0100
+++ gcc/config/i386/i386.opt 2008-05-20 16:34:37.000000000 +0200
@@ -250,6 +250,10 @@ Support SSE5 built-in functions and code
;; Instruction support
+mcld
+Target Report Mask(CLD)
+Generate cld instruction in the function prologue.
+
mabm
Target Report RejectNegative Var(x86_abm)
Support code generation of Advanced Bit Manipulation (ABM) instructions.
--- gcc/config/i386/i386.h.jj 2008-02-18 23:45:48.000000000 +0100
+++ gcc/config/i386/i386.h 2008-05-20 16:34:37.000000000 +0200
@@ -2443,8 +2443,9 @@ struct machine_function GTY(())
int save_varrargs_registers;
int accesses_prev_frame;
int optimize_mode_switching[MAX_386_ENTITIES];
- /* Set by ix86_compute_frame_layout and used by prologue/epilogue expander to
- determine the style used. */
+ int needs_cld;
+ /* Set by ix86_compute_frame_layout and used by prologue/epilogue
+ expander to determine the style used. */
int use_fast_prologue_epilogue;
/* Number of saved registers USE_FAST_PROLOGUE_EPILOGUE has been computed
for. */
@@ -2464,6 +2465,7 @@ struct machine_function GTY(())
#define ix86_stack_locals (cfun->machine->stack_locals)
#define ix86_save_varrargs_registers (cfun->machine->save_varrargs_registers)
#define ix86_optimize_mode_switching (cfun->machine->optimize_mode_switching)
+#define ix86_current_function_needs_cld (cfun->machine->needs_cld)
#define ix86_tls_descriptor_calls_expanded_in_cfun \
(cfun->machine->tls_descriptor_call_expanded_p)
/* Since tls_descriptor_call_expanded is not cleared, even if all TLS
--- gcc/config/i386/i386.c.jj 2008-05-18 22:16:19.000000000 +0200
+++ gcc/config/i386/i386.c 2008-05-20 16:44:36.000000000 +0200
@@ -2733,6 +2733,12 @@ override_options (void)
can be optimized to ap = __builtin_next_arg (0). */
if (!TARGET_64BIT || TARGET_64BIT_MS_ABI)
targetm.expand_builtin_va_start = NULL;
+
+#ifdef USE_IX86_CLD
+ /* Use -mcld by default for 32-bit code if configured with --enable-cld. */
+ if (!TARGET_64BIT && !(target_flags_explicit & MASK_CLD))
+ target_flags |= MASK_CLD;
+#endif
}
\f
/* Return true if this goes in large data/bss. */
@@ -6485,6 +6491,10 @@ ix86_expand_prologue (void)
emit_insn (gen_prologue_use (pic_offset_table_rtx));
emit_insn (gen_blockage ());
}
+
+ /* Emit cld instruction if stringops are used in the function. */
+ if (TARGET_CLD && ix86_current_function_needs_cld)
+ emit_insn (gen_cld ());
}
/* Emit code to restore saved registers using MOV insns. First register
--- gcc/configure.ac.jj 2008-03-05 17:17:21.000000000 +0100
+++ gcc/configure.ac 2008-05-20 16:34:37.000000000 +0200
@@ -1548,6 +1548,15 @@ AC_ARG_ENABLE(secureplt,
[ --enable-secureplt enable -msecure-plt by default for PowerPC],
[], [])
+AC_ARG_ENABLE(cld,
+[ --enable-cld enable -mcld by default for 32bit x86], [],
+[enable_cld=no])
+
+if test "$enable_cld" = yes ; then
+ AC_DEFINE(USE_IX86_CLD, 1,
+[Define if gcc should by default emit cld at the function entry when stringops are used.])
+fi
+
# Windows32 Registry support for specifying GCC installation paths.
AC_ARG_ENABLE(win32-registry,
[ --disable-win32-registry
--- gcc/configure.jj 2008-03-05 17:17:21.000000000 +0100
+++ gcc/configure 2008-05-20 16:34:36.000000000 +0200
@@ -898,6 +898,7 @@ Optional Features:
--enable-sjlj-exceptions
arrange to use setjmp/longjmp exception handling
--enable-secureplt enable -msecure-plt by default for PowerPC
+ --enable-cld enable -mcld by default for 32bit x86
--disable-win32-registry
disable lookup of installation paths in the
Registry on Windows hosts
@@ -12951,6 +12952,22 @@ if test "${enable_secureplt+set}" = set;
fi;
+# Check whether --enable-cld or --disable-cld was given.
+if test "${enable_cld+set}" = set; then
+ enableval="$enable_cld"
+
+else
+ enable_cld=no
+fi;
+
+if test "$enable_cld" = yes ; then
+
+cat >>confdefs.h <<\_ACEOF
+#define USE_IX86_CLD 1
+_ACEOF
+
+fi
+
# Windows32 Registry support for specifying GCC installation paths.
# Check whether --enable-win32-registry or --disable-win32-registry was given.
if test "${enable_win32_registry+set}" = set; then
--- gcc/config.in.jj 2008-02-18 23:46:08.000000000 +0100
+++ gcc/config.in 2008-05-20 16:34:37.000000000 +0200
@@ -1441,6 +1441,13 @@
#endif
+/* Define if gcc should by default emit cld at the function entry when
+ stringops are used. */
+#ifndef USED_FOR_TARGET
+#undef USE_IX86_CLD
+#endif
+
+
/* Define to 1 if the 'long long' (or '__int64') is wider than 'long' but
still efficiently supported by the host hardware. */
#ifndef USED_FOR_TARGET
Jakub
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
2008-05-20 16:29 ` Jakub Jelinek
@ 2008-05-20 16:43 ` Jakub Jelinek
2008-05-20 17:03 ` Paolo Bonzini
2008-05-21 2:58 ` Mark Mitchell
2 siblings, 0 replies; 12+ messages in thread
From: Jakub Jelinek @ 2008-05-20 16:43 UTC (permalink / raw)
To: Uros Bizjak; +Cc: GCC Patches
On Tue, May 20, 2008 at 11:10:34AM -0400, Jakub Jelinek wrote:
> On Tue, May 20, 2008 at 02:47:15PM +0200, Uros Bizjak wrote:
> --- gcc/config/i386/i386.c.jj 2008-05-18 22:16:19.000000000 +0200
> +++ gcc/config/i386/i386.c 2008-05-20 16:44:36.000000000 +0200
> @@ -2733,6 +2733,12 @@ override_options (void)
> can be optimized to ap = __builtin_next_arg (0). */
> if (!TARGET_64BIT || TARGET_64BIT_MS_ABI)
> targetm.expand_builtin_va_start = NULL;
> +
> +#ifdef USE_IX86_CLD
> + /* Use -mcld by default for 32-bit code if configured with --enable-cld. */
> + if (!TARGET_64BIT && !(target_flags_explicit & MASK_CLD))
> + target_flags |= MASK_CLD;
> +#endif
> }
> \f
> /* Return true if this goes in large data/bss. */
This hunk should actually be:
@@ -2733,6 +2733,12 @@ override_options (void)
can be optimized to ap = __builtin_next_arg (0). */
if (!TARGET_64BIT || TARGET_64BIT_MS_ABI)
targetm.expand_builtin_va_start = NULL;
+
+#ifdef USE_IX86_CLD
+ /* Use -mcld by default for 32-bit code if configured with --enable-cld. */
+ if (!TARGET_64BIT)
+ target_flags |= MASK_CLD & ~target_flags_explicit;
+#endif
}
\f
/* Return true if this goes in large data/bss. */
Jakub
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
2008-05-20 16:29 ` Jakub Jelinek
2008-05-20 16:43 ` Jakub Jelinek
@ 2008-05-20 17:03 ` Paolo Bonzini
2008-05-21 2:58 ` Mark Mitchell
2 siblings, 0 replies; 12+ messages in thread
From: Paolo Bonzini @ 2008-05-20 17:03 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: Uros Bizjak, GCC Patches
Jakub Jelinek wrote:
> On Tue, May 20, 2008 at 02:47:15PM +0200, Uros Bizjak wrote:
>> OK, here is the controversial patch that enables cld instruction with
>> -mcld. It includes --enable-cld configure option, and in this case
>> -mno-cld disables generation of cld instructions.
>
> The configury bits are IMHO too ugly
Yes, the problem is that configure.ac is polluted with target-specific
stuff. You can instead add a
if test $enable_cld = yes; then
tm_defines=${tm_defines} USE_IX86_CLD=1"
fi
in config.gcc around this place:
case ${target} in
x86_64-*-*)
tm_file="i386/biarch64.h ${tm_file}"
;;
esac
Paolo
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
2008-05-20 16:29 ` Jakub Jelinek
2008-05-20 16:43 ` Jakub Jelinek
2008-05-20 17:03 ` Paolo Bonzini
@ 2008-05-21 2:58 ` Mark Mitchell
2008-05-21 8:24 ` [PATCH v2, " Uros Bizjak
2 siblings, 1 reply; 12+ messages in thread
From: Mark Mitchell @ 2008-05-21 2:58 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: Uros Bizjak, GCC Patches
Jakub Jelinek wrote:
> option processing. I think we shouldn't ignore -mcld even for 64-bit code,
> remember the original bug has been found in a 64-bit app with weirdo hand
> written assembly, but as libc never uses std for 64-bit code, IMHO
> --enable-cld shouldn't make -mcld the default for 64-bit code.
Uros, thank you for writing this patch! Richard, Jakub, thank you for
the reviews.
Per our off-list RM discussions, let's not worry too much about the
defaults at this point. Whether --enable-cld is the default or not, and
whether --enable-cld enables -mcld by default in 64-bit mode or not, are
topics we can and should debate, but let's get the core technology in.
Thanks,
--
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH v2, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
2008-05-21 2:58 ` Mark Mitchell
@ 2008-05-21 8:24 ` Uros Bizjak
2008-05-21 8:25 ` Paolo Bonzini
2008-05-21 9:00 ` Richard Guenther
0 siblings, 2 replies; 12+ messages in thread
From: Uros Bizjak @ 2008-05-21 8:24 UTC (permalink / raw)
To: Mark Mitchell; +Cc: Jakub Jelinek, GCC Patches, Paolo Bonzini, Richard Guenther
[-- Attachment #1: Type: text/plain, Size: 1765 bytes --]
On Wed, May 21, 2008 at 3:40 AM, Mark Mitchell <mark@codesourcery.com> wrote:
>> option processing. I think we shouldn't ignore -mcld even for 64-bit
>> code,
>> remember the original bug has been found in a 64-bit app with weirdo hand
>> written assembly, but as libc never uses std for 64-bit code, IMHO
>> --enable-cld shouldn't make -mcld the default for 64-bit code.
>
> Uros, thank you for writing this patch! Richard, Jakub, thank you for the
> reviews.
>
> Per our off-list RM discussions, let's not worry too much about the defaults
> at this point. Whether --enable-cld is the default or not, and whether
> --enable-cld enables -mcld by default in 64-bit mode or not, are topics we
> can and should debate, but let's get the core technology in.
Attached patch adds all suggestions by Jakub and Paolo. Patch works as
advertised, bootstrap and regression test on i686-pc-linux with
--enable-cld is in progress.
2008-05-21 Uros Bizjak <ubizjak@gmail.com>
Jakub Jelinek <jakub@redhat.com>
PR target/36079
* configure.ac: Handle --enable-cld.
* configure: Regenerated.
* config.gcc: Add USE_IX86_CLD to tm_defines for x86 targets.
* config/i386/i386.h (struct machine_function): Add needs_cld field.
(ix86_current_function_needs_cld): New define.
* config/i386/i386.md (UNSPEC_CLD): New unspec volatile constant.
(cld): New isns pattern.
(strmov_singleop, rep_mov, strset_singleop, rep_stos, cmpstrnqi_nz_1,
cmpstrnqi_1, strlenqi_1): Set ix86_current_function_needs_cld flag.
* config/i386/i386.opt (mcld): New option.
* config/i386/i386.c (ix86_expand_prologue): Emit cld insn if
TARGET_CLD and ix86_current_function_needs_cld.
(override_options): Use -mcld by default for 32-bit code if
USE_IX86_CLD.
OK for 4.3 branch?
Uros.
[-- Attachment #2: cld.diff.txt --]
[-- Type: text/plain, Size: 7345 bytes --]
Index: gcc/configure
===================================================================
--- gcc/configure (revision 135707)
+++ gcc/configure (working copy)
@@ -898,6 +898,7 @@ Optional Features:
--enable-sjlj-exceptions
arrange to use setjmp/longjmp exception handling
--enable-secureplt enable -msecure-plt by default for PowerPC
+ --enable-cld enable -mcld by default for 32bit x86
--disable-win32-registry
disable lookup of installation paths in the
Registry on Windows hosts
@@ -12951,6 +12952,14 @@ if test "${enable_secureplt+set}" = set;
fi;
+# Check whether --enable-cld or --disable-cld was given.
+if test "${enable_cld+set}" = set; then
+ enableval="$enable_cld"
+
+else
+ enable_cld=no
+fi;
+
# Windows32 Registry support for specifying GCC installation paths.
# Check whether --enable-win32-registry or --disable-win32-registry was given.
if test "${enable_win32_registry+set}" = set; then
Index: gcc/configure.ac
===================================================================
--- gcc/configure.ac (revision 135707)
+++ gcc/configure.ac (working copy)
@@ -1548,6 +1548,10 @@ AC_ARG_ENABLE(secureplt,
[ --enable-secureplt enable -msecure-plt by default for PowerPC],
[], [])
+AC_ARG_ENABLE(cld,
+[ --enable-cld enable -mcld by default for 32bit x86], [],
+[enable_cld=no])
+
# Windows32 Registry support for specifying GCC installation paths.
AC_ARG_ENABLE(win32-registry,
[ --disable-win32-registry
Index: gcc/config.gcc
===================================================================
--- gcc/config.gcc (revision 135707)
+++ gcc/config.gcc (working copy)
@@ -397,6 +397,14 @@ x86_64-*-*)
;;
esac
+case ${target} in
+i[34567]86-*-* | x86_64-*-*)
+ if test $enable_cld = yes; then
+ tm_defines="${tm_defines} USE_IX86_CLD=1"
+ fi
+ ;;
+esac
+
# On a.out targets, we need to use collect2.
case ${target} in
*-*-*aout*)
Index: gcc/config/i386/i386.h
===================================================================
--- gcc/config/i386/i386.h (revision 135707)
+++ gcc/config/i386/i386.h (working copy)
@@ -2443,8 +2443,9 @@ struct machine_function GTY(())
int save_varrargs_registers;
int accesses_prev_frame;
int optimize_mode_switching[MAX_386_ENTITIES];
- /* Set by ix86_compute_frame_layout and used by prologue/epilogue expander to
- determine the style used. */
+ int needs_cld;
+ /* Set by ix86_compute_frame_layout and used by prologue/epilogue
+ expander to determine the style used. */
int use_fast_prologue_epilogue;
/* Number of saved registers USE_FAST_PROLOGUE_EPILOGUE has been computed
for. */
@@ -2464,6 +2465,7 @@ struct machine_function GTY(())
#define ix86_stack_locals (cfun->machine->stack_locals)
#define ix86_save_varrargs_registers (cfun->machine->save_varrargs_registers)
#define ix86_optimize_mode_switching (cfun->machine->optimize_mode_switching)
+#define ix86_current_function_needs_cld (cfun->machine->needs_cld)
#define ix86_tls_descriptor_calls_expanded_in_cfun \
(cfun->machine->tls_descriptor_call_expanded_p)
/* Since tls_descriptor_call_expanded is not cleared, even if all TLS
Index: gcc/config/i386/i386.md
===================================================================
--- gcc/config/i386/i386.md (revision 135707)
+++ gcc/config/i386/i386.md (working copy)
@@ -205,6 +205,7 @@
(UNSPECV_XCHG 12)
(UNSPECV_LOCK 13)
(UNSPECV_PROLOGUE_USE 14)
+ (UNSPECV_CLD 15)
])
;; Constants to represent pcomtrue/pcomfalse variants
@@ -18514,6 +18515,14 @@
\f
;; Block operation instructions
+(define_insn "cld"
+ [(unspec_volatile [(const_int 0)] UNSPECV_CLD)]
+ ""
+ "cld"
+ [(set_attr "length" "1")
+ (set_attr "length_immediate" "0")
+ (set_attr "modrm" "0")])
+
(define_expand "movmemsi"
[(use (match_operand:BLK 0 "memory_operand" ""))
(use (match_operand:BLK 1 "memory_operand" ""))
@@ -18586,7 +18595,7 @@
(set (match_operand 2 "register_operand" "")
(match_operand 5 "" ""))])]
"TARGET_SINGLE_STRINGOP || optimize_size"
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strmovdi_rex_1"
[(set (mem:DI (match_operand:DI 2 "register_operand" "0"))
@@ -18703,7 +18712,7 @@
(match_operand 3 "memory_operand" ""))
(use (match_dup 4))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*rep_movdi_rex64"
[(set (match_operand:DI 2 "register_operand" "=c") (const_int 0))
@@ -18863,7 +18872,7 @@
(set (match_operand 0 "register_operand" "")
(match_operand 3 "" ""))])]
"TARGET_SINGLE_STRINGOP || optimize_size"
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strsetdi_rex_1"
[(set (mem:DI (match_operand:DI 1 "register_operand" "0"))
@@ -18957,7 +18966,7 @@
(use (match_operand 3 "register_operand" ""))
(use (match_dup 1))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*rep_stosdi_rex64"
[(set (match_operand:DI 1 "register_operand" "=c") (const_int 0))
@@ -19133,7 +19142,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (match_dup 2))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*cmpstrnqi_nz_1"
[(set (reg:CC FLAGS_REG)
@@ -19180,7 +19189,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (match_dup 2))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*cmpstrnqi_1"
[(set (reg:CC FLAGS_REG)
@@ -19249,7 +19258,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (reg:CC FLAGS_REG))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strlenqi_1"
[(set (match_operand:SI 0 "register_operand" "=&c")
Index: gcc/config/i386/i386.opt
===================================================================
--- gcc/config/i386/i386.opt (revision 135707)
+++ gcc/config/i386/i386.opt (working copy)
@@ -250,6 +250,10 @@ Support SSE5 built-in functions and code
;; Instruction support
+mcld
+Target Report Mask(CLD)
+Generate cld instruction in the function prologue.
+
mabm
Target Report RejectNegative Var(x86_abm)
Support code generation of Advanced Bit Manipulation (ABM) instructions.
Index: gcc/config/i386/i386.c
===================================================================
--- gcc/config/i386/i386.c (revision 135707)
+++ gcc/config/i386/i386.c (working copy)
@@ -2733,6 +2733,12 @@ override_options (void)
can be optimized to ap = __builtin_next_arg (0). */
if (!TARGET_64BIT || TARGET_64BIT_MS_ABI)
targetm.expand_builtin_va_start = NULL;
+
+#ifdef USE_IX86_CLD
+ /* Use -mcld by default for 32-bit code if configured with --enable-cld. */
+ if (!TARGET_64BIT)
+ target_flags |= MASK_CLD & ~target_flags_explicit;
+#endif
}
\f
/* Return true if this goes in large data/bss. */
@@ -6485,6 +6491,10 @@ ix86_expand_prologue (void)
emit_insn (gen_prologue_use (pic_offset_table_rtx));
emit_insn (gen_blockage ());
}
+
+ /* Emit cld instruction if stringops are used in the function. */
+ if (TARGET_CLD && ix86_current_function_needs_cld)
+ emit_insn (gen_cld ());
}
/* Emit code to restore saved registers using MOV insns. First register
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH v2, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
2008-05-21 8:24 ` [PATCH v2, " Uros Bizjak
@ 2008-05-21 8:25 ` Paolo Bonzini
2008-05-21 9:04 ` Jakub Jelinek
2008-05-21 9:00 ` Richard Guenther
1 sibling, 1 reply; 12+ messages in thread
From: Paolo Bonzini @ 2008-05-21 8:25 UTC (permalink / raw)
To: Uros Bizjak
Cc: Mark Mitchell, Jakub Jelinek, GCC Patches, Paolo Bonzini,
Richard Guenther
> 2008-05-21 Uros Bizjak <ubizjak@gmail.com>
> Jakub Jelinek <jakub@redhat.com>
>
> PR target/36079
> * configure.ac: Handle --enable-cld.
> * configure: Regenerated.
> * config.gcc: Add USE_IX86_CLD to tm_defines for x86 targets.
> * config/i386/i386.h (struct machine_function): Add needs_cld field.
> (ix86_current_function_needs_cld): New define.
> * config/i386/i386.md (UNSPEC_CLD): New unspec volatile constant.
> (cld): New isns pattern.
> (strmov_singleop, rep_mov, strset_singleop, rep_stos, cmpstrnqi_nz_1,
> cmpstrnqi_1, strlenqi_1): Set ix86_current_function_needs_cld flag.
> * config/i386/i386.opt (mcld): New option.
> * config/i386/i386.c (ix86_expand_prologue): Emit cld insn if
> TARGET_CLD and ix86_current_function_needs_cld.
> (override_options): Use -mcld by default for 32-bit code if
> USE_IX86_CLD.
>
> OK for 4.3 branch?
Build parts are ok for 4.3 and trunk, except that you could merge the
case statement you added with the one before it (yes, that involves a
small duplication). Those parts are preapproved with this change.
The rest, you can approve yourself (though IIUC you want a word from the
branch RMs).
Paolo
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH v2, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
2008-05-21 8:24 ` [PATCH v2, " Uros Bizjak
2008-05-21 8:25 ` Paolo Bonzini
@ 2008-05-21 9:00 ` Richard Guenther
2008-05-21 9:14 ` Uros Bizjak
2008-05-21 10:52 ` Uros Bizjak
1 sibling, 2 replies; 12+ messages in thread
From: Richard Guenther @ 2008-05-21 9:00 UTC (permalink / raw)
To: Uros Bizjak
Cc: Mark Mitchell, Jakub Jelinek, GCC Patches, Paolo Bonzini,
Richard Guenther
On Wed, May 21, 2008 at 9:41 AM, Uros Bizjak <ubizjak@gmail.com> wrote:
> On Wed, May 21, 2008 at 3:40 AM, Mark Mitchell <mark@codesourcery.com> wrote:
>
>>> option processing. I think we shouldn't ignore -mcld even for 64-bit
>>> code,
>>> remember the original bug has been found in a 64-bit app with weirdo hand
>>> written assembly, but as libc never uses std for 64-bit code, IMHO
>>> --enable-cld shouldn't make -mcld the default for 64-bit code.
>>
>> Uros, thank you for writing this patch! Richard, Jakub, thank you for the
>> reviews.
>>
>> Per our off-list RM discussions, let's not worry too much about the defaults
>> at this point. Whether --enable-cld is the default or not, and whether
>> --enable-cld enables -mcld by default in 64-bit mode or not, are topics we
>> can and should debate, but let's get the core technology in.
>
> Attached patch adds all suggestions by Jakub and Paolo. Patch works as
> advertised, bootstrap and regression test on i686-pc-linux with
> --enable-cld is in progress.
>
> 2008-05-21 Uros Bizjak <ubizjak@gmail.com>
> Jakub Jelinek <jakub@redhat.com>
>
> PR target/36079
> * configure.ac: Handle --enable-cld.
> * configure: Regenerated.
> * config.gcc: Add USE_IX86_CLD to tm_defines for x86 targets.
> * config/i386/i386.h (struct machine_function): Add needs_cld field.
> (ix86_current_function_needs_cld): New define.
> * config/i386/i386.md (UNSPEC_CLD): New unspec volatile constant.
> (cld): New isns pattern.
> (strmov_singleop, rep_mov, strset_singleop, rep_stos, cmpstrnqi_nz_1,
> cmpstrnqi_1, strlenqi_1): Set ix86_current_function_needs_cld flag.
> * config/i386/i386.opt (mcld): New option.
> * config/i386/i386.c (ix86_expand_prologue): Emit cld insn if
> TARGET_CLD and ix86_current_function_needs_cld.
> (override_options): Use -mcld by default for 32-bit code if
> USE_IX86_CLD.
>
> OK for 4.3 branch?
This is ok for the mainline and the branch if you add documentation for both the
configure option and the new target option.
If I specify --enable-cld on x86_64 I get clds emitted for -m32 but
not for -m64,
correct? But if I specify -mcld I get clds on both -m32 and -m64? I think this
is reasonable behavior, but the documentation should probably explicitly
mention this.
Thanks,
Richard.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH v2, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
2008-05-21 8:25 ` Paolo Bonzini
@ 2008-05-21 9:04 ` Jakub Jelinek
0 siblings, 0 replies; 12+ messages in thread
From: Jakub Jelinek @ 2008-05-21 9:04 UTC (permalink / raw)
To: Paolo Bonzini
Cc: Uros Bizjak, Mark Mitchell, GCC Patches, Paolo Bonzini, Richard Guenther
On Wed, May 21, 2008 at 09:53:35AM +0200, Paolo Bonzini wrote:
> > PR target/36079
> > * configure.ac: Handle --enable-cld.
> > * configure: Regenerated.
> > * config.gcc: Add USE_IX86_CLD to tm_defines for x86 targets.
> > * config/i386/i386.h (struct machine_function): Add needs_cld field.
> > (ix86_current_function_needs_cld): New define.
> > * config/i386/i386.md (UNSPEC_CLD): New unspec volatile constant.
> > (cld): New isns pattern.
> > (strmov_singleop, rep_mov, strset_singleop, rep_stos, cmpstrnqi_nz_1,
> > cmpstrnqi_1, strlenqi_1): Set ix86_current_function_needs_cld flag.
> > * config/i386/i386.opt (mcld): New option.
> > * config/i386/i386.c (ix86_expand_prologue): Emit cld insn if
> > TARGET_CLD and ix86_current_function_needs_cld.
> > (override_options): Use -mcld by default for 32-bit code if
> > USE_IX86_CLD.
> >
> >OK for 4.3 branch?
It should go to trunk too.
> Build parts are ok for 4.3 and trunk, except that you could merge the
> case statement you added with the one before it (yes, that involves a
> small duplication). Those parts are preapproved with this change.
>
> The rest, you can approve yourself (though IIUC you want a word from the
> branch RMs).
Yes, it is ok for the branch.
Jakub
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH v2, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
2008-05-21 9:00 ` Richard Guenther
@ 2008-05-21 9:14 ` Uros Bizjak
2008-05-21 10:52 ` Uros Bizjak
1 sibling, 0 replies; 12+ messages in thread
From: Uros Bizjak @ 2008-05-21 9:14 UTC (permalink / raw)
To: Richard Guenther
Cc: Mark Mitchell, Jakub Jelinek, GCC Patches, Paolo Bonzini,
Richard Guenther
[-- Attachment #1: Type: text/plain, Size: 878 bytes --]
On Wed, May 21, 2008 at 10:35 AM, Richard Guenther
<richard.guenther@gmail.com> wrote:
>>> Per our off-list RM discussions, let's not worry too much about the defaults
>>> at this point. Whether --enable-cld is the default or not, and whether
>>> --enable-cld enables -mcld by default in 64-bit mode or not, are topics we
>>> can and should debate, but let's get the core technology in.
>>
>> Attached patch adds all suggestions by Jakub and Paolo. Patch works as
>> advertised, bootstrap and regression test on i686-pc-linux with
>> --enable-cld is in progress.
> This is ok for the mainline and the branch if you add documentation for both the
> configure option and the new target option.
I have committed attached patch to the branch.
Documentation will be written in a day or two in order to get core
functionality into the branch ASAP (as per Mark's request).
Uros.
[-- Attachment #2: cld.diff.txt --]
[-- Type: text/plain, Size: 7297 bytes --]
Index: configure
===================================================================
--- configure (revision 135707)
+++ configure (working copy)
@@ -898,6 +898,7 @@ Optional Features:
--enable-sjlj-exceptions
arrange to use setjmp/longjmp exception handling
--enable-secureplt enable -msecure-plt by default for PowerPC
+ --enable-cld enable -mcld by default for 32bit x86
--disable-win32-registry
disable lookup of installation paths in the
Registry on Windows hosts
@@ -12951,6 +12952,14 @@ if test "${enable_secureplt+set}" = set;
fi;
+# Check whether --enable-cld or --disable-cld was given.
+if test "${enable_cld+set}" = set; then
+ enableval="$enable_cld"
+
+else
+ enable_cld=no
+fi;
+
# Windows32 Registry support for specifying GCC installation paths.
# Check whether --enable-win32-registry or --disable-win32-registry was given.
if test "${enable_win32_registry+set}" = set; then
Index: configure.ac
===================================================================
--- configure.ac (revision 135707)
+++ configure.ac (working copy)
@@ -1548,6 +1548,10 @@ AC_ARG_ENABLE(secureplt,
[ --enable-secureplt enable -msecure-plt by default for PowerPC],
[], [])
+AC_ARG_ENABLE(cld,
+[ --enable-cld enable -mcld by default for 32bit x86], [],
+[enable_cld=no])
+
# Windows32 Registry support for specifying GCC installation paths.
AC_ARG_ENABLE(win32-registry,
[ --disable-win32-registry
Index: config.gcc
===================================================================
--- config.gcc (revision 135707)
+++ config.gcc (working copy)
@@ -392,8 +392,16 @@ then
fi
case ${target} in
+i[34567]86-*-*)
+ if test $enable_cld = yes; then
+ tm_defines="${tm_defines} USE_IX86_CLD=1"
+ fi
+ ;;
x86_64-*-*)
tm_file="i386/biarch64.h ${tm_file}"
+ if test $enable_cld = yes; then
+ tm_defines="${tm_defines} USE_IX86_CLD=1"
+ fi
;;
esac
Index: config/i386/i386.h
===================================================================
--- config/i386/i386.h (revision 135707)
+++ config/i386/i386.h (working copy)
@@ -2443,8 +2443,9 @@ struct machine_function GTY(())
int save_varrargs_registers;
int accesses_prev_frame;
int optimize_mode_switching[MAX_386_ENTITIES];
- /* Set by ix86_compute_frame_layout and used by prologue/epilogue expander to
- determine the style used. */
+ int needs_cld;
+ /* Set by ix86_compute_frame_layout and used by prologue/epilogue
+ expander to determine the style used. */
int use_fast_prologue_epilogue;
/* Number of saved registers USE_FAST_PROLOGUE_EPILOGUE has been computed
for. */
@@ -2464,6 +2465,7 @@ struct machine_function GTY(())
#define ix86_stack_locals (cfun->machine->stack_locals)
#define ix86_save_varrargs_registers (cfun->machine->save_varrargs_registers)
#define ix86_optimize_mode_switching (cfun->machine->optimize_mode_switching)
+#define ix86_current_function_needs_cld (cfun->machine->needs_cld)
#define ix86_tls_descriptor_calls_expanded_in_cfun \
(cfun->machine->tls_descriptor_call_expanded_p)
/* Since tls_descriptor_call_expanded is not cleared, even if all TLS
Index: config/i386/i386.md
===================================================================
--- config/i386/i386.md (revision 135707)
+++ config/i386/i386.md (working copy)
@@ -205,6 +205,7 @@
(UNSPECV_XCHG 12)
(UNSPECV_LOCK 13)
(UNSPECV_PROLOGUE_USE 14)
+ (UNSPECV_CLD 15)
])
;; Constants to represent pcomtrue/pcomfalse variants
@@ -18514,6 +18515,14 @@
\f
;; Block operation instructions
+(define_insn "cld"
+ [(unspec_volatile [(const_int 0)] UNSPECV_CLD)]
+ ""
+ "cld"
+ [(set_attr "length" "1")
+ (set_attr "length_immediate" "0")
+ (set_attr "modrm" "0")])
+
(define_expand "movmemsi"
[(use (match_operand:BLK 0 "memory_operand" ""))
(use (match_operand:BLK 1 "memory_operand" ""))
@@ -18586,7 +18595,7 @@
(set (match_operand 2 "register_operand" "")
(match_operand 5 "" ""))])]
"TARGET_SINGLE_STRINGOP || optimize_size"
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strmovdi_rex_1"
[(set (mem:DI (match_operand:DI 2 "register_operand" "0"))
@@ -18703,7 +18712,7 @@
(match_operand 3 "memory_operand" ""))
(use (match_dup 4))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*rep_movdi_rex64"
[(set (match_operand:DI 2 "register_operand" "=c") (const_int 0))
@@ -18863,7 +18872,7 @@
(set (match_operand 0 "register_operand" "")
(match_operand 3 "" ""))])]
"TARGET_SINGLE_STRINGOP || optimize_size"
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strsetdi_rex_1"
[(set (mem:DI (match_operand:DI 1 "register_operand" "0"))
@@ -18957,7 +18966,7 @@
(use (match_operand 3 "register_operand" ""))
(use (match_dup 1))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*rep_stosdi_rex64"
[(set (match_operand:DI 1 "register_operand" "=c") (const_int 0))
@@ -19133,7 +19142,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (match_dup 2))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*cmpstrnqi_nz_1"
[(set (reg:CC FLAGS_REG)
@@ -19180,7 +19189,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (match_dup 2))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*cmpstrnqi_1"
[(set (reg:CC FLAGS_REG)
@@ -19249,7 +19258,7 @@
(clobber (match_operand 1 "register_operand" ""))
(clobber (reg:CC FLAGS_REG))])]
""
- "")
+ "ix86_current_function_needs_cld = 1;")
(define_insn "*strlenqi_1"
[(set (match_operand:SI 0 "register_operand" "=&c")
Index: config/i386/i386.opt
===================================================================
--- config/i386/i386.opt (revision 135707)
+++ config/i386/i386.opt (working copy)
@@ -250,6 +250,10 @@ Support SSE5 built-in functions and code
;; Instruction support
+mcld
+Target Report Mask(CLD)
+Generate cld instruction in the function prologue.
+
mabm
Target Report RejectNegative Var(x86_abm)
Support code generation of Advanced Bit Manipulation (ABM) instructions.
Index: config/i386/i386.c
===================================================================
--- config/i386/i386.c (revision 135707)
+++ config/i386/i386.c (working copy)
@@ -2733,6 +2733,12 @@ override_options (void)
can be optimized to ap = __builtin_next_arg (0). */
if (!TARGET_64BIT || TARGET_64BIT_MS_ABI)
targetm.expand_builtin_va_start = NULL;
+
+#ifdef USE_IX86_CLD
+ /* Use -mcld by default for 32-bit code if configured with --enable-cld. */
+ if (!TARGET_64BIT)
+ target_flags |= MASK_CLD & ~target_flags_explicit;
+#endif
}
\f
/* Return true if this goes in large data/bss. */
@@ -6485,6 +6491,10 @@ ix86_expand_prologue (void)
emit_insn (gen_prologue_use (pic_offset_table_rtx));
emit_insn (gen_blockage ());
}
+
+ /* Emit cld instruction if stringops are used in the function. */
+ if (TARGET_CLD && ix86_current_function_needs_cld)
+ emit_insn (gen_cld ());
}
/* Emit code to restore saved registers using MOV insns. First register
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH v2, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore.
2008-05-21 9:00 ` Richard Guenther
2008-05-21 9:14 ` Uros Bizjak
@ 2008-05-21 10:52 ` Uros Bizjak
1 sibling, 0 replies; 12+ messages in thread
From: Uros Bizjak @ 2008-05-21 10:52 UTC (permalink / raw)
To: Richard Guenther
Cc: Mark Mitchell, Jakub Jelinek, GCC Patches, Paolo Bonzini,
Richard Guenther
On Wed, May 21, 2008 at 10:35 AM, Richard Guenther
<richard.guenther@gmail.com> wrote:
> If I specify --enable-cld on x86_64 I get clds emitted for -m32 but
> not for -m64,
> correct? But if I specify -mcld I get clds on both -m32 and -m64? I think this
> is reasonable behavior, but the documentation should probably explicitly
> mention this.
This is exactly how -mcld works now. Additionally, you can add
-mno-cld to suppress generation of cld on --enable-cld configured
32bit targets.
Uros.
^ permalink raw reply [flat|nested] 12+ messages in thread
end of thread, other threads:[~2008-05-21 9:39 UTC | newest]
Thread overview: 12+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2008-05-20 13:07 [PATCH, i386]: Fix PR 36079 [4.3/4.4 Regression] cld instruction is not emitted anymore Uros Bizjak
2008-05-20 16:27 ` Richard Guenther
2008-05-20 16:29 ` Jakub Jelinek
2008-05-20 16:43 ` Jakub Jelinek
2008-05-20 17:03 ` Paolo Bonzini
2008-05-21 2:58 ` Mark Mitchell
2008-05-21 8:24 ` [PATCH v2, " Uros Bizjak
2008-05-21 8:25 ` Paolo Bonzini
2008-05-21 9:04 ` Jakub Jelinek
2008-05-21 9:00 ` Richard Guenther
2008-05-21 9:14 ` Uros Bizjak
2008-05-21 10:52 ` Uros Bizjak
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).