Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
Segher, Thanks for your comments. I have modified the patch according to your advice and committed. On 24/11/2020 上午 6:25, Segher Boessenkool wrote: Hi! Sorry this took so long. On Thu, Oct 15, 2020 at 04:46:01PM +0800, HAO CHEN GUI wrote: On 29/9/2020 上午 6:46, Segher Boessenkool wrote: [ Please do not insert patches into discussions ] So, please send new patches as a new thread. This makes it much easier to handle. If you want to reply to something in the old thread just do it there, but send the new patch not as a reply. * config/rs6000/rs6000.opt (mrelative-jumptables): Add mrelative-jumptables. Just "New." or "New flag." or similar? +/* Implement TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC. + Marco relative jumpstables indicates addr_diff_vec or addr_vec. */ The documentation says Return true to generate ADDR_DIF_VEC table or false to generate ADDR_VEC table for jumps in case of -fPIC. but you can just leave out that last line completely. + if (rs6000_relative_jumptables) + { + if (TARGET_32BIT) + jump = gen_tablejumpsi_nospec (operands[0], operands[1], ccreg); + else + jump = gen_tablejumpdi_nospec (operands[0], operands[1], ccreg); + } else - jump = gen_tablejumpdi_nospec (operands[0], operands[1], ccreg); + jump = gen_tablejump_absolute_nospec (Pmode, operands[0], operands[1], + ccreg); emit_jump_insn (jump); It now is @tablejump_nospec so this code will become something like if (rs6000_relative_jumptables) jump = gen_tablejumpsi_nospec (Pmode, operands[0], operands[1], ccreg); else jump = gen_tablejump_absolute_nospec (Pmode, operands[0], operands[1], ccreg); emit_jump_insn (jump); Okay for trunk with those changes. Thank you! Segher
Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
Hi! Sorry this took so long. On Thu, Oct 15, 2020 at 04:46:01PM +0800, HAO CHEN GUI wrote: > On 29/9/2020 上午 6:46, Segher Boessenkool wrote: > >[ Please do not insert patches into discussions ] So, please send new patches as a new thread. This makes it much easier to handle. If you want to reply to something in the old thread just do it there, but send the new patch not as a reply. > * config/rs6000/rs6000.opt (mrelative-jumptables): Add > mrelative-jumptables. Just "New." or "New flag." or similar? > +/* Implement TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC. > + Marco relative jumpstables indicates addr_diff_vec or addr_vec. */ The documentation says Return true to generate ADDR_DIF_VEC table or false to generate ADDR_VEC table for jumps in case of -fPIC. but you can just leave out that last line completely. > + if (rs6000_relative_jumptables) > + { > + if (TARGET_32BIT) > + jump = gen_tablejumpsi_nospec (operands[0], operands[1], ccreg); > + else > + jump = gen_tablejumpdi_nospec (operands[0], operands[1], ccreg); > + } >else > - jump = gen_tablejumpdi_nospec (operands[0], operands[1], ccreg); > + jump = gen_tablejump_absolute_nospec (Pmode, operands[0], operands[1], > + ccreg); >emit_jump_insn (jump); It now is @tablejump_nospec so this code will become something like if (rs6000_relative_jumptables) jump = gen_tablejumpsi_nospec (Pmode, operands[0], operands[1], ccreg); else jump = gen_tablejump_absolute_nospec (Pmode, operands[0], operands[1], ccreg); emit_jump_insn (jump); Okay for trunk with those changes. Thank you! Segher
Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
Hi, Gentle ping this: https://gcc.gnu.org/pipermail/gcc-patches/2020-October/556236.html Thanks. Gui Haochen On 6/11/2020 上午 9:02, HAO CHEN GUI wrote: Hi, Gentle ping this: https://gcc.gnu.org/pipermail/gcc-patches/2020-October/556236.html Thanks. Gui Haochen On 15/10/2020 下午 4:46, HAO CHEN GUI wrote: Segher, I re-wrote the patch based on parameterized name. The attachments are the patch diff file and change log file. Bootstrapped and tested on powerpc64le-linux-gnu with no regressions. Is this okay for trunk? Any recommendations? Thanks a lot. On 29/9/2020 上午 6:46, Segher Boessenkool wrote: Hi hao Chen, On Wed, Sep 09, 2020 at 04:55:29PM +0800, HAO CHEN GUI wrote: Thanks for your advice. I removed macros defined in linux64.h and linux.h. So they take relative jump tables by default. When no-relative-jumptables is set, the absolute jump tables are taken. All things relevant to section relocations are put in another patch. Thanks again. [ Please do not insert patches into discussions ] +/* Implement TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC. + Return true if rs6000_relative_jumptables is set. */ Don't just say what the code does (we can see that ;-) ); say *why*. Of course it is terribly simple in this case, so maybe just that first line is plenty. +/* Specify the machine mode that this machine uses + for the index in the tablejump instruction. */ +#define CASE_VECTOR_MODE \ + (TARGET_32BIT || rs6000_relative_jumptables ? SImode : Pmode) If TARGET_32BIT is true, SImode and Pmode are the same, so this is simpler said as #define CASE_VECTOR_MODE (rs6000_relative_jumptables ? SImode : Pmode) I'll have the tablejump* patterns converted to paremeterized names hopefully tonight or tomorrow, which will make your patch much easier to read. It looks like it will be fine, thanks :-) Segher
Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
Hi, Gentle ping this: https://gcc.gnu.org/pipermail/gcc-patches/2020-October/556236.html Thanks. Gui Haochen On 6/11/2020 上午 9:02, HAO CHEN GUI wrote: Hi, Gentle ping this: https://gcc.gnu.org/pipermail/gcc-patches/2020-October/556236.html Thanks. Gui Haochen On 15/10/2020 下午 4:46, HAO CHEN GUI wrote: Segher, I re-wrote the patch based on parameterized name. The attachments are the patch diff file and change log file. Bootstrapped and tested on powerpc64le-linux-gnu with no regressions. Is this okay for trunk? Any recommendations? Thanks a lot. On 29/9/2020 上午 6:46, Segher Boessenkool wrote: Hi hao Chen, On Wed, Sep 09, 2020 at 04:55:29PM +0800, HAO CHEN GUI wrote: Thanks for your advice. I removed macros defined in linux64.h and linux.h. So they take relative jump tables by default. When no-relative-jumptables is set, the absolute jump tables are taken. All things relevant to section relocations are put in another patch. Thanks again. [ Please do not insert patches into discussions ] +/* Implement TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC. + Return true if rs6000_relative_jumptables is set. */ Don't just say what the code does (we can see that ;-) ); say *why*. Of course it is terribly simple in this case, so maybe just that first line is plenty. +/* Specify the machine mode that this machine uses + for the index in the tablejump instruction. */ +#define CASE_VECTOR_MODE \ + (TARGET_32BIT || rs6000_relative_jumptables ? SImode : Pmode) If TARGET_32BIT is true, SImode and Pmode are the same, so this is simpler said as #define CASE_VECTOR_MODE (rs6000_relative_jumptables ? SImode : Pmode) I'll have the tablejump* patterns converted to paremeterized names hopefully tonight or tomorrow, which will make your patch much easier to read. It looks like it will be fine, thanks :-) Segher * config/rs6000/rs6000-protos.h (rs6000_output_addr_vec_elt): Declare. * config/rs6000/rs6000.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Define. (rs6000_gen_pic_addr_diff_vec, rs6000_output_addr_vec_elt): Implement. * config/rs6000/rs6000.h (CASE_VECTOR_PC_RELATIVE, CASE_VECTOR_MODE, ASM_OUTPUT_ADDR_VEC_ELT): Define. * config/rs6000/rs6000.md (tablejump_absolute, tablejump_absolute_nospec): New expanders. * config/rs6000/rs6000.opt (mrelative-jumptables): Add mrelative-jumptables. diff --git a/gcc/config/rs6000/rs6000-protos.h b/gcc/config/rs6000/rs6000-protos.h index 25fa5dd57cd..49ab51042c5 100644 --- a/gcc/config/rs6000/rs6000-protos.h +++ b/gcc/config/rs6000/rs6000-protos.h @@ -155,6 +155,7 @@ extern void rs6000_split_logical (rtx [], enum rtx_code, bool, bool, bool); extern bool rs6000_function_pcrel_p (struct function *); extern bool rs6000_pcrel_p (void); extern bool rs6000_fndecl_pcrel_p (const_tree); +extern void rs6000_output_addr_vec_elt (FILE *, int); /* Different PowerPC instruction formats that are used by GCC. There are various other instruction formats used by the PowerPC hardware, but these diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c index b58eeae2b98..7e441f7e730 100644 --- a/gcc/config/rs6000/rs6000.c +++ b/gcc/config/rs6000/rs6000.c @@ -1367,6 +1367,9 @@ static const struct attribute_spec rs6000_attribute_table[] = #undef TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA #define TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA rs6000_output_addr_const_extra +#undef TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC +#define TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC rs6000_gen_pic_addr_diff_vec + #undef TARGET_LEGITIMIZE_ADDRESS #define TARGET_LEGITIMIZE_ADDRESS rs6000_legitimize_address @@ -27053,6 +27056,27 @@ rs6000_emit_xxspltidp_v2df (rtx dst, long value) emit_insn( gen_xxspltidp_v2df_inst (dst, GEN_INT (value))); } +/* Implement TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC. + Marco relative jumpstables indicates addr_diff_vec or addr_vec. */ + +static bool +rs6000_gen_pic_addr_diff_vec (void) +{ + return rs6000_relative_jumptables; +} + +void +rs6000_output_addr_vec_elt (FILE *file, int value) +{ + const char *directive = TARGET_64BIT ? DOUBLE_INT_ASM_OP : "\t.long\t"; + char buf[100]; + + fprintf (file, "%s", directive); + ASM_GENERATE_INTERNAL_LABEL (buf, "L", value); + assemble_name (file, buf); + fprintf (file, "\n"); +} + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-rs6000.h" diff --git a/gcc/config/rs6000/rs6000.h b/gcc/config/rs6000/rs6000.h index bbd8060e143..4a948e25f71 100644 --- a/gcc/config/rs6000/rs6000.h +++ b/gcc/config/rs6000/rs6000.h @@ -1754,15 +1754,15 @@ typedef struct rs6000_args /* #define LEGITIMATE_PIC_OPERAND_P (X) */ -/* Specify the machine mode that this machine uses - for the index in the tablejump instruction. */ -#define CASE_VECTOR_MODE SImode - /* Define as C expression which evaluates to nonzero if the tablejump instruction expects the table to contain offsets from the address of the table
Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
Hi, Gentle ping this: https://gcc.gnu.org/pipermail/gcc-patches/2020-October/556236.html Thanks. Gui Haochen On 15/10/2020 下午 4:46, HAO CHEN GUI wrote: Segher, I re-wrote the patch based on parameterized name. The attachments are the patch diff file and change log file. Bootstrapped and tested on powerpc64le-linux-gnu with no regressions. Is this okay for trunk? Any recommendations? Thanks a lot. On 29/9/2020 上午 6:46, Segher Boessenkool wrote: Hi hao Chen, On Wed, Sep 09, 2020 at 04:55:29PM +0800, HAO CHEN GUI wrote: Thanks for your advice. I removed macros defined in linux64.h and linux.h. So they take relative jump tables by default. When no-relative-jumptables is set, the absolute jump tables are taken. All things relevant to section relocations are put in another patch. Thanks again. [ Please do not insert patches into discussions ] +/* Implement TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC. + Return true if rs6000_relative_jumptables is set. */ Don't just say what the code does (we can see that ;-) ); say *why*. Of course it is terribly simple in this case, so maybe just that first line is plenty. +/* Specify the machine mode that this machine uses + for the index in the tablejump instruction. */ +#define CASE_VECTOR_MODE \ + (TARGET_32BIT || rs6000_relative_jumptables ? SImode : Pmode) If TARGET_32BIT is true, SImode and Pmode are the same, so this is simpler said as #define CASE_VECTOR_MODE (rs6000_relative_jumptables ? SImode : Pmode) I'll have the tablejump* patterns converted to paremeterized names hopefully tonight or tomorrow, which will make your patch much easier to read. It looks like it will be fine, thanks :-) Segher
Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
Segher, I re-wrote the patch based on parameterized name. The attachments are the patch diff file and change log file. Bootstrapped and tested on powerpc64le-linux-gnu with no regressions. Is this okay for trunk? Any recommendations? Thanks a lot. On 29/9/2020 上午 6:46, Segher Boessenkool wrote: Hi hao Chen, On Wed, Sep 09, 2020 at 04:55:29PM +0800, HAO CHEN GUI wrote: Thanks for your advice. I removed macros defined in linux64.h and linux.h. So they take relative jump tables by default. When no-relative-jumptables is set, the absolute jump tables are taken. All things relevant to section relocations are put in another patch. Thanks again. [ Please do not insert patches into discussions ] +/* Implement TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC. + Return true if rs6000_relative_jumptables is set. */ Don't just say what the code does (we can see that ;-) ); say *why*. Of course it is terribly simple in this case, so maybe just that first line is plenty. +/* Specify the machine mode that this machine uses + for the index in the tablejump instruction. */ +#define CASE_VECTOR_MODE \ + (TARGET_32BIT || rs6000_relative_jumptables ? SImode : Pmode) If TARGET_32BIT is true, SImode and Pmode are the same, so this is simpler said as #define CASE_VECTOR_MODE (rs6000_relative_jumptables ? SImode : Pmode) I'll have the tablejump* patterns converted to paremeterized names hopefully tonight or tomorrow, which will make your patch much easier to read. It looks like it will be fine, thanks :-) Segher * config/rs6000/rs6000-protos.h (rs6000_output_addr_vec_elt): Declare. * config/rs6000/rs6000.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Define. (rs6000_gen_pic_addr_diff_vec, rs6000_output_addr_vec_elt): Implement. * config/rs6000/rs6000.h (CASE_VECTOR_PC_RELATIVE, CASE_VECTOR_MODE, ASM_OUTPUT_ADDR_VEC_ELT): Define. * config/rs6000/rs6000.md (tablejump_absolute, tablejump_absolute_nospec): New expanders. * config/rs6000/rs6000.opt (mrelative-jumptables): Add mrelative-jumptables. diff --git a/gcc/config/rs6000/rs6000-protos.h b/gcc/config/rs6000/rs6000-protos.h index 25fa5dd57cd..49ab51042c5 100644 --- a/gcc/config/rs6000/rs6000-protos.h +++ b/gcc/config/rs6000/rs6000-protos.h @@ -155,6 +155,7 @@ extern void rs6000_split_logical (rtx [], enum rtx_code, bool, bool, bool); extern bool rs6000_function_pcrel_p (struct function *); extern bool rs6000_pcrel_p (void); extern bool rs6000_fndecl_pcrel_p (const_tree); +extern void rs6000_output_addr_vec_elt (FILE *, int); /* Different PowerPC instruction formats that are used by GCC. There are various other instruction formats used by the PowerPC hardware, but these diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c index b58eeae2b98..7e441f7e730 100644 --- a/gcc/config/rs6000/rs6000.c +++ b/gcc/config/rs6000/rs6000.c @@ -1367,6 +1367,9 @@ static const struct attribute_spec rs6000_attribute_table[] = #undef TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA #define TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA rs6000_output_addr_const_extra +#undef TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC +#define TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC rs6000_gen_pic_addr_diff_vec + #undef TARGET_LEGITIMIZE_ADDRESS #define TARGET_LEGITIMIZE_ADDRESS rs6000_legitimize_address @@ -27053,6 +27056,27 @@ rs6000_emit_xxspltidp_v2df (rtx dst, long value) emit_insn( gen_xxspltidp_v2df_inst (dst, GEN_INT (value))); } +/* Implement TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC. + Marco relative jumpstables indicates addr_diff_vec or addr_vec. */ + +static bool +rs6000_gen_pic_addr_diff_vec (void) +{ + return rs6000_relative_jumptables; +} + +void +rs6000_output_addr_vec_elt (FILE *file, int value) +{ + const char *directive = TARGET_64BIT ? DOUBLE_INT_ASM_OP : "\t.long\t"; + char buf[100]; + + fprintf (file, "%s", directive); + ASM_GENERATE_INTERNAL_LABEL (buf, "L", value); + assemble_name (file, buf); + fprintf (file, "\n"); +} + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-rs6000.h" diff --git a/gcc/config/rs6000/rs6000.h b/gcc/config/rs6000/rs6000.h index bbd8060e143..4a948e25f71 100644 --- a/gcc/config/rs6000/rs6000.h +++ b/gcc/config/rs6000/rs6000.h @@ -1754,15 +1754,15 @@ typedef struct rs6000_args /* #define LEGITIMATE_PIC_OPERAND_P (X) */ -/* Specify the machine mode that this machine uses - for the index in the tablejump instruction. */ -#define CASE_VECTOR_MODE SImode - /* Define as C expression which evaluates to nonzero if the tablejump instruction expects the table to contain offsets from the address of the table. Do not define this if the table should contain absolute addresses. */ -#define CASE_VECTOR_PC_RELATIVE 1 +#define CASE_VECTOR_PC_RELATIVE rs6000_relative_jumptables + +/* Specify the machine mode that this machine uses + for the index in the tablejump instruction. */ +#define CASE_VECTOR_MODE (rs6000_relati
Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
Hi hao Chen, On Wed, Sep 09, 2020 at 04:55:29PM +0800, HAO CHEN GUI wrote: > Thanks for your advice. I removed macros defined in linux64.h and > linux.h. So they take relative jump tables by default. When > no-relative-jumptables is set, the absolute jump tables are taken. All > things relevant to section relocations are put in another patch. Thanks > again. [ Please do not insert patches into discussions ] > +/* Implement TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC. > + Return true if rs6000_relative_jumptables is set. */ Don't just say what the code does (we can see that ;-) ); say *why*. Of course it is terribly simple in this case, so maybe just that first line is plenty. > +/* Specify the machine mode that this machine uses > + for the index in the tablejump instruction. */ > +#define CASE_VECTOR_MODE \ > + (TARGET_32BIT || rs6000_relative_jumptables ? SImode : Pmode) If TARGET_32BIT is true, SImode and Pmode are the same, so this is simpler said as #define CASE_VECTOR_MODE (rs6000_relative_jumptables ? SImode : Pmode) I'll have the tablejump* patterns converted to paremeterized names hopefully tonight or tomorrow, which will make your patch much easier to read. It looks like it will be fine, thanks :-) Segher
Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
Segher, Gentle ping this: https://gcc.gnu.org/pipermail/gcc-patches/2020-September/553486.html Thanks Gui Haochen On 9/9/2020 下午 4:55, HAO CHEN GUI wrote: Hi Segher, Thanks for your advice. I removed macros defined in linux64.h and linux.h. So they take relative jump tables by default. When no-relative-jumptables is set, the absolute jump tables are taken. All things relevant to section relocations are put in another patch. Thanks again. On 8/9/2020 上午 5:46, Segher Boessenkool wrote: On Mon, Aug 24, 2020 at 03:48:43PM +0800, HAO CHEN GUI wrote: I'll try to be quicker at reviewing iterations of this -- there is quite some way to go, without me slowing things down! Sigh :-( * config/rs6000/linux.h (rs6000_relative_jumptables): Define. That macro looks like it is variable (or function). *Make* it a variable, please? * config/rs6000/rs6000.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Define Period? (rs6000_gen_pic_addr_diff_vec, rs6000_output_addr_vec_elt): Implement. "New function." * config/rs6000/rs6000.md (absolute_tablejumpsi, absolute_tablejumpsi_nospec, absolute_tablejumpdi, absolute_tablejumpdi_nospec): Add four new expansions. "New define_expands." or "New expanders." * config/rs6000/rs6000.opt (mrelative-jumptables): Add a new option and set rs6000_relative_jumptables to true by default. "rs6000.opt: Add -mrelative-jumptables." +/* Disable relative jump tables for Power Linux. */ +#undef rs6000_relative_jumptables +#define rs6000_relative_jumptables 0 Why? +/* Disable relative jump tables for Power Linux64. */ +#undef rs6000_relative_jumptables +#define rs6000_relative_jumptables 0 (That's not what it's called... Just don't say the "for..." at all? It is clear from what file it is in.) /* Indicate that jump tables go in the text section. */ #undef JUMP_TABLES_IN_TEXT_SECTION -#define JUMP_TABLES_IN_TEXT_SECTION TARGET_64BIT +#define JUMP_TABLES_IN_TEXT_SECTION rs6000_relative_jumptables Not sure that is correct. Maybe the patch using rodata (.data.rel.ro) should be a separate patch? /* Define as C expression which evaluates to nonzero if the tablejump instruction expects the table to contain offsets from the address of the table. Do not define this if the table should contain absolute addresses. */ -#define CASE_VECTOR_PC_RELATIVE 1 +#define CASE_VECTOR_PC_RELATIVE 0 This should depend on the new flag? +/* Specify the machine mode that this machine uses + for the index in the tablejump instruction. */ +#define CASE_VECTOR_MODE \ + (TARGET_32BIT || rs6000_relative_jumptables ? SImode : DImode) rs6000_relative_jumptables ? SImode : Pmode; + if (rs6000_relative_jumptables) + { + if (TARGET_32BIT) + emit_jump_insn (gen_tablejumpsi (operands[0], operands[1])); + else + emit_jump_insn (gen_tablejumpdi (operands[0], operands[1])); + } Hrm, I guess we should make that a parameterized name (future work, don't do it now :-) ) +(define_expand "absolute_tablejumpsi" Don't prefix names; it should start with "tablejump". Segher * config/rs6000/rs6000-protos.h (rs6000_output_addr_vec_elt): Declare. * config/rs6000/rs6000.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Define. (rs6000_gen_pic_addr_diff_vec, rs6000_output_addr_vec_elt): Implement. * config/rs6000/rs6000.h (CASE_VECTOR_PC_RELATIVE, CASE_VECTOR_MODE, ASM_OUTPUT_ADDR_VEC_ELT): Define. * config/rs6000/rs6000.md (tablejumpsi_absolute, tablejumpsi_nospec_absolute, tablejumpdi_absolute, tablejumpdi_nospec_absolute): New expanders. * config/rs6000/rs6000.opt (mrelative-jumptables): Add mrelative-jumptables. diff --git a/gcc/config/rs6000/rs6000-protos.h b/gcc/config/rs6000/rs6000-protos.h index 5508484ba19..62564dd67f2 100644 --- a/gcc/config/rs6000/rs6000-protos.h +++ b/gcc/config/rs6000/rs6000-protos.h @@ -155,6 +155,8 @@ extern void rs6000_split_logical (rtx [], enum rtx_code, bool, bool, bool); extern bool rs6000_pcrel_p (struct function *); extern bool rs6000_fndecl_pcrel_p (const_tree); +extern void rs6000_output_addr_vec_elt (FILE *, int); + /* Different PowerPC instruction formats that are used by GCC. There are various other instruction formats used by the PowerPC hardware, but these formats are not currently used by GCC. */ diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c index 58f5d780603..94d1e650b94 100644 --- a/gcc/config/rs6000/rs6000.c +++ b/gcc/config/rs6000/rs6000.c @@ -1369,6 +1369,9 @@ static const struct attribute_spec rs6000_attribute_table[] = #undef TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA #define TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA rs6000_output_addr_const_extra +#undef TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC +#define TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC rs6000_gen_pic_addr_diff_vec + #undef TARGET_LEGITIMIZE_ADDRESS #define TARGET_LEGITIMIZ
Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
Hi Segher, Thanks for your advice. I removed macros defined in linux64.h and linux.h. So they take relative jump tables by default. When no-relative-jumptables is set, the absolute jump tables are taken. All things relevant to section relocations are put in another patch. Thanks again. On 8/9/2020 上午 5:46, Segher Boessenkool wrote: On Mon, Aug 24, 2020 at 03:48:43PM +0800, HAO CHEN GUI wrote: I'll try to be quicker at reviewing iterations of this -- there is quite some way to go, without me slowing things down! Sigh :-( * config/rs6000/linux.h (rs6000_relative_jumptables): Define. That macro looks like it is variable (or function). *Make* it a variable, please? * config/rs6000/rs6000.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Define Period? (rs6000_gen_pic_addr_diff_vec, rs6000_output_addr_vec_elt): Implement. "New function." * config/rs6000/rs6000.md (absolute_tablejumpsi, absolute_tablejumpsi_nospec, absolute_tablejumpdi, absolute_tablejumpdi_nospec): Add four new expansions. "New define_expands." or "New expanders." * config/rs6000/rs6000.opt (mrelative-jumptables): Add a new option and set rs6000_relative_jumptables to true by default. "rs6000.opt: Add -mrelative-jumptables." +/* Disable relative jump tables for Power Linux. */ +#undef rs6000_relative_jumptables +#define rs6000_relative_jumptables 0 Why? +/* Disable relative jump tables for Power Linux64. */ +#undef rs6000_relative_jumptables +#define rs6000_relative_jumptables 0 (That's not what it's called... Just don't say the "for..." at all? It is clear from what file it is in.) /* Indicate that jump tables go in the text section. */ #undef JUMP_TABLES_IN_TEXT_SECTION -#define JUMP_TABLES_IN_TEXT_SECTION TARGET_64BIT +#define JUMP_TABLES_IN_TEXT_SECTION rs6000_relative_jumptables Not sure that is correct. Maybe the patch using rodata (.data.rel.ro) should be a separate patch? /* Define as C expression which evaluates to nonzero if the tablejump instruction expects the table to contain offsets from the address of the table. Do not define this if the table should contain absolute addresses. */ -#define CASE_VECTOR_PC_RELATIVE 1 +#define CASE_VECTOR_PC_RELATIVE 0 This should depend on the new flag? +/* Specify the machine mode that this machine uses + for the index in the tablejump instruction. */ +#define CASE_VECTOR_MODE \ + (TARGET_32BIT || rs6000_relative_jumptables ? SImode : DImode) rs6000_relative_jumptables ? SImode : Pmode; + if (rs6000_relative_jumptables) + { + if (TARGET_32BIT) + emit_jump_insn (gen_tablejumpsi (operands[0], operands[1])); + else + emit_jump_insn (gen_tablejumpdi (operands[0], operands[1])); + } Hrm, I guess we should make that a parameterized name (future work, don't do it now :-) ) +(define_expand "absolute_tablejumpsi" Don't prefix names; it should start with "tablejump". Segher * config/rs6000/rs6000-protos.h (rs6000_output_addr_vec_elt): Declare. * config/rs6000/rs6000.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Define. (rs6000_gen_pic_addr_diff_vec, rs6000_output_addr_vec_elt): Implement. * config/rs6000/rs6000.h (CASE_VECTOR_PC_RELATIVE, CASE_VECTOR_MODE, ASM_OUTPUT_ADDR_VEC_ELT): Define. * config/rs6000/rs6000.md (tablejumpsi_absolute, tablejumpsi_nospec_absolute, tablejumpdi_absolute, tablejumpdi_nospec_absolute): New expanders. * config/rs6000/rs6000.opt (mrelative-jumptables): Add mrelative-jumptables. diff --git a/gcc/config/rs6000/rs6000-protos.h b/gcc/config/rs6000/rs6000-protos.h index 5508484ba19..62564dd67f2 100644 --- a/gcc/config/rs6000/rs6000-protos.h +++ b/gcc/config/rs6000/rs6000-protos.h @@ -155,6 +155,8 @@ extern void rs6000_split_logical (rtx [], enum rtx_code, bool, bool, bool); extern bool rs6000_pcrel_p (struct function *); extern bool rs6000_fndecl_pcrel_p (const_tree); +extern void rs6000_output_addr_vec_elt (FILE *, int); + /* Different PowerPC instruction formats that are used by GCC. There are various other instruction formats used by the PowerPC hardware, but these formats are not currently used by GCC. */ diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c index 58f5d780603..94d1e650b94 100644 --- a/gcc/config/rs6000/rs6000.c +++ b/gcc/config/rs6000/rs6000.c @@ -1369,6 +1369,9 @@ static const struct attribute_spec rs6000_attribute_table[] = #undef TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA #define TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA rs6000_output_addr_const_extra +#undef TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC +#define TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC rs6000_gen_pic_addr_diff_vec + #undef TARGET_LEGITIMIZE_ADDRESS #define TARGET_LEGITIMIZE_ADDRESS rs6000_legitimize_address @@ -26494,6 +26497,27 @@ rs6000_cannot_substitute_mem_equiv_p (rtx mem) return fal
Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
On Mon, Aug 24, 2020 at 03:48:43PM +0800, HAO CHEN GUI wrote: > >I'll try to be quicker at reviewing iterations of this -- there is quite > >some way to go, without me slowing things down! Sigh :-( > * config/rs6000/linux.h (rs6000_relative_jumptables): Define. That macro looks like it is variable (or function). *Make* it a variable, please? > * config/rs6000/rs6000.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): > Define Period? > (rs6000_gen_pic_addr_diff_vec, rs6000_output_addr_vec_elt): Implement. "New function." > * config/rs6000/rs6000.md (absolute_tablejumpsi, > absolute_tablejumpsi_nospec, absolute_tablejumpdi, > absolute_tablejumpdi_nospec): Add four new expansions. "New define_expands." or "New expanders." > * config/rs6000/rs6000.opt (mrelative-jumptables): Add a new option and > set rs6000_relative_jumptables to true by default. "rs6000.opt: Add -mrelative-jumptables." > +/* Disable relative jump tables for Power Linux. */ > +#undef rs6000_relative_jumptables > +#define rs6000_relative_jumptables 0 Why? > +/* Disable relative jump tables for Power Linux64. */ > +#undef rs6000_relative_jumptables > +#define rs6000_relative_jumptables 0 (That's not what it's called... Just don't say the "for..." at all? It is clear from what file it is in.) > /* Indicate that jump tables go in the text section. */ > #undef JUMP_TABLES_IN_TEXT_SECTION > -#define JUMP_TABLES_IN_TEXT_SECTION TARGET_64BIT > +#define JUMP_TABLES_IN_TEXT_SECTION rs6000_relative_jumptables Not sure that is correct. Maybe the patch using rodata (.data.rel.ro) should be a separate patch? > /* Define as C expression which evaluates to nonzero if the tablejump > instruction expects the table to contain offsets from the address of the > table. > Do not define this if the table should contain absolute addresses. */ > -#define CASE_VECTOR_PC_RELATIVE 1 > +#define CASE_VECTOR_PC_RELATIVE 0 This should depend on the new flag? > +/* Specify the machine mode that this machine uses > + for the index in the tablejump instruction. */ > +#define CASE_VECTOR_MODE \ > + (TARGET_32BIT || rs6000_relative_jumptables ? SImode : DImode) rs6000_relative_jumptables ? SImode : Pmode; > + if (rs6000_relative_jumptables) > + { > + if (TARGET_32BIT) > + emit_jump_insn (gen_tablejumpsi (operands[0], operands[1])); > + else > + emit_jump_insn (gen_tablejumpdi (operands[0], operands[1])); > + } Hrm, I guess we should make that a parameterized name (future work, don't do it now :-) ) > +(define_expand "absolute_tablejumpsi" Don't prefix names; it should start with "tablejump". Segher
Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
Hi, I revised the patch according to the advice. The attachment is the change log and diff file. Thanks a lot. On 20/8/2020 上午 8:14, Segher Boessenkool wrote: Hi! Sorry this took so long to review. "I lost track of this patch", what can I say :-/ On Fri, Aug 14, 2020 at 03:31:05PM +0800, HAO CHEN GUI wrote: This patch adds non-relative jump table support on Power Linux. It implements ASM_OUTPUT_ADDR_VEC_ELT and adds four new expansions for non-relative jump table in rs6000.md. It also defines a rs6000 option(mrelative-jumptables). If it's set to false, the non-relative jump table is picked up even PIC flag is set. diff --git a/gcc/config/rs6000/linux64.h b/gcc/config/rs6000/linux64.h index 34776c8421e..626f2201d96 100644 --- a/gcc/config/rs6000/linux64.h +++ b/gcc/config/rs6000/linux64.h @@ -324,7 +324,10 @@ extern int dot_symbols; /* Indicate that jump tables go in the text section. */ #undef JUMP_TABLES_IN_TEXT_SECTION -#define JUMP_TABLES_IN_TEXT_SECTION TARGET_64BIT +#define JUMP_TABLES_IN_TEXT_SECTION rs6000_relative_jumptables + +#undef JUMP_TABLES_IN_RELRO +#define JUMP_TABLES_IN_RELRO !rs6000_relative_jumptables The third choice is .rodata. .data.rel.ro can be considered to be a sub-case of that, but not the other way around. +extern void rs6000_output_addr_vec_elt(FILE *, int); (space before opening paren) +/* Implement TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC. Return true if rs6000_relative_jumptables is set. */ The maximum line length is 80 characters. Two spaces after a full stop. +/* Enable non-relative jump tables for linux ELF. */ +#ifdef TARGET_ELF +#if !TARGET_AIX +#undef rs6000_relative_jumptables +#define rs6000_relative_jumptables 0 +#endif +#endif AIX never is ELF, but many other things *are* ELF. Do all of those handle .data.rel.ro, including do they have enough OS support for it? This also changes the default for ELF systems? Let's not do that (in this patch, that is). +/* Specify the machine mode that this machine uses + for the index in the tablejump instruction. */ +#define CASE_VECTOR_MODE \ + (TARGET_32BIT || rs6000_relative_jumptables ? SImode : DImode) #define CASE_VECTOR_MODE (rs6000_relative_jumptables ? SImode : Pmode) +/* This is how to output an element of a case-vector + that is non-relative. */ (no space at end of line; two spaces after full stop, also before */) +#define ASM_OUTPUT_ADDR_VEC_ELT(FILE, VALUE) \ + rs6000_output_addr_vec_elt((FILE), (VALUE)) Space before first ( please (on the second line, on the first you cannot -- macro definitions are the only exception). diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md index 0aa5265d199..2f28da0ea1f 100644 --- a/gcc/config/rs6000/rs6000.md +++ b/gcc/config/rs6000/rs6000.md @@ -12669,18 +12669,32 @@ if (rs6000_speculate_indirect_jumps) { if (TARGET_32BIT) - emit_jump_insn (gen_tablejumpsi (operands[0], operands[1])); - else +{ + if (rs6000_relative_jumptables) +emit_jump_insn (gen_tablejumpsi (operands[0], operands[1])); + else + emit_jump_insn (gen_nonrelative_tablejumpsi (operands[0], operands[1])); Please fix the indent while you're at it (tabs shouldn't follow spaces). Oh, and line too long. +} + else if (rs6000_relative_jumptables) emit_jump_insn (gen_tablejumpdi (operands[0], operands[1])); + else +emit_jump_insn (gen_nonrelative_tablejumpdi (operands[0], operands[1])); Please put the "relative" case on the outside, and the "32 or 64" if on the inside. +(define_expand "nonrelative_tablejumpsi" Maybe call it "absolute_tablejumpsi"? Not sure :-) + "TARGET_32BIT && rs6000_speculate_indirect_jumps && !rs6000_relative_jumptables" Line too long (more follow, I won't point all out, please check :-) ) + { + operands[2] = gen_reg_rtx (SImode); + }) No spaces before { and } here, those are flush left like in outer braces in a C function are. +(define_expand "nonrelative_tablejumpsi_nospec" + [(set (match_dup 3) +(match_operand:SI 0 "gpc_reg_operand" "r")) + (parallel [(set (pc) + (match_dup 3)) + (use (label_ref (match_operand 1))) + (clobber (match_operand 2))])] Please use a leading tab instead of every 8 leading spaces. I'll try to be quicker at reviewing iterations of this -- there is quite some way to go, without me slowing things down! Segher * config/rs6000/linux.h (rs6000_relative_jumptables): Define. * config/rs6000/linux64.h (JUMP_TABLES_IN_TEXT_SECTION, rs6000_relative_jumptables): Define. * config/rs6000/rs6000-protos.h (rs6000_output_addr_vec_elt): Declare. * config/rs6000/rs6000.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Define (rs6000_gen_pic_addr_diff_vec, rs6000_output_addr_vec_elt): Implement. * config/rs6000/rs6000.h (CASE_VECTOR_PC_RELATIVE, CASE_VECTOR_MODE, ASM_OUTPUT_AD
Re: [PATCH, rs6000] Add non-relative jump table support on Power Linux
Hi! Sorry this took so long to review. "I lost track of this patch", what can I say :-/ On Fri, Aug 14, 2020 at 03:31:05PM +0800, HAO CHEN GUI wrote: > This patch adds non-relative jump table support on Power Linux. It > implements ASM_OUTPUT_ADDR_VEC_ELT and adds four new expansions for > non-relative jump table in rs6000.md. It also defines a rs6000 > option(mrelative-jumptables). If it's set to false, the non-relative > jump table is picked up even PIC flag is set. > diff --git a/gcc/config/rs6000/linux64.h b/gcc/config/rs6000/linux64.h > index 34776c8421e..626f2201d96 100644 > --- a/gcc/config/rs6000/linux64.h > +++ b/gcc/config/rs6000/linux64.h > @@ -324,7 +324,10 @@ extern int dot_symbols; > > /* Indicate that jump tables go in the text section. */ > #undef JUMP_TABLES_IN_TEXT_SECTION > -#define JUMP_TABLES_IN_TEXT_SECTION TARGET_64BIT > +#define JUMP_TABLES_IN_TEXT_SECTION rs6000_relative_jumptables > + > +#undef JUMP_TABLES_IN_RELRO > +#define JUMP_TABLES_IN_RELRO !rs6000_relative_jumptables The third choice is .rodata. .data.rel.ro can be considered to be a sub-case of that, but not the other way around. > +extern void rs6000_output_addr_vec_elt(FILE *, int); (space before opening paren) > +/* Implement TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC. Return true if > rs6000_relative_jumptables is set. */ The maximum line length is 80 characters. Two spaces after a full stop. > +/* Enable non-relative jump tables for linux ELF. */ > +#ifdef TARGET_ELF > +#if !TARGET_AIX > +#undef rs6000_relative_jumptables > +#define rs6000_relative_jumptables 0 > +#endif > +#endif AIX never is ELF, but many other things *are* ELF. Do all of those handle .data.rel.ro, including do they have enough OS support for it? This also changes the default for ELF systems? Let's not do that (in this patch, that is). > +/* Specify the machine mode that this machine uses > + for the index in the tablejump instruction. */ > +#define CASE_VECTOR_MODE \ > + (TARGET_32BIT || rs6000_relative_jumptables ? SImode : DImode) #define CASE_VECTOR_MODE (rs6000_relative_jumptables ? SImode : Pmode) > +/* This is how to output an element of a case-vector > + that is non-relative. */ (no space at end of line; two spaces after full stop, also before */) > +#define ASM_OUTPUT_ADDR_VEC_ELT(FILE, VALUE) \ > + rs6000_output_addr_vec_elt((FILE), (VALUE)) Space before first ( please (on the second line, on the first you cannot -- macro definitions are the only exception). > diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md > index 0aa5265d199..2f28da0ea1f 100644 > --- a/gcc/config/rs6000/rs6000.md > +++ b/gcc/config/rs6000/rs6000.md > @@ -12669,18 +12669,32 @@ >if (rs6000_speculate_indirect_jumps) > { >if (TARGET_32BIT) > - emit_jump_insn (gen_tablejumpsi (operands[0], operands[1])); > - else > +{ > + if (rs6000_relative_jumptables) > +emit_jump_insn (gen_tablejumpsi (operands[0], operands[1])); > + else > + emit_jump_insn (gen_nonrelative_tablejumpsi (operands[0], > operands[1])); Please fix the indent while you're at it (tabs shouldn't follow spaces). Oh, and line too long. > +} > + else if (rs6000_relative_jumptables) > emit_jump_insn (gen_tablejumpdi (operands[0], operands[1])); > + else > +emit_jump_insn (gen_nonrelative_tablejumpdi (operands[0], > operands[1])); Please put the "relative" case on the outside, and the "32 or 64" if on the inside. > +(define_expand "nonrelative_tablejumpsi" Maybe call it "absolute_tablejumpsi"? Not sure :-) > + "TARGET_32BIT && rs6000_speculate_indirect_jumps && > !rs6000_relative_jumptables" Line too long (more follow, I won't point all out, please check :-) ) > + { > + operands[2] = gen_reg_rtx (SImode); > + }) No spaces before { and } here, those are flush left like in outer braces in a C function are. > +(define_expand "nonrelative_tablejumpsi_nospec" > + [(set (match_dup 3) > +(match_operand:SI 0 "gpc_reg_operand" "r")) > + (parallel [(set (pc) > + (match_dup 3)) > + (use (label_ref (match_operand 1))) > + (clobber (match_operand 2))])] Please use a leading tab instead of every 8 leading spaces. I'll try to be quicker at reviewing iterations of this -- there is quite some way to go, without me slowing things down! Segher