This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
RE: [PATCHv2][MIPS] Implement O32 ABI extensions (GCC)
- From: Matthew Fortune <Matthew dot Fortune at imgtec dot com>
- To: "'gcc-patches at gcc dot gnu dot org' (gcc-patches at gcc dot gnu dot org)" <gcc-patches at gcc dot gnu dot org>, "Eric Christopher (echristo at gmail dot com)" <echristo at gmail dot com>
- Cc: "Moore, Catherine (Catherine_Moore at mentor dot com)" <Catherine_Moore at mentor dot com>, Richard Sandiford <rdsandiford at googlemail dot com>, Rich Fuhler <Rich dot Fuhler at imgtec dot com>, "macro at codesourcery dot com" <macro at codesourcery dot com>, "Joseph Myers (joseph at codesourcery dot com)" <joseph at codesourcery dot com>
- Date: Thu, 4 Sep 2014 15:48:55 +0000
- Subject: RE: [PATCHv2][MIPS] Implement O32 ABI extensions (GCC)
- Authentication-results: sourceware.org; auth=none
- References: <6D39441BF12EF246A7ABCE6654B0235320EF5E6F at LEMAIL01 dot le dot imgtec dot org>
Ping!
> -----Original Message-----
> From: gcc-patches-owner@gcc.gnu.org [mailto:gcc-patches-owner@gcc.gnu.org]
> On Behalf Of Matthew Fortune
> Sent: 22 August 2014 10:43
> To: 'gcc-patches@gcc.gnu.org' (gcc-patches@gcc.gnu.org); Eric Christopher
> (echristo@gmail.com)
> Cc: Moore, Catherine (Catherine_Moore@mentor.com); Richard Sandiford; Rich
> Fuhler; macro@codesourcery.com; Joseph Myers (joseph@codesourcery.com)
> Subject: [PATCHv2][MIPS] Implement O32 ABI extensions (GCC)
>
> Updated patch covering all comments from the previous thread:
> https://gcc.gnu.org/ml/gcc-patches/2014-05/msg00401.html
>
> This patch has merged together the odd-spreg work with the other ABI
> work as these two features are now inseparable due to the inclusion of
> a 4th FP ABI variant called FP64A. The wiki page describing these
> extensions has been updated and the patch is consistent with the
> features.
>
> https://dmz-portal.mips.com/wiki/MIPS_O32_ABI_-_FR0_and_FR1_Interlinking
>
> It is worth noting that LLVM 3.5 will include all ABI extensions
> described in the wiki page with consistent options and behaviour.
>
> The vast majority of this patch has been reviewed in detail already and
> testing has been ongoing for some time within teams at Imagination.
> This code has also been released for inclusion in the next Android NDK.
>
> I have addressed one of the final concerns from Richard and Maciej
> regarding an inconsistency between the prologue FP callee-save and
> fixed-regs. I opted to resolve this in a very focussed manner to just
> address the impact of fixed-regs rather than the more general issue
> of occasionally saving more state than absolutely required. If there
> is a desire to improve this further then I am very keen to leave that
> to a future patch.
>
> This work has been tested for both bare metal and linux based targets.
> There are no regressions.
>
> This patch is dependent on two patches which are awaiting approval:
>
> "Add target hook to override DWARF2 frame register size
> https://gcc.gnu.org/ml/gcc-patches/2014-08/msg01748.html
>
> "Do not reload unallocated FP_REGS pseudos via GR_REGS"
> https://gcc.gnu.org/ml/gcc-patches/2014-08/msg01745.html
>
> Regards,
> Matthew
>
> 2014-07-31 Matthew Fortune <matthew.fortune@imgtec.com>
>
> gcc/
> * common/config/mips/mips-common.c (mips_handle_option): Ensure
> that -mfp32, -mfp64 disable -mfpxx and -mfpxx disables -mfp64.
> * config.gcc (--with-fp-32): New option.
> (--with-odd-spreg-32): Likewise.
> * config.in (HAVE_AS_MODULE): New config define.
> * config/mips/mips-protos.h
> (mips_secondary_memory_needed): New prototype.
> (mips_hard_regno_caller_save_mode): Likewise.
> * config/mips/mips.c (mips_get_reg_raw_mode): New static prototype.
> (mips_get_arg_info): Assert that V2SFmode is only handled specially
> with TARGET_PAIRED_SINGLE_FLOAT.
> (mips_return_mode_in_fpr_p): Likewise.
> (mips16_call_stub_mode_suffix): Likewise.
> (mips_get_reg_raw_mode): New static function.
> (mips_return_fpr_pair): O32 return values span two registers.
> (mips16_build_call_stub): Likewise.
> (mips_function_value_regno_p): Support both FP return registers.
> (mips_output_64bit_xfer): Use mthc1 whenever TARGET_HAS_MXHC1. Add
> specific cases for TARGET_FPXX to move via memory.
> (mips_dwarf_register_span): For TARGET_FPXX pretend that modes larger
> than UNITS_PER_FPREG 'span' one register.
> (mips_dwarf_frame_reg_mode): New static function.
> (mips_file_start): Switch to using .module instead of .gnu_attribute.
> No longer support FP ABI 4 (-mips32r2 -mfp64), replace with FP ABI 6.
> Add FP ABI 5 (-mfpxx) and FP ABI 7 (-mfp64 -mno-odd-spreg).
> (mips_save_reg, mips_restore_reg): Always represent DFmode frame
> slots with two CFI directives even for O32 FP64.
> (mips_for_each_saved_gpr_and_fpr): Account for fixed_regs when
> saving/restoring callee-saved registers.
> (mips_hard_regno_mode_ok_p): Implement O32 FP64A extension.
> (mips_secondary_memory_needed): New function.
> (mips_option_override): ABI check for TARGET_FLOATXX. Disable
> odd-numbered single-precision registers when using TARGET_FLOATXX.
> Implement -modd-spreg and defaults.
> (mips_conditional_register_usage): Redefine O32 FP64 to match O32 FP32
> callee-saved behaviour.
> (mips_hard_regno_caller_save_mode): Implement.
> (TARGET_GET_RAW_RESULT_MODE): Define target hook.
> (TARGET_GET_RAW_ARG_MODE): Define target hook.
> (TARGET_DWARF_FRAME_REG_MODE): Define target hook.
> * config/mips/mips.h (TARGET_FLOAT32): New macro.
> (TARGET_CPU_CPP_BUILTINS): TARGET_FPXX is __mips_fpr==0. Add
> _MIPS_SPFPSET builtin define.
> (MIPS_FPXX_OPTION_SPEC): New macro.
> (OPTION_DEFAULT_SPECS): Pass through --with-fp-32=* to -mfp and
> --with-odd-spreg-32=* to -m[no-]odd-spreg.
> (ISA_HAS_ODD_SPREG): New macro.
> (ISA_HAS_MXHC1): True for anything other than -mfp32.
> (ASM_SPEC): Pass through mfpxx, mfp64, -mno-odd-spreg and -modd-spreg.
> (MIN_FPRS_PER_FMT): Redefine in terms of TARGET_ODD_SPREG.
> (HARD_REGNO_CALLER_SAVE_MODE): Define. Implement O32 FPXX extension
> (HARD_REGNO_CALL_PART_CLOBBERED): Likewise.
> (SECONDARY_MEMORY_NEEDED): Likewise.
> (FUNCTION_ARG_REGNO_P): Update for O32 FPXX and FP64 extensions.
> * config/mips/mips.md (define_attr enabled): Implement O32 FPXX and
> FP64A ABI extensions.
> (move_doubleword_fpr<mode>): Use ISA_HAS_MXHC1 instead of
> TARGET_FLOAT64.
> * config/mips/mips.opt (mfpxx): New target option.
> (modd-spreg): Likewise.
> * config/mips/mti-elf.h (DRIVER_SELF_SPECS): Infer FP ABI from arch.
> * config/mips/mti-linux.h (DRIVER_SELF_SPECS): Likewise and remove
> fp64 sysroot.
> * config/mips/t-mti-elf: Remove fp64 multilib.
> * config/mips/t-mti-linux: Likewise.
> * configure.ac: Detect .module support.
> * configure: Regenerate.
> * doc/invoke.texi: Document -mfpxx, -modd-spreg, -mno-odd-spreg option.
>
> gcc/testsuite/
> * gcc.target/mips/args-1.c: Handle __mips_fpr == 0.
> * gcc.target/mips/call-clobbered-1.c: New.
> * gcc.target/mips/call-clobbered-2.c: New.
> * gcc.target/mips/call-clobbered-3.c: New.
> * gcc.target/mips/call-clobbered-4.c: New.
> * gcc.target/mips/call-clobbered-5.c: New.
> * gcc.target/mips/call-saved-4.c: New.
> * gcc.target/mips/call-saved-5.c: New.
> * gcc.target/mips/call-saved-6.c: New.
> * gcc.target/mips/mips.exp: Support -mfpxx, -ffixed-f*,
> and -m[no-]odd-spreg. Use _MIPS_SPFPSET to determine default
> odd-spreg option. Account for -modd-spreg in minimum arch code.
> * gcc.target/mips/movdf-1.c: New.
> * gcc.target/mips/movdf-2.c: New.
> * gcc.target/mips/movdf-3.c: New.
> * gcc.target/mips/oddspreg-1.c: New.
> * gcc.target/mips/oddspreg-2.c: New.
> * gcc.target/mips/oddspreg-3.c: New.
> * gcc.target/mips/oddspreg-4.c: New.
> * gcc.target/mips/oddspreg-5.c: New.
> * gcc.target/mips/oddspreg-6.c: New.
>
> libgcc/
> * config/mips/mips16.S: Set .module when supported. Update O32
> FP64 calling convention and use for FPXX when possible. Add FPXX
> calling convention fallback case