public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [0/67] Add wrapper classes for machine_modes
@ 2016-12-09 12:48 Richard Sandiford
  2016-12-09 12:50 ` [1/67] Add an E_ prefix to mode names and update case statements Richard Sandiford
                   ` (69 more replies)
  0 siblings, 70 replies; 79+ messages in thread
From: Richard Sandiford @ 2016-12-09 12:48 UTC (permalink / raw)
  To: gcc-patches

This series includes most of the changes in group C from:

    https://gcc.gnu.org/ml/gcc/2016-11/msg00033.html

The idea is to add wrapper classes around machine_mode_enum
for specific groups of modes, such as scalar integers, scalar floats,
complex values, etc.  This has two main benefits: one specific to SVE
and one not.

The SVE-specific benefit is that it helps to introduce the concept
of variable-length vectors.  To do that we need to change the size
of a vector mode from being a known compile-time constant to being
(possibly) a run-time invariant.  We then need to do the same for
unconstrained machine_modes, which might or might not be vectors.
Introducing these new constrained types means that we can continue
to treat them as having a constant size.

The other benefit is that it uses static type checking to enforce
conditions that are easily forgotten otherwise.  The most common
sources of problems seem to be:

(a) using VOIDmode or BLKmode where a scalar integer was expected
    (e.g. when getting the number of bits in the value).

(b) simplifying vector operations in ways that only make sense for
    scalars.

The series helps with both of these, although we don't get the full
benefit of (b) until variable-sized modes are introduced.

I know of three specific cases in which the static type checking
forced fixes for things that turned out to be real bugs (although
we didn't know that at the time, otherwise we'd have posted patches).
They were later fixed for trunk by:

  https://gcc.gnu.org/ml/gcc-patches/2016-07/msg01783.html
  https://gcc.gnu.org/ml/gcc-patches/2016-11/msg02983.html
  https://gcc.gnu.org/ml/gcc-patches/2016-11/msg02896.html

The group C patches in ARM/sve-branch did slow compile time down a little.
I've since taken steps to avoid that:

- Make the tailcall pass handle aggregate parameters and return values
  (already in trunk).

- Turn some of the new wrapper functions into inline functions.

- Make all the machmode.h macros that used:

    __builtin_constant_p (M) ? foo_inline (M) : foo_array[M[

  forward to an ALWAYS_INLINE function, so that (a) M is only evaluated
  once and (b) __builtin_constant_p is applied to a variable, and so is
  deferred until later passes.  This helped the optimisation to fire in
  more cases and to continue firing when M is a class rather than a
  raw enum.

- In a similar vein, make sure that conditions like:

     SImode == DImode

  are treated as builtin_constant_p by gencondmd, so that .md patterns
  with those conditions are dropped.

With these changes the series is actually a very slight compile-time win.
That might seem unlikely, but there are several possible reasons:

1. The machmode.h macro change above might allow more constant folding.

2. The series has a tendency to evaluate modes once, rather than
   continually fetching them from (sometimes quite deep) rtx nests.
   Refetching a mode is a particular problem if call comes between
   two uses, since the compiler then has to re-evaluate the whole thing.

3. The series introduces many uses of new SCALAR_*TYPE_MODE macros,
   as alternatives to TYPE_MODE.  The new macros avoid the usual:

     (VECTOR_TYPE_P (TYPE_CHECK (NODE)) \
      ? vector_type_mode (NODE) : (NODE)->type_common.mode)

   and become direct field accesses in release builds.

   VECTOR_TYPE_P would be consistently false for these uses,
   but call-clobbered registers would usually be treated as clobbered
   by the condition as a whole.

Maybe (3) is the most likely reason.

I tested this by compiling the testsuite for:

    aarch64-linux-gnu alpha-linux-gnu arc-elf arm-linux-gnueabi
    arm-linux-gnueabihf avr-elf bfin-elf c6x-elf cr16-elf cris-elf
    epiphany-elf fr30-elf frv-linux-gnu ft32-elf h8300-elf
    hppa64-hp-hpux11.23 ia64-linux-gnu i686-pc-linux-gnu
    i686-apple-darwin iq2000-elf lm32-elf m32c-elf m32r-elf
    m68k-linux-gnu mcore-elf microblaze-elf mips-linux-gnu
    mipsisa64-linux-gnu mmix mn10300-elf moxie-rtems msp430-elf
    nds32le-elf nios2-linux-gnu nvptx-none pdp11 powerpc-linux-gnuspe
    powerpc-eabispe powerpc64-linux-gnu powerpc-ibm-aix7.0 rl78-elf
    rx-elf s390-linux-gnu s390x-linux-gnu sh-linux-gnu sparc-linux-gnu
    sparc64-linux-gnu sparc-wrs-vxworks spu-elf tilegx-elf tilepro-elf
    xstormy16-elf v850-elf vax-netbsdelf visium-elf x86_64-darwin
    x86_64-linux-gnu xtensa-elf

and checking that there were no changes in assembly.  Also tested
in the normal way on aarch64-linux-gnu and x86_64-linux-gnu.

The series depends on the already-posted:

  https://gcc.gnu.org/ml/gcc-patches/2016-11/msg01657.html

Sorry that this is so late, been distracted by other things.  Even if
we're too far into stage 3 for SVE itself to go in, I was hoping this
part (which was kind-of posted during stage 1) could go in independently.

Thanks,
Richard

^ permalink raw reply	[flat|nested] 79+ messages in thread
end of thread, other threads:[~2017-06-28 17:27 UTC | newest]

Thread overview: 79+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-12-09 12:48 [0/67] Add wrapper classes for machine_modes Richard Sandiford
2016-12-09 12:50 ` [1/67] Add an E_ prefix to mode names and update case statements Richard Sandiford
2016-12-09 12:52 ` [2/67] Make machine_mode a class Richard Sandiford
2016-12-15 22:34   ` Trevor Saunders
2016-12-09 12:53 ` [3/67] Add GDB pretty printer for machine mode classes Richard Sandiford
2016-12-09 12:54 ` [4/67] Add FOR_EACH iterators for modes Richard Sandiford
2016-12-09 12:55 ` [6/67] Make GET_MODE_WIDER return an opt_mode Richard Sandiford
2016-12-09 12:55 ` [5/67] Small tweak to array_value_type Richard Sandiford
2016-12-09 12:57 ` [7/67] Add scalar_float_mode Richard Sandiford
2016-12-09 14:19   ` David Malcolm
2016-12-09 14:30     ` Richard Sandiford
2016-12-09 12:57 ` [8/67] Simplify gen_trunc/extend_conv_libfunc Richard Sandiford
2016-12-09 12:58 ` [9/67] Add SCALAR_FLOAT_TYPE_MODE Richard Sandiford
2016-12-09 12:59 ` [10/67] Make assemble_real take a scalar_float_mode Richard Sandiford
2016-12-09 13:00 ` [0/67] Add wrapper classes for machine_modes Richard Biener
2016-12-09 13:54   ` Richard Sandiford
2016-12-09 13:00 ` [11/67] Add a float_mode_for_size helper function Richard Sandiford
2016-12-09 13:00 ` [12/67] Use opt_scalar_float_mode when iterating over float modes Richard Sandiford
2016-12-09 13:01 ` [13/67] Make floatn_mode return an opt_scalar_float_mode Richard Sandiford
2016-12-09 13:02 ` [14/67] Make libgcc_floating_mode_supported_p take a scalar_float_mode Richard Sandiford
2016-12-09 13:03 ` [16/67] Add scalar_int_mode_pod Richard Sandiford
2016-12-09 13:03 ` [15/67] Add scalar_int_mode Richard Sandiford
2016-12-09 13:04 ` [17/67] Add an int_mode_for_size helper function Richard Sandiford
2016-12-09 13:05 ` [19/67] Add a smallest_int_mode_for_size " Richard Sandiford
2016-12-09 13:05 ` [18/67] Make int_mode_for_mode return an opt_scalar_int_mode Richard Sandiford
2016-12-09 13:06 ` [20/67] Replace MODE_INT checks with is_int_mode Richard Sandiford
2016-12-09 13:07 ` [21/67] Replace SCALAR_INT_MODE_P checks with is_a <scalar_int_mode> Richard Sandiford
2016-12-09 13:07 ` [22/67] Replace !VECTOR_MODE_P " Richard Sandiford
2016-12-09 13:08 ` [23/67] Replace != VOIDmode checks " Richard Sandiford
2016-12-09 13:08 ` [24/67] Replace a != BLKmode check " Richard Sandiford
2016-12-09 13:22   ` Richard Biener
2016-12-09 14:42     ` Richard Sandiford
2016-12-09 13:09 ` [25/67] Use is_a <scalar_int_mode> for bitmask optimisations Richard Sandiford
2016-12-09 13:10 ` [27/67] Use is_a <scalar_int_mode> before LOAD_EXTEND_OP Richard Sandiford
2016-12-09 13:10 ` [26/67] Use is_a <scalar_int_mode> in subreg/extract simplifications Richard Sandiford
2016-12-09 13:11 ` [28/67] Use is_a <scalar_int_mode> for miscellaneous types of test Richard Sandiford
2016-12-09 13:12 ` [29/67] Make some *_loc_descriptor helpers take scalar_int_mode Richard Sandiford
2016-12-09 13:12 ` [30/67] Use scalar_int_mode for doubleword splits Richard Sandiford
2016-12-09 13:13 ` [31/67] Use scalar_int_mode for move2add Richard Sandiford
2016-12-09 13:14 ` [33/67] Add a NARROWEST_INT_MODE macro Richard Sandiford
2016-12-09 13:14 ` [32/67] Check is_a <scalar_int_mode> before calling valid_pointer_mode Richard Sandiford
2016-12-09 13:15 ` [34/67] Add a SCALAR_INT_TYPE_MODE macro Richard Sandiford
2016-12-09 13:16 ` [36/67] Use scalar_int_mode in the RTL iv routines Richard Sandiford
2016-12-09 13:16 ` [35/67] Add uses of as_a <scalar_int_mode> Richard Sandiford
2016-12-09 13:17 ` [38/67] Move SCALAR_INT_MODE_P out of strict_volatile_bitfield_p Richard Sandiford
2016-12-09 13:17 ` [37/67] Use scalar_int_mode when emitting cstores Richard Sandiford
2016-12-09 13:18 ` [39/67] Two changes to the get_best_mode interface Richard Sandiford
2016-12-09 13:19 ` [40/67] Use scalar_int_mode for extraction_insn fields Richard Sandiford
2016-12-09 13:20 ` [41/67] Split scalar integer handling out of force_to_mode Richard Sandiford
2016-12-09 13:20 ` [42/67] Use scalar_int_mode in simplify_shift_const_1 Richard Sandiford
2016-12-09 13:22 ` [43/67] Use scalar_int_mode in simplify_comparison Richard Sandiford
2016-12-09 13:22 ` [44/67] Make simplify_and_const_int take a scalar_int_mode Richard Sandiford
2016-12-09 13:23 ` [45/67] Make extract_left_shift " Richard Sandiford
2016-12-09 13:23 ` [46/67] Make widest_int_mode_for_size return " Richard Sandiford
2016-12-09 13:24 ` [47/67] Make subroutines of nonzero_bits operate on scalar_int_mode Richard Sandiford
2016-12-09 13:25 ` [49/67] Simplify nonzero/num_sign_bits hooks Richard Sandiford
2016-12-09 13:25 ` [48/67] Make subroutines of num_sign_bit_copies operate on scalar_int_mode Richard Sandiford
2016-12-09 13:28 ` [50/67] Add helper routines for SUBREG_PROMOTED_VAR_P subregs Richard Sandiford
2016-12-09 13:30 ` [51/67] Use opt_scalar_int_mode when iterating over integer modes Richard Sandiford
2016-12-09 13:30 ` [52/67] Use scalar_int_mode in extract/store_bit_field Richard Sandiford
2016-12-09 13:31 ` [54/67] Add explicit int checks for alternative optab implementations Richard Sandiford
2016-12-09 13:31 ` [53/67] Pass a mode to const_scalar_mask_from_tree Richard Sandiford
2016-12-09 13:33 ` [56/67] Use the more specific type when two modes are known to be equal Richard Sandiford
2016-12-09 13:33 ` [55/67] Use scalar_int_mode in simplify_const_unary_operation Richard Sandiford
2016-12-09 13:34 ` [57/67] Use scalar_int_mode in expand_expr_addr_expr Richard Sandiford
2016-12-09 13:35 ` [58/67] Use scalar_int_mode in a try_combine optimisation Richard Sandiford
2016-12-09 13:36 ` [60/67] Pass scalar_int_modes to do_jump_by_parts_* Richard Sandiford
2016-12-09 13:36 ` [59/67] Add a rtx_jump_table_data::get_data_mode helper Richard Sandiford
2016-12-09 13:37 ` [61/67] Use scalar_int_mode in the AArch64 port Richard Sandiford
2016-12-09 13:38 ` [63/67] Simplifications after type switch Richard Sandiford
2016-12-09 13:38 ` [62/67] Big machine_mode to scalar_int_mode replacement Richard Sandiford
2016-12-09 13:40 ` [64/67] Add a scalar_mode class Richard Sandiford
2016-12-09 13:40 ` [65/67] Use scalar_mode in the AArch64 port Richard Sandiford
2016-12-09 13:40 ` [66/67] Add a scalar_mode_pod class Richard Sandiford
2016-12-09 13:41 ` [67/67] Add a complex_mode class Richard Sandiford
2016-12-09 18:20 ` [0/67] Add wrapper classes for machine_modes Sandra Loosemore
2017-05-05  7:08 ` Jeff Law
2017-05-24 14:33   ` Richard Sandiford
2017-06-28 17:27     ` Jeff Law

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).