[Mesa-dev] [PATCH 4/6] nir: Add an optimization pass to remove trivial continues
Jason Ekstrand
jason at jlekstrand.net
Thu Dec 22 21:36:46 UTC 2016
On Wed, Dec 21, 2016 at 6:33 PM, Jason Ekstrand <jason at jlekstrand.net>
wrote:
> On Dec 21, 2016 8:11 PM, "Timothy Arceri" <timothy.arceri at collabora.com>
> wrote:
>
> On Mon, 2016-12-19 at 20:11 -0800, Jason Ekstrand wrote:
> > ---
> > src/compiler/Makefile.sources | 1 +
> > src/compiler/nir/nir.h | 2 +
> > src/compiler/nir/nir_opt_trivial_continues.c | 141
> > +++++++++++++++++++++++++++
> > 3 files changed, 144 insertions(+)
> > create mode 100644 src/compiler/nir/nir_opt_trivial_continues.c
> >
> > diff --git a/src/compiler/Makefile.sources
> > b/src/compiler/Makefile.sources
> > index ae3e5f0..a0abede 100644
> > --- a/src/compiler/Makefile.sources
> > +++ b/src/compiler/Makefile.sources
> > @@ -242,6 +242,7 @@ NIR_FILES = \
> > nir/nir_opt_loop_unroll.c \
> > nir/nir_opt_peephole_select.c \
> > nir/nir_opt_remove_phis.c \
> > + nir/nir_opt_trivial_continues.c \
> > nir/nir_opt_undef.c \
> > nir/nir_phi_builder.c \
> > nir/nir_phi_builder.h \
> > diff --git a/src/compiler/nir/nir.h b/src/compiler/nir/nir.h
> > index 89b0c70..a6c8956 100644
> > --- a/src/compiler/nir/nir.h
> > +++ b/src/compiler/nir/nir.h
> > @@ -2563,6 +2563,8 @@ bool nir_opt_peephole_select(nir_shader
> > *shader, unsigned limit);
> >
> > bool nir_opt_remove_phis(nir_shader *shader);
> >
> > +bool nir_opt_trivial_continues(nir_shader *shader);
> > +
> > bool nir_opt_undef(nir_shader *shader);
> >
> > bool nir_opt_conditional_discard(nir_shader *shader);
> > diff --git a/src/compiler/nir/nir_opt_trivial_continues.c
> > b/src/compiler/nir/nir_opt_trivial_continues.c
> > new file mode 100644
> > index 0000000..ff3205d
> > --- /dev/null
> > +++ b/src/compiler/nir/nir_opt_trivial_continues.c
> > @@ -0,0 +1,141 @@
> > +/*
> > + * Copyright © 2016 Intel Corporation
> > + *
> > + * Permission is hereby granted, free of charge, to any person
> > obtaining a
> > + * copy of this software and associated documentation files (the
> > "Software"),
> > + * to deal in the Software without restriction, including without
> > limitation
> > + * the rights to use, copy, modify, merge, publish, distribute,
> > sublicense,
> > + * and/or sell copies of the Software, and to permit persons to whom
> > the
> > + * Software is furnished to do so, subject to the following
> > conditions:
> > + *
> > + * The above copyright notice and this permission notice (including
> > the next
> > + * paragraph) shall be included in all copies or substantial
> > portions of the
> > + * Software.
> > + *
> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> > EXPRESS OR
> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> > MERCHANTABILITY,
> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO
> > EVENT SHALL
> > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES
> > OR OTHER
> > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
> > ARISING
> > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
> > OTHER DEALINGS
> > + * IN THE SOFTWARE.
> > + */
> > +
> > +#include "nir.h"
> > +
> > +static bool
> > +instr_is_continue(nir_instr *instr)
> > +{
> > + if (instr->type != nir_instr_type_jump)
> > + return false;
> > +
> > + return nir_instr_as_jump(instr)->type == nir_jump_continue;
> > +}
> > +
> > +static bool
> > +lower_trivial_continues_block(nir_block *block, nir_loop *loop)
> > +{
> > + bool progress = false;
> > + nir_instr *first_instr = nir_block_first_instr(block);
> > + if (!first_instr || instr_is_continue(first_instr)) {
> > + /* The block contains only a continue if anything */
> > + nir_cf_node *prev_node = nir_cf_node_prev(&block->cf_node);
> > + if (prev_node && prev_node->type == nir_cf_node_if) {
> > + nir_if *prev_if = nir_cf_node_as_if(prev_node);
> > + progress |= lower_trivial_continues_block(
> > + nir_if_last_then_block(prev_if), loop);
> > + progress |= lower_trivial_continues_block(
> > + nir_if_last_else_block(prev_if), loop);
> > + }
> > +
> > + /* The lower_phis_to_regs helper is smart enough to push phi
> > sources as
> > + * far up if-ladders as it possibly can. In other words, the
> > recursive
> > + * calls above shouldn't be adding instructions to this block
> > since it
> > + * follows an if statement.
> > + */
> > + first_instr = nir_block_first_instr(block);
> > + assert(!first_instr || instr_is_continue(first_instr));
> > + }
> > +
> > + nir_instr *last_instr = nir_block_last_instr(block);
> > + if (!last_instr || !instr_is_continue(last_instr))
> > + return progress;
> > +
> > + /* We're at the end of a block that goes straight through to the
> > end of the
> > + * loop and continues to the top. We can exit normally instead
> > of jump.
> > + */
> > + nir_lower_phis_to_regs_block(nir_loop_first_block(loop));
> > + nir_instr_remove(last_instr);
> > + return true;
> > +}
> > +
> > +static bool
> > +lower_trivial_continues_list(struct exec_list *cf_list,
> > + bool list_ends_at_loop_tail,
> > + nir_loop *loop)
> > +{
> > + bool progress = false;
> > + foreach_list_typed(nir_cf_node, cf_node, node, cf_list) {
> > + bool at_loop_tail = list_ends_at_loop_tail &&
> > + &cf_node->node ==
> > exec_list_get_tail(cf_list);
> > + switch (cf_node->type) {
> > + case nir_cf_node_block:
> > + break;
> > +
> > + case nir_cf_node_if: {
> > + nir_if *nif = nir_cf_node_as_if(cf_node);
> > + if (lower_trivial_continues_list(&nif->then_list,
> > at_loop_tail, loop))
> > + progress = true;
> > + if (lower_trivial_continues_list(&nif->else_list,
> > at_loop_tail, loop))
> > + progress = true;
> > + break;
> > + }
> > +
> > + case nir_cf_node_loop: {
> > + nir_loop *loop = nir_cf_node_as_loop(cf_node);
> > + if (lower_trivial_continues_list(&loop->body, true, loop))
> > + progress = true;
> > + if
> > (lower_trivial_continues_block(nir_loop_last_block(loop), loop))
> > + progress = true;
> > + break;
> > + }
> > +
> > + case nir_cf_node_function:
> > + unreachable("Invalid cf type");
> > + }
> > + }
> > +
> > + return progress;
> > +}
> > +
> > +/**
> > + * This simple pass gets rid of any "trivial" continues, i.e. those
> > that are
> > + * at the tail of a loop where we can just delete the continue
> > instruction and
> > + * control-flow will naturally and harmlessly take us back to the
> > top of the
> > + * loop.
> > + */
> > +bool
> > +nir_opt_trivial_continues(nir_shader *shader)
> > +{
> > + bool progress = false;
> > +
> > + nir_foreach_function(function, shader) {
> > + if (function->impl == NULL)
> > + continue;
> > +
> > + /* First we run the simple pass to get rid of pesky continues
> > */
> > + if (lower_trivial_continues_list(&function->impl->body, false,
> > NULL)) {
> > + nir_metadata_preserve(function->impl, nir_metadata_none);
> > +
> > + /* If that made progress, we're no longer really in SSA
> > form. Go
> > + * back into SSA form and clean up.
> > + */
> > + nir_convert_to_ssa_impl(function->impl);
> > + nir_copy_prop_impl(function->impl);
> > + nir_opt_dce_impl(function->impl);
>
> Just curious why you explicitly call these rather than just let them
> get cleaned up normally? Does it cut compile times to get rid of them
> straight away?
>
>
> The nir_opt_if pass I wrote relies on this pass. Of this pass is run, it
> also needs copy_prop and dce. This potentially allows is to drop an
> iteration of the optimization loop. Another option would be to do this in
> brw_nir.c:
>
> if (OPT(nir_opt_trivial_continues)) {
> OPT(nir_opt_copy_prop);
> OPT(nir_opt_dce);
> }
>
> I don't have a strong preference either way.
>
I've updated the brw_nir code to do this for us:
https://cgit.freedesktop.org/~jekstrand/mesa/commit/?h=wip/
nir-opt-continues-v1.1&id=dd7f6934063b27101a4face52fd8bc3c8cadea78
I've also deleted 3/6 because it's no longer needed.
>
> Anyway this looks good to me and very easy to follow:
>
> Reviewed-by: Timothy Arceri <timothy.arceri at collabora.com>
>
> > + progress = true;
> > + }
> > + }
> > +
> > + return progress;
> > +}
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20161222/2d8b9412/attachment-0001.html>
More information about the mesa-dev
mailing list