[Mesa-dev] [PATCH] nir: Add a pass for gathering transform feedback info

Alejandro Piñeiro apinheiro at igalia.com
Fri Oct 5 15:48:46 UTC 2018


On 05/10/18 17:44, Jason Ekstrand wrote:
> On Fri, Oct 5, 2018 at 10:34 AM Alejandro Piñeiro
> <apinheiro at igalia.com <mailto:apinheiro at igalia.com>> wrote:
>
>     On 05/10/18 16:13, Jason Ekstrand wrote:
>     > This is different from the GL_ARB_spirv pass because it
>     generates a much
>     > simpler data structure that isn't tied to OpenGL and mtypes.h.
>
>     I have just skimmed it (don't have time right now for a full
>     check, will
>     take a deeper look next Monday), but FWIW, the GL_ARB_spirv pass
>     does a
>     initial mtypes-free info gathering (get_active_xfb_varyings) and
>     then it
>     uses that info to fill-up the OpenGL specific mtypes.h. In fact, if we
>     just compare the initial GL_ARB_spirv gathering with this new
>     pass, your
>     pass seems more complete, and with more checks. So I was wondering
>     if it
>     would be possible to remove some code on the GL_ARB_spirv pass and use
>     this new pass instead. Did you check if that would be possible? If not
>     I'm willing to check next week.
>
>
> I didn't really look into that too much.  When drafting this one, I
> did base it somewhat on the GL_ARB_spirv pass so I"m not surprised
> it's similar.  At the time, I was just trying to get something working
> and wasn't too worried about code duplication.  If we can use this
> pass for GL_ARB_spirv, that'd be fantastic. 

Ok, then next week I will check for sure if we can do that.

> If we do go that route, however, we may want to do something better
> than assert() for the error handling.

Ok, first I will check if this new pass gather the info GL_ARB_spirv
pass needs, and see how much the GL_ARB_spirv pass would need to be
modified (as the gathering info formatting would change). Then we can
talk about what to do with the asserts.

>  
>
>     > ---
>     >  src/compiler/Makefile.sources          |   4 +-
>     >  src/compiler/nir/meson.build           |   2 +
>     >  src/compiler/nir/nir_gather_xfb_info.c | 150
>     +++++++++++++++++++++++++
>     >  src/compiler/nir/nir_xfb_info.h        |  59 ++++++++++
>     >  4 files changed, 214 insertions(+), 1 deletion(-)
>     >  create mode 100644 src/compiler/nir/nir_gather_xfb_info.c
>     >  create mode 100644 src/compiler/nir/nir_xfb_info.h
>     >
>     > diff --git a/src/compiler/Makefile.sources
>     b/src/compiler/Makefile.sources
>     > index d3b06564832..46ed5e47b46 100644
>     > --- a/src/compiler/Makefile.sources
>     > +++ b/src/compiler/Makefile.sources
>     > @@ -216,6 +216,7 @@ NIR_FILES = \
>     >       nir/nir_format_convert.h \
>     >       nir/nir_from_ssa.c \
>     >       nir/nir_gather_info.c \
>     > +     nir/nir_gather_xfb_info.c \
>     >       nir/nir_gs_count_vertices.c \
>     >       nir/nir_inline_functions.c \
>     >       nir/nir_instr_set.c \
>     > @@ -307,7 +308,8 @@ NIR_FILES = \
>     >       nir/nir_validate.c \
>     >       nir/nir_vla.h \
>     >       nir/nir_worklist.c \
>     > -     nir/nir_worklist.h
>     > +     nir/nir_worklist.h \
>     > +     nir/nir_xfb_info.h
>>     >  SPIRV_GENERATED_FILES = \
>     >       spirv/spirv_info.c \
>     > diff --git a/src/compiler/nir/meson.build
>     b/src/compiler/nir/meson.build
>     > index 090aa7a628f..b416e561eb0 100644
>     > --- a/src/compiler/nir/meson.build
>     > +++ b/src/compiler/nir/meson.build
>     > @@ -100,6 +100,7 @@ files_libnir = files(
>     >    'nir_format_convert.h',
>     >    'nir_from_ssa.c',
>     >    'nir_gather_info.c',
>     > +  'nir_gather_xfb_info.c',
>     >    'nir_gs_count_vertices.c',
>     >    'nir_inline_functions.c',
>     >    'nir_instr_set.c',
>     > @@ -192,6 +193,7 @@ files_libnir = files(
>     >    'nir_vla.h',
>     >    'nir_worklist.c',
>     >    'nir_worklist.h',
>     > +  'nir_xfb_info.h',
>     >    '../spirv/GLSL.ext.AMD.h',
>     >    '../spirv/GLSL.std.450.h',
>     >    '../spirv/gl_spirv.c',
>     > diff --git a/src/compiler/nir/nir_gather_xfb_info.c
>     b/src/compiler/nir/nir_gather_xfb_info.c
>     > new file mode 100644
>     > index 00000000000..a53703bb9bf
>     > --- /dev/null
>     > +++ b/src/compiler/nir/nir_gather_xfb_info.c
>     > @@ -0,0 +1,150 @@
>     > +/*
>     > + * Copyright © 2018 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_xfb_info.h"
>     > +
>     > +#include <util/u_math.h>
>     > +
>     > +static void
>     > +add_var_xfb_outputs(nir_xfb_info *xfb,
>     > +                    nir_variable *var,
>     > +                    unsigned *location,
>     > +                    unsigned *offset,
>     > +                    const struct glsl_type *type)
>     > +{
>     > +   if (glsl_type_is_array(type) || glsl_type_is_matrix(type)) {
>     > +      unsigned length = glsl_get_length(type);
>     > +      const struct glsl_type *child_type =
>     glsl_get_array_element(type);
>     > +      for (unsigned i = 0; i < length; i++)
>     > +         add_var_xfb_outputs(xfb, var, location, offset,
>     child_type);
>     > +   } else if (glsl_type_is_struct(type)) {
>     > +      unsigned length = glsl_get_length(type);
>     > +      for (unsigned i = 0; i < length; i++) {
>     > +         const struct glsl_type *child_type =
>     glsl_get_struct_field(type, i);
>     > +         add_var_xfb_outputs(xfb, var, location, offset,
>     child_type);
>     > +      }
>     > +   } else {
>     > +      assert(var->data.xfb_buffer < NIR_MAX_XFB_BUFFERS);
>     > +      if (xfb->buffers_written & (1 << var->data.xfb_buffer)) {
>     > +         assert(xfb->strides[var->data.xfb_buffer] ==
>     var->data.xfb_stride);
>     > +         assert(xfb->buffer_to_stream[var->data.xfb_buffer] ==
>     var->data.stream);
>     > +      } else {
>     > +         xfb->buffers_written |= (1 << var->data.xfb_buffer);
>     > +         xfb->strides[var->data.xfb_buffer] = var->data.xfb_stride;
>     > +         xfb->buffer_to_stream[var->data.xfb_buffer] =
>     var->data.stream;
>     > +      }
>     > +
>     > +      assert(var->data.stream < NIR_MAX_XFB_STREAMS);
>     > +      xfb->streams_written |= (1 << var->data.stream);
>     > +
>     > +      unsigned comp_slots = glsl_get_component_slots(type);
>     > +      unsigned attrib_slots = DIV_ROUND_UP(comp_slots, 4);
>     > +      assert(attrib_slots == glsl_count_attribute_slots(type,
>     false));
>     > +
>     > +      /* Ensure that we don't have, for instance, a dvec2 with
>     a location_frac
>     > +       * of 2 which would make it crass a location boundary
>     even though it
>     > +       * fits in a single slot.  However, you can have a dvec3
>     which crosses
>     > +       * the slot boundary with a location_frac of 2.
>     > +       */
>     > +      assert(DIV_ROUND_UP(var->data.location_frac + comp_slots,
>     4) == attrib_slots);
>     > +
>     > +      assert(var->data.location_frac + comp_slots <= 8);
>     > +      uint8_t comp_mask = ((1 << comp_slots) - 1) <<
>     var->data.location_frac;
>     > +
>     > +      assert(attrib_slots <= 2);
>     > +      for (unsigned s = 0; s < attrib_slots; s++) {
>     > +         nir_xfb_output_info *output =
>     &xfb->outputs[xfb->output_count++];
>     > +
>     > +         output->buffer = var->data.xfb_buffer;
>     > +         output->offset = *offset;
>     > +         output->location = *location;
>     > +         output->component_mask = (comp_mask >> (s * 4)) & 0xf;
>     > +
>     > +         (*location)++;
>     > +         *offset += comp_slots * 4;
>     > +      }
>     > +   }
>     > +}
>     > +
>     > +static int
>     > +compare_xfb_output_offsets(const void *_a, const void *_b)
>     > +{
>     > +   const nir_xfb_output_info *a = _a, *b = _b;
>     > +   return a->offset - b->offset;
>     > +}
>
>     We have  this same comparison on link_varyings.c and gl_nir_link_xfb.
>     Not sure how I feel about having the same function three times. I
>     would
>     really hope that my previous comment to reuse code to be true.
>
>
> Yeah, if we can re-use some stuff, that'd be great.  On the other
> hand, it's six lines of code which ammount to one interesting line so
> I'm not too worried about the duplication. :-)
>
> --Jason
>  
>
>     > +
>     > +nir_xfb_info *
>     > +nir_gather_xfb_info(const nir_shader *shader, void *mem_ctx)
>     > +{
>     > +   assert(shader->info.stage == MESA_SHADER_VERTEX ||
>     > +          shader->info.stage == MESA_SHADER_TESS_EVAL ||
>     > +          shader->info.stage == MESA_SHADER_GEOMETRY);
>     > +
>     > +   /* Compute the number of outputs we have.  This is simply
>     the number of
>     > +    * cumulative locations consumed by all the variables.  If a
>     location is
>     > +    * represented by multiple variables, then they each count
>     separately in
>     > +    * number of outputs.
>     > +    */
>     > +   unsigned num_outputs = 0;
>     > +   nir_foreach_variable(var, &shader->outputs) {
>     > +      if (var->data.explicit_xfb_buffer ||
>     > +          var->data.explicit_xfb_stride) {
>     > +         assert(var->data.explicit_xfb_buffer &&
>     > +                var->data.explicit_xfb_stride &&
>     > +                var->data.explicit_offset);
>     > +         num_outputs += glsl_count_attribute_slots(var->type,
>     false);
>     > +      }
>     > +   }
>     > +   if (num_outputs == 0)
>     > +      return NULL;
>     > +
>     > +   nir_xfb_info *xfb = rzalloc_size(mem_ctx,
>     nir_xfb_info_size(num_outputs));
>     > +
>     > +   /* Walk the list of outputs and add them to the array */
>     > +   nir_foreach_variable(var, &shader->outputs) {
>     > +      if (var->data.explicit_xfb_buffer ||
>     > +          var->data.explicit_xfb_stride) {
>     > +         unsigned logcation = var->data.location;
>     > +         unsigned offset = var->data.offset;
>     > +         add_var_xfb_outputs(xfb, var, &location, &offset,
>     var->type);
>     > +      }
>     > +   }
>     > +   assert(xfb->output_count == num_outputs);
>     > +
>     > +   /* Everything is easier in the state setup code if the list
>     is sorted in
>     > +    * order of output offset.
>     > +    */
>     > +   qsort(xfb->outputs, xfb->output_count, sizeof(xfb->outputs[0]),
>     > +         compare_xfb_output_offsets);
>     > +
>     > +   /* Finally, do a sanity check */
>     > +   unsigned max_offset[NIR_MAX_XFB_BUFFERS] = { };
>     > +   for (unsigned i = 0; i < xfb->output_count; i++) {
>     > +      assert(xfb->outputs[i].offset >=
>     max_offset[xfb->outputs[i].buffer]);
>     > +      assert(xfb->outputs[i].component_mask != 0);
>     > +      unsigned slots =
>     util_bitcount(xfb->outputs[i].component_mask);
>     > +      max_offset[xfb->outputs[i].buffer] =
>     xfb->outputs[i].offset + slots * 4;
>     > +   }
>     > +
>     > +   return xfb;
>     > +}
>     > diff --git a/src/compiler/nir/nir_xfb_info.h
>     b/src/compiler/nir/nir_xfb_info.h
>     > new file mode 100644
>     > index 00000000000..9b543df5f47
>     > --- /dev/null
>     > +++ b/src/compiler/nir/nir_xfb_info.h
>     > @@ -0,0 +1,59 @@
>     > +/*
>     > + * Copyright © 2018 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.
>     > + */
>     > +
>     > +#ifndef NIR_XFB_INFO_H
>     > +#define NIR_XFB_INFO_H
>     > +
>     > +#include "nir.h"
>     > +
>     > +#define NIR_MAX_XFB_BUFFERS 4
>     > +#define NIR_MAX_XFB_STREAMS 4
>     > +
>     > +typedef struct {
>     > +   uint8_t buffer;
>     > +   uint16_t offset;
>     > +   uint8_t location;
>     > +   uint8_t component_mask;
>     > +} nir_xfb_output_info;
>     > +
>     > +typedef struct {
>     > +   uint8_t buffers_written;
>     > +   uint8_t streams_written;
>     > +
>     > +   uint16_t strides[NIR_MAX_XFB_BUFFERS];
>     > +   uint8_t buffer_to_stream[NIR_MAX_XFB_STREAMS];
>     > +
>     > +   uint16_t output_count;
>     > +   nir_xfb_output_info outputs[0];
>     > +} nir_xfb_info;
>     > +
>     > +static inline size_t
>     > +nir_xfb_info_size(uint16_t output_count)
>     > +{
>     > +   return sizeof(nir_xfb_info) + sizeof(nir_xfb_output_info) *
>     output_count;
>     > +}
>     > +
>     > +nir_xfb_info *
>     > +nir_gather_xfb_info(const nir_shader *shader, void *mem_ctx);
>     > +
>     > +#endif /* NIR_XFB_INFO_H */
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20181005/35fcb0f4/attachment-0001.html>


More information about the mesa-dev mailing list