[Mesa-dev] [PATCH 05/15] i965/sf: Consolidate common code for setting up gen6-7 attribute overrides.
Ian Romanick
idr at freedesktop.org
Mon Sep 9 13:05:44 PDT 2013
On 09/03/2013 06:18 PM, Paul Berry wrote:
> ---
> src/mesa/drivers/dri/i965/brw_state.h | 9 +-
> src/mesa/drivers/dri/i965/gen6_sf_state.c | 153 +++++++++++++++++-------------
> src/mesa/drivers/dri/i965/gen7_sf_state.c | 64 +------------
> 3 files changed, 97 insertions(+), 129 deletions(-)
>
> diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h
> index 22e4a61..dd3e216 100644
> --- a/src/mesa/drivers/dri/i965/brw_state.h
> +++ b/src/mesa/drivers/dri/i965/brw_state.h
> @@ -223,9 +223,12 @@ void gen4_init_vtable_sampler_functions(struct brw_context *brw);
> void gen7_init_vtable_sampler_functions(struct brw_context *brw);
>
> /* gen6_sf_state.c */
> -uint32_t
> -get_attr_override(const struct brw_vue_map *vue_map, int urb_entry_read_offset,
> - int fs_attr, bool two_side_color, uint32_t *max_source_attr);
> +void
> +calculate_attr_overrides(const struct brw_context *brw,
> + uint16_t *attr_overrides,
> + uint32_t *point_sprite_enables,
> + uint32_t *flat_enables,
> + uint32_t *urb_entry_read_length);
>
> /* brw_vs_surface_state.c */
> void
> diff --git a/src/mesa/drivers/dri/i965/gen6_sf_state.c b/src/mesa/drivers/dri/i965/gen6_sf_state.c
> index dfe9a31..7094994 100644
> --- a/src/mesa/drivers/dri/i965/gen6_sf_state.c
> +++ b/src/mesa/drivers/dri/i965/gen6_sf_state.c
> @@ -52,7 +52,7 @@
> * the VUE that are not needed by the fragment shader. It is measured in
> * 256-bit increments.
> */
> -uint32_t
> +static uint32_t
> get_attr_override(const struct brw_vue_map *vue_map, int urb_entry_read_offset,
> int fs_attr, bool two_side_color, uint32_t *max_source_attr)
> {
> @@ -123,21 +123,98 @@ get_attr_override(const struct brw_vue_map *vue_map, int urb_entry_read_offset,
> return source_attr;
> }
>
> +
> +/**
> + * Create the mapping from the FS inputs we produce to the VS outputs they
Only VS outputs?
> + * source from.
> + */
> +void
> +calculate_attr_overrides(const struct brw_context *brw,
> + uint16_t *attr_overrides,
> + uint32_t *point_sprite_enables,
> + uint32_t *flat_enables,
> + uint32_t *urb_entry_read_length)
> +{
> + const int urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET;
> + uint32_t max_source_attr = 0;
> + int input_index = 0;
> +
> + /* _NEW_LIGHT */
> + bool shade_model_flat = brw->ctx.Light.ShadeModel == GL_FLAT;
> +
> + for (int attr = 0; attr < VARYING_SLOT_MAX; attr++) {
> + enum glsl_interp_qualifier interp_qualifier =
> + brw->fragment_program->InterpQualifier[attr];
> + bool is_gl_Color = attr == VARYING_SLOT_COL0 || attr == VARYING_SLOT_COL1;
> +
> + if (!(brw->fragment_program->Base.InputsRead & BITFIELD64_BIT(attr)))
> + continue;
> +
> + /* _NEW_POINT */
> + if (brw->ctx.Point.PointSprite &&
> + (attr >= VARYING_SLOT_TEX0 && attr <= VARYING_SLOT_TEX7) &&
> + brw->ctx.Point.CoordReplace[attr - VARYING_SLOT_TEX0]) {
> + *point_sprite_enables |= (1 << input_index);
> + }
> +
> + if (attr == VARYING_SLOT_PNTC)
> + *point_sprite_enables |= (1 << input_index);
> +
> + /* flat shading */
> + if (interp_qualifier == INTERP_QUALIFIER_FLAT ||
> + (shade_model_flat && is_gl_Color &&
> + interp_qualifier == INTERP_QUALIFIER_NONE))
> + *flat_enables |= (1 << input_index);
> +
> + /* The hardware can only do the overrides on 16 overrides at a
> + * time, and the other up to 16 have to be lined up so that the
> + * input index = the output index. We'll need to do some
> + * tweaking to make sure that's the case.
> + */
> + assert(input_index < 16 || attr == input_index);
> +
> + /* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_LIGHT | _NEW_PROGRAM */
> + attr_overrides[input_index++] =
> + get_attr_override(&brw->vue_map_geom_out,
> + urb_entry_read_offset, attr,
> + brw->ctx.VertexProgram._TwoSideEnabled,
> + &max_source_attr);
> + }
> +
> + for (; input_index < VARYING_SLOT_MAX; input_index++)
> + attr_overrides[input_index] = 0;
> +
> + /* From the Sandy Bridge PRM, Volume 2, Part 1, documentation for
> + * 3DSTATE_SF DWord 1 bits 15:11, "Vertex URB Entry Read Length":
> + *
> + * "This field should be set to the minimum length required to read the
> + * maximum source attribute. The maximum source attribute is indicated
> + * by the maximum value of the enabled Attribute # Source Attribute if
> + * Attribute Swizzle Enable is set, Number of Output Attributes-1 if
> + * enable is not set.
> + * read_length = ceiling((max_source_attr + 1) / 2)
> + *
> + * [errata] Corruption/Hang possible if length programmed larger than
> + * recommended"
> + *
> + * Similar text exists for Ivy Bridge.
> + */
> + *urb_entry_read_length = ALIGN(max_source_attr + 1, 2) / 2;
> +}
> +
> +
> static void
> upload_sf_state(struct brw_context *brw)
> {
> struct gl_context *ctx = &brw->ctx;
> /* BRW_NEW_FRAGMENT_PROGRAM */
> uint32_t num_outputs = _mesa_bitcount_64(brw->fragment_program->Base.InputsRead);
> - /* _NEW_LIGHT */
> - bool shade_model_flat = ctx->Light.ShadeModel == GL_FLAT;
> uint32_t dw1, dw2, dw3, dw4, dw16, dw17;
> int i;
> /* _NEW_BUFFER */
> bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer);
> bool multisampled_fbo = ctx->DrawBuffer->Visual.samples > 1;
>
> - int attr = 0, input_index = 0;
> const int urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET;
> float point_size;
> uint16_t attr_overrides[VARYING_SLOT_MAX];
> @@ -276,68 +353,12 @@ upload_sf_state(struct brw_context *brw)
> (1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT);
> }
>
> - /* Create the mapping from the FS inputs we produce to the VS outputs
> - * they source from.
> - */
> - uint32_t max_source_attr = 0;
> - for (; attr < VARYING_SLOT_MAX; attr++) {
> - enum glsl_interp_qualifier interp_qualifier =
> - brw->fragment_program->InterpQualifier[attr];
> - bool is_gl_Color = attr == VARYING_SLOT_COL0 || attr == VARYING_SLOT_COL1;
> -
> - if (!(brw->fragment_program->Base.InputsRead & BITFIELD64_BIT(attr)))
> - continue;
> -
> - /* _NEW_POINT */
> - if (ctx->Point.PointSprite &&
> - (attr >= VARYING_SLOT_TEX0 && attr <= VARYING_SLOT_TEX7) &&
> - ctx->Point.CoordReplace[attr - VARYING_SLOT_TEX0]) {
> - dw16 |= (1 << input_index);
> - }
> -
> - if (attr == VARYING_SLOT_PNTC)
> - dw16 |= (1 << input_index);
> -
> - /* flat shading */
> - if (interp_qualifier == INTERP_QUALIFIER_FLAT ||
> - (shade_model_flat && is_gl_Color &&
> - interp_qualifier == INTERP_QUALIFIER_NONE))
> - dw17 |= (1 << input_index);
> -
> - /* The hardware can only do the overrides on 16 overrides at a
> - * time, and the other up to 16 have to be lined up so that the
> - * input index = the output index. We'll need to do some
> - * tweaking to make sure that's the case.
> - */
> - assert(input_index < 16 || attr == input_index);
> -
> - /* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_LIGHT | _NEW_PROGRAM */
> - attr_overrides[input_index++] =
> - get_attr_override(&brw->vue_map_geom_out,
> - urb_entry_read_offset, attr,
> - ctx->VertexProgram._TwoSideEnabled,
> - &max_source_attr);
> - }
> -
> - for (; input_index < VARYING_SLOT_MAX; input_index++)
> - attr_overrides[input_index] = 0;
> -
> - /* From the Sandy Bridge PRM, Volume 2, Part 1, documentation for
> - * 3DSTATE_SF DWord 1 bits 15:11, "Vertex URB Entry Read Length":
> - *
> - * "This field should be set to the minimum length required to read the
> - * maximum source attribute. The maximum source attribute is indicated
> - * by the maximum value of the enabled Attribute # Source Attribute if
> - * Attribute Swizzle Enable is set, Number of Output Attributes-1 if
> - * enable is not set.
> - * read_length = ceiling((max_source_attr + 1) / 2)
> - *
> - * [errata] Corruption/Hang possible if length programmed larger than
> - * recommended"
> - */
> - uint32_t urb_entry_read_length = ALIGN(max_source_attr + 1, 2) / 2;
> - dw1 |= urb_entry_read_length << GEN6_SF_URB_ENTRY_READ_LENGTH_SHIFT |
> - urb_entry_read_offset << GEN6_SF_URB_ENTRY_READ_OFFSET_SHIFT;
> + /* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_POINT | _NEW_LIGHT | _NEW_PROGRAM */
> + uint32_t urb_entry_read_length;
> + calculate_attr_overrides(brw, attr_overrides, &dw16, &dw17,
> + &urb_entry_read_length);
> + dw1 |= (urb_entry_read_length << GEN6_SF_URB_ENTRY_READ_LENGTH_SHIFT |
> + urb_entry_read_offset << GEN6_SF_URB_ENTRY_READ_OFFSET_SHIFT);
>
> BEGIN_BATCH(20);
> OUT_BATCH(_3DSTATE_SF << 16 | (20 - 2));
> diff --git a/src/mesa/drivers/dri/i965/gen7_sf_state.c b/src/mesa/drivers/dri/i965/gen7_sf_state.c
> index 715eb6c..3f7c2e5 100644
> --- a/src/mesa/drivers/dri/i965/gen7_sf_state.c
> +++ b/src/mesa/drivers/dri/i965/gen7_sf_state.c
> @@ -35,11 +35,8 @@ upload_sbe_state(struct brw_context *brw)
> struct gl_context *ctx = &brw->ctx;
> /* BRW_NEW_FRAGMENT_PROGRAM */
> uint32_t num_outputs = _mesa_bitcount_64(brw->fragment_program->Base.InputsRead);
> - /* _NEW_LIGHT */
> - bool shade_model_flat = ctx->Light.ShadeModel == GL_FLAT;
> uint32_t dw1, dw10, dw11;
> int i;
> - int attr = 0, input_index = 0;
> const int urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET;
> uint16_t attr_overrides[VARYING_SLOT_MAX];
> /* _NEW_BUFFERS */
> @@ -65,66 +62,13 @@ upload_sbe_state(struct brw_context *brw)
> dw10 = 0;
> dw11 = 0;
>
> - /* Create the mapping from the FS inputs we produce to the VS outputs
> - * they source from.
> - */
> - uint32_t max_source_attr = 0;
> - for (; attr < VARYING_SLOT_MAX; attr++) {
> - enum glsl_interp_qualifier interp_qualifier =
> - brw->fragment_program->InterpQualifier[attr];
> - bool is_gl_Color = attr == VARYING_SLOT_COL0 || attr == VARYING_SLOT_COL1;
> -
> - if (!(brw->fragment_program->Base.InputsRead & BITFIELD64_BIT(attr)))
> - continue;
> -
> - if (ctx->Point.PointSprite &&
> - attr >= VARYING_SLOT_TEX0 && attr <= VARYING_SLOT_TEX7 &&
> - ctx->Point.CoordReplace[attr - VARYING_SLOT_TEX0]) {
> - dw10 |= (1 << input_index);
> - }
> -
> - if (attr == VARYING_SLOT_PNTC)
> - dw10 |= (1 << input_index);
> -
> - /* flat shading */
> - if (interp_qualifier == INTERP_QUALIFIER_FLAT ||
> - (shade_model_flat && is_gl_Color &&
> - interp_qualifier == INTERP_QUALIFIER_NONE))
> - dw11 |= (1 << input_index);
> -
> - /* The hardware can only do the overrides on 16 overrides at a
> - * time, and the other up to 16 have to be lined up so that the
> - * input index = the output index. We'll need to do some
> - * tweaking to make sure that's the case.
> - */
> - assert(input_index < 16 || attr == input_index);
> -
> - /* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_LIGHT | _NEW_PROGRAM */
> - attr_overrides[input_index++] =
> - get_attr_override(&brw->vue_map_geom_out,
> - urb_entry_read_offset, attr,
> - ctx->VertexProgram._TwoSideEnabled,
> - &max_source_attr);
> - }
> -
> - /* From the Ivy Bridge PRM, Volume 2, Part 1, documentation for
> - * 3DSTATE_SBE DWord 1 bits 15:11, "Vertex URB Entry Read Length":
> - *
> - * "This field should be set to the minimum length required to read the
> - * maximum source attribute. The maximum source attribute is indicated
> - * by the maximum value of the enabled Attribute # Source Attribute if
> - * Attribute Swizzle Enable is set, Number of Output Attributes-1 if
> - * enable is not set.
> - *
> - * read_length = ceiling((max_source_attr + 1) / 2)"
> - */
> - uint32_t urb_entry_read_length = ALIGN(max_source_attr + 1, 2) / 2;
> + /* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_POINT | _NEW_LIGHT | _NEW_PROGRAM */
> + uint32_t urb_entry_read_length;
> + calculate_attr_overrides(brw, attr_overrides, &dw10, &dw11,
> + &urb_entry_read_length);
> dw1 |= urb_entry_read_length << GEN7_SBE_URB_ENTRY_READ_LENGTH_SHIFT |
> urb_entry_read_offset << GEN7_SBE_URB_ENTRY_READ_OFFSET_SHIFT;
>
> - for (; input_index < VARYING_SLOT_MAX; input_index++)
> - attr_overrides[input_index] = 0;
> -
> BEGIN_BATCH(14);
> OUT_BATCH(_3DSTATE_SBE << 16 | (14 - 2));
> OUT_BATCH(dw1);
>
More information about the mesa-dev
mailing list