From 9f12d9166bedbe281e6cd866a903f8bdcc472c12 Mon Sep 17 00:00:00 2001 From: Rafael Antognolli Date: Fri, 17 Mar 2017 10:23:45 -0700 Subject: [PATCH] i965: Port gen6+ 3DSTATE_SF to genxml. Emit sf state on Gen6+ using brw_batch_emit helper, using pack structs from genxml. v3: - Reorganize code and reduce #if/#endif's (Ken) - Style fixes (Ken) - Always set AALINEDISTANCE_TRUE (Ken) Signed-off-by: Rafael Antognolli Reviewed-by: Kenneth Graunke --- src/mesa/drivers/dri/i965/brw_state.h | 3 - src/mesa/drivers/dri/i965/gen6_sf_state.c | 189 ------------ src/mesa/drivers/dri/i965/gen7_sf_state.c | 156 ---------- src/mesa/drivers/dri/i965/gen8_sf_state.c | 73 ----- src/mesa/drivers/dri/i965/genX_state_upload.c | 420 +++++++++++++++++++++++++- 5 files changed, 417 insertions(+), 424 deletions(-) diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h index 3a10a8a..594757c 100644 --- a/src/mesa/drivers/dri/i965/brw_state.h +++ b/src/mesa/drivers/dri/i965/brw_state.h @@ -119,7 +119,6 @@ extern const struct brw_tracked_state gen6_renderbuffer_surfaces; extern const struct brw_tracked_state gen6_sampler_state; extern const struct brw_tracked_state gen6_scissor_state; extern const struct brw_tracked_state gen6_sol_surface; -extern const struct brw_tracked_state gen6_sf_state; extern const struct brw_tracked_state gen6_sf_vp; extern const struct brw_tracked_state gen6_urb; extern const struct brw_tracked_state gen6_viewport_state; @@ -137,7 +136,6 @@ extern const struct brw_tracked_state gen7_ps_state; extern const struct brw_tracked_state gen7_push_constant_space; extern const struct brw_tracked_state gen7_sbe_state; extern const struct brw_tracked_state gen7_sf_clip_viewport; -extern const struct brw_tracked_state gen7_sf_state; extern const struct brw_tracked_state gen7_sol_state; extern const struct brw_tracked_state gen7_te_state; extern const struct brw_tracked_state gen7_tes_push_constants; @@ -157,7 +155,6 @@ extern const struct brw_tracked_state gen8_ps_extra; extern const struct brw_tracked_state gen8_ps_state; extern const struct brw_tracked_state gen8_wm_state; extern const struct brw_tracked_state gen8_sbe_state; -extern const struct brw_tracked_state gen8_sf_state; extern const struct brw_tracked_state gen8_sf_clip_viewport; extern const struct brw_tracked_state gen8_vertices; extern const struct brw_tracked_state gen8_vf_topology; diff --git a/src/mesa/drivers/dri/i965/gen6_sf_state.c b/src/mesa/drivers/dri/i965/gen6_sf_state.c index 0f118b6..45b5769 100644 --- a/src/mesa/drivers/dri/i965/gen6_sf_state.c +++ b/src/mesa/drivers/dri/i965/gen6_sf_state.c @@ -263,192 +263,3 @@ calculate_attr_overrides(const struct brw_context *brw, */ *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_FS_PROG_DATA */ - const struct brw_wm_prog_data *wm_prog_data = - brw_wm_prog_data(brw->wm.base.prog_data); - uint32_t num_outputs = wm_prog_data->num_varying_inputs; - uint32_t dw1, dw2, dw3, dw4; - uint32_t point_sprite_enables; - int i; - /* _NEW_BUFFER */ - bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer); - const bool multisampled_fbo = _mesa_geometric_samples(ctx->DrawBuffer) > 1; - - float point_size; - uint16_t attr_overrides[16]; - uint32_t point_sprite_origin; - - dw1 = GEN6_SF_SWIZZLE_ENABLE | num_outputs << GEN6_SF_NUM_OUTPUTS_SHIFT; - dw2 = GEN6_SF_STATISTICS_ENABLE; - dw3 = GEN6_SF_SCISSOR_ENABLE | GEN6_SF_LINE_AA_MODE_TRUE; - dw4 = 0; - - if (brw->sf.viewport_transform_enable) - dw2 |= GEN6_SF_VIEWPORT_TRANSFORM_ENABLE; - - /* _NEW_POLYGON */ - if (ctx->Polygon._FrontBit == render_to_fbo) - dw2 |= GEN6_SF_WINDING_CCW; - - if (ctx->Polygon.OffsetFill) - dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_SOLID; - - if (ctx->Polygon.OffsetLine) - dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_WIREFRAME; - - if (ctx->Polygon.OffsetPoint) - dw2 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_POINT; - - switch (ctx->Polygon.FrontMode) { - case GL_FILL: - dw2 |= GEN6_SF_FRONT_SOLID; - break; - - case GL_LINE: - dw2 |= GEN6_SF_FRONT_WIREFRAME; - break; - - case GL_POINT: - dw2 |= GEN6_SF_FRONT_POINT; - break; - - default: - unreachable("not reached"); - } - - switch (ctx->Polygon.BackMode) { - case GL_FILL: - dw2 |= GEN6_SF_BACK_SOLID; - break; - - case GL_LINE: - dw2 |= GEN6_SF_BACK_WIREFRAME; - break; - - case GL_POINT: - dw2 |= GEN6_SF_BACK_POINT; - break; - - default: - unreachable("not reached"); - } - - /* _NEW_POLYGON */ - if (ctx->Polygon.CullFlag) { - switch (ctx->Polygon.CullFaceMode) { - case GL_FRONT: - dw3 |= GEN6_SF_CULL_FRONT; - break; - case GL_BACK: - dw3 |= GEN6_SF_CULL_BACK; - break; - case GL_FRONT_AND_BACK: - dw3 |= GEN6_SF_CULL_BOTH; - break; - default: - unreachable("not reached"); - } - } else { - dw3 |= GEN6_SF_CULL_NONE; - } - - /* _NEW_LINE */ - { - uint32_t line_width_u3_7 = brw_get_line_width(brw); - dw3 |= line_width_u3_7 << GEN6_SF_LINE_WIDTH_SHIFT; - } - if (ctx->Line.SmoothFlag) { - dw3 |= GEN6_SF_LINE_AA_ENABLE; - dw3 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0; - } - /* _NEW_MULTISAMPLE */ - if (multisampled_fbo && ctx->Multisample.Enabled) - dw3 |= GEN6_SF_MSRAST_ON_PATTERN; - - /* _NEW_PROGRAM | _NEW_POINT, BRW_NEW_VUE_MAP_GEOM_OUT */ - if (use_state_point_size(brw)) - dw4 |= GEN6_SF_USE_STATE_POINT_WIDTH; - - /* _NEW_POINT - Clamp to ARB_point_parameters user limits */ - point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize); - - /* Clamp to the hardware limits and convert to fixed point */ - dw4 |= U_FIXED(CLAMP(point_size, 0.125f, 255.875f), 3); - - /* - * Window coordinates in an FBO are inverted, which means point - * sprite origin must be inverted, too. - */ - if ((ctx->Point.SpriteOrigin == GL_LOWER_LEFT) != render_to_fbo) { - point_sprite_origin = GEN6_SF_POINT_SPRITE_LOWERLEFT; - } else { - point_sprite_origin = GEN6_SF_POINT_SPRITE_UPPERLEFT; - } - dw1 |= point_sprite_origin; - - /* _NEW_LIGHT */ - if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) { - dw4 |= - (2 << GEN6_SF_TRI_PROVOKE_SHIFT) | - (2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT) | - (1 << GEN6_SF_LINE_PROVOKE_SHIFT); - } else { - dw4 |= - (1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT); - } - - /* BRW_NEW_VUE_MAP_GEOM_OUT | BRW_NEW_FRAGMENT_PROGRAM | - * _NEW_POINT | _NEW_LIGHT | _NEW_PROGRAM | BRW_NEW_FS_PROG_DATA - */ - uint32_t urb_entry_read_length; - uint32_t urb_entry_read_offset; - calculate_attr_overrides(brw, attr_overrides, &point_sprite_enables, - &urb_entry_read_length, &urb_entry_read_offset); - 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)); - OUT_BATCH(dw1); - OUT_BATCH(dw2); - OUT_BATCH(dw3); - OUT_BATCH(dw4); - OUT_BATCH_F(ctx->Polygon.OffsetUnits * 2); /* constant. copied from gen4 */ - OUT_BATCH_F(ctx->Polygon.OffsetFactor); /* scale */ - OUT_BATCH_F(ctx->Polygon.OffsetClamp); /* global depth offset clamp */ - for (i = 0; i < 8; i++) { - OUT_BATCH(attr_overrides[i * 2] | attr_overrides[i * 2 + 1] << 16); - } - OUT_BATCH(point_sprite_enables); /* dw16 */ - OUT_BATCH(wm_prog_data->flat_inputs); - OUT_BATCH(0); /* wrapshortest enables 0-7 */ - OUT_BATCH(0); /* wrapshortest enables 8-15 */ - ADVANCE_BATCH(); -} - -const struct brw_tracked_state gen6_sf_state = { - .dirty = { - .mesa = _NEW_BUFFERS | - _NEW_LIGHT | - _NEW_LINE | - _NEW_MULTISAMPLE | - _NEW_POINT | - _NEW_POLYGON | - _NEW_PROGRAM, - .brw = BRW_NEW_BLORP | - BRW_NEW_CONTEXT | - BRW_NEW_FRAGMENT_PROGRAM | - BRW_NEW_FS_PROG_DATA | - BRW_NEW_GS_PROG_DATA | - BRW_NEW_PRIMITIVE | - BRW_NEW_TES_PROG_DATA | - BRW_NEW_VUE_MAP_GEOM_OUT, - }, - .emit = upload_sf_state, -}; diff --git a/src/mesa/drivers/dri/i965/gen7_sf_state.c b/src/mesa/drivers/dri/i965/gen7_sf_state.c index d577a36..7ab8a99 100644 --- a/src/mesa/drivers/dri/i965/gen7_sf_state.c +++ b/src/mesa/drivers/dri/i965/gen7_sf_state.c @@ -107,159 +107,3 @@ const struct brw_tracked_state gen7_sbe_state = { }, .emit = upload_sbe_state, }; - -static void -upload_sf_state(struct brw_context *brw) -{ - struct gl_context *ctx = &brw->ctx; - uint32_t dw1, dw2, dw3; - float point_size; - /* _NEW_BUFFERS */ - bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer); - const bool multisampled_fbo = _mesa_geometric_samples(ctx->DrawBuffer) > 1; - - dw1 = GEN6_SF_STATISTICS_ENABLE; - - if (brw->sf.viewport_transform_enable) - dw1 |= GEN6_SF_VIEWPORT_TRANSFORM_ENABLE; - - /* _NEW_BUFFERS */ - dw1 |= (brw_depthbuffer_format(brw) << GEN7_SF_DEPTH_BUFFER_SURFACE_FORMAT_SHIFT); - - /* _NEW_POLYGON */ - if (ctx->Polygon._FrontBit == render_to_fbo) - dw1 |= GEN6_SF_WINDING_CCW; - - if (ctx->Polygon.OffsetFill) - dw1 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_SOLID; - - if (ctx->Polygon.OffsetLine) - dw1 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_WIREFRAME; - - if (ctx->Polygon.OffsetPoint) - dw1 |= GEN6_SF_GLOBAL_DEPTH_OFFSET_POINT; - - switch (ctx->Polygon.FrontMode) { - case GL_FILL: - dw1 |= GEN6_SF_FRONT_SOLID; - break; - - case GL_LINE: - dw1 |= GEN6_SF_FRONT_WIREFRAME; - break; - - case GL_POINT: - dw1 |= GEN6_SF_FRONT_POINT; - break; - - default: - unreachable("not reached"); - } - - switch (ctx->Polygon.BackMode) { - case GL_FILL: - dw1 |= GEN6_SF_BACK_SOLID; - break; - - case GL_LINE: - dw1 |= GEN6_SF_BACK_WIREFRAME; - break; - - case GL_POINT: - dw1 |= GEN6_SF_BACK_POINT; - break; - - default: - unreachable("not reached"); - } - - dw2 = GEN6_SF_SCISSOR_ENABLE; - - if (ctx->Polygon.CullFlag) { - switch (ctx->Polygon.CullFaceMode) { - case GL_FRONT: - dw2 |= GEN6_SF_CULL_FRONT; - break; - case GL_BACK: - dw2 |= GEN6_SF_CULL_BACK; - break; - case GL_FRONT_AND_BACK: - dw2 |= GEN6_SF_CULL_BOTH; - break; - default: - unreachable("not reached"); - } - } else { - dw2 |= GEN6_SF_CULL_NONE; - } - - /* _NEW_LINE */ - { - uint32_t line_width_u3_7 = brw_get_line_width(brw); - dw2 |= line_width_u3_7 << GEN6_SF_LINE_WIDTH_SHIFT; - } - if (ctx->Line.SmoothFlag) { - dw2 |= GEN6_SF_LINE_AA_ENABLE; - dw2 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0; - } - if (ctx->Line.StippleFlag && brw->is_haswell) { - dw2 |= HSW_SF_LINE_STIPPLE_ENABLE; - } - /* _NEW_MULTISAMPLE */ - if (multisampled_fbo && ctx->Multisample.Enabled) - dw2 |= GEN6_SF_MSRAST_ON_PATTERN; - - /* FINISHME: Last Pixel Enable? Vertex Sub Pixel Precision Select? - */ - - dw3 = GEN6_SF_LINE_AA_MODE_TRUE; - - /* _NEW_PROGRAM | _NEW_POINT, BRW_NEW_VUE_MAP_GEOM_OUT */ - if (use_state_point_size(brw)) - dw3 |= GEN6_SF_USE_STATE_POINT_WIDTH; - - /* _NEW_POINT - Clamp to ARB_point_parameters user limits */ - point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize); - - /* Clamp to the hardware limits and convert to fixed point */ - dw3 |= U_FIXED(CLAMP(point_size, 0.125f, 255.875f), 3); - - /* _NEW_LIGHT */ - if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) { - dw3 |= - (2 << GEN6_SF_TRI_PROVOKE_SHIFT) | - (2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT) | - (1 << GEN6_SF_LINE_PROVOKE_SHIFT); - } else { - dw3 |= (1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT); - } - - BEGIN_BATCH(7); - OUT_BATCH(_3DSTATE_SF << 16 | (7 - 2)); - OUT_BATCH(dw1); - OUT_BATCH(dw2); - OUT_BATCH(dw3); - OUT_BATCH_F(ctx->Polygon.OffsetUnits * 2); /* constant. copied from gen4 */ - OUT_BATCH_F(ctx->Polygon.OffsetFactor); /* scale */ - OUT_BATCH_F(ctx->Polygon.OffsetClamp); /* global depth offset clamp */ - ADVANCE_BATCH(); -} - -const struct brw_tracked_state gen7_sf_state = { - .dirty = { - .mesa = _NEW_BUFFERS | - _NEW_LIGHT | - _NEW_LINE | - _NEW_MULTISAMPLE | - _NEW_POINT | - _NEW_POLYGON | - _NEW_PROGRAM, - .brw = BRW_NEW_BLORP | - BRW_NEW_CONTEXT | - BRW_NEW_GS_PROG_DATA | - BRW_NEW_PRIMITIVE | - BRW_NEW_TES_PROG_DATA | - BRW_NEW_VUE_MAP_GEOM_OUT, - }, - .emit = upload_sf_state, -}; diff --git a/src/mesa/drivers/dri/i965/gen8_sf_state.c b/src/mesa/drivers/dri/i965/gen8_sf_state.c index d47adcd..3b1dd61 100644 --- a/src/mesa/drivers/dri/i965/gen8_sf_state.c +++ b/src/mesa/drivers/dri/i965/gen8_sf_state.c @@ -151,76 +151,3 @@ const struct brw_tracked_state gen8_sbe_state = { }, .emit = upload_sbe, }; - -static void -upload_sf(struct brw_context *brw) -{ - struct gl_context *ctx = &brw->ctx; - uint32_t dw1 = 0, dw2 = 0, dw3 = 0; - float point_size; - - dw1 = GEN6_SF_STATISTICS_ENABLE; - - if (brw->sf.viewport_transform_enable) - dw1 |= GEN6_SF_VIEWPORT_TRANSFORM_ENABLE; - - /* _NEW_LINE */ - uint32_t line_width_u3_7 = brw_get_line_width(brw); - if (brw->gen >= 9 || brw->is_cherryview) { - dw1 |= line_width_u3_7 << GEN9_SF_LINE_WIDTH_SHIFT; - } else { - dw2 |= line_width_u3_7 << GEN6_SF_LINE_WIDTH_SHIFT; - } - - if (ctx->Line.SmoothFlag) { - dw2 |= GEN6_SF_LINE_END_CAP_WIDTH_1_0; - } - - /* _NEW_POINT - Clamp to ARB_point_parameters user limits */ - point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize); - - /* Clamp to the hardware limits and convert to fixed point */ - dw3 |= U_FIXED(CLAMP(point_size, 0.125f, 255.875f), 3); - - /* _NEW_PROGRAM | _NEW_POINT, BRW_NEW_VUE_MAP_GEOM_OUT */ - if (use_state_point_size(brw)) - dw3 |= GEN6_SF_USE_STATE_POINT_WIDTH; - - /* _NEW_POINT | _NEW_MULTISAMPLE */ - if ((ctx->Point.SmoothFlag || _mesa_is_multisample_enabled(ctx)) && - !ctx->Point.PointSprite) { - dw3 |= GEN8_SF_SMOOTH_POINT_ENABLE; - } - - dw3 |= GEN6_SF_LINE_AA_MODE_TRUE; - - /* _NEW_LIGHT */ - if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) { - dw3 |= (2 << GEN6_SF_TRI_PROVOKE_SHIFT) | - (2 << GEN6_SF_TRIFAN_PROVOKE_SHIFT) | - (1 << GEN6_SF_LINE_PROVOKE_SHIFT); - } else { - dw3 |= (1 << GEN6_SF_TRIFAN_PROVOKE_SHIFT); - } - - BEGIN_BATCH(4); - OUT_BATCH(_3DSTATE_SF << 16 | (4 - 2)); - OUT_BATCH(dw1); - OUT_BATCH(dw2); - OUT_BATCH(dw3); - ADVANCE_BATCH(); -} - -const struct brw_tracked_state gen8_sf_state = { - .dirty = { - .mesa = _NEW_LIGHT | - _NEW_PROGRAM | - _NEW_LINE | - _NEW_MULTISAMPLE | - _NEW_POINT, - .brw = BRW_NEW_BLORP | - BRW_NEW_CONTEXT | - BRW_NEW_VUE_MAP_GEOM_OUT, - }, - .emit = upload_sf, -}; diff --git a/src/mesa/drivers/dri/i965/genX_state_upload.c b/src/mesa/drivers/dri/i965/genX_state_upload.c index ac46731..84d2271 100644 --- a/src/mesa/drivers/dri/i965/genX_state_upload.c +++ b/src/mesa/drivers/dri/i965/genX_state_upload.c @@ -28,6 +28,7 @@ #include "brw_context.h" #include "brw_state.h" +#include "brw_util.h" #include "intel_batchbuffer.h" #include "intel_fbo.h" @@ -111,6 +112,235 @@ __gen_combine_address(struct brw_context *brw, void *location, _brw_cmd_pack(cmd)(brw, (void *)_dst, &name), \ _dst = NULL) +#if GEN_GEN == 6 +/** + * Determine the appropriate attribute override value to store into the + * 3DSTATE_SF structure for a given fragment shader attribute. The attribute + * override value contains two pieces of information: the location of the + * attribute in the VUE (relative to urb_entry_read_offset, see below), and a + * flag indicating whether to "swizzle" the attribute based on the direction + * the triangle is facing. + * + * If an attribute is "swizzled", then the given VUE location is used for + * front-facing triangles, and the VUE location that immediately follows is + * used for back-facing triangles. We use this to implement the mapping from + * gl_FrontColor/gl_BackColor to gl_Color. + * + * urb_entry_read_offset is the offset into the VUE at which the SF unit is + * being instructed to begin reading attribute data. It can be set to a + * nonzero value to prevent the SF unit from wasting time reading elements of + * the VUE that are not needed by the fragment shader. It is measured in + * 256-bit increments. + */ +static void +genX(get_attr_override)(struct GENX(SF_OUTPUT_ATTRIBUTE_DETAIL) *attr, + const struct brw_vue_map *vue_map, + int urb_entry_read_offset, int fs_attr, + bool two_side_color, uint32_t *max_source_attr) +{ + /* Find the VUE slot for this attribute. */ + int slot = vue_map->varying_to_slot[fs_attr]; + + /* Viewport and Layer are stored in the VUE header. We need to override + * them to zero if earlier stages didn't write them, as GL requires that + * they read back as zero when not explicitly set. + */ + if (fs_attr == VARYING_SLOT_VIEWPORT || fs_attr == VARYING_SLOT_LAYER) { + attr->ComponentOverrideX = true; + attr->ComponentOverrideW = true; + attr->ConstantSource = CONST_0000; + + if (!(vue_map->slots_valid & VARYING_BIT_LAYER)) + attr->ComponentOverrideY = true; + if (!(vue_map->slots_valid & VARYING_BIT_VIEWPORT)) + attr->ComponentOverrideZ = true; + + return; + } + + /* If there was only a back color written but not front, use back + * as the color instead of undefined + */ + if (slot == -1 && fs_attr == VARYING_SLOT_COL0) + slot = vue_map->varying_to_slot[VARYING_SLOT_BFC0]; + if (slot == -1 && fs_attr == VARYING_SLOT_COL1) + slot = vue_map->varying_to_slot[VARYING_SLOT_BFC1]; + + if (slot == -1) { + /* This attribute does not exist in the VUE--that means that the vertex + * shader did not write to it. This means that either: + * + * (a) This attribute is a texture coordinate, and it is going to be + * replaced with point coordinates (as a consequence of a call to + * glTexEnvi(GL_POINT_SPRITE, GL_COORD_REPLACE, GL_TRUE)), so the + * hardware will ignore whatever attribute override we supply. + * + * (b) This attribute is read by the fragment shader but not written by + * the vertex shader, so its value is undefined. Therefore the + * attribute override we supply doesn't matter. + * + * (c) This attribute is gl_PrimitiveID, and it wasn't written by the + * previous shader stage. + * + * Note that we don't have to worry about the cases where the attribute + * is gl_PointCoord or is undergoing point sprite coordinate + * replacement, because in those cases, this function isn't called. + * + * In case (c), we need to program the attribute overrides so that the + * primitive ID will be stored in this slot. In every other case, the + * attribute override we supply doesn't matter. So just go ahead and + * program primitive ID in every case. + */ + attr->ComponentOverrideW = true; + attr->ComponentOverrideX = true; + attr->ComponentOverrideY = true; + attr->ComponentOverrideZ = true; + attr->ConstantSource = PRIM_ID; + return; + } + + /* Compute the location of the attribute relative to urb_entry_read_offset. + * Each increment of urb_entry_read_offset represents a 256-bit value, so + * it counts for two 128-bit VUE slots. + */ + int source_attr = slot - 2 * urb_entry_read_offset; + assert(source_attr >= 0 && source_attr < 32); + + /* If we are doing two-sided color, and the VUE slot following this one + * represents a back-facing color, then we need to instruct the SF unit to + * do back-facing swizzling. + */ + bool swizzling = two_side_color && + ((vue_map->slot_to_varying[slot] == VARYING_SLOT_COL0 && + vue_map->slot_to_varying[slot+1] == VARYING_SLOT_BFC0) || + (vue_map->slot_to_varying[slot] == VARYING_SLOT_COL1 && + vue_map->slot_to_varying[slot+1] == VARYING_SLOT_BFC1)); + + /* Update max_source_attr. If swizzling, the SF will read this slot + 1. */ + if (*max_source_attr < source_attr + swizzling) + *max_source_attr = source_attr + swizzling; + + attr->SourceAttribute = source_attr; + if (swizzling) + attr->SwizzleSelect = INPUTATTR_FACING; +} + + +static void +genX(calculate_attr_overrides)(const struct brw_context *brw, + struct GENX(SF_OUTPUT_ATTRIBUTE_DETAIL) *attr_overrides, + uint32_t *point_sprite_enables, + uint32_t *urb_entry_read_length, + uint32_t *urb_entry_read_offset) +{ + const struct gl_context *ctx = &brw->ctx; + + /* _NEW_POINT */ + const struct gl_point_attrib *point = &ctx->Point; + + /* BRW_NEW_FS_PROG_DATA */ + const struct brw_wm_prog_data *wm_prog_data = + brw_wm_prog_data(brw->wm.base.prog_data); + uint32_t max_source_attr = 0; + + *point_sprite_enables = 0; + + /* BRW_NEW_FRAGMENT_PROGRAM + * + * If the fragment shader reads VARYING_SLOT_LAYER, then we need to pass in + * the full vertex header. Otherwise, we can program the SF to start + * reading at an offset of 1 (2 varying slots) to skip unnecessary data: + * - VARYING_SLOT_PSIZ and BRW_VARYING_SLOT_NDC on gen4-5 + * - VARYING_SLOT_{PSIZ,LAYER} and VARYING_SLOT_POS on gen6+ + */ + + bool fs_needs_vue_header = brw->fragment_program->info.inputs_read & + (VARYING_BIT_LAYER | VARYING_BIT_VIEWPORT); + + *urb_entry_read_offset = fs_needs_vue_header ? 0 : 1; + + /* From the Ivybridge PRM, Vol 2 Part 1, 3DSTATE_SBE, + * description of dw10 Point Sprite Texture Coordinate Enable: + * + * "This field must be programmed to zero when non-point primitives + * are rendered." + * + * The SandyBridge PRM doesn't explicitly say that point sprite enables + * must be programmed to zero when rendering non-point primitives, but + * the IvyBridge PRM does, and if we don't, we get garbage. + * + * This is not required on Haswell, as the hardware ignores this state + * when drawing non-points -- although we do still need to be careful to + * correctly set the attr overrides. + * + * _NEW_POLYGON + * BRW_NEW_PRIMITIVE | BRW_NEW_GS_PROG_DATA | BRW_NEW_TES_PROG_DATA + */ + bool drawing_points = brw_is_drawing_points(brw); + + for (int attr = 0; attr < VARYING_SLOT_MAX; attr++) { + int input_index = wm_prog_data->urb_setup[attr]; + + if (input_index < 0) + continue; + + /* _NEW_POINT */ + bool point_sprite = false; + if (drawing_points) { + if (point->PointSprite && + (attr >= VARYING_SLOT_TEX0 && attr <= VARYING_SLOT_TEX7) && + (point->CoordReplace & (1u << (attr - VARYING_SLOT_TEX0)))) { + point_sprite = true; + } + + if (attr == VARYING_SLOT_PNTC) + point_sprite = true; + + if (point_sprite) + *point_sprite_enables |= (1 << input_index); + } + + /* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_LIGHT | _NEW_PROGRAM */ + struct GENX(SF_OUTPUT_ATTRIBUTE_DETAIL) attribute = { 0 }; + + if (!point_sprite) { + genX(get_attr_override)(&attribute, + &brw->vue_map_geom_out, + *urb_entry_read_offset, attr, + brw->ctx.VertexProgram._TwoSideEnabled, + &max_source_attr); + } + + /* 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. + */ + if (input_index < 16) + attr_overrides[input_index] = attribute; + else + assert(attribute.SourceAttribute == input_index); + } + + /* 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 = DIV_ROUND_UP(max_source_attr + 1, 2); +} +#endif + /* ---------------------------------------------------------------------- */ #if GEN_GEN >= 6 @@ -338,6 +568,190 @@ static const struct brw_tracked_state genX(clip_state) = { .emit = genX(upload_clip_state), }; +/* ---------------------------------------------------------------------- */ + +static void +genX(upload_sf)(struct brw_context *brw) +{ + struct gl_context *ctx = &brw->ctx; + float point_size; + +#if GEN_GEN <= 7 + /* _NEW_BUFFERS */ + bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer); + const bool multisampled_fbo = _mesa_geometric_samples(ctx->DrawBuffer) > 1; +#endif + + brw_batch_emit(brw, GENX(3DSTATE_SF), sf) { + sf.StatisticsEnable = true; + sf.ViewportTransformEnable = brw->sf.viewport_transform_enable; + +#if GEN_GEN == 7 + /* _NEW_BUFFERS */ + sf.DepthBufferSurfaceFormat = brw_depthbuffer_format(brw); +#endif + +#if GEN_GEN <= 7 + /* _NEW_POLYGON */ + sf.FrontWinding = ctx->Polygon._FrontBit == render_to_fbo; + sf.GlobalDepthOffsetEnableSolid = ctx->Polygon.OffsetFill; + sf.GlobalDepthOffsetEnableWireframe = ctx->Polygon.OffsetLine; + sf.GlobalDepthOffsetEnablePoint = ctx->Polygon.OffsetPoint; + + switch (ctx->Polygon.FrontMode) { + case GL_FILL: + sf.FrontFaceFillMode = FILL_MODE_SOLID; + break; + case GL_LINE: + sf.FrontFaceFillMode = FILL_MODE_WIREFRAME; + break; + case GL_POINT: + sf.FrontFaceFillMode = FILL_MODE_POINT; + break; + default: + unreachable("not reached"); + } + + switch (ctx->Polygon.BackMode) { + case GL_FILL: + sf.BackFaceFillMode = FILL_MODE_SOLID; + break; + case GL_LINE: + sf.BackFaceFillMode = FILL_MODE_WIREFRAME; + break; + case GL_POINT: + sf.BackFaceFillMode = FILL_MODE_POINT; + break; + default: + unreachable("not reached"); + } + + sf.ScissorRectangleEnable = true; + + if (ctx->Polygon.CullFlag) { + switch (ctx->Polygon.CullFaceMode) { + case GL_FRONT: + sf.CullMode = CULLMODE_FRONT; + break; + case GL_BACK: + sf.CullMode = CULLMODE_BACK; + break; + case GL_FRONT_AND_BACK: + sf.CullMode = CULLMODE_BOTH; + break; + default: + unreachable("not reached"); + } + } else { + sf.CullMode = CULLMODE_NONE; + } + +#if GEN_IS_HASWELL + sf.LineStippleEnable = ctx->Line.StippleFlag; +#endif + + if (multisampled_fbo && ctx->Multisample.Enabled) + sf.MultisampleRasterizationMode = MSRASTMODE_ON_PATTERN; + + sf.GlobalDepthOffsetConstant = ctx->Polygon.OffsetUnits * 2; + sf.GlobalDepthOffsetScale = ctx->Polygon.OffsetFactor; + sf.GlobalDepthOffsetClamp = ctx->Polygon.OffsetClamp; +#endif + + /* _NEW_LINE */ + sf.LineWidth = brw_get_line_width_float(brw); + + if (ctx->Line.SmoothFlag) { + sf.LineEndCapAntialiasingRegionWidth = _10pixels; +#if GEN_GEN <= 7 + sf.AntiAliasingEnable = true; +#endif + } + + /* _NEW_POINT - Clamp to ARB_point_parameters user limits */ + point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize); + /* Clamp to the hardware limits */ + sf.PointWidth = CLAMP(point_size, 0.125f, 255.875f); + + /* _NEW_PROGRAM | _NEW_POINT, BRW_NEW_VUE_MAP_GEOM_OUT */ + if (use_state_point_size(brw)) + sf.PointWidthSource = State; + +#if GEN_GEN >= 8 + /* _NEW_POINT | _NEW_MULTISAMPLE */ + if ((ctx->Point.SmoothFlag || _mesa_is_multisample_enabled(ctx)) && + !ctx->Point.PointSprite) + sf.SmoothPointEnable = true; +#endif + + sf.AALineDistanceMode = AALINEDISTANCE_TRUE; + + /* _NEW_LIGHT */ + if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION) { + sf.TriangleStripListProvokingVertexSelect = 2; + sf.TriangleFanProvokingVertexSelect = 2; + sf.LineStripListProvokingVertexSelect = 1; + } else { + sf.TriangleFanProvokingVertexSelect = 1; + } + +#if GEN_GEN == 6 + /* BRW_NEW_FS_PROG_DATA */ + const struct brw_wm_prog_data *wm_prog_data = + brw_wm_prog_data(brw->wm.base.prog_data); + + sf.AttributeSwizzleEnable = true; + sf.NumberofSFOutputAttributes = wm_prog_data->num_varying_inputs; + + /* + * Window coordinates in an FBO are inverted, which means point + * sprite origin must be inverted, too. + */ + if ((ctx->Point.SpriteOrigin == GL_LOWER_LEFT) != render_to_fbo) { + sf.PointSpriteTextureCoordinateOrigin = LOWERLEFT; + } else { + sf.PointSpriteTextureCoordinateOrigin = UPPERLEFT; + } + + /* BRW_NEW_VUE_MAP_GEOM_OUT | BRW_NEW_FRAGMENT_PROGRAM | + * _NEW_POINT | _NEW_LIGHT | _NEW_PROGRAM | BRW_NEW_FS_PROG_DATA + */ + uint32_t urb_entry_read_length; + uint32_t urb_entry_read_offset; + uint32_t point_sprite_enables; + genX(calculate_attr_overrides)(brw, sf.Attribute, &point_sprite_enables, + &urb_entry_read_length, + &urb_entry_read_offset); + sf.VertexURBEntryReadLength = urb_entry_read_length; + sf.VertexURBEntryReadOffset = urb_entry_read_offset; + sf.PointSpriteTextureCoordinateEnable = point_sprite_enables; + sf.ConstantInterpolationEnable = wm_prog_data->flat_inputs; +#endif + } +} + +static const struct brw_tracked_state genX(sf_state) = { + .dirty = { + .mesa = _NEW_LIGHT | + _NEW_LINE | + _NEW_MULTISAMPLE | + _NEW_POINT | + _NEW_PROGRAM | + (GEN_GEN <= 7 ? _NEW_BUFFERS | _NEW_POLYGON : 0), + .brw = BRW_NEW_BLORP | + BRW_NEW_CONTEXT | + BRW_NEW_VUE_MAP_GEOM_OUT | + (GEN_GEN <= 7 ? BRW_NEW_GS_PROG_DATA | + BRW_NEW_PRIMITIVE | + BRW_NEW_TES_PROG_DATA + : 0) | + (GEN_GEN == 6 ? BRW_NEW_FS_PROG_DATA | + BRW_NEW_FRAGMENT_PROGRAM + : 0), + }, + .emit = genX(upload_sf), +}; + #endif /* ---------------------------------------------------------------------- */ @@ -570,7 +984,7 @@ genX(init_atoms)(struct brw_context *brw) &gen6_vs_state, &gen6_gs_state, &genX(clip_state), - &gen6_sf_state, + &genX(sf_state), &gen6_wm_state, &gen6_scissor_state, @@ -659,7 +1073,7 @@ genX(init_atoms)(struct brw_context *brw) &gen7_sol_state, &genX(clip_state), &gen7_sbe_state, - &gen7_sf_state, + &genX(sf_state), &gen7_wm_state, &gen7_ps_state, @@ -747,7 +1161,7 @@ genX(init_atoms)(struct brw_context *brw) &genX(clip_state), &genX(raster_state), &gen8_sbe_state, - &gen8_sf_state, + &genX(sf_state), &gen8_ps_blend, &gen8_ps_extra, &gen8_ps_state, -- 2.7.4