X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Fdrivers%2Fdri%2Fi965%2Fgen6_sf_state.c;h=8b1b7eb3f8f8d379a6b3c6f2fa651cdf6d275433;hb=672cffee0f24a40585c6c0a77caedd3aa0af60b1;hp=24d27546cd883ecbf96f088c17af31936a7e8a3c;hpb=61838fd9ad9e5e878cbc39d3f37b2188b9ec8d11;p=mesa.git diff --git a/src/mesa/drivers/dri/i965/gen6_sf_state.c b/src/mesa/drivers/dri/i965/gen6_sf_state.c index 24d27546cd8..8b1b7eb3f8f 100644 --- a/src/mesa/drivers/dri/i965/gen6_sf_state.c +++ b/src/mesa/drivers/dri/i965/gen6_sf_state.c @@ -31,6 +31,7 @@ #include "brw_util.h" #include "main/macros.h" #include "main/fbobject.h" +#include "main/framebuffer.h" #include "intel_batchbuffer.h" /** @@ -59,6 +60,23 @@ get_attr_override(const struct brw_vue_map *vue_map, int urb_entry_read_offset, /* 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) { + unsigned override = + ATTRIBUTE_0_OVERRIDE_X | ATTRIBUTE_0_OVERRIDE_W | + ATTRIBUTE_CONST_0000 << ATTRIBUTE_0_CONST_SOURCE_SHIFT; + + if (!(vue_map->slots_valid & VARYING_BIT_LAYER)) + override |= ATTRIBUTE_0_OVERRIDE_Y; + if (!(vue_map->slots_valid & VARYING_BIT_VIEWPORT)) + override |= ATTRIBUTE_0_OVERRIDE_Z; + + return override; + } + /* If there was only a back color written but not front, use back * as the color instead of undefined */ @@ -137,17 +155,47 @@ 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) + uint32_t *urb_entry_read_length, + uint32_t *urb_entry_read_offset) { - const int urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET; uint32_t max_source_attr = 0; *point_sprite_enables = 0; - *flat_enables = 0; - /* _NEW_LIGHT */ - bool shade_model_flat = brw->ctx.Light.ShadeModel == GL_FLAT; + *urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET; + + /* 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->Base.InputsRead & + (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_GEOMETRY_PROGRAM | BRW_NEW_TES_PROG_DATA + */ + bool drawing_points = is_drawing_points(brw); /* Initialize all the attr_overrides to 0. In the loop below we'll modify * just the ones that correspond to inputs used by the fs. @@ -155,11 +203,7 @@ calculate_attr_overrides(const struct brw_context *brw, memset(attr_overrides, 0, 16*sizeof(*attr_overrides)); for (int attr = 0; attr < VARYING_SLOT_MAX; attr++) { - /* BRW_NEW_FRAGMENT_PROGRAM */ - enum glsl_interp_qualifier interp_qualifier = - brw->fragment_program->InterpQualifier[attr]; - bool is_gl_Color = attr == VARYING_SLOT_COL0 || attr == VARYING_SLOT_COL1; - /* CACHE_NEW_WM_PROG */ + /* BRW_NEW_FS_PROG_DATA */ int input_index = brw->wm.prog_data->urb_setup[attr]; if (input_index < 0) @@ -167,28 +211,24 @@ calculate_attr_overrides(const struct brw_context *brw, /* _NEW_POINT */ bool point_sprite = false; - if (brw->ctx.Point.PointSprite && - (attr >= VARYING_SLOT_TEX0 && attr <= VARYING_SLOT_TEX7) && - brw->ctx.Point.CoordReplace[attr - VARYING_SLOT_TEX0]) { - point_sprite = true; + if (drawing_points) { + if (brw->ctx.Point.PointSprite && + (attr >= VARYING_SLOT_TEX0 && attr <= VARYING_SLOT_TEX7) && + brw->ctx.Point.CoordReplace[attr - VARYING_SLOT_TEX0]) { + point_sprite = true; + } + + if (attr == VARYING_SLOT_PNTC) + point_sprite = true; + + if (point_sprite) + *point_sprite_enables |= (1 << input_index); } - if (attr == VARYING_SLOT_PNTC) - point_sprite = true; - - if (point_sprite) - *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); - /* BRW_NEW_VUE_MAP_GEOM_OUT | _NEW_LIGHT | _NEW_PROGRAM */ uint16_t attr_override = point_sprite ? 0 : get_attr_override(&brw->vue_map_geom_out, - urb_entry_read_offset, attr, + *urb_entry_read_offset, attr, brw->ctx.VertexProgram._TwoSideEnabled, &max_source_attr); @@ -226,17 +266,15 @@ static void upload_sf_state(struct brw_context *brw) { struct gl_context *ctx = &brw->ctx; - /* CACHE_NEW_WM_PROG */ + /* BRW_NEW_FS_PROG_DATA */ uint32_t num_outputs = brw->wm.prog_data->num_varying_inputs; uint32_t dw1, dw2, dw3, dw4; uint32_t point_sprite_enables; - uint32_t flat_enables; int i; /* _NEW_BUFFER */ bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer); - bool multisampled_fbo = ctx->DrawBuffer->Visual.samples > 1; + const bool multisampled_fbo = _mesa_geometric_samples(ctx->DrawBuffer) > 1; - const int urb_entry_read_offset = BRW_SF_URB_ENTRY_READ_OFFSET; float point_size; uint16_t attr_overrides[16]; uint32_t point_sprite_origin; @@ -251,7 +289,7 @@ upload_sf_state(struct brw_context *brw) dw4 = 0; /* _NEW_POLYGON */ - if ((ctx->Polygon.FrontFace == GL_CCW) ^ render_to_fbo) + if (ctx->Polygon._FrontBit == render_to_fbo) dw2 |= GEN6_SF_WINDING_CCW; if (ctx->Polygon.OffsetFill) @@ -297,8 +335,9 @@ upload_sf_state(struct brw_context *brw) unreachable("not reached"); } - /* _NEW_SCISSOR */ - if (ctx->Scissor.EnableFlags) + /* _NEW_SCISSOR _NEW_POLYGON BRW_NEW_GEOMETRY_PROGRAM BRW_NEW_PRIMITIVE */ + if (ctx->Scissor.EnableFlags || + is_drawing_points(brw) || is_drawing_lines(brw)) dw3 |= GEN6_SF_SCISSOR_ENABLE; /* _NEW_POLYGON */ @@ -322,11 +361,7 @@ upload_sf_state(struct brw_context *brw) /* _NEW_LINE */ { - uint32_t line_width_u3_7 = - U_FIXED(CLAMP(ctx->Line.Width, 0.0, ctx->Const.MaxLineWidth), 7); - /* TODO: line width of 0 is not allowed when MSAA enabled */ - if (line_width_u3_7 == 0) - line_width_u3_7 = 1; + 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) { @@ -347,7 +382,7 @@ upload_sf_state(struct brw_context *brw) 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, 0.125, 255.875), 3); + dw4 |= U_FIXED(CLAMP(point_size, 0.125f, 255.875f), 3); /* * Window coordinates in an FBO are inverted, which means point @@ -372,11 +407,12 @@ upload_sf_state(struct brw_context *brw) } /* BRW_NEW_VUE_MAP_GEOM_OUT | BRW_NEW_FRAGMENT_PROGRAM | - * _NEW_POINT | _NEW_LIGHT | _NEW_PROGRAM | CACHE_NEW_WM_PROG + * _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, - &flat_enables, &urb_entry_read_length); + &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); @@ -388,12 +424,12 @@ upload_sf_state(struct brw_context *brw) OUT_BATCH(dw4); OUT_BATCH_F(ctx->Polygon.OffsetUnits * 2); /* constant. copied from gen4 */ OUT_BATCH_F(ctx->Polygon.OffsetFactor); /* scale */ - OUT_BATCH_F(0.0); /* XXX: global depth offset clamp */ + 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(flat_enables); + OUT_BATCH(brw->wm.prog_data->flat_inputs); OUT_BATCH(0); /* wrapshortest enables 0-7 */ OUT_BATCH(0); /* wrapshortest enables 8-15 */ ADVANCE_BATCH(); @@ -401,18 +437,22 @@ upload_sf_state(struct brw_context *brw) const struct brw_tracked_state gen6_sf_state = { .dirty = { - .mesa = (_NEW_LIGHT | - _NEW_PROGRAM | - _NEW_POLYGON | - _NEW_LINE | - _NEW_SCISSOR | - _NEW_BUFFERS | - _NEW_POINT | - _NEW_MULTISAMPLE), - .brw = (BRW_NEW_CONTEXT | - BRW_NEW_FRAGMENT_PROGRAM | - BRW_NEW_VUE_MAP_GEOM_OUT), - .cache = CACHE_NEW_WM_PROG + .mesa = _NEW_BUFFERS | + _NEW_LIGHT | + _NEW_LINE | + _NEW_MULTISAMPLE | + _NEW_POINT | + _NEW_POLYGON | + _NEW_PROGRAM | + _NEW_SCISSOR, + .brw = BRW_NEW_BLORP | + BRW_NEW_CONTEXT | + BRW_NEW_FRAGMENT_PROGRAM | + BRW_NEW_FS_PROG_DATA | + BRW_NEW_GEOMETRY_PROGRAM | + BRW_NEW_PRIMITIVE | + BRW_NEW_TES_PROG_DATA | + BRW_NEW_VUE_MAP_GEOM_OUT, }, .emit = upload_sf_state, };