void brw_clip_tri_alloc_regs( struct brw_clip_compile *c,
GLuint nr_verts )
{
- struct intel_context *intel = &c->func.brw->intel;
+ struct brw_context *brw = c->func.brw;
GLuint i = 0,j;
/* Register usage is static, precompute here:
i += c->nr_regs;
}
- if (c->nr_attrs & 1) {
+ if (c->vue_map.num_slots % 2) {
+ /* The VUE has an odd number of slots so the last register is only half
+ * used. Fill the second half with zero.
+ */
for (j = 0; j < 3; j++) {
- GLuint delta = c->nr_attrs*16 + 32;
-
- if (intel->gen == 5)
- delta = c->nr_attrs * 16 + 32 * 3;
+ GLuint delta = brw_vue_slot_to_offset(c->vue_map.num_slots);
brw_MOV(&c->func, byte_offset(c->reg.vertex[j], delta), brw_imm_f(0));
}
i++;
}
- if (intel->needs_ff_sync) {
+ c->reg.vertex_src_mask = retype(brw_vec1_grf(i, 0), BRW_REGISTER_TYPE_UD);
+ c->reg.clipdistance_offset = retype(brw_vec1_grf(i, 1), BRW_REGISTER_TYPE_W);
+ i++;
+
+ if (brw->gen == 5) {
c->reg.ff_sync = retype(brw_vec1_grf(i, 0), BRW_REGISTER_TYPE_UD);
i++;
}
{
struct brw_compile *p = &c->func;
struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */
- struct brw_instruction *is_rev;
/* Initial list of indices for incoming vertexes:
*/
/* XXX: Is there an easier way to do this? Need to reverse every
* second tristrip element: Can ignore sometimes?
*/
- is_rev = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
brw_MOV(p, get_element(c->reg.inlist, 0), brw_address(c->reg.vertex[1]) );
brw_MOV(p, get_element(c->reg.inlist, 1), brw_address(c->reg.vertex[0]) );
if (c->need_direction)
brw_MOV(p, c->reg.dir, brw_imm_f(-1));
}
- is_rev = brw_ELSE(p, is_rev);
+ brw_ELSE(p);
{
brw_MOV(p, get_element(c->reg.inlist, 0), brw_address(c->reg.vertex[0]) );
brw_MOV(p, get_element(c->reg.inlist, 1), brw_address(c->reg.vertex[1]) );
if (c->need_direction)
brw_MOV(p, c->reg.dir, brw_imm_f(1));
}
- brw_ENDIF(p, is_rev);
+ brw_ENDIF(p);
brw_MOV(p, get_element(c->reg.inlist, 2), brw_address(c->reg.vertex[2]) );
brw_MOV(p, brw_vec8_grf(c->reg.outlist.nr, 0), brw_imm_f(0));
void brw_clip_tri_flat_shade( struct brw_clip_compile *c )
{
struct brw_compile *p = &c->func;
- struct brw_instruction *is_poly, *is_trifan;
struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */
brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK));
tmp0,
brw_imm_ud(_3DPRIM_POLYGON));
- is_poly = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
- brw_clip_copy_colors(c, 1, 0);
- brw_clip_copy_colors(c, 2, 0);
+ brw_clip_copy_flatshaded_attributes(c, 1, 0);
+ brw_clip_copy_flatshaded_attributes(c, 2, 0);
}
- is_poly = brw_ELSE(p, is_poly);
+ brw_ELSE(p);
{
if (c->key.pv_first) {
brw_CMP(p,
BRW_CONDITIONAL_EQ,
tmp0,
brw_imm_ud(_3DPRIM_TRIFAN));
- is_trifan = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
- brw_clip_copy_colors(c, 0, 1);
- brw_clip_copy_colors(c, 2, 1);
+ brw_clip_copy_flatshaded_attributes(c, 0, 1);
+ brw_clip_copy_flatshaded_attributes(c, 2, 1);
}
- is_trifan = brw_ELSE(p, is_trifan);
+ brw_ELSE(p);
{
- brw_clip_copy_colors(c, 1, 0);
- brw_clip_copy_colors(c, 2, 0);
+ brw_clip_copy_flatshaded_attributes(c, 1, 0);
+ brw_clip_copy_flatshaded_attributes(c, 2, 0);
}
- brw_ENDIF(p, is_trifan);
+ brw_ENDIF(p);
}
else {
- brw_clip_copy_colors(c, 0, 2);
- brw_clip_copy_colors(c, 1, 2);
+ brw_clip_copy_flatshaded_attributes(c, 0, 2);
+ brw_clip_copy_flatshaded_attributes(c, 1, 2);
}
}
- brw_ENDIF(p, is_poly);
+ brw_ENDIF(p);
}
+/**
+ * Loads the clip distance for a vertex into `dst`, and ends with
+ * a comparison of it to zero with the condition `cond`.
+ *
+ * - If using a fixed plane, the distance is dot(hpos, plane).
+ * - If using a user clip plane, the distance is directly available in the vertex.
+ */
+static inline void
+load_clip_distance(struct brw_clip_compile *c, struct brw_indirect vtx,
+ struct brw_reg dst, GLuint hpos_offset, int cond)
+{
+ struct brw_compile *p = &c->func;
+
+ dst = vec4(dst);
+ brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+ brw_AND(p, vec1(brw_null_reg()), c->reg.vertex_src_mask, brw_imm_ud(1));
+ brw_IF(p, BRW_EXECUTE_1);
+ {
+ struct brw_indirect temp_ptr = brw_indirect(7, 0);
+ brw_ADD(p, get_addr_reg(temp_ptr), get_addr_reg(vtx), c->reg.clipdistance_offset);
+ brw_MOV(p, vec1(dst), deref_1f(temp_ptr, 0));
+ }
+ brw_ELSE(p);
+ {
+ brw_MOV(p, dst, deref_4f(vtx, hpos_offset));
+ brw_DP4(p, dst, dst, c->reg.plane_equation);
+ }
+ brw_ENDIF(p);
+
+ brw_set_conditionalmod(p, cond);
+ brw_CMP(p, brw_null_reg(), cond, vec1(dst), brw_imm_f(0.0f));
+}
+
/* Use mesa's clipping algorithms, translated to GEN4 assembly.
*/
struct brw_indirect inlist_ptr = brw_indirect(4, 0);
struct brw_indirect outlist_ptr = brw_indirect(5, 0);
struct brw_indirect freelist_ptr = brw_indirect(6, 0);
- struct brw_instruction *plane_loop;
- struct brw_instruction *plane_active;
- struct brw_instruction *vertex_loop;
- struct brw_instruction *next_test;
- struct brw_instruction *prev_test;
-
+ GLuint hpos_offset = brw_varying_to_offset(&c->vue_map, VARYING_SLOT_POS);
+ GLint clipdist0_offset = c->key.nr_userclip
+ ? brw_varying_to_offset(&c->vue_map, VARYING_SLOT_CLIP_DIST0)
+ : 0;
+
brw_MOV(p, get_addr_reg(vtxPrev), brw_address(c->reg.vertex[2]) );
brw_MOV(p, get_addr_reg(plane_ptr), brw_clip_plane0_address(c));
brw_MOV(p, get_addr_reg(inlist_ptr), brw_address(c->reg.inlist));
brw_MOV(p, get_addr_reg(freelist_ptr), brw_address(c->reg.vertex[3]) );
- plane_loop = brw_DO(p, BRW_EXECUTE_1);
+ /* Set the initial vertex source mask: The first 6 planes are the bounds
+ * of the view volume; the next 8 planes are the user clipping planes.
+ */
+ brw_MOV(p, c->reg.vertex_src_mask, brw_imm_ud(0x3fc0));
+
+ /* Set the initial clipdistance offset to be 6 floats before gl_ClipDistance[0].
+ * We'll increment 6 times before we start hitting actual user clipping. */
+ brw_MOV(p, c->reg.clipdistance_offset, brw_imm_d(clipdist0_offset - 6*sizeof(float)));
+
+ brw_DO(p, BRW_EXECUTE_1);
{
/* if (planemask & 1)
*/
brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
brw_AND(p, vec1(brw_null_reg()), c->reg.planemask, brw_imm_ud(1));
- plane_active = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
/* vtxOut = freelist_ptr++
*/
brw_MOV(p, c->reg.loopcount, c->reg.nr_verts);
brw_MOV(p, c->reg.nr_verts, brw_imm_ud(0));
- vertex_loop = brw_DO(p, BRW_EXECUTE_1);
+ brw_DO(p, BRW_EXECUTE_1);
{
/* vtx = *input_ptr;
*/
brw_MOV(p, get_addr_reg(vtx), deref_1uw(inlist_ptr, 0));
+ load_clip_distance(c, vtxPrev, c->reg.dpPrev, hpos_offset, BRW_CONDITIONAL_L);
/* IS_NEGATIVE(prev) */
- brw_set_conditionalmod(p, BRW_CONDITIONAL_L);
- brw_DP4(p, vec4(c->reg.dpPrev), deref_4f(vtxPrev, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation);
- prev_test = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
+ load_clip_distance(c, vtx, c->reg.dp, hpos_offset, BRW_CONDITIONAL_GE);
/* IS_POSITIVE(next)
*/
- brw_set_conditionalmod(p, BRW_CONDITIONAL_GE);
- brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation);
- next_test = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
/* Coming back in.
brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtxPrev) );
brw_set_predicate_control(p, BRW_PREDICATE_NONE);
- brw_clip_interp_vertex(c, vtxOut, vtxPrev, vtx, c->reg.t, GL_FALSE);
+ brw_clip_interp_vertex(c, vtxOut, vtxPrev, vtx, c->reg.t, false);
/* *outlist_ptr++ = vtxOut;
* nr_verts++;
brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1));
brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) );
}
- brw_ENDIF(p, next_test);
+ brw_ENDIF(p);
}
- prev_test = brw_ELSE(p, prev_test);
+ brw_ELSE(p);
{
/* *outlist_ptr++ = vtxPrev;
* nr_verts++;
brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short)));
brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1));
+ load_clip_distance(c, vtx, c->reg.dp, hpos_offset, BRW_CONDITIONAL_L);
/* IS_NEGATIVE(next)
*/
- brw_set_conditionalmod(p, BRW_CONDITIONAL_L);
- brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation);
- next_test = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
/* Going out of bounds. Avoid division by zero as we
* know dp != dpPrev from DIFFERENT_SIGNS, above.
brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtx) );
brw_set_predicate_control(p, BRW_PREDICATE_NONE);
- brw_clip_interp_vertex(c, vtxOut, vtx, vtxPrev, c->reg.t, GL_TRUE);
+ brw_clip_interp_vertex(c, vtxOut, vtx, vtxPrev, c->reg.t, true);
/* *outlist_ptr++ = vtxOut;
* nr_verts++;
brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1));
brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) );
}
- brw_ENDIF(p, next_test);
+ brw_ENDIF(p);
}
- brw_ENDIF(p, prev_test);
+ brw_ENDIF(p);
/* vtxPrev = vtx;
* inlist_ptr++;
brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1));
}
- brw_WHILE(p, vertex_loop);
+ brw_WHILE(p);
/* vtxPrev = *(outlist_ptr-1) OR: outlist[nr_verts-1]
* inlist = outlist
brw_MOV(p, get_addr_reg(inlist_ptr), brw_address(c->reg.inlist));
brw_MOV(p, get_addr_reg(outlist_ptr), brw_address(c->reg.outlist));
}
- brw_ENDIF(p, plane_active);
+ brw_ENDIF(p);
/* plane_ptr++;
*/
*/
brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1));
+ brw_SHR(p, c->reg.vertex_src_mask, c->reg.vertex_src_mask, brw_imm_ud(1));
+ brw_ADD(p, c->reg.clipdistance_offset, c->reg.clipdistance_offset, brw_imm_w(sizeof(float)));
}
- brw_WHILE(p, plane_loop);
+ brw_WHILE(p);
}
void brw_clip_tri_emit_polygon(struct brw_clip_compile *c)
{
struct brw_compile *p = &c->func;
- struct brw_instruction *loop, *if_insn;
/* for (loopcount = nr_verts-2; loopcount > 0; loopcount--)
*/
c->reg.nr_verts,
brw_imm_d(-2));
- if_insn = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
struct brw_indirect v0 = brw_indirect(0, 0);
struct brw_indirect vptr = brw_indirect(1, 0);
brw_MOV(p, get_addr_reg(vptr), brw_address(c->reg.inlist));
brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0));
- brw_clip_emit_vue(c, v0, 1, 0, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_START));
+ brw_clip_emit_vue(c, v0, BRW_URB_WRITE_ALLOCATE_COMPLETE,
+ ((_3DPRIM_TRIFAN << URB_WRITE_PRIM_TYPE_SHIFT)
+ | URB_WRITE_PRIM_START));
brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2));
brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0));
- loop = brw_DO(p, BRW_EXECUTE_1);
+ brw_DO(p, BRW_EXECUTE_1);
{
- brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_TRIFAN << 2));
+ brw_clip_emit_vue(c, v0, BRW_URB_WRITE_ALLOCATE_COMPLETE,
+ (_3DPRIM_TRIFAN << URB_WRITE_PRIM_TYPE_SHIFT));
brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2));
brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0));
brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1));
}
- brw_WHILE(p, loop);
+ brw_WHILE(p);
- brw_clip_emit_vue(c, v0, 0, 1, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_END));
+ brw_clip_emit_vue(c, v0, BRW_URB_WRITE_EOT_COMPLETE,
+ ((_3DPRIM_TRIFAN << URB_WRITE_PRIM_TYPE_SHIFT)
+ | URB_WRITE_PRIM_END));
}
- brw_ENDIF(p, if_insn);
+ brw_ENDIF(p);
}
static void do_clip_tri( struct brw_clip_compile *c )
static void maybe_do_clip_tri( struct brw_clip_compile *c )
{
struct brw_compile *p = &c->func;
- struct brw_instruction *do_clip;
brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, c->reg.planemask, brw_imm_ud(0));
- do_clip = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
do_clip_tri(c);
}
- brw_ENDIF(p, do_clip);
+ brw_ENDIF(p);
}
static void brw_clip_test( struct brw_clip_compile *c )
struct brw_indirect vt2 = brw_indirect(2, 0);
struct brw_compile *p = &c->func;
- struct brw_instruction *is_outside;
struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */
+ GLuint hpos_offset = brw_varying_to_offset(&c->vue_map,
+ VARYING_SLOT_POS);
+
brw_MOV(p, get_addr_reg(vt0), brw_address(c->reg.vertex[0]));
brw_MOV(p, get_addr_reg(vt1), brw_address(c->reg.vertex[1]));
brw_MOV(p, get_addr_reg(vt2), brw_address(c->reg.vertex[2]));
- brw_MOV(p, v0, deref_4f(vt0, c->offset[VERT_RESULT_HPOS]));
- brw_MOV(p, v1, deref_4f(vt1, c->offset[VERT_RESULT_HPOS]));
- brw_MOV(p, v2, deref_4f(vt2, c->offset[VERT_RESULT_HPOS]));
+ brw_MOV(p, v0, deref_4f(vt0, hpos_offset));
+ brw_MOV(p, v1, deref_4f(vt1, hpos_offset));
+ brw_MOV(p, v2, deref_4f(vt2, hpos_offset));
brw_AND(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(~0x3f));
/* test nearz, xmin, ymin plane */
brw_OR(p, tmp0, tmp0, get_element(t, 2));
brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
brw_AND(p, brw_null_reg(), tmp0, brw_imm_ud(0x1));
- is_outside = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
brw_clip_kill_thread(c);
}
- brw_ENDIF(p, is_outside);
+ brw_ENDIF(p);
brw_set_predicate_control(p, BRW_PREDICATE_NONE);
/* some vertices are inside a plane, some are outside,need to clip */
brw_OR(p, tmp0, tmp0, get_element(t, 2));
brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
brw_AND(p, brw_null_reg(), tmp0, brw_imm_ud(0x1));
- is_outside = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
brw_clip_kill_thread(c);
}
- brw_ENDIF(p, is_outside);
+ brw_ENDIF(p);
brw_set_predicate_control(p, BRW_PREDICATE_NONE);
/* some vertices are inside a plane, some are outside,need to clip */
void brw_emit_tri_clip( struct brw_clip_compile *c )
{
- struct brw_instruction *neg_rhw;
struct brw_compile *p = &c->func;
struct brw_context *brw = p->brw;
brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6);
brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2),
brw_imm_ud(1<<20));
- neg_rhw = brw_IF(p, BRW_EXECUTE_1);
+ brw_IF(p, BRW_EXECUTE_1);
{
brw_clip_test(c);
}
- brw_ENDIF(p, neg_rhw);
+ brw_ENDIF(p);
}
/* Can't push into do_clip_tri because with polygon (or quad)
* flatshading, need to apply the flatshade here because we don't
* respect the PV when converting to trifan for emit:
*/
- if (c->key.do_flat_shading)
+ if (c->has_flat_shading)
brw_clip_tri_flat_shade(c);
if ((c->key.clip_mode == BRW_CLIPMODE_NORMAL) ||