static void r300FireEB(r300ContextPtr rmesa, unsigned long addr,
int vertex_count, int type, int elt_size)
{
int cmd_reserved = 0;
int cmd_written = 0;
drm_radeon_cmd_header_t *cmd = NULL;
unsigned long t_addr;
unsigned long magic_1, magic_2;
assert(elt_size == 2 || elt_size == 4);
if (addr & (elt_size - 1)) {
WARN_ONCE("Badly aligned buffer\n");
return;
}
magic_1 = (addr % 32) / 4;
t_addr = addr & ~0x1d;
magic_2 = (vertex_count + 1 + (t_addr & 0x2)) / 2 + magic_1;
start_packet3(RADEON_CP_PACKET3_3D_DRAW_INDX_2, 0);
if (elt_size == 4) {
e32(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
(vertex_count << 16) | type |
R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
} else {
e32(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
(vertex_count << 16) | type);
}
start_packet3(RADEON_CP_PACKET3_INDX_BUFFER, 2);
#ifdef OPTIMIZE_ELTS
if (elt_size == 4) {
e32(R300_EB_UNK1 | (0 << 16) | R300_EB_UNK2);
e32(addr);
} else {
e32(R300_EB_UNK1 | (magic_1 << 16) | R300_EB_UNK2);
e32(t_addr);
}
#else
e32(R300_EB_UNK1 | (0 << 16) | R300_EB_UNK2);
e32(addr);
#endif
if (elt_size == 4) {
e32(vertex_count);
} else {
#ifdef OPTIMIZE_ELTS
e32(magic_2);
#else
e32((vertex_count + 1) / 2);
#endif
}
}
static void r300FireEB(r300ContextPtr rmesa, unsigned long addr,
int vertex_count, int type)
{
int cmd_reserved = 0;
int cmd_written = 0;
drm_radeon_cmd_header_t *cmd = NULL;
start_packet3(CP_PACKET3(R300_PACKET3_3D_DRAW_INDX_2, 0), 0);
e32(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (vertex_count << 16) | type | R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
start_packet3(CP_PACKET3(R300_PACKET3_INDX_BUFFER, 2), 2);
e32(R300_EB_UNK1 | (0 << 16) | R300_EB_UNK2);
e32(addr);
e32(vertex_count);
}
static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset)
{
int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2;
int i;
int cmd_reserved = 0;
int cmd_written = 0;
drm_radeon_cmd_header_t *cmd = NULL;
if (RADEON_DEBUG & DEBUG_VERTS)
fprintf(stderr, "%s: nr=%d, ofs=0x%08x\n", __FUNCTION__, nr,
offset);
start_packet3(CP_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, sz - 1), sz - 1);
e32(nr);
for (i = 0; i + 1 < nr; i += 2) {
e32((rmesa->state.aos[i].aos_size << 0) |
(rmesa->state.aos[i].aos_stride << 8) |
(rmesa->state.aos[i + 1].aos_size << 16) |
(rmesa->state.aos[i + 1].aos_stride << 24));
e32(rmesa->state.aos[i].aos_offset + offset * 4 * rmesa->state.aos[i].aos_stride);
e32(rmesa->state.aos[i + 1].aos_offset + offset * 4 * rmesa->state.aos[i + 1].aos_stride);
}
if (nr & 1) {
e32((rmesa->state.aos[nr - 1].aos_size << 0) |
(rmesa->state.aos[nr - 1].aos_stride << 8));
e32(rmesa->state.aos[nr - 1].aos_offset + offset * 4 * rmesa->state.aos[nr - 1].aos_stride);
}
}
static void r300FireAOS(r300ContextPtr rmesa, int vertex_count, int type)
{
int cmd_reserved = 0;
int cmd_written = 0;
drm_radeon_cmd_header_t *cmd = NULL;
start_packet3(CP_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0), 0);
e32(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (vertex_count << 16) | type);
}
static void inline fire_EB(PREFIX unsigned long addr, int vertex_count, int type, int elt_size)
{
LOCAL_VARS
unsigned long addr_a;
unsigned long t_addr;
unsigned long magic_1, magic_2;
GLcontext *ctx;
ctx = rmesa->radeon.glCtx;
assert(elt_size == 2 || elt_size == 4);
if(addr & (elt_size-1)){
WARN_ONCE("Badly aligned buffer\n");
return ;
}
#ifdef OPTIMIZE_ELTS
addr_a = 0;
magic_1 = (addr % 32) / 4;
t_addr = addr & (~0x1d);
magic_2 = (vertex_count + 1 + (t_addr & 0x2)) / 2 + magic_1;
check_space(6);
start_packet3(RADEON_CP_PACKET3_3D_DRAW_INDX_2, 0);
if(elt_size == 4){
e32(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (vertex_count<<16) | type | R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
} else {
e32(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (vertex_count<<16) | type);
}
start_packet3(RADEON_CP_PACKET3_INDX_BUFFER, 2);
if(elt_size == 4){
e32(R300_EB_UNK1 | (0 << 16) | R300_EB_UNK2);
e32(addr /*& 0xffffffe3*/);
} else {
e32(R300_EB_UNK1 | (magic_1 << 16) | R300_EB_UNK2);
e32(t_addr);
}
if(elt_size == 4){
e32(vertex_count /*+ addr_a/4*/); /* Total number of dwords needed? */
} else {
e32(magic_2); /* Total number of dwords needed? */
}
//cp_delay(PASS_PREFIX 1);
#if 0
fprintf(stderr, "magic_1 %d\n", magic_1);
fprintf(stderr, "t_addr %x\n", t_addr);
fprintf(stderr, "magic_2 %d\n", magic_2);
exit(1);
#endif
#else
(void)magic_2, (void)magic_1, (void)t_addr;
addr_a = 0;
check_space(6);
start_packet3(RADEON_CP_PACKET3_3D_DRAW_INDX_2, 0);
if(elt_size == 4){
e32(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (vertex_count<<16) | type | R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
} else {
e32(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (vertex_count<<16) | type);
}
start_packet3(RADEON_CP_PACKET3_INDX_BUFFER, 2);
e32(R300_EB_UNK1 | (0 << 16) | R300_EB_UNK2);
e32(addr /*& 0xffffffe3*/);
if(elt_size == 4){
e32(vertex_count /*+ addr_a/4*/); /* Total number of dwords needed? */
} else {
e32((vertex_count+1)/2 /*+ addr_a/4*/); /* Total number of dwords needed? */
}
//cp_delay(PASS_PREFIX 1);
#endif
}
