static void
nv50_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_pushbuf *push = nv50->base.pushbuf;
struct nv50_query *q = nv50_query(pq);
/* For occlusion queries we have to change the storage, because a previous
* query might set the initial render conition to FALSE even *after* we re-
* initialized it to TRUE.
*/
if (q->type == PIPE_QUERY_OCCLUSION_COUNTER) {
q->offset += 16;
q->data += 16 / sizeof(*q->data);
if (q->offset - q->base == NV50_QUERY_ALLOC_SPACE)
nv50_query_allocate(nv50, q, NV50_QUERY_ALLOC_SPACE);
/* XXX: can we do this with the GPU, and sync with respect to a previous
* query ?
*/
q->data[0] = q->sequence; /* initialize sequence */
q->data[1] = 1; /* initial render condition = TRUE */
q->data[4] = q->sequence + 1; /* for comparison COND_MODE */
q->data[5] = 0;
}
if (!q->is64bit)
q->data[0] = q->sequence++; /* the previously used one */
switch (q->type) {
case PIPE_QUERY_OCCLUSION_COUNTER:
PUSH_SPACE(push, 4);
BEGIN_NV04(push, NV50_3D(COUNTER_RESET), 1);
PUSH_DATA (push, NV50_3D_COUNTER_RESET_SAMPLECNT);
BEGIN_NV04(push, NV50_3D(SAMPLECNT_ENABLE), 1);
PUSH_DATA (push, 1);
break;
case PIPE_QUERY_PRIMITIVES_GENERATED:
nv50_query_get(push, q, 0x10, 0x06805002);
break;
case PIPE_QUERY_PRIMITIVES_EMITTED:
nv50_query_get(push, q, 0x10, 0x05805002);
break;
case PIPE_QUERY_SO_STATISTICS:
nv50_query_get(push, q, 0x20, 0x05805002);
nv50_query_get(push, q, 0x30, 0x06805002);
break;
case PIPE_QUERY_TIME_ELAPSED:
nv50_query_get(push, q, 0x10, 0x00005002);
break;
default:
break;
}
q->ready = FALSE;
}
static void
nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_pushbuf *push = nv50->base.pushbuf;
struct nv50_query *q = nv50_query(pq);
switch (q->type) {
case PIPE_QUERY_OCCLUSION_COUNTER:
nv50_query_get(push, q, 0, 0x0100f002);
PUSH_SPACE(push, 2);
BEGIN_NV04(push, NV50_3D(SAMPLECNT_ENABLE), 1);
PUSH_DATA (push, 0);
break;
case PIPE_QUERY_PRIMITIVES_GENERATED:
nv50_query_get(push, q, 0, 0x06805002);
break;
case PIPE_QUERY_PRIMITIVES_EMITTED:
nv50_query_get(push, q, 0, 0x05805002);
break;
case PIPE_QUERY_SO_STATISTICS:
nv50_query_get(push, q, 0x00, 0x05805002);
nv50_query_get(push, q, 0x10, 0x06805002);
break;
case PIPE_QUERY_TIMESTAMP:
q->sequence++;
/* fall through */
case PIPE_QUERY_TIME_ELAPSED:
nv50_query_get(push, q, 0, 0x00005002);
break;
case PIPE_QUERY_GPU_FINISHED:
q->sequence++;
nv50_query_get(push, q, 0, 0x1000f010);
break;
case NVA0_QUERY_STREAM_OUTPUT_BUFFER_OFFSET:
nv50_query_get(push, q, 0, 0x0d005002 | (q->index << 5));
break;
case PIPE_QUERY_TIMESTAMP_DISJOINT:
break;
default:
assert(0);
break;
}
q->ready = q->flushed = FALSE;
}
static void
nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_pushbuf *push = nv50->base.pushbuf;
struct nv50_query *q = nv50_query(pq);
switch (q->type) {
case PIPE_QUERY_OCCLUSION_COUNTER:
nv50_query_get(push, q, 0, 0x0100f002);
PUSH_SPACE(push, 2);
BEGIN_NV04(push, NV50_3D(SAMPLECNT_ENABLE), 1);
PUSH_DATA (push, 0);
break;
case PIPE_QUERY_PRIMITIVES_GENERATED:
nv50_query_get(push, q, 0, 0x06805002);
break;
case PIPE_QUERY_PRIMITIVES_EMITTED:
nv50_query_get(push, q, 0, 0x05805002);
break;
case PIPE_QUERY_SO_STATISTICS:
nv50_query_get(push, q, 0x00, 0x05805002);
nv50_query_get(push, q, 0x10, 0x06805002);
break;
case PIPE_QUERY_PIPELINE_STATISTICS:
nv50_query_get(push, q, 0x00, 0x00801002); /* VFETCH, VERTICES */
nv50_query_get(push, q, 0x10, 0x01801002); /* VFETCH, PRIMS */
nv50_query_get(push, q, 0x20, 0x02802002); /* VP, LAUNCHES */
nv50_query_get(push, q, 0x30, 0x03806002); /* GP, LAUNCHES */
nv50_query_get(push, q, 0x40, 0x04806002); /* GP, PRIMS_OUT */
nv50_query_get(push, q, 0x50, 0x07804002); /* RAST, PRIMS_IN */
nv50_query_get(push, q, 0x60, 0x08804002); /* RAST, PRIMS_OUT */
nv50_query_get(push, q, 0x70, 0x0980a002); /* ROP, PIXELS */
break;
case PIPE_QUERY_TIMESTAMP:
q->sequence++;
/* fall through */
case PIPE_QUERY_TIME_ELAPSED:
nv50_query_get(push, q, 0, 0x00005002);
break;
case PIPE_QUERY_GPU_FINISHED:
q->sequence++;
nv50_query_get(push, q, 0, 0x1000f010);
break;
case NVA0_QUERY_STREAM_OUTPUT_BUFFER_OFFSET:
nv50_query_get(push, q, 0, 0x0d005002 | (q->index << 5));
break;
case PIPE_QUERY_TIMESTAMP_DISJOINT:
/* This query is not issued on GPU because disjoint is forced to FALSE */
q->ready = TRUE;
break;
default:
assert(0);
break;
}
q->ready = q->flushed = FALSE;
}
static boolean
nv50_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_pushbuf *push = nv50->base.pushbuf;
struct nv50_query *q = nv50_query(pq);
/* For occlusion queries we have to change the storage, because a previous
* query might set the initial render conition to false even *after* we re-
* initialized it to true.
*/
if (q->type == PIPE_QUERY_OCCLUSION_COUNTER) {
q->offset += 32;
q->data += 32 / sizeof(*q->data);
if (q->offset - q->base == NV50_QUERY_ALLOC_SPACE)
nv50_query_allocate(nv50, q, NV50_QUERY_ALLOC_SPACE);
/* XXX: can we do this with the GPU, and sync with respect to a previous
* query ?
*/
q->data[0] = q->sequence; /* initialize sequence */
q->data[1] = 1; /* initial render condition = true */
q->data[4] = q->sequence + 1; /* for comparison COND_MODE */
q->data[5] = 0;
}
if (!q->is64bit)
q->data[0] = q->sequence++; /* the previously used one */
switch (q->type) {
case PIPE_QUERY_OCCLUSION_COUNTER:
q->nesting = nv50->screen->num_occlusion_queries_active++;
if (q->nesting) {
nv50_query_get(push, q, 0x10, 0x0100f002);
} else {
PUSH_SPACE(push, 4);
BEGIN_NV04(push, NV50_3D(COUNTER_RESET), 1);
PUSH_DATA (push, NV50_3D_COUNTER_RESET_SAMPLECNT);
BEGIN_NV04(push, NV50_3D(SAMPLECNT_ENABLE), 1);
PUSH_DATA (push, 1);
}
break;
case PIPE_QUERY_PRIMITIVES_GENERATED:
nv50_query_get(push, q, 0x10, 0x06805002);
break;
case PIPE_QUERY_PRIMITIVES_EMITTED:
nv50_query_get(push, q, 0x10, 0x05805002);
break;
case PIPE_QUERY_SO_STATISTICS:
nv50_query_get(push, q, 0x20, 0x05805002);
nv50_query_get(push, q, 0x30, 0x06805002);
break;
case PIPE_QUERY_PIPELINE_STATISTICS:
nv50_query_get(push, q, 0x80, 0x00801002); /* VFETCH, VERTICES */
nv50_query_get(push, q, 0x90, 0x01801002); /* VFETCH, PRIMS */
nv50_query_get(push, q, 0xa0, 0x02802002); /* VP, LAUNCHES */
nv50_query_get(push, q, 0xb0, 0x03806002); /* GP, LAUNCHES */
nv50_query_get(push, q, 0xc0, 0x04806002); /* GP, PRIMS_OUT */
nv50_query_get(push, q, 0xd0, 0x07804002); /* RAST, PRIMS_IN */
nv50_query_get(push, q, 0xe0, 0x08804002); /* RAST, PRIMS_OUT */
nv50_query_get(push, q, 0xf0, 0x0980a002); /* ROP, PIXELS */
break;
case PIPE_QUERY_TIME_ELAPSED:
nv50_query_get(push, q, 0x10, 0x00005002);
break;
default:
break;
}
q->state = NV50_QUERY_STATE_ACTIVE;
return true;
}