[Mesa-dev] [PATCH] gallium: add PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE
Marek Olšák
maraeo at gmail.com
Thu Sep 14 00:56:57 UTC 2017
Reviewed-by: Marek Olšák <marek.olsak at amd.com>
Marek
On Wed, Sep 13, 2017 at 6:53 PM, Nicolai Hähnle <nhaehnle at gmail.com> wrote:
> From: Nicolai Hähnle <nicolai.haehnle at amd.com>
>
> To be able to properly distinguish between GL_ANY_SAMPLES_PASSED
> and GL_ANY_SAMPLES_PASSED_CONSERVATIVE.
>
> This patch goes through all drivers, having them treat the two
> query types identically, except:
>
> 1. radeon incorrectly enabled conservative mode on
> PIPE_QUERY_OCCLUSION_PREDICATE. We now do it correctly, only
> on PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE.
> 2. st/mesa uses the new query type.
>
> Fixes dEQP-GLES31.functional.fbo.no_attachments.*
> ---
> src/gallium/auxiliary/util/u_dump_defines.c | 1 +
> src/gallium/auxiliary/util/u_inlines.h | 1 +
> src/gallium/docs/source/context.rst | 6 ++++++
> src/gallium/drivers/freedreno/a3xx/fd3_query.c | 8 ++++++++
> src/gallium/drivers/freedreno/a4xx/fd4_query.c | 8 ++++++++
> src/gallium/drivers/freedreno/a5xx/fd5_query.c | 10 ++++++++++
> src/gallium/drivers/freedreno/freedreno_query.h | 1 +
> src/gallium/drivers/llvmpipe/lp_query.c | 3 +++
> src/gallium/drivers/llvmpipe/lp_rast.c | 2 ++
> src/gallium/drivers/llvmpipe/lp_setup.c | 3 +++
> src/gallium/drivers/nouveau/nv30/nv30_query.c | 4 +++-
> src/gallium/drivers/nouveau/nv50/nv50_query.c | 1 +
> src/gallium/drivers/nouveau/nv50/nv50_query_hw.c | 4 ++++
> src/gallium/drivers/nouveau/nvc0/nvc0_query.c | 1 +
> src/gallium/drivers/nouveau/nvc0/nvc0_query_hw.c | 7 ++++++-
> src/gallium/drivers/r300/r300_query.c | 7 +++++--
> src/gallium/drivers/radeon/r600_pipe_common.c | 3 ++-
> src/gallium/drivers/radeon/r600_query.c | 19 ++++++++++++++-----
> src/gallium/drivers/softpipe/sp_query.c | 4 ++++
> src/gallium/drivers/svga/svga_pipe_query.c | 7 ++++++-
> src/gallium/drivers/swr/swr_query.cpp | 1 +
> src/gallium/drivers/trace/tr_dump_state.c | 1 +
> src/gallium/include/pipe/p_defines.h | 2 ++
> src/mesa/state_tracker/st_cb_queryobj.c | 5 ++++-
> 24 files changed, 97 insertions(+), 12 deletions(-)
>
> diff --git a/src/gallium/auxiliary/util/u_dump_defines.c b/src/gallium/auxiliary/util/u_dump_defines.c
> index 5032974a880..e87e5301600 100644
> --- a/src/gallium/auxiliary/util/u_dump_defines.c
> +++ b/src/gallium/auxiliary/util/u_dump_defines.c
> @@ -358,20 +358,21 @@ util_tex_filter_short_names[] = {
> "linear"
> };
>
> DEFINE_UTIL_STR_CONTINUOUS(tex_filter)
>
>
> static const char *
> util_query_type_names[] = {
> "PIPE_QUERY_OCCLUSION_COUNTER",
> "PIPE_QUERY_OCCLUSION_PREDICATE",
> + "PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE",
> "PIPE_QUERY_TIMESTAMP",
> "PIPE_QUERY_TIMESTAMP_DISJOINT",
> "PIPE_QUERY_TIME_ELAPSED",
> "PIPE_QUERY_PRIMITIVES_GENERATED",
> "PIPE_QUERY_PRIMITIVES_EMITTED",
> "PIPE_QUERY_SO_STATISTICS",
> "PIPE_QUERY_SO_OVERFLOW_PREDICATE",
> "PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE",
> "PIPE_QUERY_GPU_FINISHED",
> "PIPE_QUERY_PIPELINE_STATISTICS",
> diff --git a/src/gallium/auxiliary/util/u_inlines.h b/src/gallium/auxiliary/util/u_inlines.h
> index e0ed594c9fe..79f62c32266 100644
> --- a/src/gallium/auxiliary/util/u_inlines.h
> +++ b/src/gallium/auxiliary/util/u_inlines.h
> @@ -529,20 +529,21 @@ util_get_min_point_size(const struct pipe_rasterizer_state *state)
> return !state->point_quad_rasterization &&
> !state->point_smooth &&
> !state->multisample ? 1.0f : 0.0f;
> }
>
> static inline void
> util_query_clear_result(union pipe_query_result *result, unsigned type)
> {
> switch (type) {
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
> case PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE:
> case PIPE_QUERY_GPU_FINISHED:
> result->b = FALSE;
> break;
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_TIMESTAMP:
> case PIPE_QUERY_TIME_ELAPSED:
> case PIPE_QUERY_PRIMITIVES_GENERATED:
> case PIPE_QUERY_PRIMITIVES_EMITTED:
> diff --git a/src/gallium/docs/source/context.rst b/src/gallium/docs/source/context.rst
> index 6ac45819a66..ba7fef8301d 100644
> --- a/src/gallium/docs/source/context.rst
> +++ b/src/gallium/docs/source/context.rst
> @@ -387,20 +387,26 @@ are written to the framebuffer without being culled by
> The result is an unsigned 64-bit integer.
> This query can be used with ``render_condition``.
>
> In cases where a boolean result of an occlusion query is enough,
> ``PIPE_QUERY_OCCLUSION_PREDICATE`` should be used. It is just like
> ``PIPE_QUERY_OCCLUSION_COUNTER`` except that the result is a boolean
> value of FALSE for cases where COUNTER would result in 0 and TRUE
> for all other cases.
> This query can be used with ``render_condition``.
>
> +In cases where a conservative approximation of an occlusion query is enough,
> +``PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE`` should be used. It behaves
> +like ``PIPE_QUERY_OCCLUSION_PREDICATE``, except that it may return TRUE in
> +additional, implementation-dependent cases.
> +This query can be used with ``render_condition``.
> +
> ``PIPE_QUERY_TIME_ELAPSED`` returns the amount of time, in nanoseconds,
> the context takes to perform operations.
> The result is an unsigned 64-bit integer.
>
> ``PIPE_QUERY_TIMESTAMP`` returns a device/driver internal timestamp,
> scaled to nanoseconds, recorded after all commands issued prior to
> ``end_query`` have been processed.
> This query does not require a call to ``begin_query``.
> The result is an unsigned 64-bit integer.
>
> diff --git a/src/gallium/drivers/freedreno/a3xx/fd3_query.c b/src/gallium/drivers/freedreno/a3xx/fd3_query.c
> index cde42c37313..97a95b21546 100644
> --- a/src/gallium/drivers/freedreno/a3xx/fd3_query.c
> +++ b/src/gallium/drivers/freedreno/a3xx/fd3_query.c
> @@ -124,22 +124,30 @@ static const struct fd_hw_sample_provider occlusion_counter = {
> .accumulate_result = occlusion_counter_accumulate_result,
> };
>
> static const struct fd_hw_sample_provider occlusion_predicate = {
> .query_type = PIPE_QUERY_OCCLUSION_PREDICATE,
> .active = FD_STAGE_DRAW,
> .get_sample = occlusion_get_sample,
> .accumulate_result = occlusion_predicate_accumulate_result,
> };
>
> +static const struct fd_hw_sample_provider occlusion_predicate_conservative = {
> + .query_type = PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE,
> + .active = FD_STAGE_DRAW,
> + .get_sample = occlusion_get_sample,
> + .accumulate_result = occlusion_predicate_accumulate_result,
> +};
> +
> void fd3_query_context_init(struct pipe_context *pctx)
> {
> struct fd_context *ctx = fd_context(pctx);
>
> ctx->create_query = fd_hw_create_query;
> ctx->query_prepare = fd_hw_query_prepare;
> ctx->query_prepare_tile = fd_hw_query_prepare_tile;
> ctx->query_set_stage = fd_hw_query_set_stage;
>
> fd_hw_query_register_provider(pctx, &occlusion_counter);
> fd_hw_query_register_provider(pctx, &occlusion_predicate);
> + fd_hw_query_register_provider(pctx, &occlusion_predicate_conservative);
> }
> diff --git a/src/gallium/drivers/freedreno/a4xx/fd4_query.c b/src/gallium/drivers/freedreno/a4xx/fd4_query.c
> index f7b385d552d..809e7570b48 100644
> --- a/src/gallium/drivers/freedreno/a4xx/fd4_query.c
> +++ b/src/gallium/drivers/freedreno/a4xx/fd4_query.c
> @@ -244,20 +244,27 @@ static const struct fd_hw_sample_provider occlusion_counter = {
> .accumulate_result = occlusion_counter_accumulate_result,
> };
>
> static const struct fd_hw_sample_provider occlusion_predicate = {
> .query_type = PIPE_QUERY_OCCLUSION_PREDICATE,
> .active = FD_STAGE_DRAW,
> .get_sample = occlusion_get_sample,
> .accumulate_result = occlusion_predicate_accumulate_result,
> };
>
> +static const struct fd_hw_sample_provider occlusion_predicate = {
> + .query_type = PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE,
> + .active = FD_STAGE_DRAW,
> + .get_sample = occlusion_get_sample,
> + .accumulate_result = occlusion_predicate_accumulate_result,
> +};
> +
> static const struct fd_hw_sample_provider time_elapsed = {
> .query_type = PIPE_QUERY_TIME_ELAPSED,
> .active = FD_STAGE_DRAW | FD_STAGE_CLEAR,
> .enable = time_elapsed_enable,
> .get_sample = time_elapsed_get_sample,
> .accumulate_result = time_elapsed_accumulate_result,
> };
>
> /* NOTE: timestamp query isn't going to give terribly sensible results
> * on a tiler. But it is needed by qapitrace profile heatmap. If you
> @@ -277,13 +284,14 @@ void fd4_query_context_init(struct pipe_context *pctx)
> {
> struct fd_context *ctx = fd_context(pctx);
>
> ctx->create_query = fd_hw_create_query;
> ctx->query_prepare = fd_hw_query_prepare;
> ctx->query_prepare_tile = fd_hw_query_prepare_tile;
> ctx->query_set_stage = fd_hw_query_set_stage;
>
> fd_hw_query_register_provider(pctx, &occlusion_counter);
> fd_hw_query_register_provider(pctx, &occlusion_predicate);
> + fd_hw_query_register_provider(pctx, &occlusion_predicate_conservative);
> fd_hw_query_register_provider(pctx, &time_elapsed);
> fd_hw_query_register_provider(pctx, ×tamp);
> }
> diff --git a/src/gallium/drivers/freedreno/a5xx/fd5_query.c b/src/gallium/drivers/freedreno/a5xx/fd5_query.c
> index 80b84ce54ba..87417f1f96e 100644
> --- a/src/gallium/drivers/freedreno/a5xx/fd5_query.c
> +++ b/src/gallium/drivers/freedreno/a5xx/fd5_query.c
> @@ -137,20 +137,29 @@ static const struct fd_acc_sample_provider occlusion_counter = {
>
> static const struct fd_acc_sample_provider occlusion_predicate = {
> .query_type = PIPE_QUERY_OCCLUSION_PREDICATE,
> .active = FD_STAGE_DRAW,
> .size = sizeof(struct fd5_query_sample),
> .resume = occlusion_resume,
> .pause = occlusion_pause,
> .result = occlusion_predicate_result,
> };
>
> +static const struct fd_acc_sample_provider occlusion_predicate_conservative = {
> + .query_type = PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE,
> + .active = FD_STAGE_DRAW,
> + .size = sizeof(struct fd5_query_sample),
> + .resume = occlusion_resume,
> + .pause = occlusion_pause,
> + .result = occlusion_predicate_result,
> +};
> +
> /*
> * Timestamp Queries:
> */
>
> static void
> timestamp_resume(struct fd_acc_query *aq, struct fd_batch *batch)
> {
> struct fd_ringbuffer *ring = batch->draw;
>
> OUT_PKT7(ring, CP_EVENT_WRITE, 4);
> @@ -240,14 +249,15 @@ static const struct fd_acc_sample_provider timestamp = {
> void
> fd5_query_context_init(struct pipe_context *pctx)
> {
> struct fd_context *ctx = fd_context(pctx);
>
> ctx->create_query = fd_acc_create_query;
> ctx->query_set_stage = fd_acc_query_set_stage;
>
> fd_acc_query_register_provider(pctx, &occlusion_counter);
> fd_acc_query_register_provider(pctx, &occlusion_predicate);
> + fd_acc_query_register_provider(pctx, &occlusion_predicate_conservative);
>
> fd_acc_query_register_provider(pctx, &time_elapsed);
> fd_acc_query_register_provider(pctx, ×tamp);
> }
> diff --git a/src/gallium/drivers/freedreno/freedreno_query.h b/src/gallium/drivers/freedreno/freedreno_query.h
> index 49a86803c4d..b8fa0951ded 100644
> --- a/src/gallium/drivers/freedreno/freedreno_query.h
> +++ b/src/gallium/drivers/freedreno/freedreno_query.h
> @@ -78,20 +78,21 @@ skip_begin_query(int type)
> }
>
> /* maps query_type to sample provider idx: */
> static inline
> int pidx(unsigned query_type)
> {
> switch (query_type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> return 0;
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> return 1;
> /* TODO currently queries only emitted in main pass (not in binning pass)..
> * which is fine for occlusion query, but pretty much not anything else.
> */
> case PIPE_QUERY_TIME_ELAPSED:
> return 2;
> case PIPE_QUERY_TIMESTAMP:
> return 3;
> default:
> return -1;
> diff --git a/src/gallium/drivers/llvmpipe/lp_query.c b/src/gallium/drivers/llvmpipe/lp_query.c
> index 6f8ce94e5d8..7b81903b413 100644
> --- a/src/gallium/drivers/llvmpipe/lp_query.c
> +++ b/src/gallium/drivers/llvmpipe/lp_query.c
> @@ -118,20 +118,21 @@ llvmpipe_get_query_result(struct pipe_context *pipe,
> */
> *result = 0;
>
> switch (pq->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> for (i = 0; i < num_threads; i++) {
> *result += pq->end[i];
> }
> break;
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> for (i = 0; i < num_threads; i++) {
> /* safer (still not guaranteed) when there's an overflow */
> vresult->b = vresult->b || pq->end[i];
> }
> break;
> case PIPE_QUERY_TIMESTAMP:
> for (i = 0; i < num_threads; i++) {
> if (pq->end[i] > *result) {
> *result = pq->end[i];
> }
> @@ -224,20 +225,21 @@ llvmpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q)
> /* reset our cache */
> if (llvmpipe->active_statistics_queries == 0) {
> memset(&llvmpipe->pipeline_statistics, 0,
> sizeof(llvmpipe->pipeline_statistics));
> }
> memcpy(&pq->stats, &llvmpipe->pipeline_statistics, sizeof(pq->stats));
> llvmpipe->active_statistics_queries++;
> break;
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> llvmpipe->active_occlusion_queries++;
> llvmpipe->dirty |= LP_NEW_OCCLUSION_QUERY;
> break;
> default:
> break;
> }
> return true;
> }
>
>
> @@ -287,20 +289,21 @@ llvmpipe_end_query(struct pipe_context *pipe, struct pipe_query *q)
> llvmpipe->pipeline_statistics.c_invocations - pq->stats.c_invocations;
> pq->stats.c_primitives =
> llvmpipe->pipeline_statistics.c_primitives - pq->stats.c_primitives;
> pq->stats.ps_invocations =
> llvmpipe->pipeline_statistics.ps_invocations - pq->stats.ps_invocations;
>
> llvmpipe->active_statistics_queries--;
> break;
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> assert(llvmpipe->active_occlusion_queries);
> llvmpipe->active_occlusion_queries--;
> llvmpipe->dirty |= LP_NEW_OCCLUSION_QUERY;
> break;
> default:
> break;
> }
>
> return true;
> }
> diff --git a/src/gallium/drivers/llvmpipe/lp_rast.c b/src/gallium/drivers/llvmpipe/lp_rast.c
> index b25ade3d2ec..0050655d9e3 100644
> --- a/src/gallium/drivers/llvmpipe/lp_rast.c
> +++ b/src/gallium/drivers/llvmpipe/lp_rast.c
> @@ -479,20 +479,21 @@ lp_rast_shade_quads_mask(struct lp_rasterizer_task *task,
> */
> static void
> lp_rast_begin_query(struct lp_rasterizer_task *task,
> const union lp_rast_cmd_arg arg)
> {
> struct llvmpipe_query *pq = arg.query_obj;
>
> switch (pq->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> pq->start[task->thread_index] = task->thread_data.vis_counter;
> break;
> case PIPE_QUERY_PIPELINE_STATISTICS:
> pq->start[task->thread_index] = task->ps_invocations;
> break;
> default:
> assert(0);
> break;
> }
> }
> @@ -505,20 +506,21 @@ lp_rast_begin_query(struct lp_rasterizer_task *task,
> */
> static void
> lp_rast_end_query(struct lp_rasterizer_task *task,
> const union lp_rast_cmd_arg arg)
> {
> struct llvmpipe_query *pq = arg.query_obj;
>
> switch (pq->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> pq->end[task->thread_index] +=
> task->thread_data.vis_counter - pq->start[task->thread_index];
> pq->start[task->thread_index] = 0;
> break;
> case PIPE_QUERY_TIMESTAMP:
> pq->end[task->thread_index] = os_time_get_nano();
> break;
> case PIPE_QUERY_PIPELINE_STATISTICS:
> pq->end[task->thread_index] +=
> task->ps_invocations - pq->start[task->thread_index];
> diff --git a/src/gallium/drivers/llvmpipe/lp_setup.c b/src/gallium/drivers/llvmpipe/lp_setup.c
> index 32387ab5532..2be6fc033d1 100644
> --- a/src/gallium/drivers/llvmpipe/lp_setup.c
> +++ b/src/gallium/drivers/llvmpipe/lp_setup.c
> @@ -1373,20 +1373,21 @@ no_setup:
> */
> void
> lp_setup_begin_query(struct lp_setup_context *setup,
> struct llvmpipe_query *pq)
> {
>
> set_scene_state(setup, SETUP_ACTIVE, "begin_query");
>
> if (!(pq->type == PIPE_QUERY_OCCLUSION_COUNTER ||
> pq->type == PIPE_QUERY_OCCLUSION_PREDICATE ||
> + pq->type == PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE ||
> pq->type == PIPE_QUERY_PIPELINE_STATISTICS))
> return;
>
> /* init the query to its beginning state */
> assert(setup->active_binned_queries < LP_MAX_ACTIVE_BINNED_QUERIES);
> /* exceeding list size so just ignore the query */
> if (setup->active_binned_queries >= LP_MAX_ACTIVE_BINNED_QUERIES) {
> return;
> }
> assert(setup->active_queries[setup->active_binned_queries] == NULL);
> @@ -1423,20 +1424,21 @@ lp_setup_end_query(struct lp_setup_context *setup, struct llvmpipe_query *pq)
>
> assert(setup->scene);
> if (setup->scene) {
> /* pq->fence should be the fence of the *last* scene which
> * contributed to the query result.
> */
> lp_fence_reference(&pq->fence, setup->scene->fence);
>
> if (pq->type == PIPE_QUERY_OCCLUSION_COUNTER ||
> pq->type == PIPE_QUERY_OCCLUSION_PREDICATE ||
> + pq->type == PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE ||
> pq->type == PIPE_QUERY_PIPELINE_STATISTICS ||
> pq->type == PIPE_QUERY_TIMESTAMP) {
> if (pq->type == PIPE_QUERY_TIMESTAMP &&
> !(setup->scene->tiles_x | setup->scene->tiles_y)) {
> /*
> * If there's a zero width/height framebuffer, there's no bins and
> * hence no rast task is ever run. So fill in something here instead.
> */
> pq->end[0] = os_time_get_nano();
> }
> @@ -1459,20 +1461,21 @@ lp_setup_end_query(struct lp_setup_context *setup, struct llvmpipe_query *pq)
> else {
> lp_fence_reference(&pq->fence, setup->last_fence);
> }
>
> fail:
> /* Need to do this now not earlier since it still needs to be marked as
> * active when binning it would cause a flush.
> */
> if (pq->type == PIPE_QUERY_OCCLUSION_COUNTER ||
> pq->type == PIPE_QUERY_OCCLUSION_PREDICATE ||
> + pq->type == PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE ||
> pq->type == PIPE_QUERY_PIPELINE_STATISTICS) {
> unsigned i;
>
> /* remove from active binned query list */
> for (i = 0; i < setup->active_binned_queries; i++) {
> if (setup->active_queries[i] == pq)
> break;
> }
> assert(i < setup->active_binned_queries);
> if (i == setup->active_binned_queries)
> diff --git a/src/gallium/drivers/nouveau/nv30/nv30_query.c b/src/gallium/drivers/nouveau/nv30/nv30_query.c
> index 83ea3afec02..f26728868af 100644
> --- a/src/gallium/drivers/nouveau/nv30/nv30_query.c
> +++ b/src/gallium/drivers/nouveau/nv30/nv30_query.c
> @@ -114,20 +114,21 @@ nv30_query_create(struct pipe_context *pipe, unsigned type, unsigned index)
> q->type = type;
>
> switch (q->type) {
> case PIPE_QUERY_TIMESTAMP:
> case PIPE_QUERY_TIME_ELAPSED:
> q->enable = 0x0000;
> q->report = 1;
> break;
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> q->enable = NV30_3D_QUERY_ENABLE;
> q->report = 1;
> break;
> case NV30_QUERY_ZCULL_0:
> case NV30_QUERY_ZCULL_1:
> case NV30_QUERY_ZCULL_2:
> case NV30_QUERY_ZCULL_3:
> q->enable = 0x1804;
> q->report = 2 + (q->type - NV30_QUERY_ZCULL_0);
> break;
> @@ -221,21 +222,22 @@ nv30_query_result(struct pipe_context *pipe, struct pipe_query *pq,
> break;
> default:
> q->result = ntfy1[2];
> break;
> }
>
> nv30_query_object_del(screen, &q->qo[0]);
> nv30_query_object_del(screen, &q->qo[1]);
> }
>
> - if (q->type == PIPE_QUERY_OCCLUSION_PREDICATE)
> + if (q->type == PIPE_QUERY_OCCLUSION_PREDICATE ||
> + q->type == PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE)
> result->b = !!q->result;
> else
> result->u64 = q->result;
> return true;
> }
>
> static void
> nv40_query_render_condition(struct pipe_context *pipe,
> struct pipe_query *pq,
> boolean condition, enum pipe_render_cond_flag mode)
> diff --git a/src/gallium/drivers/nouveau/nv50/nv50_query.c b/src/gallium/drivers/nouveau/nv50/nv50_query.c
> index ec6ee0f5eb3..e30380cd84d 100644
> --- a/src/gallium/drivers/nouveau/nv50/nv50_query.c
> +++ b/src/gallium/drivers/nouveau/nv50/nv50_query.c
> @@ -90,20 +90,21 @@ nv50_render_condition(struct pipe_context *pipe,
> else {
> /* NOTE: comparison of 2 queries only works if both have completed */
> switch (q->type) {
> case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
> cond = condition ? NV50_3D_COND_MODE_EQUAL :
> NV50_3D_COND_MODE_NOT_EQUAL;
> wait = true;
> break;
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> if (likely(!condition)) {
> if (unlikely(hq->nesting))
> cond = wait ? NV50_3D_COND_MODE_NOT_EQUAL :
> NV50_3D_COND_MODE_ALWAYS;
> else
> cond = NV50_3D_COND_MODE_RES_NON_ZERO;
> } else {
> cond = wait ? NV50_3D_COND_MODE_EQUAL : NV50_3D_COND_MODE_ALWAYS;
> }
> break;
> diff --git a/src/gallium/drivers/nouveau/nv50/nv50_query_hw.c b/src/gallium/drivers/nouveau/nv50/nv50_query_hw.c
> index 727b509372d..ac3e409b2d5 100644
> --- a/src/gallium/drivers/nouveau/nv50/nv50_query_hw.c
> +++ b/src/gallium/drivers/nouveau/nv50/nv50_query_hw.c
> @@ -150,20 +150,21 @@ nv50_hw_begin_query(struct nv50_context *nv50, struct nv50_query *q)
> hq->data[1] = 1; /* initial render condition = true */
> hq->data[4] = hq->sequence + 1; /* for comparison COND_MODE */
> hq->data[5] = 0;
> }
> if (!hq->is64bit)
> hq->data[0] = hq->sequence++; /* the previously used one */
>
> switch (q->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> hq->nesting = nv50->screen->num_occlusion_queries_active++;
> if (hq->nesting) {
> nv50_hw_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);
> }
> @@ -208,20 +209,21 @@ nv50_hw_end_query(struct nv50_context *nv50, struct nv50_query *q)
> if (hq->funcs && hq->funcs->end_query) {
> hq->funcs->end_query(nv50, hq);
> return;
> }
>
> hq->state = NV50_HW_QUERY_STATE_ENDED;
>
> switch (q->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> nv50_hw_query_get(push, q, 0, 0x0100f002);
> if (--nv50->screen->num_occlusion_queries_active == 0) {
> PUSH_SPACE(push, 2);
> BEGIN_NV04(push, NV50_3D(SAMPLECNT_ENABLE), 1);
> PUSH_DATA (push, 0);
> }
> break;
> case PIPE_QUERY_PRIMITIVES_GENERATED:
> nv50_hw_query_get(push, q, 0, 0x06805002);
> break;
> @@ -300,20 +302,21 @@ nv50_hw_get_query_result(struct nv50_context *nv50, struct nv50_query *q,
> hq->state = NV50_HW_QUERY_STATE_READY;
>
> switch (q->type) {
> case PIPE_QUERY_GPU_FINISHED:
> res8[0] = true;
> break;
> case PIPE_QUERY_OCCLUSION_COUNTER: /* u32 sequence, u32 count, u64 time */
> res64[0] = hq->data[1] - hq->data[5];
> break;
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> res8[0] = hq->data[1] != hq->data[5];
> break;
> case PIPE_QUERY_PRIMITIVES_GENERATED: /* u64 count, u64 time */
> case PIPE_QUERY_PRIMITIVES_EMITTED: /* u64 count, u64 time */
> res64[0] = data64[0] - data64[2];
> break;
> case PIPE_QUERY_SO_STATISTICS:
> res64[0] = data64[0] - data64[4];
> res64[1] = data64[2] - data64[6];
> break;
> @@ -371,20 +374,21 @@ nv50_hw_create_query(struct nv50_context *nv50, unsigned type, unsigned index)
> if (!hq)
> return NULL;
>
> q = &hq->base;
> q->funcs = &hw_query_funcs;
> q->type = type;
>
> switch (q->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> hq->rotate = 32;
> break;
> case PIPE_QUERY_PRIMITIVES_GENERATED:
> case PIPE_QUERY_PRIMITIVES_EMITTED:
> case PIPE_QUERY_SO_STATISTICS:
> case PIPE_QUERY_PIPELINE_STATISTICS:
> hq->is64bit = true;
> break;
> case PIPE_QUERY_TIME_ELAPSED:
> case PIPE_QUERY_TIMESTAMP:
> diff --git a/src/gallium/drivers/nouveau/nvc0/nvc0_query.c b/src/gallium/drivers/nouveau/nvc0/nvc0_query.c
> index e3090745e3b..d0a9e0c2c04 100644
> --- a/src/gallium/drivers/nouveau/nvc0/nvc0_query.c
> +++ b/src/gallium/drivers/nouveau/nvc0/nvc0_query.c
> @@ -112,20 +112,21 @@ nvc0_render_condition(struct pipe_context *pipe,
> else {
> /* NOTE: comparison of 2 queries only works if both have completed */
> switch (q->type) {
> case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
> cond = condition ? NVC0_3D_COND_MODE_EQUAL :
> NVC0_3D_COND_MODE_NOT_EQUAL;
> wait = true;
> break;
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> if (likely(!condition)) {
> if (unlikely(hq->nesting))
> cond = wait ? NVC0_3D_COND_MODE_NOT_EQUAL :
> NVC0_3D_COND_MODE_ALWAYS;
> else
> cond = NVC0_3D_COND_MODE_RES_NON_ZERO;
> } else {
> cond = wait ? NVC0_3D_COND_MODE_EQUAL : NVC0_3D_COND_MODE_ALWAYS;
> }
> break;
> diff --git a/src/gallium/drivers/nouveau/nvc0/nvc0_query_hw.c b/src/gallium/drivers/nouveau/nvc0/nvc0_query_hw.c
> index d8d82de8650..f2fdb0cd999 100644
> --- a/src/gallium/drivers/nouveau/nvc0/nvc0_query_hw.c
> +++ b/src/gallium/drivers/nouveau/nvc0/nvc0_query_hw.c
> @@ -150,20 +150,21 @@ nvc0_hw_begin_query(struct nvc0_context *nvc0, struct nvc0_query *q)
> hq->data[0] = hq->sequence; /* initialize sequence */
> hq->data[1] = 1; /* initial render condition = true */
> hq->data[4] = hq->sequence + 1; /* for comparison COND_MODE */
> hq->data[5] = 0;
> }
> hq->sequence++;
>
> switch (q->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> hq->nesting = nvc0->screen->num_occlusion_queries_active++;
> if (hq->nesting) {
> nvc0_hw_query_get(push, q, 0x10, 0x0100f002);
> } else {
> PUSH_SPACE(push, 3);
> BEGIN_NVC0(push, NVC0_3D(COUNTER_RESET), 1);
> PUSH_DATA (push, NVC0_3D_COUNTER_RESET_SAMPLECNT);
> IMMED_NVC0(push, NVC0_3D(SAMPLECNT_ENABLE), 1);
> }
> break;
> @@ -217,20 +218,21 @@ nvc0_hw_end_query(struct nvc0_context *nvc0, struct nvc0_query *q)
> /* some queries don't require 'begin' to be called (e.g. GPU_FINISHED) */
> if (hq->rotate)
> nvc0_hw_query_rotate(nvc0, q);
> hq->sequence++;
> }
> hq->state = NVC0_HW_QUERY_STATE_ENDED;
>
> switch (q->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> nvc0_hw_query_get(push, q, 0, 0x0100f002);
> if (--nvc0->screen->num_occlusion_queries_active == 0) {
> PUSH_SPACE(push, 1);
> IMMED_NVC0(push, NVC0_3D(SAMPLECNT_ENABLE), 0);
> }
> break;
> case PIPE_QUERY_PRIMITIVES_GENERATED:
> nvc0_hw_query_get(push, q, 0, 0x09005002 | (q->index << 5));
> break;
> case PIPE_QUERY_PRIMITIVES_EMITTED:
> @@ -313,20 +315,21 @@ nvc0_hw_get_query_result(struct nvc0_context *nvc0, struct nvc0_query *q,
> hq->state = NVC0_HW_QUERY_STATE_READY;
>
> switch (q->type) {
> case PIPE_QUERY_GPU_FINISHED:
> res8[0] = true;
> break;
> case PIPE_QUERY_OCCLUSION_COUNTER: /* u32 sequence, u32 count, u64 time */
> res64[0] = hq->data[1] - hq->data[5];
> break;
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> res8[0] = hq->data[1] != hq->data[5];
> break;
> case PIPE_QUERY_PRIMITIVES_GENERATED: /* u64 count, u64 time */
> case PIPE_QUERY_PRIMITIVES_EMITTED: /* u64 count, u64 time */
> res64[0] = data64[0] - data64[2];
> break;
> case PIPE_QUERY_SO_STATISTICS:
> res64[0] = data64[0] - data64[4];
> res64[1] = data64[2] - data64[6];
> break;
> @@ -401,21 +404,22 @@ nvc0_hw_get_query_result_resource(struct nvc0_context *nvc0,
> if (hq->state != NVC0_HW_QUERY_STATE_READY)
> nvc0_hw_query_update(nvc0->screen->base.client, q);
>
> if (wait && hq->state != NVC0_HW_QUERY_STATE_READY)
> nvc0_hw_query_fifo_wait(nvc0, q);
>
> nouveau_pushbuf_space(push, 32, 2, 0);
> PUSH_REFN (push, hq->bo, NOUVEAU_BO_GART | NOUVEAU_BO_RD);
> PUSH_REFN (push, buf->bo, buf->domain | NOUVEAU_BO_WR);
> BEGIN_1IC0(push, NVC0_3D(MACRO_QUERY_BUFFER_WRITE), 9);
> - if (q->type == PIPE_QUERY_OCCLUSION_PREDICATE) /* XXX what if 64-bit? */
> + if (q->type == PIPE_QUERY_OCCLUSION_PREDICATE ||
> + q->type ++ PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE) /* XXX what if 64-bit? */
> PUSH_DATA(push, 0x00000001);
> else if (result_type == PIPE_QUERY_TYPE_I32)
> PUSH_DATA(push, 0x7fffffff);
> else if (result_type == PIPE_QUERY_TYPE_U32)
> PUSH_DATA(push, 0xffffffff);
> else
> PUSH_DATA(push, 0x00000000);
>
> switch (q->type) {
> case PIPE_QUERY_SO_STATISTICS:
> @@ -506,20 +510,21 @@ nvc0_hw_create_query(struct nvc0_context *nvc0, unsigned type, unsigned index)
> if (!hq)
> return NULL;
>
> q = &hq->base;
> q->funcs = &hw_query_funcs;
> q->type = type;
>
> switch (q->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> hq->rotate = 32;
> space = NVC0_HW_QUERY_ALLOC_SPACE;
> break;
> case PIPE_QUERY_PIPELINE_STATISTICS:
> hq->is64bit = true;
> space = 512;
> break;
> case PIPE_QUERY_SO_STATISTICS:
> case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
> hq->is64bit = true;
> diff --git a/src/gallium/drivers/r300/r300_query.c b/src/gallium/drivers/r300/r300_query.c
> index d2dc7b73c05..a84c941768f 100644
> --- a/src/gallium/drivers/r300/r300_query.c
> +++ b/src/gallium/drivers/r300/r300_query.c
> @@ -32,20 +32,21 @@
> static struct pipe_query *r300_create_query(struct pipe_context *pipe,
> unsigned query_type,
> unsigned index)
> {
> struct r300_context *r300 = r300_context(pipe);
> struct r300_screen *r300screen = r300->screen;
> struct r300_query *q;
>
> if (query_type != PIPE_QUERY_OCCLUSION_COUNTER &&
> query_type != PIPE_QUERY_OCCLUSION_PREDICATE &&
> + query_type != PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE &&
> query_type != PIPE_QUERY_GPU_FINISHED) {
> return NULL;
> }
>
> q = CALLOC_STRUCT(r300_query);
> if (!q)
> return NULL;
>
> q->type = query_type;
>
> @@ -164,21 +165,22 @@ static boolean r300_get_query_result(struct pipe_context* pipe,
> return FALSE;
>
> /* Sum up the results. */
> temp = 0;
> for (i = 0; i < q->num_results; i++) {
> /* Convert little endian values written by GPU to CPU byte order */
> temp += util_le32_to_cpu(*map);
> map++;
> }
>
> - if (q->type == PIPE_QUERY_OCCLUSION_PREDICATE) {
> + if (q->type == PIPE_QUERY_OCCLUSION_PREDICATE ||
> + q->type == PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE) {
> vresult->b = temp != 0;
> } else {
> vresult->u64 = temp;
> }
> return TRUE;
> }
>
> static void r300_render_condition(struct pipe_context *pipe,
> struct pipe_query *query,
> boolean condition,
> @@ -188,21 +190,22 @@ static void r300_render_condition(struct pipe_context *pipe,
> union pipe_query_result result;
> boolean wait;
>
> r300->skip_rendering = FALSE;
>
> if (query) {
> wait = mode == PIPE_RENDER_COND_WAIT ||
> mode == PIPE_RENDER_COND_BY_REGION_WAIT;
>
> if (r300_get_query_result(pipe, query, wait, &result)) {
> - if (r300_query(query)->type == PIPE_QUERY_OCCLUSION_PREDICATE) {
> + if (r300_query(query)->type == PIPE_QUERY_OCCLUSION_PREDICATE ||
> + r300_query(query)->type == PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE) {
> r300->skip_rendering = condition == result.b;
> } else {
> r300->skip_rendering = condition == !!result.u64;
> }
> }
> }
> }
>
> static void
> r300_set_active_query_state(struct pipe_context *pipe, boolean enable)
> diff --git a/src/gallium/drivers/radeon/r600_pipe_common.c b/src/gallium/drivers/radeon/r600_pipe_common.c
> index b283bc65b5b..1183e181a82 100644
> --- a/src/gallium/drivers/radeon/r600_pipe_common.c
> +++ b/src/gallium/drivers/radeon/r600_pipe_common.c
> @@ -119,21 +119,22 @@ void r600_gfx_write_event_eop(struct r600_common_context *ctx,
> if (ctx->chip_class >= GFX9) {
> /* A ZPASS_DONE or PIXEL_STAT_DUMP_EVENT (of the DB occlusion
> * counters) must immediately precede every timestamp event to
> * prevent a GPU hang on GFX9.
> *
> * Occlusion queries don't need to do it here, because they
> * always do ZPASS_DONE before the timestamp.
> */
> if (ctx->chip_class == GFX9 &&
> query_type != PIPE_QUERY_OCCLUSION_COUNTER &&
> - query_type != PIPE_QUERY_OCCLUSION_PREDICATE) {
> + query_type != PIPE_QUERY_OCCLUSION_PREDICATE &&
> + query_type != PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE) {
> struct r600_resource *scratch = ctx->eop_bug_scratch;
>
> assert(16 * ctx->screen->info.num_render_backends <=
> scratch->b.b.width0);
> radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 2, 0));
> radeon_emit(cs, EVENT_TYPE(EVENT_TYPE_ZPASS_DONE) | EVENT_INDEX(1));
> radeon_emit(cs, scratch->gpu_address);
> radeon_emit(cs, scratch->gpu_address >> 32);
>
> radeon_add_to_buffer_list(ctx, &ctx->gfx, scratch,
> diff --git a/src/gallium/drivers/radeon/r600_query.c b/src/gallium/drivers/radeon/r600_query.c
> index 76307ca0662..e655bbdaa23 100644
> --- a/src/gallium/drivers/radeon/r600_query.c
> +++ b/src/gallium/drivers/radeon/r600_query.c
> @@ -544,21 +544,22 @@ static bool r600_query_hw_prepare_buffer(struct r600_common_screen *rscreen,
> /* Callers ensure that the buffer is currently unused by the GPU. */
> uint32_t *results = rscreen->ws->buffer_map(buffer->buf, NULL,
> PIPE_TRANSFER_WRITE |
> PIPE_TRANSFER_UNSYNCHRONIZED);
> if (!results)
> return false;
>
> memset(results, 0, buffer->b.b.width0);
>
> if (query->b.type == PIPE_QUERY_OCCLUSION_COUNTER ||
> - query->b.type == PIPE_QUERY_OCCLUSION_PREDICATE) {
> + query->b.type == PIPE_QUERY_OCCLUSION_PREDICATE ||
> + query->b.type == PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE) {
> unsigned max_rbs = rscreen->info.num_render_backends;
> unsigned enabled_rb_mask = rscreen->info.enabled_rb_mask;
> unsigned num_results;
> unsigned i, j;
>
> /* Set top bits for unused backends. */
> num_results = buffer->b.b.width0 / query->result_size;
> for (j = 0; j < num_results; j++) {
> for (i = 0; i < max_rbs; i++) {
> if (!(enabled_rb_mask & (1<<i))) {
> @@ -629,20 +630,21 @@ static struct pipe_query *r600_query_hw_create(struct r600_common_screen *rscree
> if (!query)
> return NULL;
>
> query->b.type = query_type;
> query->b.ops = &query_hw_ops;
> query->ops = &query_hw_default_hw_ops;
>
> switch (query_type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> query->result_size = 16 * rscreen->info.num_render_backends;
> query->result_size += 16; /* for the fence + alignment */
> query->num_cs_dw_begin = 6;
> query->num_cs_dw_end = 6 + r600_gfx_write_fence_dwords(rscreen);
> break;
> case PIPE_QUERY_TIME_ELAPSED:
> query->result_size = 24;
> query->num_cs_dw_begin = 8;
> query->num_cs_dw_end = 8 + r600_gfx_write_fence_dwords(rscreen);
> break;
> @@ -685,30 +687,31 @@ static struct pipe_query *r600_query_hw_create(struct r600_common_screen *rscree
> return NULL;
> }
>
> return (struct pipe_query *)query;
> }
>
> static void r600_update_occlusion_query_state(struct r600_common_context *rctx,
> unsigned type, int diff)
> {
> if (type == PIPE_QUERY_OCCLUSION_COUNTER ||
> - type == PIPE_QUERY_OCCLUSION_PREDICATE) {
> + type == PIPE_QUERY_OCCLUSION_PREDICATE ||
> + type == PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE) {
> bool old_enable = rctx->num_occlusion_queries != 0;
> bool old_perfect_enable =
> rctx->num_perfect_occlusion_queries != 0;
> bool enable, perfect_enable;
>
> rctx->num_occlusion_queries += diff;
> assert(rctx->num_occlusion_queries >= 0);
>
> - if (type == PIPE_QUERY_OCCLUSION_COUNTER) {
> + if (type != PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE) {
> rctx->num_perfect_occlusion_queries += diff;
> assert(rctx->num_perfect_occlusion_queries >= 0);
> }
>
> enable = rctx->num_occlusion_queries != 0;
> perfect_enable = rctx->num_perfect_occlusion_queries != 0;
>
> if (enable != old_enable || perfect_enable != old_perfect_enable) {
> rctx->set_occlusion_query_state(&rctx->b, enable);
> }
> @@ -738,20 +741,21 @@ static void emit_sample_streamout(struct radeon_winsys_cs *cs, uint64_t va,
> static void r600_query_hw_do_emit_start(struct r600_common_context *ctx,
> struct r600_query_hw *query,
> struct r600_resource *buffer,
> uint64_t va)
> {
> struct radeon_winsys_cs *cs = ctx->gfx.cs;
>
> switch (query->b.type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 2, 0));
> radeon_emit(cs, EVENT_TYPE(EVENT_TYPE_ZPASS_DONE) | EVENT_INDEX(1));
> radeon_emit(cs, va);
> radeon_emit(cs, va >> 32);
> break;
> case PIPE_QUERY_PRIMITIVES_EMITTED:
> case PIPE_QUERY_PRIMITIVES_GENERATED:
> case PIPE_QUERY_SO_STATISTICS:
> case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
> emit_sample_streamout(cs, va, query->stream);
> @@ -832,20 +836,21 @@ static void r600_query_hw_do_emit_stop(struct r600_common_context *ctx,
> struct r600_query_hw *query,
> struct r600_resource *buffer,
> uint64_t va)
> {
> struct radeon_winsys_cs *cs = ctx->gfx.cs;
> uint64_t fence_va = 0;
>
> switch (query->b.type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> va += 8;
> radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 2, 0));
> radeon_emit(cs, EVENT_TYPE(EVENT_TYPE_ZPASS_DONE) | EVENT_INDEX(1));
> radeon_emit(cs, va);
> radeon_emit(cs, va >> 32);
>
> fence_va = va + ctx->screen->info.num_render_backends * 16 - 8;
> break;
> case PIPE_QUERY_PRIMITIVES_EMITTED:
> case PIPE_QUERY_PRIMITIVES_GENERATED:
> @@ -954,20 +959,21 @@ static void r600_emit_query_predication(struct r600_common_context *ctx,
> invert = ctx->render_cond_invert;
> flag_wait = ctx->render_cond_mode == PIPE_RENDER_COND_WAIT ||
> ctx->render_cond_mode == PIPE_RENDER_COND_BY_REGION_WAIT;
>
> if (query->workaround_buf) {
> op = PRED_OP(PREDICATION_OP_BOOL64);
> } else {
> switch (query->b.type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> op = PRED_OP(PREDICATION_OP_ZPASS);
> break;
> case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
> case PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE:
> op = PRED_OP(PREDICATION_OP_PRIMCOUNT);
> invert = !invert;
> break;
> default:
> assert(0);
> return;
> @@ -1132,20 +1138,21 @@ static void r600_get_hw_query_params(struct r600_common_context *rctx,
> struct r600_hw_query_params *params)
> {
> unsigned max_rbs = rctx->screen->info.num_render_backends;
>
> params->pair_stride = 0;
> params->pair_count = 1;
>
> switch (rquery->b.type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> params->start_offset = 0;
> params->end_offset = 8;
> params->fence_offset = max_rbs * 16;
> params->pair_stride = 16;
> params->pair_count = max_rbs;
> break;
> case PIPE_QUERY_TIME_ELAPSED:
> params->start_offset = 0;
> params->end_offset = 8;
> params->fence_offset = 16;
> @@ -1224,21 +1231,22 @@ static void r600_query_hw_add_result(struct r600_common_screen *rscreen,
>
> switch (query->b.type) {
> case PIPE_QUERY_OCCLUSION_COUNTER: {
> for (unsigned i = 0; i < max_rbs; ++i) {
> unsigned results_base = i * 16;
> result->u64 +=
> r600_query_read_result(buffer + results_base, 0, 2, true);
> }
> break;
> }
> - case PIPE_QUERY_OCCLUSION_PREDICATE: {
> + case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE: {
> for (unsigned i = 0; i < max_rbs; ++i) {
> unsigned results_base = i * 16;
> result->b = result->b ||
> r600_query_read_result(buffer + results_base, 0, 2, true) != 0;
> }
> break;
> }
> case PIPE_QUERY_TIME_ELAPSED:
> result->u64 += r600_query_read_result(buffer, 0, 2, false);
> break;
> @@ -1704,21 +1712,22 @@ static void r600_query_hw_get_result_resource(struct r600_common_context *rctx,
> grid.block[0] = 1;
> grid.block[1] = 1;
> grid.block[2] = 1;
> grid.grid[0] = 1;
> grid.grid[1] = 1;
> grid.grid[2] = 1;
>
> consts.config = 0;
> if (index < 0)
> consts.config |= 4;
> - if (query->b.type == PIPE_QUERY_OCCLUSION_PREDICATE)
> + if (query->b.type == PIPE_QUERY_OCCLUSION_PREDICATE ||
> + query->b.type == PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE)
> consts.config |= 8;
> else if (query->b.type == PIPE_QUERY_SO_OVERFLOW_PREDICATE ||
> query->b.type == PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE)
> consts.config |= 8 | 256;
> else if (query->b.type == PIPE_QUERY_TIMESTAMP ||
> query->b.type == PIPE_QUERY_TIME_ELAPSED)
> consts.config |= 32;
>
> switch (result_type) {
> case PIPE_QUERY_TYPE_U64:
> diff --git a/src/gallium/drivers/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c
> index 63f6c4be042..267c99977fe 100644
> --- a/src/gallium/drivers/softpipe/sp_query.c
> +++ b/src/gallium/drivers/softpipe/sp_query.c
> @@ -53,20 +53,21 @@ static struct softpipe_query *softpipe_query( struct pipe_query *p )
>
> static struct pipe_query *
> softpipe_create_query(struct pipe_context *pipe,
> unsigned type,
> unsigned index)
> {
> struct softpipe_query* sq;
>
> assert(type == PIPE_QUERY_OCCLUSION_COUNTER ||
> type == PIPE_QUERY_OCCLUSION_PREDICATE ||
> + type == PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE ||
> type == PIPE_QUERY_TIME_ELAPSED ||
> type == PIPE_QUERY_SO_STATISTICS ||
> type == PIPE_QUERY_SO_OVERFLOW_PREDICATE ||
> type == PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE ||
> type == PIPE_QUERY_PRIMITIVES_EMITTED ||
> type == PIPE_QUERY_PRIMITIVES_GENERATED ||
> type == PIPE_QUERY_PIPELINE_STATISTICS ||
> type == PIPE_QUERY_GPU_FINISHED ||
> type == PIPE_QUERY_TIMESTAMP ||
> type == PIPE_QUERY_TIMESTAMP_DISJOINT);
> @@ -86,20 +87,21 @@ softpipe_destroy_query(struct pipe_context *pipe, struct pipe_query *q)
>
> static boolean
> softpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q)
> {
> struct softpipe_context *softpipe = softpipe_context( pipe );
> struct softpipe_query *sq = softpipe_query(q);
>
> switch (sq->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> sq->start = softpipe->occlusion_count;
> break;
> case PIPE_QUERY_TIME_ELAPSED:
> sq->start = os_time_get_nano();
> break;
> case PIPE_QUERY_SO_STATISTICS:
> sq->so.num_primitives_written = softpipe->so_stats.num_primitives_written;
> sq->so.primitives_storage_needed = softpipe->so_stats.primitives_storage_needed;
> break;
> case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
> @@ -140,20 +142,21 @@ softpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q)
> static bool
> softpipe_end_query(struct pipe_context *pipe, struct pipe_query *q)
> {
> struct softpipe_context *softpipe = softpipe_context( pipe );
> struct softpipe_query *sq = softpipe_query(q);
>
> softpipe->active_query_count--;
> switch (sq->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> sq->end = softpipe->occlusion_count;
> break;
> case PIPE_QUERY_TIMESTAMP:
> sq->start = 0;
> /* fall through */
> case PIPE_QUERY_TIME_ELAPSED:
> sq->end = os_time_get_nano();
> break;
> case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
> case PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE:
> @@ -245,20 +248,21 @@ softpipe_get_query_result(struct pipe_context *pipe,
> td->disjoint = FALSE;
> }
> break;
> case PIPE_QUERY_PRIMITIVES_EMITTED:
> *result = sq->so.num_primitives_written;
> break;
> case PIPE_QUERY_PRIMITIVES_GENERATED:
> *result = sq->so.primitives_storage_needed;
> break;
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> vresult->b = sq->end - sq->start != 0;
> break;
> default:
> *result = sq->end - sq->start;
> break;
> }
> return TRUE;
> }
>
>
> diff --git a/src/gallium/drivers/svga/svga_pipe_query.c b/src/gallium/drivers/svga/svga_pipe_query.c
> index 0490a4ab5fc..26924524372 100644
> --- a/src/gallium/drivers/svga/svga_pipe_query.c
> +++ b/src/gallium/drivers/svga/svga_pipe_query.c
> @@ -700,20 +700,21 @@ svga_create_query(struct pipe_context *pipe,
> * query of occlusion predicate type will be used
> * in the SetPredication command.
> */
> sq->predicate = svga_create_query(pipe, PIPE_QUERY_OCCLUSION_PREDICATE, index);
>
> } else {
> define_query_vgpu9(svga, sq);
> }
> break;
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> if (svga_have_vgpu10(svga)) {
> sq->svga_type = SVGA3D_QUERYTYPE_OCCLUSIONPREDICATE;
> define_query_vgpu10(svga, sq, sizeof(SVGADXOcclusionPredicateQueryResult));
> } else {
> sq->svga_type = SVGA3D_QUERYTYPE_OCCLUSION;
> define_query_vgpu9(svga, sq);
> }
> break;
> case PIPE_QUERY_PRIMITIVES_GENERATED:
> case PIPE_QUERY_PRIMITIVES_EMITTED:
> @@ -782,20 +783,21 @@ svga_destroy_query(struct pipe_context *pipe, struct pipe_query *q)
> }
>
> sq = svga_query(q);
>
> SVGA_DBG(DEBUG_QUERY, "%s sq=0x%x id=%d\n", __FUNCTION__,
> sq, sq->id);
>
> switch (sq->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> if (svga_have_vgpu10(svga)) {
> /* make sure to also destroy any associated predicate query */
> if (sq->predicate)
> svga_destroy_query(pipe, sq->predicate);
> destroy_query_vgpu10(svga, sq);
> } else {
> sws->buffer_destroy(sws, sq->hwbuf);
> }
> sws->fence_reference(sws, &sq->fence, NULL);
> break;
> @@ -857,20 +859,21 @@ svga_begin_query(struct pipe_context *pipe, struct pipe_query *q)
> sq, sq->id);
>
> /* Need to flush out buffered drawing commands so that they don't
> * get counted in the query results.
> */
> svga_hwtnl_flush_retry(svga);
>
> switch (sq->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> if (svga_have_vgpu10(svga)) {
> ret = begin_query_vgpu10(svga, sq);
> /* also need to start the associated occlusion predicate query */
> if (sq->predicate) {
> enum pipe_error status;
> status = begin_query_vgpu10(svga, svga_query(sq->predicate));
> assert(status == PIPE_OK);
> (void) status;
> }
> } else {
> @@ -970,20 +973,21 @@ svga_end_query(struct pipe_context *pipe, struct pipe_query *q)
> if (sq->type == PIPE_QUERY_TIMESTAMP && svga->sq[sq->type] != sq)
> svga_begin_query(pipe, q);
>
> svga_hwtnl_flush_retry(svga);
>
> assert(svga->sq[sq->type] == sq);
>
> switch (sq->type) {
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> if (svga_have_vgpu10(svga)) {
> ret = end_query_vgpu10(svga, sq);
> /* also need to end the associated occlusion predicate query */
> if (sq->predicate) {
> enum pipe_error status;
> status = end_query_vgpu10(svga, svga_query(sq->predicate));
> assert(status == PIPE_OK);
> (void) status;
> }
> } else {
> @@ -1086,21 +1090,22 @@ svga_get_query_result(struct pipe_context *pipe,
> case PIPE_QUERY_OCCLUSION_COUNTER:
> if (svga_have_vgpu10(svga)) {
> SVGADXOcclusionQueryResult occResult;
> ret = get_query_result_vgpu10(svga, sq, wait,
> (void *)&occResult, sizeof(occResult));
> *result = (uint64_t)occResult.samplesRendered;
> } else {
> ret = get_query_result_vgpu9(svga, sq, wait, result);
> }
> break;
> - case PIPE_QUERY_OCCLUSION_PREDICATE: {
> + case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE: {
> if (svga_have_vgpu10(svga)) {
> SVGADXOcclusionPredicateQueryResult occResult;
> ret = get_query_result_vgpu10(svga, sq, wait,
> (void *)&occResult, sizeof(occResult));
> vresult->b = occResult.anySamplesRendered != 0;
> } else {
> uint64_t count = 0;
> ret = get_query_result_vgpu9(svga, sq, wait, &count);
> vresult->b = count != 0;
> }
> diff --git a/src/gallium/drivers/swr/swr_query.cpp b/src/gallium/drivers/swr/swr_query.cpp
> index 4c14c5206b0..e3697304e05 100644
> --- a/src/gallium/drivers/swr/swr_query.cpp
> +++ b/src/gallium/drivers/swr/swr_query.cpp
> @@ -87,20 +87,21 @@ swr_get_query_result(struct pipe_context *pipe,
>
> swr_fence_finish(pipe->screen, NULL, pq->fence, 0);
> swr_fence_reference(pipe->screen, &pq->fence, NULL);
> }
>
> /* All values are reset to 0 at swr_begin_query, except starting timestamp.
> * Counters become simply end values. */
> switch (pq->type) {
> /* Booleans */
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> result->b = pq->result.core.DepthPassCount != 0;
> break;
> case PIPE_QUERY_GPU_FINISHED:
> result->b = TRUE;
> break;
> /* Counters */
> case PIPE_QUERY_OCCLUSION_COUNTER:
> result->u64 = pq->result.core.DepthPassCount;
> break;
> case PIPE_QUERY_TIMESTAMP:
> diff --git a/src/gallium/drivers/trace/tr_dump_state.c b/src/gallium/drivers/trace/tr_dump_state.c
> index 0650f63d266..e7e32237c4c 100644
> --- a/src/gallium/drivers/trace/tr_dump_state.c
> +++ b/src/gallium/drivers/trace/tr_dump_state.c
> @@ -871,20 +871,21 @@ trace_dump_query_result(unsigned query_type,
> if (!trace_dumping_enabled_locked())
> return;
>
> if (!result) {
> trace_dump_null();
> return;
> }
>
> switch (query_type) {
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
> case PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE:
> case PIPE_QUERY_GPU_FINISHED:
> trace_dump_bool(result->b);
> break;
>
> case PIPE_QUERY_OCCLUSION_COUNTER:
> case PIPE_QUERY_TIMESTAMP:
> case PIPE_QUERY_TIME_ELAPSED:
> case PIPE_QUERY_PRIMITIVES_GENERATED:
> diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
> index da7d5da7347..9b428a6ad76 100644
> --- a/src/gallium/include/pipe/p_defines.h
> +++ b/src/gallium/include/pipe/p_defines.h
> @@ -523,20 +523,21 @@ enum pipe_tess_spacing {
> PIPE_TESS_SPACING_FRACTIONAL_EVEN,
> PIPE_TESS_SPACING_EQUAL,
> };
>
> /**
> * Query object types
> */
> enum pipe_query_type {
> PIPE_QUERY_OCCLUSION_COUNTER,
> PIPE_QUERY_OCCLUSION_PREDICATE,
> + PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE,
> PIPE_QUERY_TIMESTAMP,
> PIPE_QUERY_TIMESTAMP_DISJOINT,
> PIPE_QUERY_TIME_ELAPSED,
> PIPE_QUERY_PRIMITIVES_GENERATED,
> PIPE_QUERY_PRIMITIVES_EMITTED,
> PIPE_QUERY_SO_STATISTICS,
> PIPE_QUERY_SO_OVERFLOW_PREDICATE,
> PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE,
> PIPE_QUERY_GPU_FINISHED,
> PIPE_QUERY_PIPELINE_STATISTICS,
> @@ -938,20 +939,21 @@ union pipe_numeric_type_union
> uint32_t u32;
> float f;
> };
>
> /**
> * Query result (returned by pipe_context::get_query_result).
> */
> union pipe_query_result
> {
> /* PIPE_QUERY_OCCLUSION_PREDICATE */
> + /* PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE */
> /* PIPE_QUERY_SO_OVERFLOW_PREDICATE */
> /* PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE */
> /* PIPE_QUERY_GPU_FINISHED */
> boolean b;
>
> /* PIPE_QUERY_OCCLUSION_COUNTER */
> /* PIPE_QUERY_TIMESTAMP */
> /* PIPE_QUERY_TIME_ELAPSED */
> /* PIPE_QUERY_PRIMITIVES_GENERATED */
> /* PIPE_QUERY_PRIMITIVES_EMITTED */
> diff --git a/src/mesa/state_tracker/st_cb_queryobj.c b/src/mesa/state_tracker/st_cb_queryobj.c
> index 4c25724b5dc..a470ff6b8b5 100644
> --- a/src/mesa/state_tracker/st_cb_queryobj.c
> +++ b/src/mesa/state_tracker/st_cb_queryobj.c
> @@ -95,23 +95,25 @@ st_BeginQuery(struct gl_context *ctx, struct gl_query_object *q)
> struct pipe_context *pipe = st->pipe;
> struct st_query_object *stq = st_query_object(q);
> unsigned type;
> bool ret = false;
>
> st_flush_bitmap_cache(st_context(ctx));
>
> /* convert GL query type to Gallium query type */
> switch (q->Target) {
> case GL_ANY_SAMPLES_PASSED:
> - case GL_ANY_SAMPLES_PASSED_CONSERVATIVE:
> type = PIPE_QUERY_OCCLUSION_PREDICATE;
> break;
> + case GL_ANY_SAMPLES_PASSED_CONSERVATIVE:
> + type = PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE;
> + break;
> case GL_SAMPLES_PASSED_ARB:
> type = PIPE_QUERY_OCCLUSION_COUNTER;
> break;
> case GL_PRIMITIVES_GENERATED:
> type = PIPE_QUERY_PRIMITIVES_GENERATED;
> break;
> case GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN:
> type = PIPE_QUERY_PRIMITIVES_EMITTED;
> break;
> case GL_TRANSFORM_FEEDBACK_STREAM_OVERFLOW_ARB:
> @@ -253,20 +255,21 @@ get_query_result(struct pipe_context *pipe,
> break;
> case GL_CLIPPING_INPUT_PRIMITIVES_ARB:
> stq->base.Result = data.pipeline_statistics.c_invocations;
> break;
> case GL_CLIPPING_OUTPUT_PRIMITIVES_ARB:
> stq->base.Result = data.pipeline_statistics.c_primitives;
> break;
> default:
> switch (stq->type) {
> case PIPE_QUERY_OCCLUSION_PREDICATE:
> + case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
> case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
> case PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE:
> stq->base.Result = !!data.b;
> break;
> default:
> stq->base.Result = data.u64;
> break;
> }
> break;
> }
> --
> 2.11.0
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
More information about the mesa-dev
mailing list