r300: OQ rework

Move to common code base so radeon/r200 can add support for this.
Make OQ start a state emitted like all normal state, and make no-tcl
flushing work in proper places.

Really need a generic post emit space reservation mechanism like max_state
so we can reserve some space for the emit

this code passes demos/arbocclude, piglit occlusion query and
glean occlusion query with TCL and NO-TCL on my rv530.

5 files changed, 298 insertions, 1 deletions

diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c
index 0614c89459..0e9a1ae5d8 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common.c
@@ -83,6 +83,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_lock.h"
 #include "radeon_drm.h"
 #include "radeon_mipmap_tree.h"
+#include "radeon_queryobj.h"
 
 #define DEBUG_CMDBUF         0
 
@@ -1072,6 +1073,9 @@ void radeonFlush(GLcontext *ctx)
 			}
 		}
 	}
+
+	make_empty_list(&radeon->query.not_flushed_head);
+
 }
 
 /* Make sure all commands have been sent to the hardware and have
@@ -1128,6 +1132,8 @@ int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller)
 			__FUNCTION__, caller, rmesa->numClipRects);
 	}
 
+	radeonEmitQueryEnd(rmesa->glCtx);
+
 	if (rmesa->cmdbuf.cs->cdw) {
 		ret = radeon_cs_emit(rmesa->cmdbuf.cs);
 		rmesa->hw.all_dirty = GL_TRUE;
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c
index c0abcbfa21..35622099bf 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c
@@ -263,6 +263,8 @@ GLboolean radeonInitContext(radeonContextPtr radeon,
 		radeon->texture_compressed_row_align = 64;
 	}
 
+	make_empty_list(&radeon->query.not_flushed_head);
+
 	return GL_TRUE;
 }
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h
index ee46c1f81a..a9480cd2e4 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common_context.h
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h
@@ -163,6 +163,7 @@ struct radeon_hw_state {
   	/* Head of the linked list of state atoms. */
 	struct radeon_state_atom atomlist;
 	int max_state_size;	/* Number of bytes necessary for a full state emit. */
+	int max_post_flush_size; /* Number of bytes necessary for post flushing emits */
 	GLboolean is_dirty, all_dirty;
 };
 
@@ -254,6 +255,17 @@ static INLINE radeonTexObj* radeon_tex_obj(struct gl_texture_object *texObj)
 	return (radeonTexObj*)texObj;
 }
 
+/* occlusion query */
+struct radeon_query_object {
+	struct gl_query_object Base;
+	struct radeon_bo *bo;
+	int curr_offset;
+	GLboolean emitted_begin;
+
+	/* Double linked list of not flushed query objects */
+	struct radeon_query_object *prev, *next;
+};
+
 /* Need refcounting on dma buffers:
  */
 struct radeon_dma_buffer {
@@ -500,6 +512,12 @@ struct radeon_context {
    struct dri_metaops meta;
 
    struct {
+	struct radeon_query_object *current;
+	struct radeon_query_object not_flushed_head;
+	struct radeon_state_atom queryobj;
+   } query;
+
+   struct {
 	   void (*get_lock)(radeonContextPtr radeon);
 	   void (*update_viewport_offset)(GLcontext *ctx);
 	   void (*emit_cs_header)(struct radeon_cs *cs, radeonContextPtr rmesa);
@@ -508,6 +526,7 @@ struct radeon_context {
 	   void (*pre_emit_state)(radeonContextPtr rmesa);
 	   void (*fallback)(GLcontext *ctx, GLuint bit, GLboolean mode);
 	   void (*free_context)(GLcontext *ctx);
+	   void (*emit_query_finish)(radeonContextPtr radeon);
    } vtbl;
 };
 
@@ -523,7 +542,6 @@ static inline __DRIdrawablePrivate* radeon_get_readable(radeonContextPtr radeon)
 	return radeon->dri.context->driReadablePriv;
 }
 
-
 /**
  * This function takes a float and packs it into a uint32_t
  */
diff --git a/src/mesa/drivers/dri/radeon/radeon_queryobj.c b/src/mesa/drivers/dri/radeon/radeon_queryobj.c
new file mode 100644
index 0000000000..70251946df
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_queryobj.c
@@ -0,0 +1,216 @@
+/*
+ * Copyright © 2008-2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Maciej Cencora <m.cencora@gmail.com>
+ *
+ */
+#include "radeon_common.h"
+#include "radeon_queryobj.h"
+
+#include "main/imports.h"
+#include "main/simple_list.h"
+
+#define DDEBUG 0
+
+#define PAGE_SIZE 4096
+
+static void radeonQueryGetResult(GLcontext *ctx, struct gl_query_object *q)
+{
+	struct radeon_query_object *query = (struct radeon_query_object *)q;
+	uint32_t *result;
+	int i;
+
+	if (DDEBUG) fprintf(stderr, "%s: query id %d, result %d\n", __FUNCTION__, query->Base.Id, (int) query->Base.Result);
+
+	radeon_bo_map(query->bo, GL_FALSE);
+
+	result = query->bo->ptr;
+
+	query->Base.Result = 0;
+	for (i = 0; i < query->curr_offset/sizeof(uint32_t); ++i) {
+		query->Base.Result += result[i];
+		if (DDEBUG) fprintf(stderr, "result[%d] = %d\n", i, result[i]);
+	}
+
+	radeon_bo_unmap(query->bo);
+}
+
+static struct gl_query_object * radeonNewQueryObject(GLcontext *ctx, GLuint id)
+{
+	struct radeon_query_object *query;
+
+	query = _mesa_calloc(sizeof(struct radeon_query_object));
+
+	query->Base.Id = id;
+	query->Base.Result = 0;
+	query->Base.Active = GL_FALSE;
+	query->Base.Ready = GL_TRUE;
+
+	if (DDEBUG) fprintf(stderr, "%s: query id %d\n", __FUNCTION__, query->Base.Id);
+
+	return &query->Base;
+}
+
+static void radeonDeleteQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+	struct radeon_query_object *query = (struct radeon_query_object *)q;
+
+	if (DDEBUG) fprintf(stderr, "%s: query id %d\n", __FUNCTION__, q->Id);
+
+	if (query->bo) {
+		radeon_bo_unref(query->bo);
+	}
+
+	_mesa_free(query);
+}
+
+static void radeonWaitQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct radeon_query_object *tmp, *query = (struct radeon_query_object *)q;
+
+	/* If the cmdbuf with packets for this query hasn't been flushed yet, do it now */
+	{
+		GLboolean found = GL_FALSE;
+		foreach(tmp, &radeon->query.not_flushed_head) {
+			if (tmp == query) {
+				found = GL_TRUE;
+				break;
+			}
+		}
+
+		if (found)
+			ctx->Driver.Flush(ctx);
+	}
+
+	if (DDEBUG) fprintf(stderr, "%s: query id %d, bo %p, offset %d\n", __FUNCTION__, q->Id, query->bo, query->curr_offset);
+
+	radeonQueryGetResult(ctx, q);
+
+	query->Base.Ready = GL_TRUE;
+}
+
+
+static void radeonBeginQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct radeon_query_object *query = (struct radeon_query_object *)q;
+
+	if (DDEBUG) fprintf(stderr, "%s: query id %d\n", __FUNCTION__, q->Id);
+
+	assert(radeon->query.current == NULL);
+
+	if (radeon->dma.flush)
+		radeon->dma.flush(radeon->glCtx);
+
+	if (!query->bo) {
+		query->bo = radeon_bo_open(radeon->radeonScreen->bom, 0, PAGE_SIZE, PAGE_SIZE, RADEON_GEM_DOMAIN_GTT, 0);
+	}
+	query->curr_offset = 0;
+
+	radeon->query.current = query;
+
+	radeon->query.queryobj.dirty = GL_TRUE;
+	insert_at_tail(&radeon->query.not_flushed_head, query);
+
+}
+
+void radeonEmitQueryEnd(GLcontext *ctx)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct radeon_query_object *query = radeon->query.current;
+
+	if (!query)
+		return;
+
+	if (query->emitted_begin == GL_FALSE)
+		return;
+
+	if (DDEBUG) fprintf(stderr, "%s: query id %d, bo %p, offset %d\n", __FUNCTION__, query->Base.Id, query->bo, query->curr_offset);
+
+	radeon_cs_space_check_with_bo(radeon->cmdbuf.cs,
+				      query->bo,
+				      0, RADEON_GEM_DOMAIN_GTT);
+
+	radeon->vtbl.emit_query_finish(radeon);
+}
+
+static void radeonEndQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+
+	if (DDEBUG) fprintf(stderr, "%s: query id %d\n", __FUNCTION__, q->Id);
+
+	if (radeon->dma.flush)
+		radeon->dma.flush(radeon->glCtx);
+	radeonEmitQueryEnd(ctx);
+
+	radeon->query.current = NULL;
+}
+
+/**
+ * TODO:
+ * should check if bo is idle, bo there's no interface to do it
+ * just wait for result now
+ */
+static void radeonCheckQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+	if (DDEBUG) fprintf(stderr, "%s: query id %d\n", __FUNCTION__, q->Id);
+
+	radeonWaitQuery(ctx, q);
+}
+
+void radeonInitQueryObjFunctions(struct dd_function_table *functions)
+{
+	functions->NewQueryObject = radeonNewQueryObject;
+	functions->DeleteQuery = radeonDeleteQuery;
+	functions->BeginQuery = radeonBeginQuery;
+	functions->EndQuery = radeonEndQuery;
+	functions->CheckQuery = radeonCheckQuery;
+	functions->WaitQuery = radeonWaitQuery;
+}
+
+int radeon_check_query_active(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct radeon_query_object *query = radeon->query.current;
+
+	if (!query || query->emitted_begin)
+		return 0;
+	return atom->cmd_size;
+}
+
+void radeon_emit_queryobj(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	BATCH_LOCALS(radeon);
+	int dwords;
+
+	dwords = (*atom->check) (ctx, atom);
+
+	BEGIN_BATCH_NO_AUTOSTATE(dwords);
+	OUT_BATCH_TABLE(atom->cmd, dwords);
+	END_BATCH();
+
+	radeon->query.current->emitted_begin = GL_TRUE;
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_queryobj.h b/src/mesa/drivers/dri/radeon/radeon_queryobj.h
new file mode 100644
index 0000000000..5e705838fe
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_queryobj.h
@@ -0,0 +1,55 @@
+/*
+ * Copyright © 2008 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Maciej Cencora <m.cencora@gmail.com>
+ *
+ */
+
+#include "main/imports.h"
+#include "main/simple_list.h"
+#include "radeon_context.h"
+
+extern void radeonEmitQueryBegin(GLcontext *ctx);
+extern void radeonEmitQueryEnd(GLcontext *ctx);
+
+extern void radeonInitQueryObjFunctions(struct dd_function_table *functions);
+
+#define RADEON_QUERY_PAGE_SIZE 4096
+
+int radeon_check_query_active(GLcontext *ctx, struct radeon_state_atom *atom);
+void radeon_emit_queryobj(GLcontext *ctx, struct radeon_state_atom *atom);
+
+static inline void radeon_init_query_stateobj(radeonContextPtr radeon, int SZ)
+{
+	radeon->query.queryobj.cmd_size = (SZ);
+	radeon->query.queryobj.cmd = (uint32_t*)CALLOC((SZ) * sizeof(uint32_t));
+	radeon->query.queryobj.name = "queryobj";
+	radeon->query.queryobj.idx = 0;
+	radeon->query.queryobj.check = radeon_check_query_active;
+	radeon->query.queryobj.dirty = GL_FALSE;
+	radeon->query.queryobj.emit = radeon_emit_queryobj;
+
+	radeon->hw.max_state_size += (SZ);
+	insert_at_tail(&radeon->hw.atomlist, &radeon->query.queryobj);
+}
+