Merge branch 'master' of ssh://git.freedesktop.org/git/mesa/mesa into r600_state_predict

Conflicts:
	src/mesa/drivers/dri/r300/r300_cmdbuf.c
	src/mesa/drivers/dri/radeon/radeon_cmdbuf.h

9 files changed, 97 insertions, 23 deletions

diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h
index 8789e3ab09b..71413716333 100644
--- a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h
+++ b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h
@@ -73,6 +73,7 @@ struct radeon_bo_funcs {
 			  uint32_t pitch);
     int (*bo_get_tiling)(struct radeon_bo *bo, uint32_t *tiling_flags,
 			  uint32_t *pitch);
+    int (*bo_is_busy)(struct radeon_bo *bo, uint32_t *domain);
 };
 
 struct radeon_bo_manager {
@@ -170,6 +171,15 @@ static inline int _radeon_bo_wait(struct radeon_bo *bo,
     return bo->bom->funcs->bo_wait(bo);
 }
 
+static inline int _radeon_bo_is_busy(struct radeon_bo *bo,
+				     uint32_t *domain,
+                                     const char *file,
+                                     const char *func,
+                                     int line)
+{
+    return bo->bom->funcs->bo_is_busy(bo, domain);
+}
+
 static inline int radeon_bo_set_tiling(struct radeon_bo *bo,
 				       uint32_t tiling_flags, uint32_t pitch)
 {
@@ -203,5 +213,7 @@ static inline int radeon_bo_is_static(struct radeon_bo *bo)
     _radeon_bo_debug(bo, opcode, __FILE__, __FUNCTION__, __LINE__)
 #define radeon_bo_wait(bo) \
     _radeon_bo_wait(bo, __FILE__, __func__, __LINE__)
+#define radeon_bo_is_busy(bo, domain) \
+    _radeon_bo_is_busy(bo, domain, __FILE__, __func__, __LINE__)
 
 #endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c
index b1cc155f71a..a10c6b73abc 100644
--- a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c
+++ b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c
@@ -542,6 +542,18 @@ static int bo_unmap(struct radeon_bo *bo)
     return 0;
 }
 
+static int bo_is_busy(struct radeon_bo *bo, uint32_t *domain)
+{
+    *domain = 0;
+    if (bo->domains & RADEON_GEM_DOMAIN_GTT)
+        *domain = RADEON_GEM_DOMAIN_GTT;
+    else
+        *domain = RADEON_GEM_DOMAIN_CPU;
+    if (legacy_is_pending(bo))
+        return -EBUSY;
+    else
+        return 0;
+}
 
 static int bo_is_static(struct radeon_bo *bo)
 {
@@ -559,6 +571,7 @@ static struct radeon_bo_funcs bo_legacy_funcs = {
     bo_is_static,
     NULL,
     NULL,
+    bo_is_busy
 };
 
 static int bo_vram_validate(struct radeon_bo *bo,
diff --git a/src/mesa/drivers/dri/radeon/radeon_chipset.h b/src/mesa/drivers/dri/radeon/radeon_chipset.h
index a275c8fb143..0da1c0f9ee3 100644
--- a/src/mesa/drivers/dri/radeon/radeon_chipset.h
+++ b/src/mesa/drivers/dri/radeon/radeon_chipset.h
@@ -335,6 +335,12 @@
 #define PCI_CHIP_RS780_9615             0x9615
 #define PCI_CHIP_RS780_9616             0x9616
 
+#define PCI_CHIP_RS880_9710             0x9710
+#define PCI_CHIP_RS880_9711             0x9711
+#define PCI_CHIP_RS880_9712             0x9712
+#define PCI_CHIP_RS880_9713             0x9713
+#define PCI_CHIP_RS880_9714             0x9714
+
 #define PCI_CHIP_RV770_9440             0x9440
 #define PCI_CHIP_RV770_9441             0x9441
 #define PCI_CHIP_RV770_9442             0x9442
@@ -421,6 +427,7 @@ enum {
    CHIP_FAMILY_RV620,
    CHIP_FAMILY_RV635,
    CHIP_FAMILY_RS780,
+   CHIP_FAMILY_RS880,
    CHIP_FAMILY_RV770,
    CHIP_FAMILY_RV730,
    CHIP_FAMILY_RV710,
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c
index ad4584a2bde..ef296e491eb 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c
@@ -92,6 +92,7 @@ static const char* get_chip_family_name(int chip_family)
 	case CHIP_FAMILY_RV620: return "RV620";
 	case CHIP_FAMILY_RV635: return "RV635";
 	case CHIP_FAMILY_RS780: return "RS780";
+	case CHIP_FAMILY_RS880: return "RS880";
 	case CHIP_FAMILY_RV770: return "RV770";
 	case CHIP_FAMILY_RV730: return "RV730";
 	case CHIP_FAMILY_RV710: return "RV710";
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_drm.h b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h
index ee403d173cc..ab4eca31a3c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_cs_drm.h
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h
@@ -201,6 +201,15 @@ static inline void radeon_cs_write_qword(struct radeon_cs *cs, uint64_t qword)
     }
 }
 
+static inline void radeon_cs_write_table(struct radeon_cs *cs, void *data, uint32_t size)
+{
+    memcpy(cs->packets + cs->cdw, data, size * 4);
+    cs->cdw += size;
+    if (cs->section) {
+	    cs->section_cdw += size;
+    }
+}
+
 static inline void radeon_cs_space_set_flush(struct radeon_cs *cs, void (*fn)(void *), void *data)
 {
     cs->space_flush_fn = fn;
diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c
index 7e8f50e3df2..f306befec47 100644
--- a/src/mesa/drivers/dri/radeon/radeon_dma.c
+++ b/src/mesa/drivers/dri/radeon/radeon_dma.c
@@ -30,6 +30,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 **************************************************************************/
 
+#include <errno.h>
 #include "radeon_common.h"
 #include "main/simple_list.h"
 
@@ -302,7 +303,13 @@ void radeonReturnDmaRegion(radeonContextPtr rmesa, int return_bytes)
 
 static int radeon_bo_is_idle(struct radeon_bo* bo)
 {
-	return bo->cref == 1;
+	uint32_t domain;
+	int ret = radeon_bo_is_busy(bo, &domain);
+	if (ret == -EINVAL) {
+		WARN_ONCE("Your libdrm or kernel doesn't have support for busy query.\n"
+			"This may cause small performance drop for you.\n");
+	}
+	return ret != -EBUSY;
 }
 
 void radeonReleaseDmaRegions(radeonContextPtr rmesa)
diff --git a/src/mesa/drivers/dri/radeon/radeon_queryobj.c b/src/mesa/drivers/dri/radeon/radeon_queryobj.c
index 70251946dff..7eef4faaf61 100644
--- a/src/mesa/drivers/dri/radeon/radeon_queryobj.c
+++ b/src/mesa/drivers/dri/radeon/radeon_queryobj.c
@@ -32,7 +32,19 @@
 
 #define DDEBUG 0
 
-#define PAGE_SIZE 4096
+static int radeonQueryIsFlushed(GLcontext *ctx, struct gl_query_object *q)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct radeon_query_object *tmp, *query = (struct radeon_query_object *)q;
+
+	foreach(tmp, &radeon->query.not_flushed_head) {
+		if (tmp == query) {
+			return 0;
+		}
+	}
+
+	return 1;
+}
 
 static void radeonQueryGetResult(GLcontext *ctx, struct gl_query_object *q)
 {
@@ -86,22 +98,11 @@ static void radeonDeleteQuery(GLcontext *ctx, struct gl_query_object *q)
 
 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;
+	struct radeon_query_object *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 (!radeonQueryIsFlushed(ctx, q))
+		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);
 
@@ -124,7 +125,7 @@ static void radeonBeginQuery(GLcontext *ctx, struct gl_query_object *q)
 		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->bo = radeon_bo_open(radeon->radeonScreen->bom, 0, RADEON_QUERY_PAGE_SIZE, RADEON_QUERY_PAGE_SIZE, RADEON_GEM_DOMAIN_GTT, 0);
 	}
 	query->curr_offset = 0;
 
@@ -168,16 +169,32 @@ static void radeonEndQuery(GLcontext *ctx, struct gl_query_object *q)
 	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);
 
+#ifdef DRM_RADEON_GEM_BUSY
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+
+	if (radeon->radeonScreen->kernel_mm) {
+		struct radeon_query_object *query = (struct radeon_query_object *)q;
+		uint32_t domain;
+
+		/* Need to perform a flush, as per ARB_occlusion_query spec */
+		if (!radeonQueryIsFlushed(ctx, q)) {
+			ctx->Driver.Flush(ctx);
+		}
+
+		if (radeon_bo_is_busy(query->bo, &domain) == 0) {
+			radeonQueryGetResult(ctx, q);
+			query->Base.Ready = GL_TRUE;
+		}
+	} else {
+		radeonWaitQuery(ctx, q);
+	}
+#else
 	radeonWaitQuery(ctx, q);
+#endif
 }
 
 void radeonInitQueryObjFunctions(struct dd_function_table *functions)
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c
index e28543d855a..10afe527d3d 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.c
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.c
@@ -853,6 +853,14 @@ static int radeon_set_screen_flags(radeonScreenPtr screen, int device_id)
       screen->chip_family = CHIP_FAMILY_RS780;
       screen->chip_flags = RADEON_CHIPSET_TCL;
       break;
+   case PCI_CHIP_RS880_9710:
+   case PCI_CHIP_RS880_9711:
+   case PCI_CHIP_RS880_9712:
+   case PCI_CHIP_RS880_9713:
+   case PCI_CHIP_RS880_9714:
+      screen->chip_family = CHIP_FAMILY_RS880;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
 
    case PCI_CHIP_RV770_9440:
    case PCI_CHIP_RV770_9441:
diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c
index e70dbbe06f2..bed3935130b 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state_init.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c
@@ -876,7 +876,7 @@ void radeonInitState( r100ContextPtr rmesa )
       rmesa->hw.glt.emit = vec_emit;
       rmesa->hw.eye.emit = vec_emit;
       
-      for (i = 0; i <= 6; i++)
+      for (i = 0; i < 6; i++)
 	 rmesa->hw.mat[i].emit = vec_emit;
 
       for (i = 0; i < 8; i++)