add blend state

6 files changed, 196 insertions, 21 deletions

diff --git a/r600_atom.c b/r600_atom.c
index aa986ea..3d20503 100644
--- a/r600_atom.c
+++ b/r600_atom.c
@@ -84,18 +84,17 @@ static int r600_cb_emit(struct radeon_device *rdev,
 	cb->pkts[16] = radeon_ib_reloc(ib, cb->bo, cb->placements[0] | cb->placements[1]);
 	cb->pkts[21] = radeon_ib_reloc(ib, cb->bo, cb->placements[0] | cb->placements[1]);
 	cb->pkts[26] = radeon_ib_reloc(ib, cb->bo, cb->placements[0] | cb->placements[1]);
-	cb->pkts[259] = radeon_ib_reloc(ib, cb->vsshader1, RADEON_GEM_DOMAIN_GTT);
-	cb->pkts[264] = radeon_ib_reloc(ib, cb->vsshader1, RADEON_GEM_DOMAIN_GTT);
-	cb->pkts[277] = radeon_ib_reloc(ib, cb->vsshader1, RADEON_GEM_DOMAIN_GTT);
-	cb->pkts[282] = radeon_ib_reloc(ib, cb->vsshader1, RADEON_GEM_DOMAIN_GTT);
-	cb->pkts[298] = radeon_ib_reloc(ib, cb->psshader1, RADEON_GEM_DOMAIN_GTT);
-	cb->pkts[303] = radeon_ib_reloc(ib, cb->psshader1, RADEON_GEM_DOMAIN_GTT);
-	cb->pkts[346] = radeon_ib_reloc(ib, cb->vbo1, RADEON_GEM_DOMAIN_GTT);
-	cb->pkts[357] = radeon_ib_reloc(ib, cb->vbo1, RADEON_GEM_DOMAIN_GTT);
-	cb->pkts[364] = radeon_ib_reloc(ib, cb->vbo1, RADEON_GEM_DOMAIN_GTT);
-	cb->pkts[375] = radeon_ib_reloc(ib, cb->vbo1, RADEON_GEM_DOMAIN_GTT);
-	cb->pkts[397] = radeon_ib_reloc(ib, cb->bo, cb->placements[0] | cb->placements[1]);
-
+	cb->pkts[250] = radeon_ib_reloc(ib, cb->vsshader1, RADEON_GEM_DOMAIN_GTT);
+	cb->pkts[255] = radeon_ib_reloc(ib, cb->vsshader1, RADEON_GEM_DOMAIN_GTT);
+	cb->pkts[268] = radeon_ib_reloc(ib, cb->vsshader1, RADEON_GEM_DOMAIN_GTT);
+	cb->pkts[273] = radeon_ib_reloc(ib, cb->vsshader1, RADEON_GEM_DOMAIN_GTT);
+	cb->pkts[289] = radeon_ib_reloc(ib, cb->psshader1, RADEON_GEM_DOMAIN_GTT);
+	cb->pkts[294] = radeon_ib_reloc(ib, cb->psshader1, RADEON_GEM_DOMAIN_GTT);
+	cb->pkts[337] = radeon_ib_reloc(ib, cb->vbo1, RADEON_GEM_DOMAIN_GTT);
+	cb->pkts[348] = radeon_ib_reloc(ib, cb->vbo1, RADEON_GEM_DOMAIN_GTT);
+	cb->pkts[355] = radeon_ib_reloc(ib, cb->vbo1, RADEON_GEM_DOMAIN_GTT);
+	cb->pkts[366] = radeon_ib_reloc(ib, cb->vbo1, RADEON_GEM_DOMAIN_GTT);
+	cb->pkts[388] = radeon_ib_reloc(ib, cb->bo, cb->placements[0] | cb->placements[1]);
 	r = radeon_ib_copy(ib, cb->pkts, atom->npkts);
 	return r;
 }
@@ -655,6 +654,70 @@ out_err:
 	return r;
 }
 
+/*
+ * r600_blend
+ */
+struct r600_blend {
+	struct radeon_atom	atom;
+	u32			pkts[16];
+	struct r600_atoms	*atoms;
+};
+
+static void r600_blend_release(struct kref *kref)
+{
+	struct radeon_atom *atom = container_of(kref, struct radeon_atom, kref);
+	struct r600_blend *blend = container_of(atom, struct r600_blend, atom);
+
+	mutex_lock(&blend->atoms->mutex);
+	list_del_init(&blend->atom.list);
+	mutex_unlock(&blend->atoms->mutex);
+	kfree(blend);
+}
+
+static int r600_blend_create(struct radeon_device *rdev,
+				struct r600_atoms *atoms,
+				struct drm_radeon_atom *patom,
+				struct radeon_atom **atom)
+{
+	struct drm_r600_blend pblend;
+	struct r600_blend *blend;
+	int r;
+
+	blend = kmalloc(sizeof(*blend), GFP_KERNEL);
+	if (blend == NULL)
+		return -ENOMEM;
+	/* make sure structure properly initialized */
+	memset(blend, 0, sizeof(*blend));
+	r = radeon_atom_init(&blend->atom, &atoms->idr, &r600_blend_release,
+			&radeon_atom_emit_default, &r600_atom_process_default,
+			blend->pkts);
+	if (r)
+		goto out_err;
+	/* KERNEL use get user data */
+	memcpy(&pblend, (void*)(unsigned long)patom->data, sizeof(struct drm_r600_blend));
+	/* CB_BLEND0_CONTROL */
+	blend->pkts[blend->atom.npkts++] = PKT3(PKT3_SET_CONTEXT_REG, 8);
+	blend->pkts[blend->atom.npkts++] = 0x000001E0;
+	blend->pkts[blend->atom.npkts++] = pblend.cb_blend0_control;
+	blend->pkts[blend->atom.npkts++] = pblend.cb_blend1_control;
+	blend->pkts[blend->atom.npkts++] = pblend.cb_blend2_control;
+	blend->pkts[blend->atom.npkts++] = pblend.cb_blend3_control;
+	blend->pkts[blend->atom.npkts++] = pblend.cb_blend4_control;
+	blend->pkts[blend->atom.npkts++] = pblend.cb_blend5_control;
+	blend->pkts[blend->atom.npkts++] = pblend.cb_blend6_control;
+	blend->pkts[blend->atom.npkts++] = pblend.cb_blend7_control;
+	/* CB_BLEND_CONTROL */
+	blend->pkts[blend->atom.npkts++] = PKT3(PKT3_SET_CONTEXT_REG, 1);
+	blend->pkts[blend->atom.npkts++] = 0x00000201;
+	blend->pkts[blend->atom.npkts++] = pblend.cb_blend_control;
+	*atom = &blend->atom;
+fprintf(stderr, "%s %d pkts\n", __func__, (*atom)->npkts);
+	return 0;
+out_err:
+	radeon_atom_put(&blend->atom);
+	*atom = NULL;
+	return r;
+}
 
 /*
  * r600 atom core functions
@@ -704,6 +767,13 @@ int r600_atom_create(struct radeon_device *rdev,
 		atom->type = patom->type;
 		list_add_tail(&atom->list, &atoms->vport_atoms);
 		break;
+	case R600_ATOM_BLEND:
+		r = r600_blend_create(rdev, atoms, patom, &atom);
+		if (r)
+			return r;
+		atom->type = patom->type;
+		list_add_tail(&atom->list, &atoms->blend_atoms);
+		break;
 	default:
 		dev_err(rdev->dev, "unknown R600 atom type 0x%08X\n", patom->type);
 		return -EINVAL;
@@ -727,7 +797,8 @@ static int r600_batch_alloc(struct r600_batch **batch)
 	INIT_LIST_HEAD(&rbatch->list);
 	INIT_LIST_HEAD(&rbatch->pre_flushes);
 	INIT_LIST_HEAD(&rbatch->post_flushes);
-	rbatch->npkts = 0;
+	/* blend color (6dw) + clear color (6dw) + fog color (5 dw) */
+	rbatch->npkts = 17;
 	*batch = rbatch;
 	return 0;
 }
@@ -756,6 +827,23 @@ static int r600_batches_flush_locked(struct radeon_device *rdev, struct r600_bat
 	struct r600_batch *batch;
 	int r, i;
 
+	batches->ib->ptr[batches->ib->cpkts++] = PKT3(PKT3_SET_CONTEXT_REG, 4);
+	batches->ib->ptr[batches->ib->cpkts++] = 0x00000048;
+	batches->ib->ptr[batches->ib->cpkts++] = batch->cb_clear_red;
+	batches->ib->ptr[batches->ib->cpkts++] = batch->cb_clear_green;
+	batches->ib->ptr[batches->ib->cpkts++] = batch->cb_clear_blue;
+	batches->ib->ptr[batches->ib->cpkts++] = batch->cb_clear_alpha;
+	batches->ib->ptr[batches->ib->cpkts++] = PKT3(PKT3_SET_CONTEXT_REG, 4);
+	batches->ib->ptr[batches->ib->cpkts++] = 0x00000105;
+	batches->ib->ptr[batches->ib->cpkts++] = batch->cb_blend_red;
+	batches->ib->ptr[batches->ib->cpkts++] = batch->cb_blend_green;
+	batches->ib->ptr[batches->ib->cpkts++] = batch->cb_blend_blue;
+	batches->ib->ptr[batches->ib->cpkts++] = batch->cb_blend_alpha;
+	batches->ib->ptr[batches->ib->cpkts++] = PKT3(PKT3_SET_CONTEXT_REG, 3);
+	batches->ib->ptr[batches->ib->cpkts++] = 0x00000109;
+	batches->ib->ptr[batches->ib->cpkts++] = batch->cb_fog_red;
+	batches->ib->ptr[batches->ib->cpkts++] = batch->cb_fog_green;
+	batches->ib->ptr[batches->ib->cpkts++] = batch->cb_fog_blue;
 	list_for_each_entry(batch, &batches->batches, list) {
 		for (i = 0; i < R600_BATCH_NATOMS; i++) {
 			r = batch->atoms[i]->emit(rdev, batch->atoms[i], batch, batches->ib);
@@ -784,7 +872,22 @@ int r600_batches_queue(struct radeon_device *rdev,
 	if (r)
 		return r;
 	mutex_lock(&atoms->mutex);
+	rbatch->cb_clear_alpha = batch->cb_clear_alpha;
+	rbatch->cb_clear_blue = batch->cb_clear_blue;
+	rbatch->cb_clear_green = batch->cb_clear_green;
+	rbatch->cb_clear_red = batch->cb_clear_red;
+	rbatch->cb_blend_alpha = batch->cb_blend_alpha;
+	rbatch->cb_blend_blue = batch->cb_blend_blue;
+	rbatch->cb_blend_green = batch->cb_blend_green;
+	rbatch->cb_blend_red = batch->cb_blend_red;
+	rbatch->cb_fog_blue = batch->cb_fog_blue;
+	rbatch->cb_fog_green = batch->cb_fog_green;
+	rbatch->cb_fog_red = batch->cb_fog_red;
 	i = 0;
+	r = radeon_atom_find_locked(&atoms->blend_atoms, batch->blend_id,
+					R600_ATOM_BLEND, &rbatch->atoms[i++]);
+	if (r)
+		goto out_err;
 	r = radeon_atom_find_locked(&atoms->cb_cntl_atoms, batch->cb_cntl_id,
 					R600_ATOM_CB_CNTL, &rbatch->atoms[i++]);
 	if (r)
@@ -875,6 +978,7 @@ int r600_batches_flush(struct radeon_device *rdev, struct r600_atoms *atoms)
 int r600_atoms_init(struct radeon_device *rdev, struct r600_atoms *atoms)
 {
 	mutex_init(&atoms->mutex);
+	INIT_LIST_HEAD(&atoms->blend_atoms);
 	INIT_LIST_HEAD(&atoms->cb_atoms);
 	INIT_LIST_HEAD(&atoms->cb_cntl_atoms);
 	INIT_LIST_HEAD(&atoms->pa_atoms);
@@ -986,6 +1090,7 @@ void r600_tflat(struct radeon_atom *atom)
     WPKT(PKT3(PKT3_SET_CONTEXT_REG, 1));
     WPKT(0x000001E8);
     WPKT(0x00000003);
+#if 0
     WPKT(PKT3(PKT3_SET_CONTEXT_REG, 1));
     WPKT(0x000001E0);
     WPKT(0x00010001);
@@ -995,6 +1100,7 @@ void r600_tflat(struct radeon_atom *atom)
     WPKT(0x00000000);
     WPKT(0x00000000);
     WPKT(0x00000000);
+#endif
     WPKT(PKT3(PKT3_SET_CONTEXT_REG, 1));
     WPKT(0x000000D4);
     WPKT(0x00000000);
diff --git a/r600_atom_api.h b/r600_atom_api.h
index 8b2522c..16938c7 100644
--- a/r600_atom_api.h
+++ b/r600_atom_api.h
@@ -32,6 +32,7 @@ struct drm_radeon_atom {
 #define R600_ATOM_PA		3
 #define R600_ATOM_CB_CNTL	4
 #define R600_ATOM_VPORT		5
+#define R600_ATOM_BLEND		6
 
 struct drm_r600_cb {
 	u32			pitch;
@@ -114,12 +115,37 @@ struct drm_r600_vport {
 	u32			pa_sc_generic_scissor_br;
 };
 
+/* blend - blending */
+struct drm_r600_blend {
+	u32			cb_blend0_control;
+	u32			cb_blend1_control;
+	u32			cb_blend2_control;
+	u32			cb_blend3_control;
+	u32			cb_blend4_control;
+	u32			cb_blend5_control;
+	u32			cb_blend6_control;
+	u32			cb_blend7_control;
+	u32			cb_blend_control;
+};
+
 struct drm_r600_batch {
+	u32			blend_id;
 	u32			cb_id;
 	u32			cb_cntl_id;
 	u32			pa_id;
 	u32			tp_id;
 	u32			vport_id;
+	u32			cb_blend_alpha;
+	u32			cb_blend_blue;
+	u32			cb_blend_green;
+	u32			cb_blend_red;
+	u32			cb_clear_alpha;
+	u32			cb_clear_blue;
+	u32			cb_clear_green;
+	u32			cb_clear_red;
+	u32			cb_fog_blue;
+	u32			cb_fog_green;
+	u32			cb_fog_red;
 };
 
 
diff --git a/radeon_atom.h b/radeon_atom.h
index 2491113..9f696ba 100644
--- a/radeon_atom.h
+++ b/radeon_atom.h
@@ -53,13 +53,24 @@ struct radeon_atom {
 };
 
 /* R600 */
-#define R600_BATCH_NATOMS	5
+#define R600_BATCH_NATOMS	6
 struct r600_batch {
 	struct list_head	list;
 	struct list_head	pre_flushes;
 	struct list_head	post_flushes;
 	struct radeon_atom	*atoms[R600_BATCH_NATOMS];
 	u32			npkts;
+	u32			cb_blend_alpha;
+	u32			cb_blend_blue;
+	u32			cb_blend_green;
+	u32			cb_blend_red;
+	u32			cb_clear_alpha;
+	u32			cb_clear_blue;
+	u32			cb_clear_green;
+	u32			cb_clear_red;
+	u32			cb_fog_blue;
+	u32			cb_fog_green;
+	u32			cb_fog_red;
 };
 
 struct r600_batches {
@@ -70,6 +81,7 @@ struct r600_batches {
 };
 
 struct r600_atoms {
+	struct list_head	blend_atoms;
 	struct list_head	cb_atoms;
 	struct list_head	cb_cntl_atoms;
 	struct list_head	pa_atoms;
diff --git a/radeon_device.c b/radeon_device.c
index 3ff29c2..21ec4d6 100644
--- a/radeon_device.c
+++ b/radeon_device.c
@@ -57,8 +57,8 @@ int radeon_ib_get(struct radeon_device *rdev, struct radeon_ib **ib)
 	if (lib == NULL)
 		return -ENOMEM;
 	memset(lib, sizeof(*lib), 0);
-	lib->pkts = malloc(64 * 1024);
-	if (lib->pkts == NULL) {
+	lib->ptr = malloc(64 * 1024);
+	if (lib->ptr == NULL) {
 		free(lib);
 		return -ENOMEM;
 	}
@@ -66,7 +66,7 @@ int radeon_ib_get(struct radeon_device *rdev, struct radeon_ib **ib)
 	lib->length_dw = 64 * 1024 / 4;
 	lib->relocs = malloc(64 * 1024);
 	if (lib->relocs == NULL) {
-		free(lib->pkts);
+		free(lib->ptr);
 		free(lib);
 		return -ENOMEM;
 	}
@@ -80,7 +80,7 @@ void radeon_ib_free(struct radeon_ib *ib)
 {
 	if (ib == NULL)
 		return;
-	free(ib->pkts);
+	free(ib->ptr);
 	free(ib->relocs);
 	free(ib);
 }
@@ -96,7 +96,7 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib)
 	drmib.chunks = (uint64_t)(uintptr_t)chunk_array;
 	chunks[0].chunk_id = RADEON_CHUNK_ID_IB;
 	chunks[0].length_dw = ib->cpkts;
-	chunks[0].chunk_data = (uint64_t)(uintptr_t)ib->pkts;
+	chunks[0].chunk_data = (uint64_t)(uintptr_t)ib->ptr;
 	chunks[1].chunk_id = RADEON_CHUNK_ID_RELOCS;
 	chunks[1].length_dw = ib->crelocs * 4;
 	chunks[1].chunk_data = (uint64_t)(uintptr_t)ib->relocs;
diff --git a/radeon_device.h b/radeon_device.h
index 036e176..c0e2ba8 100644
--- a/radeon_device.h
+++ b/radeon_device.h
@@ -44,7 +44,7 @@ struct radeon_device {
 };
 
 struct radeon_ib {
-	u32	*pkts;
+	u32	*ptr;
 	u32	cpkts;
 	u32	length_dw;
 	u32	*relocs;
@@ -68,7 +68,7 @@ static inline int radeon_ib_copy(struct radeon_ib *ib, u32 *pkts, u32 ndw)
 {
 	if ((ib->cpkts + ndw) > ib->length_dw)
 		return -ENOMEM;
-	memcpy(&ib->pkts[ib->cpkts], pkts, ndw * 4);
+	memcpy(&ib->ptr[ib->cpkts], pkts, ndw * 4);
 	ib->cpkts += ndw;
 	return 0;
 }
diff --git a/test.c b/test.c
index 1503bfb..5bda074 100644
--- a/test.c
+++ b/test.c
@@ -47,6 +47,7 @@ int main(void)
 int r600_tri_flat(struct radeon *radeon)
 {
 	struct radeon_device *rdev;
+	struct drm_r600_blend blend;
 	struct drm_r600_cb cb;
 	struct drm_r600_cb_cntl cb_cntl;
 	struct drm_r600_pa pa;
@@ -163,6 +164,36 @@ int r600_tri_flat(struct radeon *radeon)
 	if (r)
 		return r;
 	batch.vport_id = atom.id;
+	/* blend */
+	blend.cb_blend0_control = 0x00010001;
+	blend.cb_blend1_control = 0x00000000;
+	blend.cb_blend2_control = 0x00000000;
+	blend.cb_blend3_control = 0x00000000;
+	blend.cb_blend4_control = 0x00000000;
+	blend.cb_blend5_control = 0x00000000;
+	blend.cb_blend6_control = 0x00000000;
+	blend.cb_blend7_control = 0x00000000;
+	blend.cb_blend_control = 0x00010001;
+	atom.type = R600_ATOM_BLEND;
+	atom.id = 0;
+	atom.data = (uint64_t)(uintptr_t)&blend;
+	r = radeon_atom_create(rdev, &atom);
+	if (r)
+		return r;
+	batch.blend_id = atom.id;
+
+	/* batch */
+	batch.cb_clear_alpha = 0x00000000;
+	batch.cb_clear_blue = 0x00000000;
+	batch.cb_clear_green = 0x00000000;
+	batch.cb_clear_red = 0x00000000;
+	batch.cb_blend_alpha = 0x00000000;
+	batch.cb_blend_blue = 0x00000000;
+	batch.cb_blend_green = 0x00000000;
+	batch.cb_blend_red = 0x00000000;
+	batch.cb_fog_blue = 0x00000000;
+	batch.cb_fog_green = 0x00000000;
+	batch.cb_fog_red = 0x00000000;
 
 	r =  radeon_batches_queue(rdev, &batch);
 	if (r)