radeonsi: move sampler descriptors from IB to memory

Sampler descriptors are now represented by si_descriptors.
This also adds support for fine-grained sampler state updates and
the border color update is now isolated in a separate function.

Border colors have been broken if texturing from multiple shader stages is
used. This patch doesn't change that.

BTW, blitting already makes use of fine-grained state updates.
u_blitter uses 2 textures at most, so we only have to save 2.

Reviewed-by: Michel Dänzer <[email protected]>

5 files changed, 82 insertions, 82 deletions

diff --git a/src/gallium/drivers/radeonsi/si_blit.c b/src/gallium/drivers/radeonsi/si_blit.c
index 072024aa323..a76d9052494 100644
--- a/src/gallium/drivers/radeonsi/si_blit.c
+++ b/src/gallium/drivers/radeonsi/si_blit.c
@@ -71,8 +71,8 @@ static void si_blitter_begin(struct pipe_context *ctx, enum si_blitter_op op)
 
 	if (op & SI_SAVE_TEXTURES) {
 		util_blitter_save_fragment_sampler_states(
-			sctx->blitter, sctx->samplers[PIPE_SHADER_FRAGMENT].n_samplers,
-			(void**)sctx->samplers[PIPE_SHADER_FRAGMENT].samplers);
+			sctx->blitter, 2,
+			sctx->samplers[PIPE_SHADER_FRAGMENT].states.saved_states);
 
 		util_blitter_save_fragment_sampler_views(sctx->blitter,
 			util_last_bit(sctx->samplers[PIPE_SHADER_FRAGMENT].views.desc.enabled_mask &
diff --git a/src/gallium/drivers/radeonsi/si_descriptors.c b/src/gallium/drivers/radeonsi/si_descriptors.c
index 4dd831a7aad..01fcec3e13d 100644
--- a/src/gallium/drivers/radeonsi/si_descriptors.c
+++ b/src/gallium/drivers/radeonsi/si_descriptors.c
@@ -363,6 +363,52 @@ void si_set_sampler_view(struct si_context *sctx, unsigned shader,
 	si_update_descriptors(sctx, &views->desc);
 }
 
+/* SAMPLER STATES */
+
+static void si_emit_sampler_states(struct si_context *sctx, struct r600_atom *atom)
+{
+	struct si_sampler_states *states = (struct si_sampler_states*)atom;
+
+	si_emit_descriptors(sctx, &states->desc, states->desc_data);
+}
+
+static void si_sampler_states_begin_new_cs(struct si_context *sctx,
+					   struct si_sampler_states *states)
+{
+	r600_context_bo_reloc(&sctx->b, &sctx->b.rings.gfx, states->desc.buffer,
+			      RADEON_USAGE_READWRITE, RADEON_PRIO_SHADER_DATA);
+	si_emit_shader_pointer(sctx, &states->desc);
+}
+
+void si_set_sampler_descriptors(struct si_context *sctx, unsigned shader,
+				unsigned start, unsigned count, void **states)
+{
+	struct si_sampler_states *samplers = &sctx->samplers[shader].states;
+	struct si_pipe_sampler_state **sstates = (struct si_pipe_sampler_state**)states;
+	int i;
+
+	if (start == 0)
+		samplers->saved_states[0] = states[0];
+	if (start == 1)
+		samplers->saved_states[1] = states[0];
+	else if (start == 0 && count >= 2)
+		samplers->saved_states[1] = states[1];
+
+	for (i = 0; i < count; i++) {
+		unsigned slot = start + i;
+
+		if (!sstates[i]) {
+			samplers->desc.dirty_mask &= ~(1 << slot);
+			continue;
+		}
+
+		samplers->desc_data[slot] = sstates[i]->val;
+		samplers->desc.dirty_mask |= 1 << slot;
+	}
+
+	si_update_descriptors(sctx, &samplers->desc);
+}
+
 /* BUFFER RESOURCES */
 
 static void si_emit_buffer_resources(struct si_context *sctx, struct r600_atom *atom)
@@ -985,9 +1031,14 @@ void si_init_all_descriptors(struct si_context *sctx)
 
 		si_init_sampler_views(sctx, &sctx->samplers[i].views, i);
 
+		si_init_descriptors(sctx, &sctx->samplers[i].states.desc,
+				    si_get_shader_user_data_base(i) + SI_SGPR_SAMPLER * 4,
+				    4, SI_NUM_SAMPLER_STATES, si_emit_sampler_states);
+
 		sctx->atoms.s.const_buffers[i] = &sctx->const_buffers[i].desc.atom;
 		sctx->atoms.s.rw_buffers[i] = &sctx->rw_buffers[i].desc.atom;
 		sctx->atoms.s.sampler_views[i] = &sctx->samplers[i].views.desc.atom;
+		sctx->atoms.s.sampler_states[i] = &sctx->samplers[i].states.desc.atom;
 	}
 
 
@@ -1006,6 +1057,7 @@ void si_release_all_descriptors(struct si_context *sctx)
 		si_release_buffer_resources(&sctx->const_buffers[i]);
 		si_release_buffer_resources(&sctx->rw_buffers[i]);
 		si_release_sampler_views(&sctx->samplers[i].views);
+		si_release_descriptors(&sctx->samplers[i].states.desc);
 	}
 }
 
@@ -1017,5 +1069,6 @@ void si_all_descriptors_begin_new_cs(struct si_context *sctx)
 		si_buffer_resources_begin_new_cs(sctx, &sctx->const_buffers[i]);
 		si_buffer_resources_begin_new_cs(sctx, &sctx->rw_buffers[i]);
 		si_sampler_views_begin_new_cs(sctx, &sctx->samplers[i].views);
+		si_sampler_states_begin_new_cs(sctx, &sctx->samplers[i].states);
 	}
 }
diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index c95774d6f5e..901beb279a0 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -63,11 +63,10 @@ struct si_cs_shader_state {
 
 struct si_textures_info {
 	struct si_sampler_views		views;
-	struct si_pipe_sampler_state	*samplers[SI_NUM_USER_SAMPLERS];
+	struct si_sampler_states	states;
 	unsigned			n_views;
 	uint32_t			depth_texture_mask; /* which textures are depth */
 	uint32_t			compressed_colortex_mask;
-	unsigned			n_samplers;
 };
 
 struct si_framebuffer {
@@ -102,6 +101,7 @@ struct si_context {
 			struct r600_atom *const_buffers[SI_NUM_SHADERS];
 			struct r600_atom *rw_buffers[SI_NUM_SHADERS];
 			struct r600_atom *sampler_views[SI_NUM_SHADERS];
+			struct r600_atom *sampler_states[SI_NUM_SHADERS];
 			/* Caches must be flushed after resource descriptors are
 			 * updated in memory. */
 			struct r600_atom *cache_flush;
diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index c106747a67d..44c1af844fc 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -2738,22 +2738,18 @@ static void si_set_sampler_views(struct pipe_context *ctx,
 	sctx->b.flags |= R600_CONTEXT_INV_TEX_CACHE;
 }
 
-static void si_set_sampler_states(struct si_context *sctx,
-				  struct si_pm4_state *pm4,
-				  unsigned count, void **states,
-				  struct si_textures_info *samplers,
-				  unsigned user_data_reg)
+/* Upload border colors and update the pointers in resource descriptors.
+ * There can only be 4096 border colors per context.
+ *
+ * XXX: This is broken if the buffer gets reallocated.
+ */
+static void si_set_border_colors(struct si_context *sctx, unsigned count,
+				 void **states)
 {
 	struct si_pipe_sampler_state **rstates = (struct si_pipe_sampler_state **)states;
 	uint32_t *border_color_table = NULL;
 	int i, j;
 
-	if (!count)
-		goto out;
-
-	sctx->b.flags |= R600_CONTEXT_INV_TEX_CACHE;
-
-	si_pm4_sh_data_begin(pm4);
 	for (i = 0; i < count; i++) {
 		if (rstates[i] &&
 		    G_008F3C_BORDER_COLOR_TYPE(rstates[i]->val[3]) ==
@@ -2786,14 +2782,11 @@ static void si_set_sampler_states(struct si_context *sctx,
 			rstates[i]->val[3] &= C_008F3C_BORDER_COLOR_PTR;
 			rstates[i]->val[3] |= S_008F3C_BORDER_COLOR_PTR(sctx->border_color_offset++);
 		}
-
-		for (j = 0; j < Elements(rstates[i]->val); ++j) {
-			si_pm4_sh_data_add(pm4, rstates[i] ? rstates[i]->val[j] : 0);
-		}
 	}
-	si_pm4_sh_data_end(pm4, user_data_reg, SI_SGPR_SAMPLER);
 
 	if (border_color_table) {
+		struct si_pm4_state *pm4 = si_pm4_alloc_state(sctx);
+
 		uint64_t va_offset =
 			r600_resource_va(&sctx->screen->b.b,
 					 (void*)sctx->border_color_table);
@@ -2801,76 +2794,24 @@ static void si_set_sampler_states(struct si_context *sctx,
 		si_pm4_set_reg(pm4, R_028080_TA_BC_BASE_ADDR, va_offset >> 8);
 		if (sctx->b.chip_class >= CIK)
 			si_pm4_set_reg(pm4, R_028084_TA_BC_BASE_ADDR_HI, va_offset >> 40);
-		sctx->b.ws->buffer_unmap(sctx->border_color_table->cs_buf);
 		si_pm4_add_bo(pm4, sctx->border_color_table, RADEON_USAGE_READ,
 			      RADEON_PRIO_SHADER_DATA);
+		si_pm4_set_state(sctx, ta_bordercolor_base, pm4);
 	}
-
-	memcpy(samplers->samplers, states, sizeof(void*) * count);
-
-out:
-	samplers->n_samplers = count;
-}
-
-static void si_bind_vs_sampler_states(struct pipe_context *ctx, unsigned count, void **states)
-{
-	struct si_context *sctx = (struct si_context *)ctx;
-	struct si_pm4_state *pm4 = si_pm4_alloc_state(sctx);
-
-	si_set_sampler_states(sctx, pm4, count, states,
-			      &sctx->samplers[PIPE_SHADER_VERTEX],
-			      R_00B130_SPI_SHADER_USER_DATA_VS_0);
-	si_set_sampler_states(sctx, pm4, count, states,
-			      &sctx->samplers[PIPE_SHADER_VERTEX],
-			      R_00B330_SPI_SHADER_USER_DATA_ES_0);
-	si_pm4_set_state(sctx, vs_sampler, pm4);
-}
-
-static void si_bind_gs_sampler_states(struct pipe_context *ctx, unsigned count, void **states)
-{
-	struct si_context *sctx = (struct si_context *)ctx;
-	struct si_pm4_state *pm4 = si_pm4_alloc_state(sctx);
-
-	si_set_sampler_states(sctx, pm4, count, states,
-			      &sctx->samplers[PIPE_SHADER_GEOMETRY],
-			      R_00B230_SPI_SHADER_USER_DATA_GS_0);
-	si_pm4_set_state(sctx, gs_sampler, pm4);
 }
 
-static void si_bind_ps_sampler_states(struct pipe_context *ctx, unsigned count, void **states)
-{
-	struct si_context *sctx = (struct si_context *)ctx;
-	struct si_pm4_state *pm4 = si_pm4_alloc_state(sctx);
-
-	si_set_sampler_states(sctx, pm4, count, states,
-			      &sctx->samplers[PIPE_SHADER_FRAGMENT],
-			      R_00B030_SPI_SHADER_USER_DATA_PS_0);
-	si_pm4_set_state(sctx, ps_sampler, pm4);
-}
-
-
 static void si_bind_sampler_states(struct pipe_context *ctx, unsigned shader,
                                    unsigned start, unsigned count,
                                    void **states)
 {
-   assert(start == 0);
-
-   switch (shader) {
-   case PIPE_SHADER_VERTEX:
-      si_bind_vs_sampler_states(ctx, count, states);
-      break;
-   case PIPE_SHADER_GEOMETRY:
-      si_bind_gs_sampler_states(ctx, count, states);
-      break;
-   case PIPE_SHADER_FRAGMENT:
-      si_bind_ps_sampler_states(ctx, count, states);
-      break;
-   default:
-      ;
-   }
-}
+	struct si_context *sctx = (struct si_context *)ctx;
 
+	if (!count || shader >= SI_NUM_SHADERS)
+		return;
 
+	si_set_border_colors(sctx, count, states);
+	si_set_sampler_descriptors(sctx, shader, start, count, states);
+}
 
 static void si_set_sample_mask(struct pipe_context *ctx, unsigned sample_mask)
 {
diff --git a/src/gallium/drivers/radeonsi/si_state.h b/src/gallium/drivers/radeonsi/si_state.h
index ae42e6673b5..7a0433c1b9d 100644
--- a/src/gallium/drivers/radeonsi/si_state.h
+++ b/src/gallium/drivers/radeonsi/si_state.h
@@ -89,15 +89,13 @@ union si_state {
 		struct si_pm4_state		*fb_rs;
 		struct si_pm4_state		*fb_blend;
 		struct si_pm4_state		*dsa_stencil_ref;
+		struct si_pm4_state		*ta_bordercolor_base;
 		struct si_pm4_state		*es;
 		struct si_pm4_state		*gs;
 		struct si_pm4_state		*gs_rings;
-		struct si_pm4_state		*gs_sampler;
 		struct si_pm4_state		*gs_onoff;
 		struct si_pm4_state		*vs;
-		struct si_pm4_state		*vs_sampler;
 		struct si_pm4_state		*ps;
-		struct si_pm4_state		*ps_sampler;
 		struct si_pm4_state		*spi;
 		struct si_pm4_state		*vertex_buffers;
 		struct si_pm4_state		*draw_info;
@@ -174,6 +172,12 @@ struct si_sampler_views {
 	uint32_t			*desc_data[SI_NUM_SAMPLER_VIEWS];
 };
 
+struct si_sampler_states {
+	struct si_descriptors		desc;
+	uint32_t			*desc_data[SI_NUM_SAMPLER_STATES];
+	void				*saved_states[2]; /* saved for u_blitter */
+};
+
 struct si_buffer_resources {
 	struct si_descriptors		desc;
 	unsigned			num_buffers;
@@ -218,6 +222,8 @@ struct si_buffer_resources {
 void si_set_sampler_view(struct si_context *sctx, unsigned shader,
 			 unsigned slot, struct pipe_sampler_view *view,
 			 unsigned *view_desc);
+void si_set_sampler_descriptors(struct si_context *sctx, unsigned shader,
+				unsigned start, unsigned count, void **states);
 void si_set_ring_buffer(struct pipe_context *ctx, uint shader, uint slot,
 			struct pipe_constant_buffer *input,
 			unsigned stride, unsigned num_records,