This patch implements the binding table enable command which is also
used to allocate a binding table pool where where hardware-generated
binding table entries are flushed into. Each binding table offset in
the binding table pool is unique per each shader stage that are
enabled within a batch.
Also insert the required brw_tracked_state objects to enable
hw-generated binding tables in normal render path.
v2: Clarify start of binding table pool offsets. (Topi)
Signed-off-by: Abdiel Janulgue <abdiel.janul...@linux.intel.com>
---
src/mesa/drivers/dri/i965/brw_binding_tables.c | 101 +++++++++++++++++++++++++
src/mesa/drivers/dri/i965/brw_context.c | 4 +
src/mesa/drivers/dri/i965/brw_context.h | 6 ++
src/mesa/drivers/dri/i965/brw_state.h | 6 ++
src/mesa/drivers/dri/i965/brw_state_upload.c | 4 +
src/mesa/drivers/dri/i965/intel_batchbuffer.c | 4 +
6 files changed, 125 insertions(+)
diff --git a/src/mesa/drivers/dri/i965/brw_binding_tables.c
b/src/mesa/drivers/dri/i965/brw_binding_tables.c
index 98ff0dd..7554818 100644
--- a/src/mesa/drivers/dri/i965/brw_binding_tables.c
+++ b/src/mesa/drivers/dri/i965/brw_binding_tables.c
@@ -45,6 +45,23 @@
#include "intel_batchbuffer.h"
/**
+ * We are required to start at this offset for binding table pointer state when
+ * HW-generated binding table is enabled otherwise the GPU will hung. Note that
+ * the binding table offsets are now relative to the binding tabe pool base
+ * address instead of from the state batch.
+ *
+ * From the Bspec 3DSTATE_BINDING_TABLE_POINTERS_{PS/VS/GS/DS/HS} > Pointer to
+ * PS Binding Table section lists the format as:
+ *
+ * "SurfaceStateOffset[16:6]BINDING_TABLE_STATE*256 When
+ * HW-generated binding table is enabled"
+ *
+ * When HW-generated binding tables are enabled, Surface State Offsets are
+ * 16-bit entries.
+ */
+static const uint32_t hw_bt_start_offset = 256 * sizeof(uint16_t);
+
+/**
* Upload a shader stage's binding table as indirect state.
*
* This copies brw_stage_state::surf_offset[] into the indirect state section
@@ -170,6 +187,90 @@ const struct brw_tracked_state brw_gs_binding_table = {
.emit = brw_gs_upload_binding_table,
};
+/**
+ * Hardware-generated binding tables for the resource streamer
+ */
+void
+gen7_disable_hw_binding_tables(struct brw_context *brw)
+{
+ int pkt_len = brw->gen >= 8 ? 4 : 3;
+
+ BEGIN_BATCH(3);
+ OUT_BATCH(_3DSTATE_BINDING_TABLE_POOL_ALLOC << 16 | (pkt_len - 2));
+ OUT_BATCH(brw->is_haswell ? HSW_HW_BINDING_TABLE_RESERVED : 0);
+ OUT_BATCH(0);
+ ADVANCE_BATCH();
+
+ /* From the BSpec, 3D Pipeline > Resource Streamer > Hardware Binding
+ * Tables > Programming note
+
+ * "When switching between HW and SW binding table generation, SW must
+ * issue a state cache invalidate."
+ */
+ brw_emit_pipe_control_flush(brw, PIPE_CONTROL_STATE_CACHE_INVALIDATE);
+}
+
+void
+gen7_enable_hw_binding_tables(struct brw_context *brw)
+{
+ if (!brw->has_resource_streamer) {
+ gen7_disable_hw_binding_tables(brw);
+ return;
+ }
+
+ if (!brw->hw_bt_pool.bo) {
+ /* We use a single re-usable buffer object for the lifetime of the
+ * context and size it to maximum allowed binding tables that can be
+ * programmed per batch:
+ *
+ * BSpec, 3D Pipeline > Resource Streamer > Hardware Binding Tables:
+ * "A maximum of 16,383 Binding tables are allowed in any batch buffer"
+ */
+ static const int max_size = 16383 * 4;
+ brw->hw_bt_pool.bo = drm_intel_bo_alloc(brw->bufmgr, "hw_bt",
+ max_size, 64);
+ brw->hw_bt_pool.next_offset = hw_bt_start_offset;
+ }
+
+ int pkt_len = brw->gen >= 8 ? 4 : 3;
+ uint32_t dw1 = BRW_HW_BINDING_TABLE_ENABLE;
+ if (brw->is_haswell)
+ dw1 |= SET_FIELD(GEN7_MOCS_L3, GEN7_HW_BT_MOCS) |
+ HSW_HW_BINDING_TABLE_RESERVED;
+
+ BEGIN_BATCH(3);
+ OUT_BATCH(_3DSTATE_BINDING_TABLE_POOL_ALLOC << 16 | (pkt_len - 2));
+ if (brw->gen >= 8)
+ OUT_RELOC64(brw->hw_bt_pool.bo, I915_GEM_DOMAIN_SAMPLER, 0, dw1);
+ else
+ OUT_RELOC(brw->hw_bt_pool.bo, I915_GEM_DOMAIN_SAMPLER, 0, dw1);
+ OUT_RELOC(brw->hw_bt_pool.bo, I915_GEM_DOMAIN_SAMPLER, 0,
+ brw->hw_bt_pool.bo->size);
+ ADVANCE_BATCH();
+
+ /* From the BSpec, 3D Pipeline > Resource Streamer > Hardware Binding
+ * Tables > Programming note
+
+ * "When switching between HW and SW binding table generation, SW must
+ * issue a state cache invalidate."
+ */
+ brw_emit_pipe_control_flush(brw, PIPE_CONTROL_STATE_CACHE_INVALIDATE);
+}
+
+void
+gen7_reset_rs_pool_offsets(struct brw_context *brw)
+{
+ brw->hw_bt_pool.next_offset = hw_bt_start_offset;
+}
+
+const struct brw_tracked_state gen7_hw_binding_tables = {
+ .dirty = {
+ .mesa = 0,
+ .brw = BRW_NEW_BATCH,
+ },
+ .emit = gen7_enable_hw_binding_tables
+};
+
/** @} */
/**
diff --git a/src/mesa/drivers/dri/i965/brw_context.c
b/src/mesa/drivers/dri/i965/brw_context.c
index 274a237..e17046a 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -958,6 +958,10 @@ intelDestroyContext(__DRIcontext * driContextPriv)
if (brw->wm.base.scratch_bo)
drm_intel_bo_unreference(brw->wm.base.scratch_bo);
+ gen7_reset_rs_pool_offsets(brw);
+ drm_intel_bo_unreference(brw->hw_bt_pool.bo);
+ brw->hw_bt_pool.bo = NULL;
+
drm_intel_gem_context_destroy(brw->hw_ctx);
if (ctx->swrast_context) {
diff --git a/src/mesa/drivers/dri/i965/brw_context.h
b/src/mesa/drivers/dri/i965/brw_context.h
index 3f8e59d..94127b6 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -1404,6 +1404,12 @@ struct brw_context
struct brw_cs_prog_data *prog_data;
} cs;
+ /* RS hardware binding table */
+ struct {
+ drm_intel_bo *bo;
+ uint32_t next_offset;
+ } hw_bt_pool;
+
struct {
uint32_t state_offset;
uint32_t blend_state_offset;
diff --git a/src/mesa/drivers/dri/i965/brw_state.h
b/src/mesa/drivers/dri/i965/brw_state.h
index 987672f..622ce3f 100644
--- a/src/mesa/drivers/dri/i965/brw_state.h
+++ b/src/mesa/drivers/dri/i965/brw_state.h
@@ -132,6 +132,7 @@ extern const struct brw_tracked_state gen7_sol_state;
extern const struct brw_tracked_state gen7_urb;
extern const struct brw_tracked_state gen7_vs_state;
extern const struct brw_tracked_state gen7_wm_state;
+extern const struct brw_tracked_state gen7_hw_binding_tables;
extern const struct brw_tracked_state haswell_cut_index;
extern const struct brw_tracked_state gen8_blend_state;
extern const struct brw_tracked_state gen8_disable_stages;
@@ -372,6 +373,11 @@ gen7_upload_constant_state(struct brw_context *brw,
const struct brw_stage_state *stage_state,
bool active, unsigned opcode);
+void gen7_rs_control(struct brw_context *brw, int enable);
+void gen7_enable_hw_binding_tables(struct brw_context *brw);
+void gen7_disable_hw_binding_tables(struct brw_context *brw);
+void gen7_reset_rs_pool_offsets(struct brw_context *brw);
+
#ifdef __cplusplus
}
#endif
diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c
b/src/mesa/drivers/dri/i965/brw_state_upload.c
index 84b0861..a2a63f3 100644
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -191,6 +191,8 @@ static const struct brw_tracked_state *gen7_render_atoms[] =
&gen6_color_calc_state, /* must do before cc unit */
&gen6_depth_stencil_state, /* must do before cc unit */
+ &gen7_hw_binding_tables, /* Enable hw-generated binding tables for Haswell
*/
+
&gen6_vs_push_constants, /* Before vs_state */
&gen6_gs_push_constants, /* Before gs_state */
&gen6_wm_push_constants, /* Before wm_surfaces and constant_buffer */
@@ -267,6 +269,8 @@ static const struct brw_tracked_state *gen8_render_atoms[] =
&gen8_blend_state,
&gen6_color_calc_state,
+ &gen7_hw_binding_tables, /* Enable hw-generated binding tables for
Broadwell */
+
&gen6_vs_push_constants, /* Before vs_state */
&gen6_gs_push_constants, /* Before gs_state */
&gen6_wm_push_constants, /* Before wm_surfaces and constant_buffer */
diff --git a/src/mesa/drivers/dri/i965/intel_batchbuffer.c
b/src/mesa/drivers/dri/i965/intel_batchbuffer.c
index a2a3a95..caeb31b 100644
--- a/src/mesa/drivers/dri/i965/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/i965/intel_batchbuffer.c
@@ -32,6 +32,7 @@
#include "intel_buffers.h"
#include "intel_fbo.h"
#include "brw_context.h"
+#include "brw_state.h"
#include <xf86drm.h>
#include <i915_drm.h>
@@ -379,6 +380,9 @@ _intel_batchbuffer_flush(struct brw_context *brw,
drm_intel_bo_wait_rendering(brw->batch.bo);
}
+ if (brw->gen >= 7)
+ gen7_reset_rs_pool_offsets(brw);
+
/* Start a new batch buffer. */
brw_new_batch(brw);
--
1.9.1
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
