summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorDanylo Piliaiev <danylo.piliaiev@globallogic.com>2018-10-05 17:54:07 +0300
committerJason Ekstrand <jason@jlekstrand.net>2019-01-18 18:31:44 +0000
commit1952fd8d2ce90586bc20c0f24593b00604eb1592 (patch)
tree23387d90da1c80cad214cba633ad5d09151be5bc
parented6e2bf263b0f087b2d1265d7450a2a3a22ef8ce (diff)
anv: Implement VK_EXT_conditional_rendering for gen 7.5+
Conditional rendering affects next functions: - vkCmdDraw, vkCmdDrawIndexed, vkCmdDrawIndirect, vkCmdDrawIndexedIndirect - vkCmdDrawIndirectCountKHR, vkCmdDrawIndexedIndirectCountKHR - vkCmdDispatch, vkCmdDispatchIndirect, vkCmdDispatchBase - vkCmdClearAttachments Value from conditional buffer is cached into designated register, MI_PREDICATE is emitted every time conditional rendering is enabled and command requires it. v2: by Jason Ekstrand - Use vk_find_struct_const instead of manually looping - Move draw count loading to prepare function - Zero the top 32-bits of MI_ALU_REG15 v3: Apply pipeline flush before accessing conditional buffer (The issue was found by Samuel Iglesias) v4: - Remove support of Haswell due to possible hardware bug - Made TMP_REG_PREDICATE and TMP_REG_DRAW_COUNT defines to define registers in one place. v5: thanks to Jason Ekstrand and Lionel Landwerlin - Workaround the fact that MI_PREDICATE_RESULT is not accessible on Haswell by manually calculating MI_PREDICATE_RESULT and re-emitting MI_PREDICATE when necessary. v6: suggested by Lionel Landwerlin - Instead of calculating the result of predicate once - re-emit MI_PREDICATE to make it easier to investigate error states. v7: suggested by Jason - Make anv_pipe_invalidate_bits_for_access_flag add CS_STALL if VK_ACCESS_CONDITIONAL_RENDERING_READ_BIT is set. v8: suggested by Lionel - Precompute conditional predicate's result to support secondary command buffers. - Make prepare_for_draw_count_predicate more readable. Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com> Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com> Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
-rw-r--r--src/intel/vulkan/anv_blorp.c8
-rw-r--r--src/intel/vulkan/anv_cmd_buffer.c8
-rw-r--r--src/intel/vulkan/anv_device.c12
-rw-r--r--src/intel/vulkan/anv_extensions.py1
-rw-r--r--src/intel/vulkan/anv_genX.h2
-rw-r--r--src/intel/vulkan/anv_private.h12
-rw-r--r--src/intel/vulkan/genX_cmd_buffer.c236
7 files changed, 265 insertions, 14 deletions
diff --git a/src/intel/vulkan/anv_blorp.c b/src/intel/vulkan/anv_blorp.c
index 67e8d5abb04..827f224c278 100644
--- a/src/intel/vulkan/anv_blorp.c
+++ b/src/intel/vulkan/anv_blorp.c
@@ -1144,8 +1144,12 @@ void anv_CmdClearAttachments(
* trash our depth and stencil buffers.
*/
struct blorp_batch batch;
- blorp_batch_init(&cmd_buffer->device->blorp, &batch, cmd_buffer,
- BLORP_BATCH_NO_EMIT_DEPTH_STENCIL);
+ enum blorp_batch_flags flags = BLORP_BATCH_NO_EMIT_DEPTH_STENCIL;
+ if (cmd_buffer->state.conditional_render_enabled) {
+ anv_cmd_emit_conditional_render_predicate(cmd_buffer);
+ flags |= BLORP_BATCH_PREDICATE_ENABLE;
+ }
+ blorp_batch_init(&cmd_buffer->device->blorp, &batch, cmd_buffer, flags);
for (uint32_t a = 0; a < attachmentCount; ++a) {
if (pAttachments[a].aspectMask & VK_IMAGE_ASPECT_ANY_COLOR_BIT_ANV) {
diff --git a/src/intel/vulkan/anv_cmd_buffer.c b/src/intel/vulkan/anv_cmd_buffer.c
index 67e777b74b3..7b3a3a47dd5 100644
--- a/src/intel/vulkan/anv_cmd_buffer.c
+++ b/src/intel/vulkan/anv_cmd_buffer.c
@@ -377,6 +377,14 @@ anv_cmd_buffer_mark_image_written(struct anv_cmd_buffer *cmd_buffer,
level, base_layer, layer_count);
}
+void
+anv_cmd_emit_conditional_render_predicate(struct anv_cmd_buffer *cmd_buffer)
+{
+ anv_genX_call(&cmd_buffer->device->info,
+ cmd_emit_conditional_render_predicate,
+ cmd_buffer);
+}
+
void anv_CmdBindPipeline(
VkCommandBuffer commandBuffer,
VkPipelineBindPoint pipelineBindPoint,
diff --git a/src/intel/vulkan/anv_device.c b/src/intel/vulkan/anv_device.c
index 45b1ca86859..9d8a982110b 100644
--- a/src/intel/vulkan/anv_device.c
+++ b/src/intel/vulkan/anv_device.c
@@ -916,6 +916,18 @@ void anv_GetPhysicalDeviceFeatures2(
break;
}
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CONDITIONAL_RENDERING_FEATURES_EXT: {
+ VkPhysicalDeviceConditionalRenderingFeaturesEXT *features =
+ (VkPhysicalDeviceConditionalRenderingFeaturesEXT*)ext;
+ ANV_FROM_HANDLE(anv_physical_device, pdevice, physicalDevice);
+
+ features->conditionalRendering = pdevice->info.gen >= 8 ||
+ pdevice->info.is_haswell;
+ features->inheritedConditionalRendering = pdevice->info.gen >= 8 ||
+ pdevice->info.is_haswell;
+ break;
+ }
+
case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES: {
VkPhysicalDeviceMultiviewFeatures *features =
(VkPhysicalDeviceMultiviewFeatures *)ext;
diff --git a/src/intel/vulkan/anv_extensions.py b/src/intel/vulkan/anv_extensions.py
index 79aef132840..eb69264668a 100644
--- a/src/intel/vulkan/anv_extensions.py
+++ b/src/intel/vulkan/anv_extensions.py
@@ -118,6 +118,7 @@ EXTENSIONS = [
Extension('VK_KHR_xlib_surface', 6, 'VK_USE_PLATFORM_XLIB_KHR'),
Extension('VK_EXT_acquire_xlib_display', 1, 'VK_USE_PLATFORM_XLIB_XRANDR_EXT'),
Extension('VK_EXT_calibrated_timestamps', 1, True),
+ Extension('VK_EXT_conditional_rendering', 1, 'device->info.gen >= 8 || device->info.is_haswell'),
Extension('VK_EXT_debug_report', 8, True),
Extension('VK_EXT_direct_mode_display', 1, 'VK_USE_PLATFORM_DISPLAY_KHR'),
Extension('VK_EXT_display_control', 1, 'VK_USE_PLATFORM_DISPLAY_KHR'),
diff --git a/src/intel/vulkan/anv_genX.h b/src/intel/vulkan/anv_genX.h
index 7921e0674a0..8fd32cabf1e 100644
--- a/src/intel/vulkan/anv_genX.h
+++ b/src/intel/vulkan/anv_genX.h
@@ -66,6 +66,8 @@ void genX(cmd_buffer_mark_image_written)(struct anv_cmd_buffer *cmd_buffer,
uint32_t base_layer,
uint32_t layer_count);
+void genX(cmd_emit_conditional_render_predicate)(struct anv_cmd_buffer *cmd_buffer);
+
void
genX(emit_urb_setup)(struct anv_device *device, struct anv_batch *batch,
const struct gen_l3_config *l3_config,
diff --git a/src/intel/vulkan/anv_private.h b/src/intel/vulkan/anv_private.h
index ca109959135..0299febc1d5 100644
--- a/src/intel/vulkan/anv_private.h
+++ b/src/intel/vulkan/anv_private.h
@@ -183,6 +183,11 @@ struct gen_l3_config;
#define ANV_SVGS_VB_INDEX MAX_VBS
#define ANV_DRAWID_VB_INDEX (MAX_VBS + 1)
+/* We reserve this MI ALU register for the purpose of handling predication.
+ * Other code which uses the MI ALU should leave it alone.
+ */
+#define ANV_PREDICATE_RESULT_REG MI_ALU_REG15
+
#define anv_printflike(a, b) __attribute__((__format__(__printf__, a, b)))
static inline uint32_t
@@ -1870,6 +1875,9 @@ anv_pipe_invalidate_bits_for_access_flags(VkAccessFlags flags)
case VK_ACCESS_MEMORY_WRITE_BIT:
pipe_bits |= ANV_PIPE_FLUSH_BITS;
break;
+ case VK_ACCESS_CONDITIONAL_RENDERING_READ_BIT_EXT:
+ pipe_bits |= ANV_PIPE_CS_STALL_BIT;
+ break;
default:
break; /* Nothing to do */
}
@@ -2104,6 +2112,8 @@ struct anv_cmd_state {
*/
bool hiz_enabled;
+ bool conditional_render_enabled;
+
/**
* Array length is anv_cmd_state::pass::attachment_count. Array content is
* valid only when recording a render pass instance.
@@ -2261,6 +2271,8 @@ anv_cmd_buffer_alloc_blorp_binding_table(struct anv_cmd_buffer *cmd_buffer,
void anv_cmd_buffer_dump(struct anv_cmd_buffer *cmd_buffer);
+void anv_cmd_emit_conditional_render_predicate(struct anv_cmd_buffer *cmd_buffer);
+
enum anv_fence_type {
ANV_FENCE_TYPE_NONE = 0,
ANV_FENCE_TYPE_BO,
diff --git a/src/intel/vulkan/genX_cmd_buffer.c b/src/intel/vulkan/genX_cmd_buffer.c
index 6c45acda6f5..e69c169b9ba 100644
--- a/src/intel/vulkan/genX_cmd_buffer.c
+++ b/src/intel/vulkan/genX_cmd_buffer.c
@@ -479,8 +479,9 @@ transition_depth_buffer(struct anv_cmd_buffer *cmd_buffer,
0, 0, 1, hiz_op);
}
-#define MI_PREDICATE_SRC0 0x2400
-#define MI_PREDICATE_SRC1 0x2408
+#define MI_PREDICATE_SRC0 0x2400
+#define MI_PREDICATE_SRC1 0x2408
+#define MI_PREDICATE_RESULT 0x2418
static void
set_image_compressed_bit(struct anv_cmd_buffer *cmd_buffer,
@@ -1411,6 +1412,19 @@ genX(BeginCommandBuffer)(
cmd_buffer->state.gfx.dirty |= ANV_CMD_DIRTY_RENDER_TARGETS;
}
+#if GEN_GEN >= 8 || GEN_IS_HASWELL
+ if (cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_SECONDARY) {
+ const VkCommandBufferInheritanceConditionalRenderingInfoEXT *conditional_rendering_info =
+ vk_find_struct_const(pBeginInfo->pInheritanceInfo->pNext, COMMAND_BUFFER_INHERITANCE_CONDITIONAL_RENDERING_INFO_EXT);
+
+ /* If secondary buffer supports conditional rendering
+ * we should emit commands as if conditional rendering is enabled.
+ */
+ cmd_buffer->state.conditional_render_enabled =
+ conditional_rendering_info && conditional_rendering_info->conditionalRenderingEnable;
+ }
+#endif
+
return result;
}
@@ -1515,6 +1529,19 @@ genX(CmdExecuteCommands)(
assert(secondary->level == VK_COMMAND_BUFFER_LEVEL_SECONDARY);
assert(!anv_batch_has_error(&secondary->batch));
+#if GEN_GEN >= 8 || GEN_IS_HASWELL
+ if (secondary->state.conditional_render_enabled) {
+ if (!primary->state.conditional_render_enabled) {
+ /* Secondary buffer is constructed as if it will be executed
+ * with conditional rendering, we should satisfy this dependency
+ * regardless of conditional rendering being enabled in primary.
+ */
+ emit_lri(&primary->batch, CS_GPR(ANV_PREDICATE_RESULT_REG), UINT32_MAX);
+ emit_lri(&primary->batch, CS_GPR(ANV_PREDICATE_RESULT_REG) + 4, UINT32_MAX);
+ }
+ }
+#endif
+
if (secondary->usage_flags &
VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT) {
/* If we're continuing a render pass from the primary, we need to
@@ -2777,6 +2804,9 @@ void genX(CmdDraw)(
genX(cmd_buffer_flush_state)(cmd_buffer);
+ if (cmd_buffer->state.conditional_render_enabled)
+ genX(cmd_emit_conditional_render_predicate)(cmd_buffer);
+
if (vs_prog_data->uses_firstvertex ||
vs_prog_data->uses_baseinstance)
emit_base_vertex_instance(cmd_buffer, firstVertex, firstInstance);
@@ -2789,6 +2819,7 @@ void genX(CmdDraw)(
instanceCount *= anv_subpass_view_count(cmd_buffer->state.subpass);
anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE), prim) {
+ prim.PredicateEnable = cmd_buffer->state.conditional_render_enabled;
prim.VertexAccessType = SEQUENTIAL;
prim.PrimitiveTopologyType = pipeline->topology;
prim.VertexCountPerInstance = vertexCount;
@@ -2818,6 +2849,9 @@ void genX(CmdDrawIndexed)(
genX(cmd_buffer_flush_state)(cmd_buffer);
+ if (cmd_buffer->state.conditional_render_enabled)
+ genX(cmd_emit_conditional_render_predicate)(cmd_buffer);
+
if (vs_prog_data->uses_firstvertex ||
vs_prog_data->uses_baseinstance)
emit_base_vertex_instance(cmd_buffer, vertexOffset, firstInstance);
@@ -2830,6 +2864,7 @@ void genX(CmdDrawIndexed)(
instanceCount *= anv_subpass_view_count(cmd_buffer->state.subpass);
anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE), prim) {
+ prim.PredicateEnable = cmd_buffer->state.conditional_render_enabled;
prim.VertexAccessType = RANDOM;
prim.PrimitiveTopologyType = pipeline->topology;
prim.VertexCountPerInstance = indexCount;
@@ -2954,6 +2989,9 @@ void genX(CmdDrawIndirect)(
genX(cmd_buffer_flush_state)(cmd_buffer);
+ if (cmd_buffer->state.conditional_render_enabled)
+ genX(cmd_emit_conditional_render_predicate)(cmd_buffer);
+
for (uint32_t i = 0; i < drawCount; i++) {
struct anv_address draw = anv_address_add(buffer->address, offset);
@@ -2967,6 +3005,7 @@ void genX(CmdDrawIndirect)(
anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE), prim) {
prim.IndirectParameterEnable = true;
+ prim.PredicateEnable = cmd_buffer->state.conditional_render_enabled;
prim.VertexAccessType = SEQUENTIAL;
prim.PrimitiveTopologyType = pipeline->topology;
}
@@ -2994,6 +3033,9 @@ void genX(CmdDrawIndexedIndirect)(
genX(cmd_buffer_flush_state)(cmd_buffer);
+ if (cmd_buffer->state.conditional_render_enabled)
+ genX(cmd_emit_conditional_render_predicate)(cmd_buffer);
+
for (uint32_t i = 0; i < drawCount; i++) {
struct anv_address draw = anv_address_add(buffer->address, offset);
@@ -3008,6 +3050,7 @@ void genX(CmdDrawIndexedIndirect)(
anv_batch_emit(&cmd_buffer->batch, GENX(3DPRIMITIVE), prim) {
prim.IndirectParameterEnable = true;
+ prim.PredicateEnable = cmd_buffer->state.conditional_render_enabled;
prim.VertexAccessType = RANDOM;
prim.PrimitiveTopologyType = pipeline->topology;
}
@@ -3018,17 +3061,27 @@ void genX(CmdDrawIndexedIndirect)(
cmd_buffer->state.pending_pipe_bits |= ANV_PIPE_RENDER_TARGET_WRITES;
}
+#define TMP_DRAW_COUNT_REG MI_ALU_REG14
+
static void
prepare_for_draw_count_predicate(struct anv_cmd_buffer *cmd_buffer,
- struct anv_address count_address)
+ struct anv_address count_address,
+ const bool conditional_render_enabled)
{
- /* Upload the current draw count from the draw parameters buffer to
- * MI_PREDICATE_SRC0.
- */
- emit_lrm(&cmd_buffer->batch, MI_PREDICATE_SRC0, count_address);
- emit_lri(&cmd_buffer->batch, MI_PREDICATE_SRC0 + 4, 0);
+ if (conditional_render_enabled) {
+#if GEN_GEN >= 8 || GEN_IS_HASWELL
+ emit_lrm(&cmd_buffer->batch, CS_GPR(TMP_DRAW_COUNT_REG), count_address);
+ emit_lri(&cmd_buffer->batch, CS_GPR(TMP_DRAW_COUNT_REG) + 4, 0);
+#endif
+ } else {
+ /* Upload the current draw count from the draw parameters buffer to
+ * MI_PREDICATE_SRC0.
+ */
+ emit_lrm(&cmd_buffer->batch, MI_PREDICATE_SRC0, count_address);
+ emit_lri(&cmd_buffer->batch, MI_PREDICATE_SRC0 + 4, 0);
- emit_lri(&cmd_buffer->batch, MI_PREDICATE_SRC1 + 4, 0);
+ emit_lri(&cmd_buffer->batch, MI_PREDICATE_SRC1 + 4, 0);
+ }
}
static void
@@ -3060,6 +3113,54 @@ emit_draw_count_predicate(struct anv_cmd_buffer *cmd_buffer,
}
}
+#if GEN_GEN >= 8 || GEN_IS_HASWELL
+static void
+emit_draw_count_predicate_with_conditional_render(
+ struct anv_cmd_buffer *cmd_buffer,
+ uint32_t draw_index)
+{
+ const int draw_index_reg = MI_ALU_REG0;
+ const int tmp_result_reg = MI_ALU_REG1;
+
+ emit_lri(&cmd_buffer->batch, CS_GPR(draw_index_reg), draw_index);
+ emit_lri(&cmd_buffer->batch, CS_GPR(draw_index_reg) + 4, 0);
+
+ uint32_t *dw;
+ /* Compute (draw_index < draw_count).
+ * We do this by subtracting and storing the carry bit.
+ */
+ dw = anv_batch_emitn(&cmd_buffer->batch, 9, GENX(MI_MATH));
+ dw[1] = mi_alu(MI_ALU_LOAD, MI_ALU_SRCA, draw_index_reg);
+ dw[2] = mi_alu(MI_ALU_LOAD, MI_ALU_SRCB, TMP_DRAW_COUNT_REG);
+ dw[3] = mi_alu(MI_ALU_SUB, 0, 0);
+ dw[4] = mi_alu(MI_ALU_STORE, tmp_result_reg, MI_ALU_CF);
+ /* & condition */
+ dw[5] = mi_alu(MI_ALU_LOAD, MI_ALU_SRCA, tmp_result_reg);
+ dw[6] = mi_alu(MI_ALU_LOAD, MI_ALU_SRCB, ANV_PREDICATE_RESULT_REG);
+ dw[7] = mi_alu(MI_ALU_AND, 0, 0);
+ dw[8] = mi_alu(MI_ALU_STORE, tmp_result_reg, MI_ALU_ACCU);
+
+#if GEN_GEN >= 8
+ emit_lrr(&cmd_buffer->batch, MI_PREDICATE_RESULT, CS_GPR(tmp_result_reg));
+#else
+ /* MI_PREDICATE_RESULT is not whitelisted in i915 command parser
+ * so we emit MI_PREDICATE to set it.
+ */
+
+ emit_lrr(&cmd_buffer->batch, MI_PREDICATE_SRC0, CS_GPR(tmp_result_reg));
+ emit_lri(&cmd_buffer->batch, MI_PREDICATE_SRC0 + 4, 0);
+ emit_lri(&cmd_buffer->batch, MI_PREDICATE_SRC1, 0);
+ emit_lri(&cmd_buffer->batch, MI_PREDICATE_SRC1 + 4, 0);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(MI_PREDICATE), mip) {
+ mip.LoadOperation = LOAD_LOADINV;
+ mip.CombineOperation = COMBINE_SET;
+ mip.CompareOperation = COMPARE_SRCS_EQUAL;
+ }
+#endif
+}
+#endif
+
void genX(CmdDrawIndirectCountKHR)(
VkCommandBuffer commandBuffer,
VkBuffer _buffer,
@@ -3084,12 +3185,21 @@ void genX(CmdDrawIndirectCountKHR)(
struct anv_address count_address =
anv_address_add(count_buffer->address, countBufferOffset);
- prepare_for_draw_count_predicate(cmd_buffer, count_address);
+ prepare_for_draw_count_predicate(cmd_buffer, count_address,
+ cmd_state->conditional_render_enabled);
for (uint32_t i = 0; i < maxDrawCount; i++) {
struct anv_address draw = anv_address_add(buffer->address, offset);
+#if GEN_GEN >= 8 || GEN_IS_HASWELL
+ if (cmd_state->conditional_render_enabled) {
+ emit_draw_count_predicate_with_conditional_render(cmd_buffer, i);
+ } else {
+ emit_draw_count_predicate(cmd_buffer, i);
+ }
+#else
emit_draw_count_predicate(cmd_buffer, i);
+#endif
if (vs_prog_data->uses_firstvertex ||
vs_prog_data->uses_baseinstance)
@@ -3136,12 +3246,21 @@ void genX(CmdDrawIndexedIndirectCountKHR)(
struct anv_address count_address =
anv_address_add(count_buffer->address, countBufferOffset);
- prepare_for_draw_count_predicate(cmd_buffer, count_address);
+ prepare_for_draw_count_predicate(cmd_buffer, count_address,
+ cmd_state->conditional_render_enabled);
for (uint32_t i = 0; i < maxDrawCount; i++) {
struct anv_address draw = anv_address_add(buffer->address, offset);
+#if GEN_GEN >= 8 || GEN_IS_HASWELL
+ if (cmd_state->conditional_render_enabled) {
+ emit_draw_count_predicate_with_conditional_render(cmd_buffer, i);
+ } else {
+ emit_draw_count_predicate(cmd_buffer, i);
+ }
+#else
emit_draw_count_predicate(cmd_buffer, i);
+#endif
/* TODO: We need to stomp base vertex to 0 somehow */
if (vs_prog_data->uses_firstvertex ||
@@ -3369,7 +3488,11 @@ void genX(CmdDispatchBase)(
genX(cmd_buffer_flush_compute_state)(cmd_buffer);
+ if (cmd_buffer->state.conditional_render_enabled)
+ genX(cmd_emit_conditional_render_predicate)(cmd_buffer);
+
anv_batch_emit(&cmd_buffer->batch, GENX(GPGPU_WALKER), ggw) {
+ ggw.PredicateEnable = cmd_buffer->state.conditional_render_enabled;
ggw.SIMDSize = prog_data->simd_size / 16;
ggw.ThreadDepthCounterMaximum = 0;
ggw.ThreadHeightCounterMaximum = 0;
@@ -3463,11 +3586,28 @@ void genX(CmdDispatchIndirect)(
mip.CombineOperation = COMBINE_OR;
mip.CompareOperation = COMPARE_FALSE;
}
+
+#if GEN_IS_HASWELL
+ if (cmd_buffer->state.conditional_render_enabled) {
+ emit_lrr(batch, MI_PREDICATE_SRC0, CS_GPR(ANV_PREDICATE_RESULT_REG));
+ /* predicate &= !(conditional_rendering_predicate == 0); */
+ anv_batch_emit(batch, GENX(MI_PREDICATE), mip) {
+ mip.LoadOperation = LOAD_LOADINV;
+ mip.CombineOperation = COMBINE_AND;
+ mip.CompareOperation = COMPARE_SRCS_EQUAL;
+ }
+ }
+#endif
+
+#else /* GEN_GEN > 7 */
+ if (cmd_buffer->state.conditional_render_enabled)
+ genX(cmd_emit_conditional_render_predicate)(cmd_buffer);
#endif
anv_batch_emit(batch, GENX(GPGPU_WALKER), ggw) {
ggw.IndirectParameterEnable = true;
- ggw.PredicateEnable = GEN_GEN <= 7;
+ ggw.PredicateEnable = GEN_GEN <= 7 ||
+ cmd_buffer->state.conditional_render_enabled;
ggw.SIMDSize = prog_data->simd_size / 16;
ggw.ThreadDepthCounterMaximum = 0;
ggw.ThreadHeightCounterMaximum = 0;
@@ -4423,3 +4563,75 @@ void genX(CmdEndRenderPass2KHR)(
{
genX(CmdEndRenderPass)(commandBuffer);
}
+
+void
+genX(cmd_emit_conditional_render_predicate)(struct anv_cmd_buffer *cmd_buffer)
+{
+#if GEN_GEN >= 8 || GEN_IS_HASWELL
+ emit_lrr(&cmd_buffer->batch, MI_PREDICATE_SRC0, CS_GPR(ANV_PREDICATE_RESULT_REG));
+ emit_lri(&cmd_buffer->batch, MI_PREDICATE_SRC0 + 4, 0);
+ emit_lri(&cmd_buffer->batch, MI_PREDICATE_SRC1, 0);
+ emit_lri(&cmd_buffer->batch, MI_PREDICATE_SRC1 + 4, 0);
+
+ anv_batch_emit(&cmd_buffer->batch, GENX(MI_PREDICATE), mip) {
+ mip.LoadOperation = LOAD_LOADINV;
+ mip.CombineOperation = COMBINE_SET;
+ mip.CompareOperation = COMPARE_SRCS_EQUAL;
+ }
+#endif
+}
+
+#if GEN_GEN >= 8 || GEN_IS_HASWELL
+void genX(CmdBeginConditionalRenderingEXT)(
+ VkCommandBuffer commandBuffer,
+ const VkConditionalRenderingBeginInfoEXT* pConditionalRenderingBegin)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ ANV_FROM_HANDLE(anv_buffer, buffer, pConditionalRenderingBegin->buffer);
+ struct anv_cmd_state *cmd_state = &cmd_buffer->state;
+ struct anv_address value_address =
+ anv_address_add(buffer->address, pConditionalRenderingBegin->offset);
+
+ const bool isInverted = pConditionalRenderingBegin->flags &
+ VK_CONDITIONAL_RENDERING_INVERTED_BIT_EXT;
+
+ cmd_state->conditional_render_enabled = true;
+
+ genX(cmd_buffer_apply_pipe_flushes)(cmd_buffer);
+
+ /* Section 19.4 of the Vulkan 1.1.85 spec says:
+ *
+ * If the value of the predicate in buffer memory changes
+ * while conditional rendering is active, the rendering commands
+ * may be discarded in an implementation-dependent way.
+ * Some implementations may latch the value of the predicate
+ * upon beginning conditional rendering while others
+ * may read it before every rendering command.
+ *
+ * So it's perfectly fine to read a value from the buffer once.
+ */
+ emit_lrm(&cmd_buffer->batch, CS_GPR(MI_ALU_REG0), value_address);
+ /* Zero the top 32-bits of MI_PREDICATE_SRC0 */
+ emit_lri(&cmd_buffer->batch, CS_GPR(MI_ALU_REG0) + 4, 0);
+
+ /* Precompute predicate result, it is necessary to support secondary
+ * command buffers since it is unknown if conditional rendering is
+ * inverted when populating them.
+ */
+ uint32_t *dw = anv_batch_emitn(&cmd_buffer->batch, 5, GENX(MI_MATH));
+ dw[1] = mi_alu(MI_ALU_LOAD0, MI_ALU_SRCA, 0);
+ dw[2] = mi_alu(MI_ALU_LOAD, MI_ALU_SRCB, MI_ALU_REG0);
+ dw[3] = mi_alu(MI_ALU_SUB, 0, 0);
+ dw[4] = mi_alu(isInverted ? MI_ALU_STOREINV : MI_ALU_STORE,
+ ANV_PREDICATE_RESULT_REG, MI_ALU_CF);
+}
+
+void genX(CmdEndConditionalRenderingEXT)(
+ VkCommandBuffer commandBuffer)
+{
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
+ struct anv_cmd_state *cmd_state = &cmd_buffer->state;
+
+ cmd_state->conditional_render_enabled = false;
+}
+#endif