#include "anv_private.h"
+#include "gen7_pack.h"
+#include "gen75_pack.h"
-void
-gen7_cmd_buffer_emit_state_base_address(struct anv_cmd_buffer *cmd_buffer)
+static uint32_t
+cmd_buffer_flush_push_constants(struct anv_cmd_buffer *cmd_buffer)
{
- struct anv_device *device = cmd_buffer->device;
- struct anv_bo *scratch_bo = NULL;
-
- cmd_buffer->state.scratch_size =
- anv_block_pool_size(&device->scratch_block_pool);
- if (cmd_buffer->state.scratch_size > 0)
- scratch_bo = &device->scratch_block_pool.bo;
-
- anv_batch_emit(&cmd_buffer->batch, GEN7_STATE_BASE_ADDRESS,
- .GeneralStateBaseAddress = { scratch_bo, 0 },
- .GeneralStateMemoryObjectControlState = GEN7_MOCS,
- .GeneralStateBaseAddressModifyEnable = true,
- .GeneralStateAccessUpperBound = { scratch_bo, scratch_bo->size },
- .GeneralStateAccessUpperBoundModifyEnable = true,
-
- .SurfaceStateBaseAddress = anv_cmd_buffer_surface_base_address(cmd_buffer),
- .SurfaceStateMemoryObjectControlState = GEN7_MOCS,
- .SurfaceStateBaseAddressModifyEnable = true,
-
- .DynamicStateBaseAddress = { &device->dynamic_state_block_pool.bo, 0 },
- .DynamicStateMemoryObjectControlState = GEN7_MOCS,
- .DynamicStateBaseAddressModifyEnable = true,
- .DynamicStateAccessUpperBound = { &device->dynamic_state_block_pool.bo,
- device->dynamic_state_block_pool.bo.size },
- .DynamicStateAccessUpperBoundModifyEnable = true,
-
- .IndirectObjectBaseAddress = { NULL, 0 },
- .IndirectObjectMemoryObjectControlState = GEN7_MOCS,
- .IndirectObjectBaseAddressModifyEnable = true,
-
- .IndirectObjectAccessUpperBound = { NULL, 0xffffffff },
- .IndirectObjectAccessUpperBoundModifyEnable = true,
-
- .InstructionBaseAddress = { &device->instruction_block_pool.bo, 0 },
- .InstructionMemoryObjectControlState = GEN7_MOCS,
- .InstructionBaseAddressModifyEnable = true,
- .InstructionAccessUpperBound = { &device->instruction_block_pool.bo,
- device->instruction_block_pool.bo.size },
- .InstructionAccessUpperBoundModifyEnable = true);
-
- /* After re-setting the surface state base address, we have to do some
- * cache flusing so that the sampler engine will pick up the new
- * SURFACE_STATE objects and binding tables. From the Broadwell PRM,
- * Shared Function > 3D Sampler > State > State Caching (page 96):
- *
- * Coherency with system memory in the state cache, like the texture
- * cache is handled partially by software. It is expected that the
- * command stream or shader will issue Cache Flush operation or
- * Cache_Flush sampler message to ensure that the L1 cache remains
- * coherent with system memory.
- *
- * [...]
- *
- * Whenever the value of the Dynamic_State_Base_Addr,
- * Surface_State_Base_Addr are altered, the L1 state cache must be
- * invalidated to ensure the new surface or sampler state is fetched
- * from system memory.
- *
- * The PIPE_CONTROL command has a "State Cache Invalidation Enable" bit
- * which, according the PIPE_CONTROL instruction documentation in the
- * Broadwell PRM:
- *
- * Setting this bit is independent of any other bit in this packet.
- * This bit controls the invalidation of the L1 and L2 state caches
- * at the top of the pipe i.e. at the parsing time.
- *
- * Unfortunately, experimentation seems to indicate that state cache
- * invalidation through a PIPE_CONTROL does nothing whatsoever in
- * regards to surface state and binding tables. In stead, it seems that
- * invalidating the texture cache is what is actually needed.
- *
- * XXX: As far as we have been able to determine through
- * experimentation, shows that flush the texture cache appears to be
- * sufficient. The theory here is that all of the sampling/rendering
- * units cache the binding table in the texture cache. However, we have
- * yet to be able to actually confirm this.
- */
- anv_batch_emit(&cmd_buffer->batch, GEN7_PIPE_CONTROL,
- .TextureCacheInvalidationEnable = true);
+ static const uint32_t push_constant_opcodes[] = {
+ [MESA_SHADER_VERTEX] = 21,
+ [MESA_SHADER_TESS_CTRL] = 25, /* HS */
+ [MESA_SHADER_TESS_EVAL] = 26, /* DS */
+ [MESA_SHADER_GEOMETRY] = 22,
+ [MESA_SHADER_FRAGMENT] = 23,
+ [MESA_SHADER_COMPUTE] = 0,
+ };
+
+ VkShaderStageFlags flushed = 0;
+
+ anv_foreach_stage(stage, cmd_buffer->state.push_constants_dirty) {
+ struct anv_state state = anv_cmd_buffer_push_constants(cmd_buffer, stage);
+
+ if (state.offset == 0)
+ continue;
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_CONSTANT_VS,
+ ._3DCommandSubOpcode = push_constant_opcodes[stage],
+ .ConstantBody = {
+ .PointerToConstantBuffer0 = { .offset = state.offset },
+ .ConstantBuffer0ReadLength = DIV_ROUND_UP(state.alloc_size, 32),
+ });
+
+ flushed |= mesa_to_vk_shader_stage(stage);
+ }
+
+ cmd_buffer->state.push_constants_dirty &= ~flushed;
+
+ return flushed;
+}
+
+GENX_FUNC(GEN7, GEN7) void
+genX(cmd_buffer_emit_descriptor_pointers)(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t stages)
+{
+ static const uint32_t sampler_state_opcodes[] = {
+ [MESA_SHADER_VERTEX] = 43,
+ [MESA_SHADER_TESS_CTRL] = 44, /* HS */
+ [MESA_SHADER_TESS_EVAL] = 45, /* DS */
+ [MESA_SHADER_GEOMETRY] = 46,
+ [MESA_SHADER_FRAGMENT] = 47,
+ [MESA_SHADER_COMPUTE] = 0,
+ };
+
+ static const uint32_t binding_table_opcodes[] = {
+ [MESA_SHADER_VERTEX] = 38,
+ [MESA_SHADER_TESS_CTRL] = 39,
+ [MESA_SHADER_TESS_EVAL] = 40,
+ [MESA_SHADER_GEOMETRY] = 41,
+ [MESA_SHADER_FRAGMENT] = 42,
+ [MESA_SHADER_COMPUTE] = 0,
+ };
+
+ anv_foreach_stage(s, stages) {
+ if (cmd_buffer->state.samplers[s].alloc_size > 0) {
+ anv_batch_emit(&cmd_buffer->batch,
+ GEN7_3DSTATE_SAMPLER_STATE_POINTERS_VS,
+ ._3DCommandSubOpcode = sampler_state_opcodes[s],
+ .PointertoVSSamplerState = cmd_buffer->state.samplers[s].offset);
+ }
+
+ /* Always emit binding table pointers if we're asked to, since on SKL
+ * this is what flushes push constants. */
+ anv_batch_emit(&cmd_buffer->batch,
+ GEN7_3DSTATE_BINDING_TABLE_POINTERS_VS,
+ ._3DCommandSubOpcode = binding_table_opcodes[s],
+ .PointertoVSBindingTable = cmd_buffer->state.binding_tables[s].offset);
+ }
+}
+
+GENX_FUNC(GEN7, GEN7) uint32_t
+genX(cmd_buffer_flush_descriptor_sets)(struct anv_cmd_buffer *cmd_buffer)
+{
+ VkShaderStageFlags dirty = cmd_buffer->state.descriptors_dirty &
+ cmd_buffer->state.pipeline->active_stages;
+
+ VkResult result = VK_SUCCESS;
+ anv_foreach_stage(s, dirty) {
+ result = anv_cmd_buffer_emit_samplers(cmd_buffer, s,
+ &cmd_buffer->state.samplers[s]);
+ if (result != VK_SUCCESS)
+ break;
+ result = anv_cmd_buffer_emit_binding_table(cmd_buffer, s,
+ &cmd_buffer->state.binding_tables[s]);
+ if (result != VK_SUCCESS)
+ break;
+ }
+
+ if (result != VK_SUCCESS) {
+ assert(result == VK_ERROR_OUT_OF_DEVICE_MEMORY);
+
+ result = anv_cmd_buffer_new_binding_table_block(cmd_buffer);
+ assert(result == VK_SUCCESS);
+
+ /* Re-emit state base addresses so we get the new surface state base
+ * address before we start emitting binding tables etc.
+ */
+ anv_cmd_buffer_emit_state_base_address(cmd_buffer);
+
+ /* Re-emit all active binding tables */
+ dirty |= cmd_buffer->state.pipeline->active_stages;
+ anv_foreach_stage(s, dirty) {
+ result = anv_cmd_buffer_emit_samplers(cmd_buffer, s,
+ &cmd_buffer->state.samplers[s]);
+ if (result != VK_SUCCESS)
+ return result;
+ result = anv_cmd_buffer_emit_binding_table(cmd_buffer, s,
+ &cmd_buffer->state.binding_tables[s]);
+ if (result != VK_SUCCESS)
+ return result;
+ }
+ }
+
+ cmd_buffer->state.descriptors_dirty &= ~dirty;
+
+ return dirty;
+}
+
+static inline int64_t
+clamp_int64(int64_t x, int64_t min, int64_t max)
+{
+ if (x < min)
+ return min;
+ else if (x < max)
+ return x;
+ else
+ return max;
+}
+
+static void
+emit_scissor_state(struct anv_cmd_buffer *cmd_buffer,
+ uint32_t count, const VkRect2D *scissors)
+{
+ struct anv_state scissor_state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, count * 32, 32);
+
+ for (uint32_t i = 0; i < count; i++) {
+ const VkRect2D *s = &scissors[i];
+
+ /* Since xmax and ymax are inclusive, we have to have xmax < xmin or
+ * ymax < ymin for empty clips. In case clip x, y, width height are all
+ * 0, the clamps below produce 0 for xmin, ymin, xmax, ymax, which isn't
+ * what we want. Just special case empty clips and produce a canonical
+ * empty clip. */
+ static const struct GEN7_SCISSOR_RECT empty_scissor = {
+ .ScissorRectangleYMin = 1,
+ .ScissorRectangleXMin = 1,
+ .ScissorRectangleYMax = 0,
+ .ScissorRectangleXMax = 0
+ };
+
+ const int max = 0xffff;
+ struct GEN7_SCISSOR_RECT scissor = {
+ /* Do this math using int64_t so overflow gets clamped correctly. */
+ .ScissorRectangleYMin = clamp_int64(s->offset.y, 0, max),
+ .ScissorRectangleXMin = clamp_int64(s->offset.x, 0, max),
+ .ScissorRectangleYMax = clamp_int64((uint64_t) s->offset.y + s->extent.height - 1, 0, max),
+ .ScissorRectangleXMax = clamp_int64((uint64_t) s->offset.x + s->extent.width - 1, 0, max)
+ };
+
+ if (s->extent.width <= 0 || s->extent.height <= 0) {
+ GEN7_SCISSOR_RECT_pack(NULL, scissor_state.map + i * 32,
+ &empty_scissor);
+ } else {
+ GEN7_SCISSOR_RECT_pack(NULL, scissor_state.map + i * 32, &scissor);
+ }
+ }
+
+ anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_SCISSOR_STATE_POINTERS,
+ .ScissorRectPointer = scissor_state.offset);
+
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(scissor_state);
+}
+
+GENX_FUNC(GEN7, GEN7) void
+genX(cmd_buffer_emit_scissor)(struct anv_cmd_buffer *cmd_buffer)
+{
+ if (cmd_buffer->state.dynamic.scissor.count > 0) {
+ emit_scissor_state(cmd_buffer, cmd_buffer->state.dynamic.scissor.count,
+ cmd_buffer->state.dynamic.scissor.scissors);
+ } else {
+ /* Emit a default scissor based on the currently bound framebuffer */
+ emit_scissor_state(cmd_buffer, 1,
+ &(VkRect2D) {
+ .offset = { .x = 0, .y = 0, },
+ .extent = {
+ .width = cmd_buffer->state.framebuffer->width,
+ .height = cmd_buffer->state.framebuffer->height,
+ },
+ });
+ }
}
static const uint32_t vk_to_gen_index_type[] = {
[VK_INDEX_TYPE_UINT32] = INDEX_DWORD,
};
-void gen7_CmdBindIndexBuffer(
- VkCmdBuffer cmdBuffer,
+static const uint32_t restart_index_for_type[] = {
+ [VK_INDEX_TYPE_UINT16] = UINT16_MAX,
+ [VK_INDEX_TYPE_UINT32] = UINT32_MAX,
+};
+
+void genX(CmdBindIndexBuffer)(
+ VkCommandBuffer commandBuffer,
VkBuffer _buffer,
VkDeviceSize offset,
VkIndexType indexType)
{
- ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, cmdBuffer);
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
- cmd_buffer->state.dirty |= ANV_CMD_BUFFER_INDEX_BUFFER_DIRTY;
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_INDEX_BUFFER;
+ if (ANV_IS_HASWELL)
+ cmd_buffer->state.restart_index = restart_index_for_type[indexType];
cmd_buffer->state.gen7.index_buffer = buffer;
cmd_buffer->state.gen7.index_type = vk_to_gen_index_type[indexType];
cmd_buffer->state.gen7.index_offset = offset;
}
static VkResult
-gen7_flush_compute_descriptor_set(struct anv_cmd_buffer *cmd_buffer)
+flush_compute_descriptor_set(struct anv_cmd_buffer *cmd_buffer)
{
struct anv_device *device = cmd_buffer->device;
struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
VkResult result;
result = anv_cmd_buffer_emit_samplers(cmd_buffer,
- VK_SHADER_STAGE_COMPUTE, &samplers);
+ MESA_SHADER_COMPUTE, &samplers);
if (result != VK_SUCCESS)
return result;
result = anv_cmd_buffer_emit_binding_table(cmd_buffer,
- VK_SHADER_STAGE_COMPUTE, &surfaces);
+ MESA_SHADER_COMPUTE, &surfaces);
if (result != VK_SUCCESS)
return result;
- struct GEN7_INTERFACE_DESCRIPTOR_DATA desc = {
- .KernelStartPointer = pipeline->cs_simd,
- .BindingTablePointer = surfaces.offset,
- .SamplerStatePointer = samplers.offset,
- .NumberofThreadsinGPGPUThreadGroup = 0 /* FIXME: Really? */
- };
+ const struct brw_cs_prog_data *cs_prog_data = &pipeline->cs_prog_data;
- uint32_t size = GEN7_INTERFACE_DESCRIPTOR_DATA_length * sizeof(uint32_t);
struct anv_state state =
- anv_state_pool_alloc(&device->dynamic_state_pool, size, 64);
-
- GEN7_INTERFACE_DESCRIPTOR_DATA_pack(NULL, state.map, &desc);
-
+ anv_state_pool_emit(&device->dynamic_state_pool,
+ GEN7_INTERFACE_DESCRIPTOR_DATA, 64,
+ .KernelStartPointer = pipeline->cs_simd,
+ .BindingTablePointer = surfaces.offset,
+ .SamplerStatePointer = samplers.offset,
+ .BarrierEnable = cs_prog_data->uses_barrier,
+ .NumberofThreadsinGPGPUThreadGroup =
+ pipeline->cs_thread_width_max);
+
+ const uint32_t size = GEN7_INTERFACE_DESCRIPTOR_DATA_length * sizeof(uint32_t);
anv_batch_emit(&cmd_buffer->batch, GEN7_MEDIA_INTERFACE_DESCRIPTOR_LOAD,
.InterfaceDescriptorTotalLength = size,
.InterfaceDescriptorDataStartAddress = state.offset);
}
static void
-gen7_cmd_buffer_flush_compute_state(struct anv_cmd_buffer *cmd_buffer)
+cmd_buffer_flush_compute_state(struct anv_cmd_buffer *cmd_buffer)
{
struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
VkResult result;
cmd_buffer->state.current_pipeline = GPGPU;
}
- if (cmd_buffer->state.compute_dirty & ANV_CMD_BUFFER_PIPELINE_DIRTY)
+ if (cmd_buffer->state.compute_dirty & ANV_CMD_DIRTY_PIPELINE)
anv_batch_emit_batch(&cmd_buffer->batch, &pipeline->batch);
if ((cmd_buffer->state.descriptors_dirty & VK_SHADER_STAGE_COMPUTE_BIT) ||
- (cmd_buffer->state.compute_dirty & ANV_CMD_BUFFER_PIPELINE_DIRTY)) {
+ (cmd_buffer->state.compute_dirty & ANV_CMD_DIRTY_PIPELINE)) {
/* FIXME: figure out descriptors for gen7 */
- result = gen7_flush_compute_descriptor_set(cmd_buffer);
+ result = flush_compute_descriptor_set(cmd_buffer);
assert(result == VK_SUCCESS);
- cmd_buffer->state.descriptors_dirty &= ~VK_SHADER_STAGE_COMPUTE;
+ cmd_buffer->state.descriptors_dirty &= ~VK_SHADER_STAGE_COMPUTE_BIT;
}
cmd_buffer->state.compute_dirty = 0;
}
static void
-gen7_cmd_buffer_flush_state(struct anv_cmd_buffer *cmd_buffer)
+cmd_buffer_flush_state(struct anv_cmd_buffer *cmd_buffer)
{
struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
uint32_t *p;
}
}
- if (cmd_buffer->state.dirty & ANV_CMD_BUFFER_PIPELINE_DIRTY) {
+ if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_PIPELINE) {
/* If somebody compiled a pipeline after starting a command buffer the
* scratch bo may have grown since we started this cmd buffer (and
* emitted STATE_BASE_ADDRESS). If we're binding that pipeline now,
anv_batch_emit_batch(&cmd_buffer->batch, &pipeline->batch);
}
- if (cmd_buffer->state.descriptors_dirty)
- anv_flush_descriptor_sets(cmd_buffer);
-
- if (cmd_buffer->state.dirty & ANV_CMD_BUFFER_VP_DIRTY) {
- struct anv_dynamic_vp_state *vp_state = cmd_buffer->state.vp_state;
- anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_SCISSOR_STATE_POINTERS,
- .ScissorRectPointer = vp_state->scissor.offset);
- anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_CC,
- .CCViewportPointer = vp_state->cc_vp.offset);
- anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP,
- .SFClipViewportPointer = vp_state->sf_clip_vp.offset);
+ if (cmd_buffer->state.descriptors_dirty & VK_SHADER_STAGE_VERTEX_BIT ||
+ cmd_buffer->state.push_constants_dirty & VK_SHADER_STAGE_VERTEX_BIT) {
+ /* From the IVB PRM Vol. 2, Part 1, Section 3.2.1:
+ *
+ * "A PIPE_CONTROL with Post-Sync Operation set to 1h and a depth
+ * stall needs to be sent just prior to any 3DSTATE_VS,
+ * 3DSTATE_URB_VS, 3DSTATE_CONSTANT_VS,
+ * 3DSTATE_BINDING_TABLE_POINTER_VS,
+ * 3DSTATE_SAMPLER_STATE_POINTER_VS command. Only one
+ * PIPE_CONTROL needs to be sent before any combination of VS
+ * associated 3DSTATE."
+ */
+ anv_batch_emit(&cmd_buffer->batch, GEN7_PIPE_CONTROL,
+ .DepthStallEnable = true,
+ .PostSyncOperation = WriteImmediateData,
+ .Address = { &cmd_buffer->device->workaround_bo, 0 });
}
- if (cmd_buffer->state.dirty &
- (ANV_CMD_BUFFER_PIPELINE_DIRTY | ANV_CMD_BUFFER_RS_DIRTY)) {
- anv_batch_emit_merge(&cmd_buffer->batch,
- cmd_buffer->state.rs_state->gen7.sf,
- pipeline->gen7.sf);
+ uint32_t dirty = 0;
+ if (cmd_buffer->state.descriptors_dirty) {
+ dirty = gen7_cmd_buffer_flush_descriptor_sets(cmd_buffer);
+ gen7_cmd_buffer_emit_descriptor_pointers(cmd_buffer, dirty);
}
- if (cmd_buffer->state.dirty &
- (ANV_CMD_BUFFER_PIPELINE_DIRTY | ANV_CMD_BUFFER_DS_DIRTY)) {
- struct anv_state state;
-
- if (cmd_buffer->state.ds_state == NULL)
- state = anv_cmd_buffer_emit_dynamic(cmd_buffer,
- pipeline->gen7.depth_stencil_state,
- GEN7_COLOR_CALC_STATE_length, 64);
- else
- state = anv_cmd_buffer_merge_dynamic(cmd_buffer,
- cmd_buffer->state.ds_state->gen7.depth_stencil_state,
- pipeline->gen7.depth_stencil_state,
- GEN7_DEPTH_STENCIL_STATE_length, 64);
- anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS,
- .PointertoDEPTH_STENCIL_STATE = state.offset);
+ if (cmd_buffer->state.push_constants_dirty)
+ cmd_buffer_flush_push_constants(cmd_buffer);
+
+ /* We use the gen8 state here because it only contains the additional
+ * min/max fields and, since they occur at the end of the packet and
+ * don't change the stride, they work on gen7 too.
+ */
+ if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_VIEWPORT)
+ gen8_cmd_buffer_emit_viewport(cmd_buffer);
+
+ if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_SCISSOR)
+ gen7_cmd_buffer_emit_scissor(cmd_buffer);
+
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_DYNAMIC_LINE_WIDTH |
+ ANV_CMD_DIRTY_DYNAMIC_DEPTH_BIAS)) {
+
+ bool enable_bias = cmd_buffer->state.dynamic.depth_bias.bias != 0.0f ||
+ cmd_buffer->state.dynamic.depth_bias.slope != 0.0f;
+
+ uint32_t sf_dw[GEN7_3DSTATE_SF_length];
+ struct GEN7_3DSTATE_SF sf = {
+ GEN7_3DSTATE_SF_header,
+ .LineWidth = cmd_buffer->state.dynamic.line_width,
+ .GlobalDepthOffsetEnableSolid = enable_bias,
+ .GlobalDepthOffsetEnableWireframe = enable_bias,
+ .GlobalDepthOffsetEnablePoint = enable_bias,
+ .GlobalDepthOffsetConstant = cmd_buffer->state.dynamic.depth_bias.bias,
+ .GlobalDepthOffsetScale = cmd_buffer->state.dynamic.depth_bias.slope,
+ .GlobalDepthOffsetClamp = cmd_buffer->state.dynamic.depth_bias.clamp
+ };
+ GEN7_3DSTATE_SF_pack(NULL, sf_dw, &sf);
+
+ anv_batch_emit_merge(&cmd_buffer->batch, sf_dw, pipeline->gen7.sf);
}
- if (cmd_buffer->state.dirty &
- (ANV_CMD_BUFFER_CB_DIRTY | ANV_CMD_BUFFER_DS_DIRTY)) {
- struct anv_state state;
- if (cmd_buffer->state.ds_state == NULL)
- state = anv_cmd_buffer_emit_dynamic(cmd_buffer,
- cmd_buffer->state.cb_state->color_calc_state,
- GEN7_COLOR_CALC_STATE_length, 64);
- else if (cmd_buffer->state.cb_state == NULL)
- state = anv_cmd_buffer_emit_dynamic(cmd_buffer,
- cmd_buffer->state.ds_state->gen7.color_calc_state,
- GEN7_COLOR_CALC_STATE_length, 64);
- else
- state = anv_cmd_buffer_merge_dynamic(cmd_buffer,
- cmd_buffer->state.ds_state->gen7.color_calc_state,
- cmd_buffer->state.cb_state->color_calc_state,
- GEN7_COLOR_CALC_STATE_length, 64);
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE)) {
+ struct anv_state cc_state =
+ anv_cmd_buffer_alloc_dynamic_state(cmd_buffer,
+ GEN7_COLOR_CALC_STATE_length * 4,
+ 64);
+ struct GEN7_COLOR_CALC_STATE cc = {
+ .BlendConstantColorRed = cmd_buffer->state.dynamic.blend_constants[0],
+ .BlendConstantColorGreen = cmd_buffer->state.dynamic.blend_constants[1],
+ .BlendConstantColorBlue = cmd_buffer->state.dynamic.blend_constants[2],
+ .BlendConstantColorAlpha = cmd_buffer->state.dynamic.blend_constants[3],
+ .StencilReferenceValue =
+ cmd_buffer->state.dynamic.stencil_reference.front,
+ .BackFaceStencilReferenceValue =
+ cmd_buffer->state.dynamic.stencil_reference.back,
+ };
+ GEN7_COLOR_CALC_STATE_pack(NULL, cc_state.map, &cc);
+ if (!cmd_buffer->device->info.has_llc)
+ anv_state_clflush(cc_state);
anv_batch_emit(&cmd_buffer->batch,
GEN7_3DSTATE_CC_STATE_POINTERS,
- .ColorCalcStatePointer = state.offset);
+ .ColorCalcStatePointer = cc_state.offset);
+ }
+
+ if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_RENDER_TARGETS |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK |
+ ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK)) {
+ uint32_t depth_stencil_dw[GEN7_DEPTH_STENCIL_STATE_length];
+
+ const struct anv_image_view *iview =
+ anv_cmd_buffer_get_depth_stencil_view(cmd_buffer);
+
+ struct GEN7_DEPTH_STENCIL_STATE depth_stencil = {
+ .StencilBufferWriteEnable = iview && (iview->aspect_mask & VK_IMAGE_ASPECT_STENCIL_BIT),
+
+ .StencilTestMask =
+ cmd_buffer->state.dynamic.stencil_compare_mask.front & 0xff,
+ .StencilWriteMask =
+ cmd_buffer->state.dynamic.stencil_write_mask.front & 0xff,
+
+ .BackfaceStencilTestMask =
+ cmd_buffer->state.dynamic.stencil_compare_mask.back & 0xff,
+ .BackfaceStencilWriteMask =
+ cmd_buffer->state.dynamic.stencil_write_mask.back & 0xff,
+ };
+ GEN7_DEPTH_STENCIL_STATE_pack(NULL, depth_stencil_dw, &depth_stencil);
+
+ struct anv_state ds_state =
+ anv_cmd_buffer_merge_dynamic(cmd_buffer, depth_stencil_dw,
+ pipeline->gen7.depth_stencil_state,
+ GEN7_DEPTH_STENCIL_STATE_length, 64);
+
+ anv_batch_emit(&cmd_buffer->batch,
+ GEN7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS,
+ .PointertoDEPTH_STENCIL_STATE = ds_state.offset);
}
if (cmd_buffer->state.gen7.index_buffer &&
- cmd_buffer->state.dirty & (ANV_CMD_BUFFER_PIPELINE_DIRTY |
- ANV_CMD_BUFFER_INDEX_BUFFER_DIRTY)) {
+ cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE |
+ ANV_CMD_DIRTY_INDEX_BUFFER)) {
struct anv_buffer *buffer = cmd_buffer->state.gen7.index_buffer;
uint32_t offset = cmd_buffer->state.gen7.index_offset;
+ if (ANV_IS_HASWELL) {
+ anv_batch_emit(&cmd_buffer->batch, GEN75_3DSTATE_VF,
+ .IndexedDrawCutIndexEnable = pipeline->primitive_restart,
+ .CutIndex = cmd_buffer->state.restart_index);
+ }
+
anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_INDEX_BUFFER,
.CutIndexEnable = pipeline->primitive_restart,
.IndexFormat = cmd_buffer->state.gen7.index_type,
cmd_buffer->state.dirty = 0;
}
-void gen7_CmdDraw(
- VkCmdBuffer cmdBuffer,
+void genX(CmdDraw)(
+ VkCommandBuffer commandBuffer,
uint32_t vertexCount,
uint32_t instanceCount,
uint32_t firstVertex,
uint32_t firstInstance)
{
- ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, cmdBuffer);
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
- gen7_cmd_buffer_flush_state(cmd_buffer);
+ cmd_buffer_flush_state(cmd_buffer);
anv_batch_emit(&cmd_buffer->batch, GEN7_3DPRIMITIVE,
.VertexAccessType = SEQUENTIAL,
.BaseVertexLocation = 0);
}
-void gen7_CmdDrawIndexed(
- VkCmdBuffer cmdBuffer,
+void genX(CmdDrawIndexed)(
+ VkCommandBuffer commandBuffer,
uint32_t indexCount,
uint32_t instanceCount,
uint32_t firstIndex,
int32_t vertexOffset,
uint32_t firstInstance)
{
- ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, cmdBuffer);
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
- gen7_cmd_buffer_flush_state(cmd_buffer);
+ cmd_buffer_flush_state(cmd_buffer);
anv_batch_emit(&cmd_buffer->batch, GEN7_3DPRIMITIVE,
.VertexAccessType = RANDOM,
#define GEN7_3DPRIM_START_INSTANCE 0x243C
#define GEN7_3DPRIM_BASE_VERTEX 0x2440
-void gen7_CmdDrawIndirect(
- VkCmdBuffer cmdBuffer,
+void genX(CmdDrawIndirect)(
+ VkCommandBuffer commandBuffer,
VkBuffer _buffer,
VkDeviceSize offset,
- uint32_t count,
+ uint32_t drawCount,
uint32_t stride)
{
- ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, cmdBuffer);
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
struct anv_bo *bo = buffer->bo;
uint32_t bo_offset = buffer->offset + offset;
- gen7_cmd_buffer_flush_state(cmd_buffer);
+ cmd_buffer_flush_state(cmd_buffer);
gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_VERTEX_COUNT, bo, bo_offset);
gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_INSTANCE_COUNT, bo, bo_offset + 4);
.PrimitiveTopologyType = pipeline->topology);
}
-void gen7_CmdDrawIndexedIndirect(
- VkCmdBuffer cmdBuffer,
+void genX(CmdDrawIndexedIndirect)(
+ VkCommandBuffer commandBuffer,
VkBuffer _buffer,
VkDeviceSize offset,
- uint32_t count,
+ uint32_t drawCount,
uint32_t stride)
{
- ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, cmdBuffer);
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
struct anv_pipeline *pipeline = cmd_buffer->state.pipeline;
struct anv_bo *bo = buffer->bo;
uint32_t bo_offset = buffer->offset + offset;
- gen7_cmd_buffer_flush_state(cmd_buffer);
+ cmd_buffer_flush_state(cmd_buffer);
gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_VERTEX_COUNT, bo, bo_offset);
gen7_batch_lrm(&cmd_buffer->batch, GEN7_3DPRIM_INSTANCE_COUNT, bo, bo_offset + 4);
.PrimitiveTopologyType = pipeline->topology);
}
-void gen7_CmdDispatch(
- VkCmdBuffer cmdBuffer,
+void genX(CmdDispatch)(
+ VkCommandBuffer commandBuffer,
uint32_t x,
uint32_t y,
uint32_t z)
{
- ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, cmdBuffer);
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
struct brw_cs_prog_data *prog_data = &pipeline->cs_prog_data;
- gen7_cmd_buffer_flush_compute_state(cmd_buffer);
+ cmd_buffer_flush_compute_state(cmd_buffer);
anv_batch_emit(&cmd_buffer->batch, GEN7_GPGPU_WALKER,
.SIMDSize = prog_data->simd_size / 16,
.ThreadDepthCounterMaximum = 0,
.ThreadHeightCounterMaximum = 0,
- .ThreadWidthCounterMaximum = pipeline->cs_thread_width_max,
+ .ThreadWidthCounterMaximum = pipeline->cs_thread_width_max - 1,
.ThreadGroupIDXDimension = x,
.ThreadGroupIDYDimension = y,
.ThreadGroupIDZDimension = z,
#define GPGPU_DISPATCHDIMY 0x2504
#define GPGPU_DISPATCHDIMZ 0x2508
-void gen7_CmdDispatchIndirect(
- VkCmdBuffer cmdBuffer,
+void genX(CmdDispatchIndirect)(
+ VkCommandBuffer commandBuffer,
VkBuffer _buffer,
VkDeviceSize offset)
{
- ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, cmdBuffer);
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
ANV_FROM_HANDLE(anv_buffer, buffer, _buffer);
struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
struct brw_cs_prog_data *prog_data = &pipeline->cs_prog_data;
struct anv_bo *bo = buffer->bo;
uint32_t bo_offset = buffer->offset + offset;
- gen7_cmd_buffer_flush_compute_state(cmd_buffer);
+ cmd_buffer_flush_compute_state(cmd_buffer);
gen7_batch_lrm(&cmd_buffer->batch, GPGPU_DISPATCHDIMX, bo, bo_offset);
gen7_batch_lrm(&cmd_buffer->batch, GPGPU_DISPATCHDIMY, bo, bo_offset + 4);
.SIMDSize = prog_data->simd_size / 16,
.ThreadDepthCounterMaximum = 0,
.ThreadHeightCounterMaximum = 0,
- .ThreadWidthCounterMaximum = pipeline->cs_thread_width_max,
+ .ThreadWidthCounterMaximum = pipeline->cs_thread_width_max - 1,
.RightExecutionMask = pipeline->cs_right_mask,
.BottomExecutionMask = 0xffffffff);
anv_batch_emit(&cmd_buffer->batch, GEN7_MEDIA_STATE_FLUSH);
}
-void gen7_CmdPipelineBarrier(
- VkCmdBuffer cmdBuffer,
- VkPipelineStageFlags srcStageMask,
- VkPipelineStageFlags destStageMask,
- VkBool32 byRegion,
- uint32_t memBarrierCount,
- const void* const* ppMemBarriers)
-{
- stub();
-}
-
static void
-gen7_cmd_buffer_emit_depth_stencil(struct anv_cmd_buffer *cmd_buffer)
+cmd_buffer_emit_depth_stencil(struct anv_cmd_buffer *cmd_buffer)
{
const struct anv_framebuffer *fb = cmd_buffer->state.framebuffer;
const struct anv_image_view *iview =
anv_cmd_buffer_get_depth_stencil_view(cmd_buffer);
const struct anv_image *image = iview ? iview->image : NULL;
- const bool has_depth = iview && iview->format->depth_format;
- const bool has_stencil = iview && iview->format->has_stencil;
+
+ /* XXX: isl needs to grow depth format support */
+ const struct anv_format *anv_format =
+ iview ? anv_format_for_vk_format(iview->vk_format) : NULL;
+
+ const bool has_depth = iview && anv_format->depth_format;
+ const bool has_stencil = iview && anv_format->has_stencil;
/* Emit 3DSTATE_DEPTH_BUFFER */
if (has_depth) {
- anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_DEPTH_BUFFER,
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_DEPTH_BUFFER),
.SurfaceType = SURFTYPE_2D,
- .DepthWriteEnable = iview->format->depth_format,
+ .DepthWriteEnable = true,
.StencilWriteEnable = has_stencil,
.HierarchicalDepthBufferEnable = false,
- .SurfaceFormat = iview->format->depth_format,
- .SurfacePitch = image->depth_surface.stride - 1,
+ .SurfaceFormat = anv_format->depth_format,
+ .SurfacePitch = image->depth_surface.isl.row_pitch - 1,
.SurfaceBaseAddress = {
.bo = image->bo,
.offset = image->depth_surface.offset,
.LOD = 0,
.Depth = 1 - 1,
.MinimumArrayElement = 0,
- .DepthBufferObjectControlState = GEN7_MOCS,
+ .DepthBufferObjectControlState = GENX(MOCS),
.RenderTargetViewExtent = 1 - 1);
} else {
/* Even when no depth buffer is present, the hardware requires that
* actual framebuffer's width and height, even when neither depth buffer
* nor stencil buffer is present.
*/
- anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_DEPTH_BUFFER,
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_DEPTH_BUFFER),
.SurfaceType = SURFTYPE_2D,
.SurfaceFormat = D16_UNORM,
.Width = fb->width - 1,
/* Emit 3DSTATE_STENCIL_BUFFER */
if (has_stencil) {
- anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_STENCIL_BUFFER,
- .StencilBufferObjectControlState = GEN7_MOCS,
+ anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_STENCIL_BUFFER),
+# if (ANV_IS_HASWELL)
+ .StencilBufferEnable = true,
+# endif
+ .StencilBufferObjectControlState = GENX(MOCS),
/* Stencil buffers have strange pitch. The PRM says:
*
* The pitch must be set to 2x the value computed based on width,
* as the stencil buffer is stored with two rows interleaved.
*/
- .SurfacePitch = 2 * image->stencil_surface.stride - 1,
+ .SurfacePitch = 2 * image->stencil_surface.isl.row_pitch - 1,
.SurfaceBaseAddress = {
.bo = image->bo,
anv_batch_emit(&cmd_buffer->batch, GEN7_3DSTATE_CLEAR_PARAMS);
}
-void
-gen7_cmd_buffer_begin_subpass(struct anv_cmd_buffer *cmd_buffer,
- struct anv_subpass *subpass)
+GENX_FUNC(GEN7, GEN7) void
+genX(cmd_buffer_begin_subpass)(struct anv_cmd_buffer *cmd_buffer,
+ struct anv_subpass *subpass)
{
cmd_buffer->state.subpass = subpass;
cmd_buffer->state.descriptors_dirty |= VK_SHADER_STAGE_FRAGMENT_BIT;
+ cmd_buffer->state.dirty |= ANV_CMD_DIRTY_RENDER_TARGETS;
- gen7_cmd_buffer_emit_depth_stencil(cmd_buffer);
+ cmd_buffer_emit_depth_stencil(cmd_buffer);
}
static void
pRenderPassBegin->pClearValues);
}
-void gen7_CmdBeginRenderPass(
- VkCmdBuffer cmdBuffer,
+void genX(CmdBeginRenderPass)(
+ VkCommandBuffer commandBuffer,
const VkRenderPassBeginInfo* pRenderPassBegin,
- VkRenderPassContents contents)
+ VkSubpassContents contents)
{
- ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, cmdBuffer);
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
ANV_FROM_HANDLE(anv_render_pass, pass, pRenderPassBegin->renderPass);
begin_render_pass(cmd_buffer, pRenderPassBegin);
gen7_cmd_buffer_begin_subpass(cmd_buffer, pass->subpasses);
}
-void gen7_CmdNextSubpass(
- VkCmdBuffer cmdBuffer,
- VkRenderPassContents contents)
+void genX(CmdNextSubpass)(
+ VkCommandBuffer commandBuffer,
+ VkSubpassContents contents)
{
- ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, cmdBuffer);
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
- assert(cmd_buffer->level == VK_CMD_BUFFER_LEVEL_PRIMARY);
+ assert(cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_PRIMARY);
gen7_cmd_buffer_begin_subpass(cmd_buffer, cmd_buffer->state.subpass + 1);
}
-void gen7_CmdEndRenderPass(
- VkCmdBuffer cmdBuffer)
+void genX(CmdEndRenderPass)(
+ VkCommandBuffer commandBuffer)
{
- ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, cmdBuffer);
+ ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
/* Emit a flushing pipe control at the end of a pass. This is kind of a
* hack but it ensures that render targets always actually get written.
.TextureCacheInvalidationEnable = true,
.CommandStreamerStallEnable = true);
}
+
+void genX(CmdSetEvent)(
+ VkCommandBuffer commandBuffer,
+ VkEvent event,
+ VkPipelineStageFlags stageMask)
+{
+ stub();
+}
+
+void genX(CmdResetEvent)(
+ VkCommandBuffer commandBuffer,
+ VkEvent event,
+ VkPipelineStageFlags stageMask)
+{
+ stub();
+}
+
+void genX(CmdWaitEvents)(
+ VkCommandBuffer commandBuffer,
+ uint32_t eventCount,
+ const VkEvent* pEvents,
+ VkPipelineStageFlags srcStageMask,
+ VkPipelineStageFlags destStageMask,
+ uint32_t memBarrierCount,
+ const void* const* ppMemBarriers)
+{
+ stub();
+}