#include "tu_private.h"
#include "vk_util.h"
+#include "vk_format.h"
+
+static void update_samples(struct tu_subpass *subpass,
+ VkSampleCountFlagBits samples)
+{
+ assert(subpass->samples == 0 || subpass->samples == samples);
+ subpass->samples = samples;
+}
+
+#define GMEM_ALIGN 0x4000
+
+static void
+create_render_pass_common(struct tu_render_pass *pass,
+ const struct tu_physical_device *phys_dev)
+{
+ /* calculate total bytes per pixel */
+ uint32_t cpp_total = 0;
+ for (uint32_t i = 0; i < pass->attachment_count; i++) {
+ struct tu_render_pass_attachment *att = &pass->attachments[i];
+ if (att->gmem_offset >= 0)
+ cpp_total += att->cpp;
+ }
+
+ /* no gmem attachments */
+ if (cpp_total == 0) {
+ /* any value non-zero value so tiling config works with no attachments */
+ pass->gmem_pixels = 1024*1024;
+ return;
+ }
+
+ /* TODO: using ccu_offset_gmem so that BLIT_OP_SCALE resolve path
+ * doesn't break things. maybe there is a better solution?
+ * TODO: this algorithm isn't optimal
+ * for example, two attachments with cpp = {1, 4}
+ * result: nblocks = {12, 52}, pixels = 196608
+ * optimal: nblocks = {13, 51}, pixels = 208896
+ */
+ uint32_t gmem_blocks = phys_dev->ccu_offset_gmem / GMEM_ALIGN;
+ uint32_t offset = 0, pixels = ~0u;
+ for (uint32_t i = 0; i < pass->attachment_count; i++) {
+ struct tu_render_pass_attachment *att = &pass->attachments[i];
+ if (att->gmem_offset < 0)
+ continue;
+
+ att->gmem_offset = offset;
+
+ /* Note: divide by 16 is for GMEM_ALIGN=16k, tile align w=64/h=16 */
+ uint32_t align = MAX2(1, att->cpp / 16);
+ uint32_t nblocks = MAX2((gmem_blocks * att->cpp / cpp_total) & ~(align - 1), align);
+
+ gmem_blocks -= nblocks;
+ cpp_total -= att->cpp;
+ offset += nblocks * GMEM_ALIGN;
+ pixels = MIN2(pixels, nblocks * GMEM_ALIGN / att->cpp);
+ }
+
+ pass->gmem_pixels = pixels;
+
+ for (uint32_t i = 0; i < pass->subpass_count; i++) {
+ struct tu_subpass *subpass = &pass->subpasses[i];
+
+ subpass->srgb_cntl = 0;
+ subpass->render_components = 0;
+
+ for (uint32_t i = 0; i < subpass->color_count; ++i) {
+ uint32_t a = subpass->color_attachments[i].attachment;
+ if (a == VK_ATTACHMENT_UNUSED)
+ continue;
+
+ subpass->render_components |= 0xf << (i * 4);
+
+ if (vk_format_is_srgb(pass->attachments[a].format))
+ subpass->srgb_cntl |= 1 << i;
+ }
+ }
+
+ /* disable unused attachments */
+ for (uint32_t i = 0; i < pass->attachment_count; i++) {
+ struct tu_render_pass_attachment *att = &pass->attachments[i];
+ if (att->gmem_offset < 0) {
+ att->clear_mask = 0;
+ att->load = false;
+ }
+ }
+}
+
+static void
+attachment_set_ops(struct tu_render_pass_attachment *att,
+ VkAttachmentLoadOp load_op,
+ VkAttachmentLoadOp stencil_load_op,
+ VkAttachmentStoreOp store_op,
+ VkAttachmentStoreOp stencil_store_op)
+{
+ /* load/store ops */
+ att->clear_mask =
+ (load_op == VK_ATTACHMENT_LOAD_OP_CLEAR) ? VK_IMAGE_ASPECT_COLOR_BIT : 0;
+ att->load = (load_op == VK_ATTACHMENT_LOAD_OP_LOAD);
+ att->store = (store_op == VK_ATTACHMENT_STORE_OP_STORE);
+
+ bool stencil_clear = (stencil_load_op == VK_ATTACHMENT_LOAD_OP_CLEAR);
+ bool stencil_load = (stencil_load_op == VK_ATTACHMENT_LOAD_OP_LOAD);
+ bool stencil_store = (stencil_store_op == VK_ATTACHMENT_STORE_OP_STORE);
+
+ switch (att->format) {
+ case VK_FORMAT_D24_UNORM_S8_UINT: /* || stencil load/store */
+ if (att->clear_mask)
+ att->clear_mask = VK_IMAGE_ASPECT_DEPTH_BIT;
+ if (stencil_clear)
+ att->clear_mask |= VK_IMAGE_ASPECT_STENCIL_BIT;
+ if (stencil_load)
+ att->load = true;
+ if (stencil_store)
+ att->store = true;
+ break;
+ case VK_FORMAT_S8_UINT: /* replace load/store with stencil load/store */
+ att->clear_mask = stencil_clear ? VK_IMAGE_ASPECT_COLOR_BIT : 0;
+ att->load = stencil_load;
+ att->store = stencil_store;
+ break;
+ default:
+ break;
+ }
+}
VkResult
tu_CreateRenderPass(VkDevice _device,
struct tu_render_pass *pass;
size_t size;
size_t attachments_offset;
- VkRenderPassMultiviewCreateInfoKHR *multiview_info = NULL;
assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO);
pass->subpass_count = pCreateInfo->subpassCount;
pass->attachments = (void *) pass + attachments_offset;
- vk_foreach_struct(ext, pCreateInfo->pNext)
- {
- switch (ext->sType) {
- case VK_STRUCTURE_TYPE_RENDER_PASS_MULTIVIEW_CREATE_INFO_KHR:
- multiview_info = (VkRenderPassMultiviewCreateInfoKHR *) ext;
- break;
- default:
- break;
- }
- }
-
for (uint32_t i = 0; i < pCreateInfo->attachmentCount; i++) {
struct tu_render_pass_attachment *att = &pass->attachments[i];
att->format = pCreateInfo->pAttachments[i].format;
att->samples = pCreateInfo->pAttachments[i].samples;
- att->load_op = pCreateInfo->pAttachments[i].loadOp;
- att->stencil_load_op = pCreateInfo->pAttachments[i].stencilLoadOp;
- att->initial_layout = pCreateInfo->pAttachments[i].initialLayout;
- att->final_layout = pCreateInfo->pAttachments[i].finalLayout;
- // att->store_op = pCreateInfo->pAttachments[i].storeOp;
- // att->stencil_store_op = pCreateInfo->pAttachments[i].stencilStoreOp;
+ att->cpp = vk_format_get_blocksize(att->format) * att->samples;
+ att->gmem_offset = -1;
+
+ attachment_set_ops(att,
+ pCreateInfo->pAttachments[i].loadOp,
+ pCreateInfo->pAttachments[i].stencilLoadOp,
+ pCreateInfo->pAttachments[i].storeOp,
+ pCreateInfo->pAttachments[i].stencilStoreOp);
}
+
uint32_t subpass_attachment_count = 0;
struct tu_subpass_attachment *p;
for (uint32_t i = 0; i < pCreateInfo->subpassCount; i++) {
subpass_attachment_count +=
desc->inputAttachmentCount + desc->colorAttachmentCount +
- (desc->pResolveAttachments ? desc->colorAttachmentCount : 0) +
- (desc->pDepthStencilAttachment != NULL);
+ (desc->pResolveAttachments ? desc->colorAttachmentCount : 0);
}
if (subpass_attachment_count) {
p = pass->subpass_attachments;
for (uint32_t i = 0; i < pCreateInfo->subpassCount; i++) {
const VkSubpassDescription *desc = &pCreateInfo->pSubpasses[i];
- uint32_t color_sample_count = 1, depth_sample_count = 1;
struct tu_subpass *subpass = &pass->subpasses[i];
subpass->input_count = desc->inputAttachmentCount;
subpass->color_count = desc->colorAttachmentCount;
- if (multiview_info)
- subpass->view_mask = multiview_info->pViewMasks[i];
+ subpass->samples = 0;
if (desc->inputAttachmentCount > 0) {
subpass->input_attachments = p;
p += desc->inputAttachmentCount;
for (uint32_t j = 0; j < desc->inputAttachmentCount; j++) {
- subpass->input_attachments[j] = (struct tu_subpass_attachment) {
- .attachment = desc->pInputAttachments[j].attachment,
- .layout = desc->pInputAttachments[j].layout,
- };
- if (desc->pInputAttachments[j].attachment != VK_ATTACHMENT_UNUSED)
- pass->attachments[desc->pInputAttachments[j].attachment]
- .view_mask |= subpass->view_mask;
+ uint32_t a = desc->pInputAttachments[j].attachment;
+ subpass->input_attachments[j].attachment = a;
+ if (a != VK_ATTACHMENT_UNUSED)
+ pass->attachments[a].gmem_offset = 0;
}
}
p += desc->colorAttachmentCount;
for (uint32_t j = 0; j < desc->colorAttachmentCount; j++) {
- subpass->color_attachments[j] = (struct tu_subpass_attachment) {
- .attachment = desc->pColorAttachments[j].attachment,
- .layout = desc->pColorAttachments[j].layout,
- };
- if (desc->pColorAttachments[j].attachment !=
- VK_ATTACHMENT_UNUSED) {
- pass->attachments[desc->pColorAttachments[j].attachment]
- .view_mask |= subpass->view_mask;
- color_sample_count =
- pCreateInfo
- ->pAttachments[desc->pColorAttachments[j].attachment]
- .samples;
+ uint32_t a = desc->pColorAttachments[j].attachment;
+ subpass->color_attachments[j].attachment = a;
+
+ if (a != VK_ATTACHMENT_UNUSED) {
+ pass->attachments[a].gmem_offset = 0;
+ update_samples(subpass, pCreateInfo->pAttachments[a].samples);
}
}
}
- subpass->has_resolve = false;
+ subpass->resolve_attachments = desc->pResolveAttachments ? p : NULL;
if (desc->pResolveAttachments) {
- subpass->resolve_attachments = p;
p += desc->colorAttachmentCount;
-
for (uint32_t j = 0; j < desc->colorAttachmentCount; j++) {
- uint32_t a = desc->pResolveAttachments[j].attachment;
- subpass->resolve_attachments[j] = (struct tu_subpass_attachment) {
- .attachment = desc->pResolveAttachments[j].attachment,
- .layout = desc->pResolveAttachments[j].layout,
- };
- if (a != VK_ATTACHMENT_UNUSED) {
- subpass->has_resolve = true;
- pass->attachments[desc->pResolveAttachments[j].attachment]
- .view_mask |= subpass->view_mask;
- }
+ subpass->resolve_attachments[j].attachment =
+ desc->pResolveAttachments[j].attachment;
}
}
- if (desc->pDepthStencilAttachment) {
- subpass->depth_stencil_attachment = (struct tu_subpass_attachment) {
- .attachment = desc->pDepthStencilAttachment->attachment,
- .layout = desc->pDepthStencilAttachment->layout,
- };
- if (desc->pDepthStencilAttachment->attachment !=
- VK_ATTACHMENT_UNUSED) {
- pass->attachments[desc->pDepthStencilAttachment->attachment]
- .view_mask |= subpass->view_mask;
- depth_sample_count =
- pCreateInfo
- ->pAttachments[desc->pDepthStencilAttachment->attachment]
- .samples;
- }
- } else {
- subpass->depth_stencil_attachment.attachment = VK_ATTACHMENT_UNUSED;
+ uint32_t a = desc->pDepthStencilAttachment ?
+ desc->pDepthStencilAttachment->attachment : VK_ATTACHMENT_UNUSED;
+ subpass->depth_stencil_attachment.attachment = a;
+ if (a != VK_ATTACHMENT_UNUSED) {
+ pass->attachments[a].gmem_offset = 0;
+ update_samples(subpass, pCreateInfo->pAttachments[a].samples);
}
- subpass->max_sample_count =
- MAX2(color_sample_count, depth_sample_count);
- }
-
- for (unsigned i = 0; i < pCreateInfo->dependencyCount; ++i) {
- uint32_t dst = pCreateInfo->pDependencies[i].dstSubpass;
- if (dst == VK_SUBPASS_EXTERNAL) {
- pass->end_barrier.src_stage_mask =
- pCreateInfo->pDependencies[i].srcStageMask;
- pass->end_barrier.src_access_mask =
- pCreateInfo->pDependencies[i].srcAccessMask;
- pass->end_barrier.dst_access_mask =
- pCreateInfo->pDependencies[i].dstAccessMask;
- } else {
- pass->subpasses[dst].start_barrier.src_stage_mask =
- pCreateInfo->pDependencies[i].srcStageMask;
- pass->subpasses[dst].start_barrier.src_access_mask =
- pCreateInfo->pDependencies[i].srcAccessMask;
- pass->subpasses[dst].start_barrier.dst_access_mask =
- pCreateInfo->pDependencies[i].dstAccessMask;
- }
+ subpass->samples = subpass->samples ?: 1;
}
*pRenderPass = tu_render_pass_to_handle(pass);
+ create_render_pass_common(pass, device->physical_device);
+
return VK_SUCCESS;
}
VkResult
-tu_CreateRenderPass2KHR(VkDevice _device,
- const VkRenderPassCreateInfo2KHR *pCreateInfo,
- const VkAllocationCallbacks *pAllocator,
- VkRenderPass *pRenderPass)
+tu_CreateRenderPass2(VkDevice _device,
+ const VkRenderPassCreateInfo2KHR *pCreateInfo,
+ const VkAllocationCallbacks *pAllocator,
+ VkRenderPass *pRenderPass)
{
TU_FROM_HANDLE(tu_device, device, _device);
struct tu_render_pass *pass;
size_t size;
size_t attachments_offset;
- assert(pCreateInfo->sType ==
- VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO_2_KHR);
+ assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO_2_KHR);
size = sizeof(*pass);
size += pCreateInfo->subpassCount * sizeof(pass->subpasses[0]);
att->format = pCreateInfo->pAttachments[i].format;
att->samples = pCreateInfo->pAttachments[i].samples;
- att->load_op = pCreateInfo->pAttachments[i].loadOp;
- att->stencil_load_op = pCreateInfo->pAttachments[i].stencilLoadOp;
- att->initial_layout = pCreateInfo->pAttachments[i].initialLayout;
- att->final_layout = pCreateInfo->pAttachments[i].finalLayout;
- // att->store_op = pCreateInfo->pAttachments[i].storeOp;
- // att->stencil_store_op = pCreateInfo->pAttachments[i].stencilStoreOp;
+ att->cpp = vk_format_get_blocksize(att->format) * att->samples;
+ att->gmem_offset = -1;
+
+ attachment_set_ops(att,
+ pCreateInfo->pAttachments[i].loadOp,
+ pCreateInfo->pAttachments[i].stencilLoadOp,
+ pCreateInfo->pAttachments[i].storeOp,
+ pCreateInfo->pAttachments[i].stencilStoreOp);
}
uint32_t subpass_attachment_count = 0;
struct tu_subpass_attachment *p;
subpass_attachment_count +=
desc->inputAttachmentCount + desc->colorAttachmentCount +
- (desc->pResolveAttachments ? desc->colorAttachmentCount : 0) +
- (desc->pDepthStencilAttachment != NULL);
+ (desc->pResolveAttachments ? desc->colorAttachmentCount : 0);
}
if (subpass_attachment_count) {
p = pass->subpass_attachments;
for (uint32_t i = 0; i < pCreateInfo->subpassCount; i++) {
const VkSubpassDescription2KHR *desc = &pCreateInfo->pSubpasses[i];
- uint32_t color_sample_count = 1, depth_sample_count = 1;
struct tu_subpass *subpass = &pass->subpasses[i];
subpass->input_count = desc->inputAttachmentCount;
subpass->color_count = desc->colorAttachmentCount;
- subpass->view_mask = desc->viewMask;
+ subpass->samples = 0;
if (desc->inputAttachmentCount > 0) {
subpass->input_attachments = p;
p += desc->inputAttachmentCount;
for (uint32_t j = 0; j < desc->inputAttachmentCount; j++) {
- subpass->input_attachments[j] = (struct tu_subpass_attachment) {
- .attachment = desc->pInputAttachments[j].attachment,
- .layout = desc->pInputAttachments[j].layout,
- };
- if (desc->pInputAttachments[j].attachment != VK_ATTACHMENT_UNUSED)
- pass->attachments[desc->pInputAttachments[j].attachment]
- .view_mask |= subpass->view_mask;
+ uint32_t a = desc->pInputAttachments[j].attachment;
+ subpass->input_attachments[j].attachment = a;
+ if (a != VK_ATTACHMENT_UNUSED)
+ pass->attachments[a].gmem_offset = 0;
}
}
p += desc->colorAttachmentCount;
for (uint32_t j = 0; j < desc->colorAttachmentCount; j++) {
- subpass->color_attachments[j] = (struct tu_subpass_attachment) {
- .attachment = desc->pColorAttachments[j].attachment,
- .layout = desc->pColorAttachments[j].layout,
- };
- if (desc->pColorAttachments[j].attachment !=
- VK_ATTACHMENT_UNUSED) {
- pass->attachments[desc->pColorAttachments[j].attachment]
- .view_mask |= subpass->view_mask;
- color_sample_count =
- pCreateInfo
- ->pAttachments[desc->pColorAttachments[j].attachment]
- .samples;
+ uint32_t a = desc->pColorAttachments[j].attachment;
+ subpass->color_attachments[j].attachment = a;
+
+ if (a != VK_ATTACHMENT_UNUSED) {
+ pass->attachments[a].gmem_offset = 0;
+ update_samples(subpass, pCreateInfo->pAttachments[a].samples);
}
}
}
- subpass->has_resolve = false;
+ subpass->resolve_attachments = desc->pResolveAttachments ? p : NULL;
if (desc->pResolveAttachments) {
- subpass->resolve_attachments = p;
p += desc->colorAttachmentCount;
-
for (uint32_t j = 0; j < desc->colorAttachmentCount; j++) {
- uint32_t a = desc->pResolveAttachments[j].attachment;
- subpass->resolve_attachments[j] = (struct tu_subpass_attachment) {
- .attachment = desc->pResolveAttachments[j].attachment,
- .layout = desc->pResolveAttachments[j].layout,
- };
- if (a != VK_ATTACHMENT_UNUSED) {
- subpass->has_resolve = true;
- pass->attachments[desc->pResolveAttachments[j].attachment]
- .view_mask |= subpass->view_mask;
- }
+ subpass->resolve_attachments[j].attachment =
+ desc->pResolveAttachments[j].attachment;
}
}
- if (desc->pDepthStencilAttachment) {
- subpass->depth_stencil_attachment = (struct tu_subpass_attachment) {
- .attachment = desc->pDepthStencilAttachment->attachment,
- .layout = desc->pDepthStencilAttachment->layout,
- };
- if (desc->pDepthStencilAttachment->attachment !=
- VK_ATTACHMENT_UNUSED) {
- pass->attachments[desc->pDepthStencilAttachment->attachment]
- .view_mask |= subpass->view_mask;
- depth_sample_count =
- pCreateInfo
- ->pAttachments[desc->pDepthStencilAttachment->attachment]
- .samples;
- }
- } else {
- subpass->depth_stencil_attachment.attachment = VK_ATTACHMENT_UNUSED;
- }
-
- subpass->max_sample_count =
- MAX2(color_sample_count, depth_sample_count);
- }
- for (unsigned i = 0; i < pCreateInfo->dependencyCount; ++i) {
- uint32_t dst = pCreateInfo->pDependencies[i].dstSubpass;
- if (dst == VK_SUBPASS_EXTERNAL) {
- pass->end_barrier.src_stage_mask =
- pCreateInfo->pDependencies[i].srcStageMask;
- pass->end_barrier.src_access_mask =
- pCreateInfo->pDependencies[i].srcAccessMask;
- pass->end_barrier.dst_access_mask =
- pCreateInfo->pDependencies[i].dstAccessMask;
- } else {
- pass->subpasses[dst].start_barrier.src_stage_mask =
- pCreateInfo->pDependencies[i].srcStageMask;
- pass->subpasses[dst].start_barrier.src_access_mask =
- pCreateInfo->pDependencies[i].srcAccessMask;
- pass->subpasses[dst].start_barrier.dst_access_mask =
- pCreateInfo->pDependencies[i].dstAccessMask;
+ uint32_t a = desc->pDepthStencilAttachment ?
+ desc->pDepthStencilAttachment->attachment : VK_ATTACHMENT_UNUSED;
+ subpass->depth_stencil_attachment.attachment = a;
+ if (a != VK_ATTACHMENT_UNUSED) {
+ pass->attachments[a].gmem_offset = 0;
+ update_samples(subpass, pCreateInfo->pAttachments[a].samples);
}
+
+ subpass->samples = subpass->samples ?: 1;
}
*pRenderPass = tu_render_pass_to_handle(pass);
+ create_render_pass_common(pass, device->physical_device);
+
return VK_SUCCESS;
}
if (!_pass)
return;
+
vk_free2(&device->alloc, pAllocator, pass->subpass_attachments);
vk_free2(&device->alloc, pAllocator, pass);
}
void
-tu_GetRenderAreaGranularity(VkDevice device,
+tu_GetRenderAreaGranularity(VkDevice _device,
VkRenderPass renderPass,
VkExtent2D *pGranularity)
{
- pGranularity->width = 1;
- pGranularity->height = 1;
+ pGranularity->width = GMEM_ALIGN_W;
+ pGranularity->height = GMEM_ALIGN_H;
}
projects / mesa.git / blobdiff