I am working on a deferred renderer. I tried to set up the lighting pass (not an accurate term in the context of dynamic rendering but you get the idea) with dynamic rendering:
VkRenderingAttachmentInfo color_attachment_0 = {};
color_attachment_0.sType = VK_STRUCTURE_TYPE_RENDERING_ATTACHMENT_INFO;
color_attachment_0.imageView = _lit_image->vk_view;
color_attachment_0.imageLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
color_attachment_0.resolveMode = VK_RESOLVE_MODE_NONE;
color_attachment_0.resolveImageView = VK_NULL_HANDLE;
color_attachment_0.resolveImageLayout = VK_IMAGE_LAYOUT_UNDEFINED;
color_attachment_0.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR;
color_attachment_0.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
color_attachment_0.clearValue.color.float32[0] = 0.0f;
color_attachment_0.clearValue.color.float32[1] = 0.0f;
color_attachment_0.clearValue.color.float32[2] = 0.0f;
color_attachment_0.clearValue.color.float32[3] = 1.0f;
VkRenderingInfo pass_begin = {};
pass_begin.sType = VK_STRUCTURE_TYPE_RENDERING_INFO;
pass_begin.renderArea = { { 0, 0 }, { window_width, window_height } };
pass_begin.layerCount = 1;
pass_begin.viewMask = 0;
pass_begin.colorAttachmentCount = 1;
pass_begin.pColorAttachments = &color_attachment_0;
pass_begin.pDepthAttachment = VK_NULL_HANDLE;
pass_begin.pStencilAttachment = VK_NULL_HANDLE;
vkCmdBeginRendering(cmd_buf->vk_command_buffer, &pass_begin);
// bind pipeline
// ...
vkCmdEndRendering();
Visual Studio debugger throws exception 0xC0000005: Access violation reading location 0x0000000000000068at vkCmdBeginRendering. Validation layer did not give any error.
I checked _lit_image->vk_view at runtime, the handle was valid. I don't think any pointer, any object in the snippet above accidentally got invalidated either.
What could possibly be the problem?
Before the lighting pass, I had G-pass carried out with dynamic rendering. The G-pass had zero issue because I managed to show every G-buffer on screen. So I don't think the exception was caused by vkCmdBeginRendering() function pointer.
I tried pointing pDepthAttachment to a depth attachment, the exception is gone. This doesn't make much sense to me as why would a null depth attachment pointer cause a crash
