panvk: Emit sample count and tile size when emitting framebuffer/tiler descriptors

This move the logic around tile size budget to be able to handle
rasterization_samples properly in case no render target is defined.

We now select the tile size right before emitting the tiler and framebuffer
descriptors.

This fixes "dEQP-VK.pipeline.monolithic.multisample.mixed_count.*" on
v12 when 64x64 tile size is allowed.

Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34674>

src/panfrost/vulkan/csf/panvk_vX_cmd_draw.c

@@ -835,6 +835,10 @@ get_tiler_desc(struct panvk_cmd_buffer *cmdbuf)
 
    const struct pan_fb_info *fbinfo = &cmdbuf->state.gfx.render.fb.info;
 
+   /* At this point, we should know sample count and the tile size should have
+    * been calculated */
+   assert(fbinfo->nr_samples > 0 && fbinfo->tile_size > 0);
+
    pan_pack(&tiler_tmpl, TILER_CONTEXT, cfg) {
       unsigned max_levels = tiler_features.max_levels;
       assert(max_levels >= 2);
@@ -1119,6 +1123,11 @@ get_fb_descs(struct panvk_cmd_buffer *cmdbuf)
 
    struct panvk_device *dev = to_panvk_device(cmdbuf->vk.base.device);
    struct pan_fb_info *fbinfo = &cmdbuf->state.gfx.render.fb.info;
+
+   /* At this point, we should know sample count and the tile size should have
+    * been calculated */
+   assert(fbinfo->nr_samples > 0 && fbinfo->tile_size > 0);
+
    bool simul_use =
       cmdbuf->flags & VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT;
 
@@ -1305,6 +1314,8 @@ set_provoking_vertex_mode(struct panvk_cmd_buffer *cmdbuf)
 static VkResult
 get_render_ctx(struct panvk_cmd_buffer *cmdbuf)
 {
+   panvk_per_arch(cmd_select_tile_size)(cmdbuf);
+
    VkResult result = get_tiler_desc(cmdbuf);
    if (result != VK_SUCCESS)
       return result;
@@ -1889,6 +1900,31 @@ prepare_draw(struct panvk_cmd_buffer *cmdbuf, struct panvk_draw_info *draw)
    if (result != VK_SUCCESS)
       return result;
 
+   if (!cmdbuf->vk.dynamic_graphics_state.rs.rasterizer_discard_enable) {
+      const struct pan_fb_info *fbinfo = &cmdbuf->state.gfx.render.fb.info;
+      uint32_t *nr_samples = &cmdbuf->state.gfx.render.fb.nr_samples;
+      uint32_t rasterization_samples =
+         cmdbuf->vk.dynamic_graphics_state.ms.rasterization_samples;
+
+      /* If there's no attachment, we patch nr_samples to match
+       * rasterization_samples, otherwise, we make sure those two numbers match.
+       */
+      if (!cmdbuf->state.gfx.render.bound_attachments) {
+         assert(rasterization_samples > 0);
+         *nr_samples = rasterization_samples;
+      } else {
+         assert(rasterization_samples == *nr_samples);
+      }
+
+      /* In case we already emitted tiler/framebuffer descriptors, we ensure
+       * that the sample count didn't change
+       * XXX: This currently can happen in case we resume a render pass with no
+       * attachements and without any draw as the FBD is emitted when suspending.
+       */
+      assert(fbinfo->nr_samples == 0 ||
+             fbinfo->nr_samples == cmdbuf->state.gfx.render.fb.nr_samples);
+   }
+
    if (!inherits_render_ctx(cmdbuf)) {
       result = get_render_ctx(cmdbuf);
       if (result != VK_SUCCESS)
@@ -1984,22 +2020,6 @@ panvk_cmd_draw(struct panvk_cmd_buffer *cmdbuf, struct panvk_draw_info *draw)
    cmdbuf->state.gfx.fs.required =
       fs_required(&cmdbuf->state.gfx, &cmdbuf->vk.dynamic_graphics_state);
 
-   if (!cmdbuf->vk.dynamic_graphics_state.rs.rasterizer_discard_enable) {
-      struct pan_fb_info *fbinfo = &cmdbuf->state.gfx.render.fb.info;
-      uint32_t rasterization_samples =
-         cmdbuf->vk.dynamic_graphics_state.ms.rasterization_samples;
-
-      /* If there's no attachment, we patch nr_samples to match
-       * rasterization_samples, otherwise, we make sure those two numbers match.
-       */
-      if (!cmdbuf->state.gfx.render.bound_attachments) {
-         assert(rasterization_samples > 0);
-         fbinfo->nr_samples = rasterization_samples;
-      } else {
-         assert(rasterization_samples == fbinfo->nr_samples);
-      }
-   }
-
    result = prepare_draw(cmdbuf, draw);
    if (result != VK_SUCCESS)
       return;
@@ -2162,22 +2182,6 @@ panvk_cmd_draw_indirect(struct panvk_cmd_buffer *cmdbuf,
    cmdbuf->state.gfx.fs.required =
       fs_required(&cmdbuf->state.gfx, &cmdbuf->vk.dynamic_graphics_state);
 
-   if (!cmdbuf->vk.dynamic_graphics_state.rs.rasterizer_discard_enable) {
-      struct pan_fb_info *fbinfo = &cmdbuf->state.gfx.render.fb.info;
-      uint32_t rasterization_samples =
-         cmdbuf->vk.dynamic_graphics_state.ms.rasterization_samples;
-
-      /* If there's no attachment, we patch nr_samples to match
-       * rasterization_samples, otherwise, we make sure those two numbers match.
-       */
-      if (!cmdbuf->state.gfx.render.bound_attachments) {
-         assert(rasterization_samples > 0);
-         fbinfo->nr_samples = rasterization_samples;
-      } else {
-         assert(rasterization_samples == fbinfo->nr_samples);
-      }
-   }
-
    /* Layered indirect draw (VK_EXT_shader_viewport_index_layer) needs
     * additional changes. We allow layer_count == 0 because that happens
     * when mixing dynamic rendering and secondary command buffers. Once
@@ -2409,12 +2413,18 @@ panvk_per_arch(cmd_inherit_render_state)(
    cmdbuf->state.gfx.render.layer_count = inheritance_info->viewMask ?
       util_last_bit(inheritance_info->viewMask) :
       0;
+
+   /* If a draw was performed, the inherited sample count should match our current sample count */
+   assert(fbinfo->nr_samples == 0 || inheritance_info->rasterizationSamples == fbinfo->nr_samples);
    *fbinfo = (struct pan_fb_info){
       .tile_buf_budget = panfrost_query_optimal_tib_size(phys_dev->model),
       .z_tile_buf_budget = panfrost_query_optimal_z_tib_size(phys_dev->model),
+      .tile_size = fbinfo->tile_size,
+      .cbuf_allocation = fbinfo->cbuf_allocation,
       .nr_samples = inheritance_info->rasterizationSamples,
       .rt_count = inheritance_info->colorAttachmentCount,
    };
+   cmdbuf->state.gfx.render.fb.nr_samples = inheritance_info->rasterizationSamples;
 
    assert(inheritance_info->colorAttachmentCount <= ARRAY_SIZE(fbinfo->rts));
 
@@ -2890,6 +2900,10 @@ panvk_per_arch(CmdEndRendering)(VkCommandBuffer commandBuffer)
 
    if (!suspending) {
       struct pan_fb_info *fbinfo = &cmdbuf->state.gfx.render.fb.info;
+
+      /* If no draw was performed, we should ensure sample count is valid and that we emit tile size */
+      panvk_per_arch(cmd_select_tile_size)(cmdbuf);
+
       bool clear = fbinfo->zs.clear.z | fbinfo->zs.clear.s;
       for (unsigned i = 0; i < fbinfo->rt_count; i++)
          clear |= fbinfo->rts[i].clear;

src/panfrost/vulkan/jm/panvk_vX_cmd_buffer.c

@@ -137,6 +137,12 @@ panvk_per_arch(cmd_close_batch)(struct panvk_cmd_buffer *cmdbuf)
       GENX(pan_emit_tls)(&batch->tlsinfo, batch->tls.cpu);
 
    if (batch->fb.desc.cpu) {
+      panvk_per_arch(cmd_select_tile_size)(cmdbuf);
+
+      /* At this point, we should know sample count and the tile size should have
+       * been calculated */
+      assert(fbinfo->nr_samples > 0 && fbinfo->tile_size > 0);
+
       fbinfo->sample_positions = dev->sample_positions->addr.dev +
                                  panfrost_sample_positions_offset(
                                     pan_sample_pattern(fbinfo->nr_samples));

src/panfrost/vulkan/jm/panvk_vX_cmd_draw.c

@@ -15,6 +15,7 @@
 #include "panvk_cmd_alloc.h"
 #include "panvk_cmd_buffer.h"
 #include "panvk_cmd_desc_state.h"
+#include "panvk_cmd_draw.h"
 #include "panvk_cmd_meta.h"
 #include "panvk_device.h"
 #include "panvk_entrypoints.h"
@@ -1181,7 +1182,8 @@ panvk_cmd_draw(struct panvk_cmd_buffer *cmdbuf, struct panvk_draw_data *draw)
    set_provoking_vertex_mode(cmdbuf);
 
    if (!rs->rasterizer_discard_enable) {
-      struct pan_fb_info *fbinfo = &cmdbuf->state.gfx.render.fb.info;
+      const struct pan_fb_info *fbinfo = &cmdbuf->state.gfx.render.fb.info;
+      uint32_t *nr_samples = &cmdbuf->state.gfx.render.fb.nr_samples;
       uint32_t rasterization_samples =
          cmdbuf->vk.dynamic_graphics_state.ms.rasterization_samples;
 
@@ -1190,16 +1192,26 @@ panvk_cmd_draw(struct panvk_cmd_buffer *cmdbuf, struct panvk_draw_data *draw)
        * make sure those two numbers match. */
       if (!batch->fb.desc.gpu && !cmdbuf->state.gfx.render.bound_attachments) {
          assert(rasterization_samples > 0);
-         fbinfo->nr_samples = rasterization_samples;
+         *nr_samples = rasterization_samples;
       } else {
-         assert(rasterization_samples == fbinfo->nr_samples);
+         assert(rasterization_samples == *nr_samples);
       }
 
+      /* In case we already emitted tiler/framebuffer descriptors, we ensure
+       * that the sample count didn't change
+       * XXX: This currently can happen in case we resume a render pass with no
+       * attachements and without any draw as the FBD is emitted when suspending.
+       */
+      assert(fbinfo->nr_samples == 0 ||
+             fbinfo->nr_samples == cmdbuf->state.gfx.render.fb.nr_samples);
+
       result = panvk_per_arch(cmd_alloc_fb_desc)(cmdbuf);
       if (result != VK_SUCCESS)
          return;
    }
 
+   panvk_per_arch(cmd_select_tile_size)(cmdbuf);
+
    result = panvk_per_arch(cmd_alloc_tls_desc)(cmdbuf, true);
    if (result != VK_SUCCESS)
       return;

src/panfrost/vulkan/panvk_cmd_draw.h

@@ -64,6 +64,9 @@ struct panvk_rendering_state {
       struct pan_fb_info info;
       bool crc_valid[MAX_RTS];
 
+      /* nr_samples to be used before framebuffer / tiler descriptor are emitted */
+      uint32_t nr_samples;
+
 #if PAN_ARCH <= 7
       uint32_t bo_count;
       struct pan_kmod_bo *bos[MAX_RTS + 2];
@@ -326,6 +329,7 @@ panvk_per_arch(cmd_preload_render_area_border)(struct panvk_cmd_buffer *cmdbuf,
                                                const VkRenderingInfo *render_info);
 
 void panvk_per_arch(cmd_resolve_attachments)(struct panvk_cmd_buffer *cmdbuf);
+void panvk_per_arch(cmd_select_tile_size)(struct panvk_cmd_buffer *cmdbuf);
 
 struct panvk_draw_info {
    struct {

src/panfrost/vulkan/panvk_vX_cmd_draw.c

@@ -36,8 +36,9 @@ render_state_set_color_attachment(struct panvk_cmd_buffer *cmdbuf,
 
    fbinfo->rts[index].view = &iview->pview;
    fbinfo->rts[index].crc_valid = &state->render.fb.crc_valid[index];
-   fbinfo->nr_samples =
-      MAX2(fbinfo->nr_samples, pan_image_view_get_nr_samples(&iview->pview));
+   state->render.fb.nr_samples =
+      MAX2(state->render.fb.nr_samples,
+           pan_image_view_get_nr_samples(&iview->pview));
 
    if (att->loadOp == VK_ATTACHMENT_LOAD_OP_CLEAR) {
       enum pipe_format fmt = vk_format_to_pipe_format(iview->vk.format);
@@ -89,8 +90,9 @@ render_state_set_z_attachment(struct panvk_cmd_buffer *cmdbuf,
 
    state->render.zs_pview.planes[0] = &img->planes[0];
    state->render.zs_pview.planes[1] = NULL;
-   fbinfo->nr_samples =
-      MAX2(fbinfo->nr_samples, pan_image_view_get_nr_samples(&iview->pview));
+   state->render.fb.nr_samples =
+      MAX2(state->render.fb.nr_samples,
+           pan_image_view_get_nr_samples(&iview->pview));
    state->render.z_attachment.iview = iview;
 
    /* D24S8 is a single plane format where the depth/stencil are interleaved.
@@ -157,8 +159,9 @@ render_state_set_s_attachment(struct panvk_cmd_buffer *cmdbuf,
       state->render.s_pview.planes[1] = NULL;
    }
 
-   fbinfo->nr_samples =
-      MAX2(fbinfo->nr_samples, pan_image_view_get_nr_samples(&iview->pview));
+   state->render.fb.nr_samples =
+      MAX2(state->render.fb.nr_samples,
+           pan_image_view_get_nr_samples(&iview->pview));
    state->render.s_attachment.iview = iview;
 
    /* If the depth and stencil attachments point to the same image,
@@ -234,9 +237,10 @@ panvk_per_arch(cmd_init_render_state)(struct panvk_cmd_buffer *cmdbuf,
    *fbinfo = (struct pan_fb_info){
       .tile_buf_budget = panfrost_query_optimal_tib_size(phys_dev->model),
       .z_tile_buf_budget = panfrost_query_optimal_z_tib_size(phys_dev->model),
-      .nr_samples = 1,
+      .nr_samples = 0,
       .rt_count = pRenderingInfo->colorAttachmentCount,
    };
+   cmdbuf->state.gfx.render.fb.nr_samples = 1;
 
    assert(pRenderingInfo->colorAttachmentCount <= ARRAY_SIZE(fbinfo->rts));
 
@@ -295,15 +299,30 @@ panvk_per_arch(cmd_init_render_state)(struct panvk_cmd_buffer *cmdbuf,
    }
 
    assert(fbinfo->width && fbinfo->height);
+}
 
-   GENX(pan_select_tile_size)(fbinfo);
+void
+panvk_per_arch(cmd_select_tile_size)(struct panvk_cmd_buffer *cmdbuf)
+{
+   struct pan_fb_info *fbinfo = &cmdbuf->state.gfx.render.fb.info;
+
+   /* In case we never emitted tiler/framebuffer descriptors, we emit the
+    * current sample count and compute tile size */
+   if (fbinfo->nr_samples == 0) {
+      fbinfo->nr_samples = cmdbuf->state.gfx.render.fb.nr_samples;
+      GENX(pan_select_tile_size)(fbinfo);
 
 #if PAN_ARCH != 6
-   if (fbinfo->cbuf_allocation > fbinfo->tile_buf_budget) {
-      vk_perf(VK_LOG_OBJS(&cmdbuf->vk.base),
-              "Using too much tile-memory, disabling pipelining");
-   }
+      if (fbinfo->cbuf_allocation > fbinfo->tile_buf_budget) {
+         vk_perf(VK_LOG_OBJS(&cmdbuf->vk.base),
+                 "Using too much tile-memory, disabling pipelining");
+      }
 #endif
+   } else {
+      /* In case we already emitted tiler/framebuffer descriptors, we ensure
+       * that the sample count didn't change (this should never happen) */
+      assert(fbinfo->nr_samples == cmdbuf->state.gfx.render.fb.nr_samples);
+   }
 }
 
 void

src/panfrost/vulkan/panvk_vX_cmd_meta.c

@@ -192,13 +192,12 @@ panvk_per_arch(CmdClearAttachments)(VkCommandBuffer commandBuffer,
                                     const VkClearRect *pRects)
 {
    VK_FROM_HANDLE(panvk_cmd_buffer, cmdbuf, commandBuffer);
-   const struct pan_fb_info *fbinfo = &cmdbuf->state.gfx.render.fb.info;
    struct panvk_device *dev = to_panvk_device(cmdbuf->vk.base.device);
    struct panvk_cmd_meta_graphics_save_ctx save = {0};
    struct vk_meta_rendering_info render = {
       .view_mask = cmdbuf->state.gfx.render.view_mask,
-      .samples = fbinfo->nr_samples,
-      .color_attachment_count = fbinfo->rt_count,
+      .samples = cmdbuf->state.gfx.render.fb.nr_samples,
+      .color_attachment_count = cmdbuf->state.gfx.render.fb.info.rt_count,
       .depth_attachment_format = cmdbuf->state.gfx.render.z_attachment.fmt,
       .stencil_attachment_format = cmdbuf->state.gfx.render.s_attachment.fmt,
    };