[Mesa-dev] [PATCH 04/11] nvc0: bind constant buffers for compute on Kepler

[Samuel Pitoiset]

Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
---
 src/gallium/drivers/nouveau/nvc0/nve4_compute.c | 70 +++++++++++++++++++++++--
 src/gallium/drivers/nouveau/nvc0/nve4_compute.h |  2 +-
 2 files changed, 66 insertions(+), 6 deletions(-)

diff --git a/src/gallium/drivers/nouveau/nvc0/nve4_compute.c b/src/gallium/drivers/nouveau/nvc0/nve4_compute.c
index 47e3cfe..557dbdc 100644
--- a/src/gallium/drivers/nouveau/nvc0/nve4_compute.c
+++ b/src/gallium/drivers/nouveau/nvc0/nve4_compute.c
@@ -309,6 +309,52 @@ nve4_compute_set_tex_handles(struct nvc0_context *nvc0)
 }
 
 static void
+nve4_compute_validate_constbufs(struct nvc0_context *nvc0)
+{
+   struct nouveau_pushbuf *push = nvc0->base.pushbuf;
+   const int s = 5;
+
+   while (nvc0->constbuf_dirty[s]) {
+      int i = ffs(nvc0->constbuf_dirty[s]) - 1;
+      nvc0->constbuf_dirty[s] &= ~(1 << i);
+
+      if (nvc0->constbuf[s][i].user) {
+         struct nouveau_bo *bo = nvc0->screen->uniform_bo;
+         const unsigned base = NVC0_CB_USR_INFO(s);
+         const unsigned size = nvc0->constbuf[s][0].size;
+         assert(i == 0); /* we really only want OpenGL uniforms here */
+         assert(nvc0->constbuf[s][0].u.data);
+
+         BEGIN_NVC0(push, NVE4_CP(UPLOAD_DST_ADDRESS_HIGH), 2);
+         PUSH_DATAh(push, bo->offset + base);
+         PUSH_DATA (push, bo->offset + base);
+         BEGIN_NVC0(push, NVE4_CP(UPLOAD_LINE_LENGTH_IN), 2);
+         PUSH_DATA (push, size);
+         PUSH_DATA (push, 0x1);
+         BEGIN_1IC0(push, NVE4_CP(UPLOAD_EXEC), 1 + (size / 4));
+         PUSH_DATA (push, NVE4_COMPUTE_UPLOAD_EXEC_LINEAR | (0x20 << 1));
+         PUSH_DATAp(push, nvc0->constbuf[s][0].u.data, size / 4);
+      }
+      else {
+         struct nv04_resource *res =
+            nv04_resource(nvc0->constbuf[s][i].u.buf);
+         if (res) {
+            /* UBOs don't need to be bound but they have to be set up in the
+             * launch descriptor. */
+            BCTX_REFN(nvc0->bufctx_cp, CP_CB(i), res, RD);
+
+            res->cb_bindings[s] |= 1 << i;
+         }
+         if (i == 0)
+            nvc0->state.uniform_buffer_bound[s] = 0;
+      }
+   }
+
+   BEGIN_NVC0(push, NVE4_CP(FLUSH), 1);
+   PUSH_DATA (push, NVE4_COMPUTE_FLUSH_CB);
+}
+
+static void
 nve4_compute_validate_buffers(struct nvc0_context *nvc0)
 {
    struct nouveau_pushbuf *push = nvc0->base.pushbuf;
@@ -359,6 +405,8 @@ nve4_compute_state_validate(struct nvc0_context *nvc0)
       nve4_compute_validate_surfaces(nvc0);
    if (nvc0->dirty_cp & NVC0_NEW_CP_GLOBALS)
       nvc0_compute_validate_globals(nvc0);
+   if (nvc0->dirty_cp & NVC0_NEW_CP_CONSTBUF)
+      nve4_compute_validate_constbufs(nvc0);
    if (nvc0->dirty_cp & NVC0_NEW_CP_BUFFERS)
       nve4_compute_validate_buffers(nvc0);
 
@@ -375,7 +423,9 @@ nve4_compute_state_validate(struct nvc0_context *nvc0)
 
 
 static void
-nve4_compute_upload_input(struct nvc0_context *nvc0, const void *input,
+nve4_compute_upload_input(struct nvc0_context *nvc0,
+                          struct nve4_cp_launch_desc *desc,
+                          const void *input,
                           const uint *block_layout,
                           const uint *grid_layout)
 {
@@ -396,6 +446,11 @@ nve4_compute_upload_input(struct nvc0_context *nvc0, const void *input,
       BEGIN_1IC0(push, NVE4_CP(UPLOAD_EXEC), 1 + (cp->parm_size / 4));
       PUSH_DATA (push, NVE4_COMPUTE_UPLOAD_EXEC_LINEAR | (0x20 << 1));
       PUSH_DATAp(push, input, cp->parm_size / 4);
+
+      /* Bind user parameters coming from clover. */
+      /* TODO: This should be harmonized with uniform_bo. */
+      assert(!(desc->cb_mask & (1 << 0)));
+      nve4_cp_launch_desc_set_cb(desc, 0, screen->parm, 0, 1 << 12);
    }
    BEGIN_NVC0(push, NVE4_CP(UPLOAD_DST_ADDRESS_HIGH), 2);
    PUSH_DATAh(push, address + NVC0_CB_AUX_GRID_INFO);
@@ -456,10 +511,14 @@ nve4_compute_setup_launch_desc(struct nvc0_context *nvc0,
 
    for (i = 0; i < 7; ++i) {
       const unsigned s = 5;
+      if (nvc0->constbuf[s][i].user) {
+         assert(i == 0);
+         nve4_cp_launch_desc_set_cb(desc, 0, screen->uniform_bo,
+                                    NVC0_CB_USR_INFO(s), 1 << 16);
+      } else
       if (nvc0->constbuf[s][i].u.buf)
-         nve4_cp_launch_desc_set_ctx_cb(desc, i + 1, &nvc0->constbuf[s][i]);
+         nve4_cp_launch_desc_set_ctx_cb(desc, i, &nvc0->constbuf[s][i]);
    }
-   nve4_cp_launch_desc_set_cb(desc, 0, screen->parm, 0, NVE4_CP_INPUT_SIZE);
    nve4_cp_launch_desc_set_cb(desc, 7, screen->uniform_bo,
                               NVC0_CB_AUX_INFO(5), 1 << 10);
 }
@@ -503,13 +562,14 @@ nve4_launch_grid(struct pipe_context *pipe, const struct pipe_grid_info *info)
 
    nve4_compute_setup_launch_desc(nvc0, desc, info->pc,
                                   info->block, info->grid);
+
+   nve4_compute_upload_input(nvc0, desc, info->input, info->block, info->grid);
+
 #ifdef DEBUG
    if (debug_get_num_option("NV50_PROG_DEBUG", 0))
       nve4_compute_dump_launch_desc(desc);
 #endif
 
-   nve4_compute_upload_input(nvc0, info->input, info->block, info->grid);
-
    /* upload descriptor and flush */
 #if 0
    BEGIN_NVC0(push, NVE4_CP(UPLOAD_DST_ADDRESS_HIGH), 2);
diff --git a/src/gallium/drivers/nouveau/nvc0/nve4_compute.h b/src/gallium/drivers/nouveau/nvc0/nve4_compute.h
index dcafbed..8f08248 100644
--- a/src/gallium/drivers/nouveau/nvc0/nve4_compute.h
+++ b/src/gallium/drivers/nouveau/nvc0/nve4_compute.h
@@ -56,7 +56,7 @@ static inline void
 nve4_cp_launch_desc_set_cb(struct nve4_cp_launch_desc *desc,
                            unsigned index,
                            struct nouveau_bo *bo,
-                           uint32_t base, uint16_t size)
+                           uint32_t base, uint32_t size)
 {
    uint64_t address = bo->offset + base;
 
-- 
2.7.1