lima: delay plbu head command generation to flush stage (v2)

Prepare for multi submit in which case only know the plb_index when
final flush. Another need for this is proper reload detection: only
when flush stage can we know if need to do reload, because user may
first clear depth, then color individually, so we don't know if need
to pack repload plbu cmd when first clear depth.

v2:
move lima_update_submit_wb before ctx->resolve change for lima_ctx_dirty
to work in lima_submit_wb.

Reviewed-by: Vasily Khoruzhick <[email protected]>
Signed-off-by: Qiang Yu <[email protected]>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3755>

3 files changed, 17 insertions, 17 deletions

diff --git a/src/gallium/drivers/lima/lima_context.c b/src/gallium/drivers/lima/lima_context.c
index 5e8ac448001..dc9a1026235 100644
--- a/src/gallium/drivers/lima/lima_context.c
+++ b/src/gallium/drivers/lima/lima_context.c
@@ -231,6 +231,7 @@ lima_context_create(struct pipe_screen *pscreen, void *priv, unsigned flags)
 
    util_dynarray_init(&ctx->vs_cmd_array, ctx);
    util_dynarray_init(&ctx->plbu_cmd_array, ctx);
+   util_dynarray_init(&ctx->plbu_cmd_head, ctx);
 
    ctx->plb_size = screen->plb_max_blk * LIMA_CTX_PLB_BLK_SIZE;
    ctx->plb_gp_size = screen->plb_max_blk * 4;
diff --git a/src/gallium/drivers/lima/lima_context.h b/src/gallium/drivers/lima/lima_context.h
index 436a2ab1339..69bb5bd093a 100644
--- a/src/gallium/drivers/lima/lima_context.h
+++ b/src/gallium/drivers/lima/lima_context.h
@@ -239,6 +239,7 @@ struct lima_context {
 
    struct util_dynarray vs_cmd_array;
    struct util_dynarray plbu_cmd_array;
+   struct util_dynarray plbu_cmd_head;
 
    struct lima_submit *gp_submit;
    struct lima_submit *pp_submit;
diff --git a/src/gallium/drivers/lima/lima_draw.c b/src/gallium/drivers/lima/lima_draw.c
index d392df0221f..e351b4d268c 100644
--- a/src/gallium/drivers/lima/lima_draw.c
+++ b/src/gallium/drivers/lima/lima_draw.c
@@ -212,7 +212,7 @@ struct lima_render_state {
 static inline bool
 lima_ctx_dirty(struct lima_context *ctx)
 {
-   return ctx->plbu_cmd_array.size;
+   return !!ctx->resolve;
 }
 
 static inline struct lima_damage_region *
@@ -323,7 +323,7 @@ lima_pack_reload_plbu_cmd(struct lima_context *ctx)
    lima_submit_add_bo(ctx->pp_submit, res->bo, LIMA_SUBMIT_BO_READ);
    pipe_resource_reference(&pres, NULL);
 
-   PLBU_CMD_BEGIN(&ctx->plbu_cmd_array, 20);
+   PLBU_CMD_BEGIN(&ctx->plbu_cmd_head, 20);
 
    PLBU_CMD_VIEWPORT_LEFT(0);
    PLBU_CMD_VIEWPORT_RIGHT(fui(fb->base.width));
@@ -347,13 +347,9 @@ lima_pack_reload_plbu_cmd(struct lima_context *ctx)
 static void
 lima_pack_head_plbu_cmd(struct lima_context *ctx)
 {
-   /* first draw need create a PLBU command header */
-   if (lima_ctx_dirty(ctx))
-      return;
-
    struct lima_context_framebuffer *fb = &ctx->framebuffer;
 
-   PLBU_CMD_BEGIN(&ctx->plbu_cmd_array, 10);
+   PLBU_CMD_BEGIN(&ctx->plbu_cmd_head, 10);
 
    PLBU_CMD_UNKNOWN2();
    PLBU_CMD_BLOCK_STEP(fb->shift_min, fb->shift_h, fb->shift_w);
@@ -689,6 +685,8 @@ lima_clear(struct pipe_context *pctx, unsigned buffers,
 
    lima_flush(ctx);
 
+   lima_update_submit_wb(ctx);
+
    ctx->resolve |= buffers;
 
    /* no need to reload if cleared */
@@ -720,10 +718,6 @@ lima_clear(struct pipe_context *pctx, unsigned buffers,
    if (buffers & PIPE_CLEAR_STENCIL)
       clear->stencil = stencil;
 
-   lima_update_submit_wb(ctx);
-
-   lima_pack_head_plbu_cmd(ctx);
-
    ctx->dirty |= LIMA_CONTEXT_DIRTY_CLEAR;
 
    lima_damage_rect_union(ctx, 0, ctx->framebuffer.base.width,
@@ -840,8 +834,6 @@ lima_pack_plbu_cmd(struct lima_context *ctx, const struct pipe_draw_info *info)
    struct lima_vs_shader_state *vs = ctx->vs;
    unsigned minx, maxx, miny, maxy;
 
-   lima_pack_head_plbu_cmd(ctx);
-
    /* If it's zero scissor, we skip adding all other commands */
    if (lima_is_scissor_zero(ctx))
       return;
@@ -1775,14 +1767,13 @@ _lima_flush(struct lima_context *ctx, bool end_of_frame)
 {
    #define pp_stack_pp_size 0x400
 
+   lima_pack_head_plbu_cmd(ctx);
    lima_finish_plbu_cmd(ctx);
 
    lima_update_submit_bo(ctx);
 
    int vs_cmd_size = ctx->vs_cmd_array.size;
-   int plbu_cmd_size = ctx->plbu_cmd_array.size;
    uint32_t vs_cmd_va = 0;
-   uint32_t plbu_cmd_va;
 
    if (vs_cmd_size) {
       void *vs_cmd =
@@ -1796,11 +1787,18 @@ _lima_flush(struct lima_context *ctx, bool end_of_frame)
       lima_dump_vs_command_stream_print(vs_cmd, vs_cmd_size, vs_cmd_va);
    }
 
+   int plbu_cmd_size = ctx->plbu_cmd_array.size + ctx->plbu_cmd_head.size;
    void *plbu_cmd =
       lima_ctx_buff_alloc(ctx, lima_ctx_buff_gp_plbu_cmd, plbu_cmd_size);
-   memcpy(plbu_cmd, util_dynarray_begin(&ctx->plbu_cmd_array), plbu_cmd_size);
+   memcpy(plbu_cmd,
+          util_dynarray_begin(&ctx->plbu_cmd_head),
+          ctx->plbu_cmd_head.size);
+   memcpy(plbu_cmd + ctx->plbu_cmd_head.size,
+          util_dynarray_begin(&ctx->plbu_cmd_array),
+          ctx->plbu_cmd_array.size);
    util_dynarray_clear(&ctx->plbu_cmd_array);
-   plbu_cmd_va = lima_ctx_buff_va(ctx, lima_ctx_buff_gp_plbu_cmd);
+   util_dynarray_clear(&ctx->plbu_cmd_head);
+   uint32_t plbu_cmd_va = lima_ctx_buff_va(ctx, lima_ctx_buff_gp_plbu_cmd);
 
    lima_dump_command_stream_print(
       plbu_cmd, plbu_cmd_size, false, "flush plbu cmd at va %x\n", plbu_cmd_va);