Mesa (staging/20.2): radeonsi: increase minimum NGG vertex count requirement per workgroup on gfx 10.3

[GitLab Mirror]

Module: Mesa
Branch: staging/20.2
Commit: 31836102281038d57e099af3c46fec7dd19dcefb
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=31836102281038d57e099af3c46fec7dd19dcefb

Author: Marek Olšák <marek.olsak at amd.com>
Date:   Wed Jul 29 19:04:38 2020 -0400

radeonsi: increase minimum NGG vertex count requirement per workgroup on gfx 10.3

Fixes: a23802bcb9a - ac,radeonsi: start adding support for gfx10.3

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer at amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6137>
(cherry picked from commit 7a468fc0f6a02d2ee71b50225459de335ed689b4)

---

 .pick_status.json                               | 2 +-
 src/gallium/drivers/radeonsi/gfx10_shader_ngg.c | 7 ++++---
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/.pick_status.json b/.pick_status.json
index ba473424408..9b6e1066f6a 100644
--- a/.pick_status.json
+++ b/.pick_status.json
@@ -58,7 +58,7 @@
         "description": "radeonsi: increase minimum NGG vertex count requirement per workgroup on gfx 10.3",
         "nominated": true,
         "nomination_type": 1,
-        "resolution": 0,
+        "resolution": 1,
         "master_sha": null,
         "because_sha": "a23802bcb9a42a02d34a5a36d6e66d6532813a0d"
     },
diff --git a/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c b/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c
index 4b639eecf5f..ee242da7ed1 100644
--- a/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c
+++ b/src/gallium/drivers/radeonsi/gfx10_shader_ngg.c
@@ -1920,6 +1920,7 @@ bool gfx10_ngg_calculate_subgroup_info(struct si_shader *shader)
    unsigned gsprim_lds_size = 0;
 
    /* All these are per subgroup: */
+   const unsigned min_esverts = gs_sel->screen->info.chip_class >= GFX10_3 ? 29 : 24;
    bool max_vert_out_per_gs_instance = false;
    unsigned max_gsprims_base = 128; /* default prim group size clamp */
    unsigned max_esverts_base = 128;
@@ -2033,7 +2034,7 @@ retry_select_mode:
    }
 
    /* Hardware restriction: minimum value of max_esverts */
-   max_esverts = MAX2(max_esverts, 23 + max_verts_per_prim);
+   max_esverts = MAX2(max_esverts, min_esverts - 1 + max_verts_per_prim);
 
    unsigned max_out_vertices =
       max_vert_out_per_gs_instance
@@ -2064,10 +2065,10 @@ retry_select_mode:
    shader->gs_info.esgs_ring_size = max_esverts * esvert_lds_size;
    shader->ngg.ngg_emit_size = max_gsprims * gsprim_lds_size;
 
-   assert(shader->ngg.hw_max_esverts >= 24); /* HW limitation */
+   assert(shader->ngg.hw_max_esverts >= min_esverts); /* HW limitation */
 
    /* If asserts are disabled, we use the same conditions to return false */
    return max_esverts >= max_verts_per_prim && max_gsprims >= 1 &&
           max_out_vertices <= 256 &&
-          shader->ngg.hw_max_esverts >= 24;
+          shader->ngg.hw_max_esverts >= min_esverts;
 }