From 09cc736c42465c5b4ba2e970b680048339139dba Mon Sep 17 00:00:00 2001
From: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Date: Wed, 14 Sep 2022 22:06:56 -0400
Subject: [PATCH] asahi: Identify shared memory fields

For compute kernels, this encodes how much workgroup-local memory is
used ("shared memory" or "threadgroup memory" or "local memory"). This
memory is partitioned by the hardware.

For fragment shaders, this... encodes exactly the same thing. There is
no traditional tilebuffer in AGX, instead local memory is interpreted as
an imageblock, where each workgroup is a tile. This is a nifty design.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18623>
---
 src/asahi/lib/cmdbuf.xml              | 8 ++++++--
 src/gallium/drivers/asahi/agx_state.c | 6 ++++--
 2 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/src/asahi/lib/cmdbuf.xml b/src/asahi/lib/cmdbuf.xml
index 49ffd5b5c72..2e15582023a 100644
--- a/src/asahi/lib/cmdbuf.xml
+++ b/src/asahi/lib/cmdbuf.xml
@@ -502,7 +502,9 @@
 
   <struct name="Set shader" size="24">
     <field name="Tag" size="8" start="0:0" type="hex" default="0x4d"/>
-    <field name="Unk 1" size="24" start="0:8" type="hex" default="0x90"/>
+    <field name="Uses shared memory" size="1" start="0:8" type="bool"/>
+    <field name="Unk 1" size="16" start="0:8" type="hex" default="0x90"/>
+    <field name="Shared memory per threadgroup in 256 bytes" size="8" start="0:24" type="uint"/>
     <field name="Unk 2" size="8" start="1:0" type="hex" default="0x0d"/> <!-- TODO differs with stage -->
     <!--- Seemingly texture count related?,
           or number of registers in the preshader? 8 with <4 in the preshader,
@@ -535,7 +537,9 @@
 
   <struct name="Set shader extended" size="28">
     <field name="Tag" size="8" start="0:0" type="hex" default="0x4d"/>
-    <field name="Unk 1" size="24" start="0:8" type="hex" default="0x2010bd"/>
+    <field name="Uses shared memory" size="1" start="0:8" type="bool"/>
+    <field name="Unk 1" size="16" start="0:8" type="hex" default="0x10bd"/>
+    <field name="Shared memory per threadgroup in 256 bytes" size="8" start="0:24" type="uint" default="0x20"/>
     <field name="Unk 2" size="8" start="1:0" type="hex" default="0x0d"/>
     <field name="Loads varyings" size="1" start="1:8" type="bool"/>
     <field name="Unk 1 9" size="1" start="1:9" type="bool" default="0"/>
diff --git a/src/gallium/drivers/asahi/agx_state.c b/src/gallium/drivers/asahi/agx_state.c
index ffeb4edc346..23be68e6e82 100644
--- a/src/gallium/drivers/asahi/agx_state.c
+++ b/src/gallium/drivers/asahi/agx_state.c
@@ -1279,7 +1279,8 @@ agx_build_pipeline(struct agx_context *ctx, struct agx_compiled_shader *cs, enum
          cfg.code = cs->bo->ptr.gpu;
          cfg.register_quadwords = 0;
          cfg.unk_3 = 0x8d;
-         cfg.unk_1 = 0x2010bd;
+         cfg.unk_1 = 0x10bd;
+         cfg.shared_memory_per_threadgroup_in_256_bytes = 32;
          cfg.unk_2 = 0x0d;
          cfg.loads_varyings = true;
          cfg.fragment_parameters.early_z_testing = !writes_sample_mask;
@@ -1328,7 +1329,8 @@ agx_build_clear_pipeline(struct agx_context *ctx, uint32_t code, uint64_t clear_
    /* TODO: Can we prepack this? */
    agx_pack(record, SET_SHADER, cfg) {
       cfg.code = code;
-      cfg.unk_1 = 0x2010bd;
+      cfg.unk_1 = 0x10bd;
+      cfg.shared_memory_per_threadgroup_in_256_bytes = 32;
       cfg.unk_2 = 0x0d;
       cfg.unk_3 = 0x8d;
       cfg.register_quadwords = 1;