pan/va: Improve texture instructions

[Alyssa: Related improvements.]

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15588>

src/panfrost/bifrost/valhall/ISA.xml

@@ -586,6 +586,17 @@
     <value>32</value>
   </enum>
 
+  <enum name="Varying texture register width">
+    <desc>
+      Size of results for varying texture instructions. For dual 16-bit results
+      use "16-bit".
+    </desc>
+    <value desc="16-bit">16</value>
+    <value desc="32-bit">32</value>
+    <value desc="16-bit, 32-bit">16.32</value>
+    <value desc="32-bit, 32-bit">32.32</value>
+  </enum>
+
   <enum name="Vector size">
     <desc>Number of channels loaded/stored for general memory access.</desc>
     <value default="true" desc="Scalar">none</value>
@@ -706,6 +717,21 @@
     <value desc="Clobber saved position">clobber</value>
   </enum>
 
+  <enum name="Sample and update mode">
+    <desc>
+      For fused varying/texture instructions, only the following specific
+      combinations of sample and update modes are permitted.
+    </desc>
+    <value desc="Center, store">center_store</value>
+    <value desc="Centroid, store">centroid_store</value>
+    <value desc="Sample, store">sample_store</value>
+    <value desc="Explicit, store">explicit_store</value>
+    <value desc="Center, clobber">center_clobber</value>
+    <reserved/>
+    <value desc="Sample, clobber">sample_clobber</value>
+    <value desc="Retrieve previous state">retrieve</value>
+  </enum>
+
   <enum name="Source format">
     <desc>
       In-memory format of varyings.
@@ -2401,6 +2427,7 @@
     <texel_offset/>
     <integer_coordinates/>
     <fetch_component/>
+    <shadow/>
 
     <!-- Leave secondary_register_width as 0 -->
     <sr_count/>
@@ -2423,6 +2450,8 @@
     <wide_indices/>
     <array_enable/>
     <texel_offset/>
+    <shadow/>
+    <lod_mode/>
 
     <sr_count/>
     <sr_write_count/>
@@ -2432,17 +2461,71 @@
     <src>Image to read from</src>
   </ins>
 
-  <ins name="TODO.VAR_TEX" title="Fused varying-texturing" opcode="0x130" unit="VT">
+  <ins name="VAR_TEX_BUF_SINGLE" title="Fused varying-texturing" opcode="0x130" unit="VT">
     <desc>
       Only works for FP32 varyings. Performance characteristics are similar
-      to LD_VAR_IMM_F32.v2.f32 followed by TEX, using both V and T units.
+      to LD_VAR_BUF_IMM_F32.v2.f32 followed by TEX, using both V and T units.
     </desc>
-    <sr write="true" count="4"/>
-    <mod name="dimension" start="28" size="2"/>
-    <mod name="skip" start="39" size="1"/>
     <slot/>
+    <skip/>
+    <sample_and_update/>
+    <register_type/>
+    <vartex_register_width/>
+    <dimension/>
+    <array_enable/>
+    <shadow/>
+    <lod_mode/>
+
     <sr_write_count/>
-    <src>Image to read from</src>
+
+    <sr write="true"/>
+    <src size="64">Image to read from</src>
+    <src>Varying offset</src>
+  </ins>
+
+  <ins name="VAR_TEX_BUF_GATHER" title="Fused varying-texturing" opcode="0x131" unit="VT">
+    <desc>
+      Only works for FP32 varyings. Performance characteristics are similar
+      to LD_VAR_BUF_IMM_F32.v2.f32 followed by TEX, using both V and T units.
+    </desc>
+    <slot/>
+    <skip/>
+    <sample_and_update/>
+    <register_type/>
+    <vartex_register_width/>
+    <dimension/>
+    <array_enable/>
+    <integer_coordinates/>
+    <fetch_component/>
+    <shadow/>
+
+    <sr_write_count/>
+
+    <sr write="true"/>
+    <src size="64">Image to read from</src>
+    <src>Varying offset</src>
+  </ins>
+
+  <ins name="VAR_TEX_BUF_DUAL" title="Fused varying-texturing" opcode="0x137" unit="VT">
+    <desc>
+      Only works for FP32 varyings. Performance characteristics are similar
+      to LD_VAR_BUF_IMM_F32.v2.f32 followed by TEX_DUAL, using both V and T units.
+    </desc>
+    <slot/>
+    <skip/>
+    <sample_and_update/>
+    <register_type/>
+    <vartex_register_width/>
+    <dimension/>
+    <array_enable/>
+    <shadow/>
+    <lod_mode/>
+
+    <sr_write_count/>
+
+    <sr write="true"/>
+    <src size="64">Image to read from</src>
+    <src>Varying offset</src>
   </ins>
 
   <ins name="FMA_RSCALE.f32" title="Fused floating-point multiply add with exponent bias" dests="1" opcode="0x160" unit="FMA">

src/panfrost/bifrost/valhall/valhall.c.py

@@ -65,6 +65,11 @@ SKIP = set([
         "CLPER.v2u16",
         "CLPER.v4u8",
 
+        # VAR_TEX
+        "VAR_TEX_BUF_SINGLE",
+        "VAR_TEX_BUF_GATHER",
+        "VAR_TEX_BUF_DUAL",
+
         # Special cased
         "FMA_RSCALE_N.f32",
         "FMA_RSCALE_LEFT.f32",
@@ -80,7 +85,6 @@ SKIP = set([
         "FATAN_ASSIST.f32",
         "SEG_ADD.u64",
         "TEX_DUAL",
-        "TODO.VAR_TEX",
     ])
 
 template = """

src/panfrost/bifrost/valhall/valhall.py

@@ -386,6 +386,7 @@ MODIFIERS = {
     "skip": Flag("skip", 39),
     "register_width": Modifier("register_width", 46, 1, force_enum = "register_width"),
     "secondary_register_width": Modifier("secondary_register_width", 47, 1, force_enum = "register_width"),
+    "vartex_register_width": Modifier("varying_texture_register_width", 24, 2),
 
     "atom_opc": Modifier("atomic_operation", 22, 4),
     "atom_opc_1": Modifier("atomic_operation_with_1", 22, 4),
@@ -405,6 +406,7 @@ MODIFIERS = {
     "cmp": Modifier("condition", 32, 3),
     "clamp": Modifier("clamp", 32, 2),
     "sr_count": Modifier("staging_register_count", 33, 3, implied = True),
+    "sample_and_update": Modifier("sample_and_update_mode", 33, 3),
     "sr_write_count": Modifier("staging_register_write_count", 36, 3, implied = True),
 
     "conservative": Flag("conservative", 35),