turnip: implement VK_EXT_filter_cubic
[mesa.git] / src / freedreno / registers / a6xx.xml
index ad3c69d05d1f4c13a1bbefb353cc03dce1455b5e..5d6ddcf1785099e62534f1172b7130223526cef3 100644 (file)
@@ -6,57 +6,6 @@ xsi:schemaLocation="http://nouveau.freedesktop.org/ rules-ng.xsd">
 <import file="adreno/adreno_common.xml"/>
 <import file="adreno/adreno_pm4.xml"/>
 
-<!-- these might be same as a5xx -->
-<enum name="a6xx_color_fmt">
-       <value value="0x02" name="RB6_A8_UNORM"/>
-       <value value="0x03" name="RB6_R8_UNORM"/>
-       <value value="0x04" name="RB6_R8_SNORM"/>
-       <value value="0x05" name="RB6_R8_UINT"/>
-       <value value="0x06" name="RB6_R8_SINT"/>
-       <value value="0x08" name="RB6_R4G4B4A4_UNORM"/>
-       <value value="0x0a" name="RB6_R5G5B5A1_UNORM"/>
-       <value value="0x0e" name="RB6_R5G6B5_UNORM"/>
-       <value value="0x0f" name="RB6_R8G8_UNORM"/>
-       <value value="0x10" name="RB6_R8G8_SNORM"/>
-       <value value="0x11" name="RB6_R8G8_UINT"/>
-       <value value="0x12" name="RB6_R8G8_SINT"/>
-       <value value="0x15" name="RB6_R16_UNORM"/>
-       <value value="0x16" name="RB6_R16_SNORM"/>
-       <value value="0x17" name="RB6_R16_FLOAT"/>
-       <value value="0x18" name="RB6_R16_UINT"/>
-       <value value="0x19" name="RB6_R16_SINT"/>
-       <value value="0x30" name="RB6_R8G8B8A8_UNORM"/>
-       <value value="0x31" name="RB6_R8G8B8_UNORM"/>
-       <value value="0x32" name="RB6_R8G8B8A8_SNORM"/>
-       <value value="0x33" name="RB6_R8G8B8A8_UINT"/>
-       <value value="0x34" name="RB6_R8G8B8A8_SINT"/>
-       <value value="0x36" name="RB6_R10G10B10A2_FLOAT16"/>  <!-- float16 for 2d blit? -->
-       <value value="0x37" name="RB6_R10G10B10A2_UNORM"/>  <!-- GL_RGB10_A2 -->
-       <value value="0x3a" name="RB6_R10G10B10A2_UINT"/>   <!-- GL_RGB10_A2UI -->
-       <value value="0x42" name="RB6_R11G11B10_FLOAT"/>    <!-- GL_R11F_G11F_B10F -->
-       <value value="0x43" name="RB6_R16G16_UNORM"/>
-       <value value="0x44" name="RB6_R16G16_SNORM"/>
-       <value value="0x45" name="RB6_R16G16_FLOAT"/>
-       <value value="0x46" name="RB6_R16G16_UINT"/>
-       <value value="0x47" name="RB6_R16G16_SINT"/>
-       <value value="0x4a" name="RB6_R32_FLOAT"/>
-       <value value="0x4b" name="RB6_R32_UINT"/>
-       <value value="0x4c" name="RB6_R32_SINT"/>
-       <value value="0x60" name="RB6_R16G16B16A16_UNORM"/>
-       <value value="0x61" name="RB6_R16G16B16A16_SNORM"/>
-       <value value="0x62" name="RB6_R16G16B16A16_FLOAT"/>
-       <value value="0x63" name="RB6_R16G16B16A16_UINT"/>
-       <value value="0x64" name="RB6_R16G16B16A16_SINT"/>
-       <value value="0x67" name="RB6_R32G32_FLOAT"/>
-       <value value="0x68" name="RB6_R32G32_UINT"/>
-       <value value="0x69" name="RB6_R32G32_SINT"/>
-       <value value="0x82" name="RB6_R32G32B32A32_FLOAT"/>
-       <value value="0x83" name="RB6_R32G32B32A32_UINT"/>
-       <value value="0x84" name="RB6_R32G32B32A32_SINT"/>
-       <value value="0x91" name="RB6_Z24_UNORM_S8_UINT_AS_R8G8B8A8"/>
-       <value value="0xa0" name="RB6_Z24_UNORM_S8_UINT"/>
-</enum>
-
 <!-- these might be same as a5xx -->
 <enum name="a6xx_tile_mode">
        <value name="TILE6_LINEAR" value="0"/>
@@ -64,173 +13,135 @@ xsi:schemaLocation="http://nouveau.freedesktop.org/ rules-ng.xsd">
        <value name="TILE6_3" value="3"/>
 </enum>
 
-<!-- these might be same as a5xx -->
-<enum name="a6xx_vtx_fmt" prefix="chipset">
-       <value value="0x03" name="VFMT6_8_UNORM"/>
-       <value value="0x04" name="VFMT6_8_SNORM"/>
-       <value value="0x05" name="VFMT6_8_UINT"/>
-       <value value="0x06" name="VFMT6_8_SINT"/>
-
-       <value value="0x0f" name="VFMT6_8_8_UNORM"/>
-       <value value="0x10" name="VFMT6_8_8_SNORM"/>
-       <value value="0x11" name="VFMT6_8_8_UINT"/>
-       <value value="0x12" name="VFMT6_8_8_SINT"/>
-
-       <value value="0x15" name="VFMT6_16_UNORM"/>
-       <value value="0x16" name="VFMT6_16_SNORM"/>
-       <value value="0x17" name="VFMT6_16_FLOAT"/>
-       <value value="0x18" name="VFMT6_16_UINT"/>
-       <value value="0x19" name="VFMT6_16_SINT"/>
-
-       <value value="0x21" name="VFMT6_8_8_8_UNORM"/>
-       <value value="0x22" name="VFMT6_8_8_8_SNORM"/>
-       <value value="0x23" name="VFMT6_8_8_8_UINT"/>
-       <value value="0x24" name="VFMT6_8_8_8_SINT"/>
-
-       <value value="0x30" name="VFMT6_8_8_8_8_UNORM"/>
-       <value value="0x32" name="VFMT6_8_8_8_8_SNORM"/>
-       <value value="0x33" name="VFMT6_8_8_8_8_UINT"/>
-       <value value="0x34" name="VFMT6_8_8_8_8_SINT"/>
-
-       <value value="0x36" name="VFMT6_10_10_10_2_UNORM"/>
-       <value value="0x39" name="VFMT6_10_10_10_2_SNORM"/>
-       <value value="0x3a" name="VFMT6_10_10_10_2_UINT"/>
-       <value value="0x3b" name="VFMT6_10_10_10_2_SINT"/>
-
-       <value value="0x42" name="VFMT6_11_11_10_FLOAT"/>
-
-       <value value="0x43" name="VFMT6_16_16_UNORM"/>
-       <value value="0x44" name="VFMT6_16_16_SNORM"/>
-       <value value="0x45" name="VFMT6_16_16_FLOAT"/>
-       <value value="0x46" name="VFMT6_16_16_UINT"/>
-       <value value="0x47" name="VFMT6_16_16_SINT"/>
-
-       <value value="0x48" name="VFMT6_32_UNORM"/>
-       <value value="0x49" name="VFMT6_32_SNORM"/>
-       <value value="0x4a" name="VFMT6_32_FLOAT"/>
-       <value value="0x4b" name="VFMT6_32_UINT"/>
-       <value value="0x4c" name="VFMT6_32_SINT"/>
-       <value value="0x4d" name="VFMT6_32_FIXED"/>
-
-       <value value="0x58" name="VFMT6_16_16_16_UNORM"/>
-       <value value="0x59" name="VFMT6_16_16_16_SNORM"/>
-       <value value="0x5a" name="VFMT6_16_16_16_FLOAT"/>
-       <value value="0x5b" name="VFMT6_16_16_16_UINT"/>
-       <value value="0x5c" name="VFMT6_16_16_16_SINT"/>
-
-       <value value="0x60" name="VFMT6_16_16_16_16_UNORM"/>
-       <value value="0x61" name="VFMT6_16_16_16_16_SNORM"/>
-       <value value="0x62" name="VFMT6_16_16_16_16_FLOAT"/>
-       <value value="0x63" name="VFMT6_16_16_16_16_UINT"/>
-       <value value="0x64" name="VFMT6_16_16_16_16_SINT"/>
-
-       <value value="0x65" name="VFMT6_32_32_UNORM"/>
-       <value value="0x66" name="VFMT6_32_32_SNORM"/>
-       <value value="0x67" name="VFMT6_32_32_FLOAT"/>
-       <value value="0x68" name="VFMT6_32_32_UINT"/>
-       <value value="0x69" name="VFMT6_32_32_SINT"/>
-       <value value="0x6a" name="VFMT6_32_32_FIXED"/>
-
-       <value value="0x70" name="VFMT6_32_32_32_UNORM"/>
-       <value value="0x71" name="VFMT6_32_32_32_SNORM"/>
-       <value value="0x72" name="VFMT6_32_32_32_UINT"/>
-       <value value="0x73" name="VFMT6_32_32_32_SINT"/>
-       <value value="0x74" name="VFMT6_32_32_32_FLOAT"/>
-       <value value="0x75" name="VFMT6_32_32_32_FIXED"/>
-
-       <value value="0x80" name="VFMT6_32_32_32_32_UNORM"/>
-       <value value="0x81" name="VFMT6_32_32_32_32_SNORM"/>
-       <value value="0x82" name="VFMT6_32_32_32_32_FLOAT"/>
-       <value value="0x83" name="VFMT6_32_32_32_32_UINT"/>
-       <value value="0x84" name="VFMT6_32_32_32_32_SINT"/>
-       <value value="0x85" name="VFMT6_32_32_32_32_FIXED"/>
-</enum>
-
-<enum name="a6xx_tex_fmt">
-       <value value="0x02" name="TFMT6_A8_UNORM"/>
-       <value value="0x03" name="TFMT6_8_UNORM"/>
-       <value value="0x04" name="TFMT6_8_SNORM"/>
-       <value value="0x05" name="TFMT6_8_UINT"/>
-       <value value="0x06" name="TFMT6_8_SINT"/>
-       <value value="0x08" name="TFMT6_4_4_4_4_UNORM"/>
-       <value value="0x0a" name="TFMT6_5_5_5_1_UNORM"/>
-       <value value="0x0e" name="TFMT6_5_6_5_UNORM"/>
-       <value value="0x0f" name="TFMT6_8_8_UNORM"/>
-       <value value="0x10" name="TFMT6_8_8_SNORM"/>
-       <value value="0x11" name="TFMT6_8_8_UINT"/>
-       <value value="0x12" name="TFMT6_8_8_SINT"/>
-       <value value="0x13" name="TFMT6_L8_A8_UNORM"/>
-       <value value="0x15" name="TFMT6_16_UNORM"/>
-       <value value="0x16" name="TFMT6_16_SNORM"/>
-       <value value="0x17" name="TFMT6_16_FLOAT"/>
-       <value value="0x18" name="TFMT6_16_UINT"/>
-       <value value="0x19" name="TFMT6_16_SINT"/>
-       <value value="0x30" name="TFMT6_8_8_8_8_UNORM"/>
-       <value value="0x31" name="TFMT6_8_8_8_UNORM"/>
-       <value value="0x32" name="TFMT6_8_8_8_8_SNORM"/>
-       <value value="0x33" name="TFMT6_8_8_8_8_UINT"/>
-       <value value="0x34" name="TFMT6_8_8_8_8_SINT"/>
-       <value value="0x35" name="TFMT6_9_9_9_E5_FLOAT"/>
-       <value value="0x36" name="TFMT6_10_10_10_2_UNORM"/>
-       <value value="0x3a" name="TFMT6_10_10_10_2_UINT"/>
-       <value value="0x42" name="TFMT6_11_11_10_FLOAT"/>
-       <value value="0x43" name="TFMT6_16_16_UNORM"/>
-       <value value="0x44" name="TFMT6_16_16_SNORM"/>
-       <value value="0x45" name="TFMT6_16_16_FLOAT"/>
-       <value value="0x46" name="TFMT6_16_16_UINT"/>
-       <value value="0x47" name="TFMT6_16_16_SINT"/>
-       <value value="0x4a" name="TFMT6_32_FLOAT"/>
-       <value value="0x4b" name="TFMT6_32_UINT"/>
-       <value value="0x4c" name="TFMT6_32_SINT"/>
-       <value value="0x60" name="TFMT6_16_16_16_16_UNORM"/>
-       <value value="0x61" name="TFMT6_16_16_16_16_SNORM"/>
-       <value value="0x62" name="TFMT6_16_16_16_16_FLOAT"/>
-       <value value="0x63" name="TFMT6_16_16_16_16_UINT"/>
-       <value value="0x64" name="TFMT6_16_16_16_16_SINT"/>
-       <value value="0x67" name="TFMT6_32_32_FLOAT"/>
-       <value value="0x68" name="TFMT6_32_32_UINT"/>
-       <value value="0x69" name="TFMT6_32_32_SINT"/>
-       <value value="0x72" name="TFMT6_32_32_32_UINT"/>
-       <value value="0x73" name="TFMT6_32_32_32_SINT"/>
-       <value value="0x74" name="TFMT6_32_32_32_FLOAT"/>
-       <value value="0x82" name="TFMT6_32_32_32_32_FLOAT"/>
-       <value value="0x83" name="TFMT6_32_32_32_32_UINT"/>
-       <value value="0x84" name="TFMT6_32_32_32_32_SINT"/>
-       <value value="0x91" name="TFMT6_Z24_UNORM_S8_UINT"/>
-       <value value="0xa0" name="TFMT6_X8Z24_UNORM"/>
-
-       <value value="0xab" name="TFMT6_ETC2_RG11_UNORM"/>
-       <value value="0xac" name="TFMT6_ETC2_RG11_SNORM"/>
-       <value value="0xad" name="TFMT6_ETC2_R11_UNORM"/>
-       <value value="0xae" name="TFMT6_ETC2_R11_SNORM"/>
-       <value value="0xaf" name="TFMT6_ETC1"/>
-       <value value="0xb0" name="TFMT6_ETC2_RGB8"/>
-       <value value="0xb1" name="TFMT6_ETC2_RGBA8"/>
-       <value value="0xb2" name="TFMT6_ETC2_RGB8A1"/>
-       <value value="0xb3" name="TFMT6_DXT1"/>
-       <value value="0xb4" name="TFMT6_DXT3"/>
-       <value value="0xb5" name="TFMT6_DXT5"/>
-       <value value="0xb7" name="TFMT6_RGTC1_UNORM"/>
-       <value value="0xb8" name="TFMT6_RGTC1_SNORM"/>
-       <value value="0xbb" name="TFMT6_RGTC2_UNORM"/>
-       <value value="0xbc" name="TFMT6_RGTC2_SNORM"/>
-       <value value="0xbe" name="TFMT6_BPTC_UFLOAT"/>
-       <value value="0xbf" name="TFMT6_BPTC_FLOAT"/>
-       <value value="0xc0" name="TFMT6_BPTC"/>
-       <value value="0xc1" name="TFMT6_ASTC_4x4"/>
-       <value value="0xc2" name="TFMT6_ASTC_5x4"/>
-       <value value="0xc3" name="TFMT6_ASTC_5x5"/>
-       <value value="0xc4" name="TFMT6_ASTC_6x5"/>
-       <value value="0xc5" name="TFMT6_ASTC_6x6"/>
-       <value value="0xc6" name="TFMT6_ASTC_8x5"/>
-       <value value="0xc7" name="TFMT6_ASTC_8x6"/>
-       <value value="0xc8" name="TFMT6_ASTC_8x8"/>
-       <value value="0xc9" name="TFMT6_ASTC_10x5"/>
-       <value value="0xca" name="TFMT6_ASTC_10x6"/>
-       <value value="0xcb" name="TFMT6_ASTC_10x8"/>
-       <value value="0xcc" name="TFMT6_ASTC_10x10"/>
-       <value value="0xcd" name="TFMT6_ASTC_12x10"/>
-       <value value="0xce" name="TFMT6_ASTC_12x12"/>
+<enum name="a6xx_format">
+       <value value="0x02" name="FMT6_A8_UNORM"/>
+       <value value="0x03" name="FMT6_8_UNORM"/>
+       <value value="0x04" name="FMT6_8_SNORM"/>
+       <value value="0x05" name="FMT6_8_UINT"/>
+       <value value="0x06" name="FMT6_8_SINT"/>
+
+       <value value="0x08" name="FMT6_4_4_4_4_UNORM"/>
+       <value value="0x0a" name="FMT6_5_5_5_1_UNORM"/>
+       <value value="0x0c" name="FMT6_1_5_5_5_UNORM"/> <!-- read only -->
+       <value value="0x0e" name="FMT6_5_6_5_UNORM"/>
+
+       <value value="0x0f" name="FMT6_8_8_UNORM"/>
+       <value value="0x10" name="FMT6_8_8_SNORM"/>
+       <value value="0x11" name="FMT6_8_8_UINT"/>
+       <value value="0x12" name="FMT6_8_8_SINT"/>
+       <value value="0x13" name="FMT6_L8_A8_UNORM"/>
+
+       <value value="0x15" name="FMT6_16_UNORM"/>
+       <value value="0x16" name="FMT6_16_SNORM"/>
+       <value value="0x17" name="FMT6_16_FLOAT"/>
+       <value value="0x18" name="FMT6_16_UINT"/>
+       <value value="0x19" name="FMT6_16_SINT"/>
+
+       <value value="0x21" name="FMT6_8_8_8_UNORM"/>
+       <value value="0x22" name="FMT6_8_8_8_SNORM"/>
+       <value value="0x23" name="FMT6_8_8_8_UINT"/>
+       <value value="0x24" name="FMT6_8_8_8_SINT"/>
+
+       <value value="0x30" name="FMT6_8_8_8_8_UNORM"/>
+       <value value="0x31" name="FMT6_8_8_8_X8_UNORM"/> <!-- samples 1 for alpha -->
+       <value value="0x32" name="FMT6_8_8_8_8_SNORM"/>
+       <value value="0x33" name="FMT6_8_8_8_8_UINT"/>
+       <value value="0x34" name="FMT6_8_8_8_8_SINT"/>
+
+       <value value="0x35" name="FMT6_9_9_9_E5_FLOAT"/>
+
+       <value value="0x36" name="FMT6_10_10_10_2_UNORM"/>
+       <value value="0x37" name="FMT6_10_10_10_2_UNORM_DEST"/>
+       <value value="0x39" name="FMT6_10_10_10_2_SNORM"/>
+       <value value="0x3a" name="FMT6_10_10_10_2_UINT"/>
+       <value value="0x3b" name="FMT6_10_10_10_2_SINT"/>
+
+       <value value="0x42" name="FMT6_11_11_10_FLOAT"/>
+
+       <value value="0x43" name="FMT6_16_16_UNORM"/>
+       <value value="0x44" name="FMT6_16_16_SNORM"/>
+       <value value="0x45" name="FMT6_16_16_FLOAT"/>
+       <value value="0x46" name="FMT6_16_16_UINT"/>
+       <value value="0x47" name="FMT6_16_16_SINT"/>
+
+       <value value="0x48" name="FMT6_32_UNORM"/>
+       <value value="0x49" name="FMT6_32_SNORM"/>
+       <value value="0x4a" name="FMT6_32_FLOAT"/>
+       <value value="0x4b" name="FMT6_32_UINT"/>
+       <value value="0x4c" name="FMT6_32_SINT"/>
+       <value value="0x4d" name="FMT6_32_FIXED"/>
+
+       <value value="0x58" name="FMT6_16_16_16_UNORM"/>
+       <value value="0x59" name="FMT6_16_16_16_SNORM"/>
+       <value value="0x5a" name="FMT6_16_16_16_FLOAT"/>
+       <value value="0x5b" name="FMT6_16_16_16_UINT"/>
+       <value value="0x5c" name="FMT6_16_16_16_SINT"/>
+
+       <value value="0x60" name="FMT6_16_16_16_16_UNORM"/>
+       <value value="0x61" name="FMT6_16_16_16_16_SNORM"/>
+       <value value="0x62" name="FMT6_16_16_16_16_FLOAT"/>
+       <value value="0x63" name="FMT6_16_16_16_16_UINT"/>
+       <value value="0x64" name="FMT6_16_16_16_16_SINT"/>
+
+       <value value="0x65" name="FMT6_32_32_UNORM"/>
+       <value value="0x66" name="FMT6_32_32_SNORM"/>
+       <value value="0x67" name="FMT6_32_32_FLOAT"/>
+       <value value="0x68" name="FMT6_32_32_UINT"/>
+       <value value="0x69" name="FMT6_32_32_SINT"/>
+       <value value="0x6a" name="FMT6_32_32_FIXED"/>
+
+       <value value="0x70" name="FMT6_32_32_32_UNORM"/>
+       <value value="0x71" name="FMT6_32_32_32_SNORM"/>
+       <value value="0x72" name="FMT6_32_32_32_UINT"/>
+       <value value="0x73" name="FMT6_32_32_32_SINT"/>
+       <value value="0x74" name="FMT6_32_32_32_FLOAT"/>
+       <value value="0x75" name="FMT6_32_32_32_FIXED"/>
+
+       <value value="0x80" name="FMT6_32_32_32_32_UNORM"/>
+       <value value="0x81" name="FMT6_32_32_32_32_SNORM"/>
+       <value value="0x82" name="FMT6_32_32_32_32_FLOAT"/>
+       <value value="0x83" name="FMT6_32_32_32_32_UINT"/>
+       <value value="0x84" name="FMT6_32_32_32_32_SINT"/>
+       <value value="0x85" name="FMT6_32_32_32_32_FIXED"/>
+
+       <value value="0x91" name="FMT6_Z24_UNORM_S8_UINT_AS_R8G8B8A8"/>
+       <value value="0xa0" name="FMT6_Z24_UNORM_S8_UINT"/>
+
+       <value value="0xab" name="FMT6_ETC2_RG11_UNORM"/>
+       <value value="0xac" name="FMT6_ETC2_RG11_SNORM"/>
+       <value value="0xad" name="FMT6_ETC2_R11_UNORM"/>
+       <value value="0xae" name="FMT6_ETC2_R11_SNORM"/>
+       <value value="0xaf" name="FMT6_ETC1"/>
+       <value value="0xb0" name="FMT6_ETC2_RGB8"/>
+       <value value="0xb1" name="FMT6_ETC2_RGBA8"/>
+       <value value="0xb2" name="FMT6_ETC2_RGB8A1"/>
+       <value value="0xb3" name="FMT6_DXT1"/>
+       <value value="0xb4" name="FMT6_DXT3"/>
+       <value value="0xb5" name="FMT6_DXT5"/>
+       <value value="0xb7" name="FMT6_RGTC1_UNORM"/>
+       <value value="0xb8" name="FMT6_RGTC1_SNORM"/>
+       <value value="0xbb" name="FMT6_RGTC2_UNORM"/>
+       <value value="0xbc" name="FMT6_RGTC2_SNORM"/>
+       <value value="0xbe" name="FMT6_BPTC_UFLOAT"/>
+       <value value="0xbf" name="FMT6_BPTC_FLOAT"/>
+       <value value="0xc0" name="FMT6_BPTC"/>
+       <value value="0xc1" name="FMT6_ASTC_4x4"/>
+       <value value="0xc2" name="FMT6_ASTC_5x4"/>
+       <value value="0xc3" name="FMT6_ASTC_5x5"/>
+       <value value="0xc4" name="FMT6_ASTC_6x5"/>
+       <value value="0xc5" name="FMT6_ASTC_6x6"/>
+       <value value="0xc6" name="FMT6_ASTC_8x5"/>
+       <value value="0xc7" name="FMT6_ASTC_8x6"/>
+       <value value="0xc8" name="FMT6_ASTC_8x8"/>
+       <value value="0xc9" name="FMT6_ASTC_10x5"/>
+       <value value="0xca" name="FMT6_ASTC_10x6"/>
+       <value value="0xcb" name="FMT6_ASTC_10x8"/>
+       <value value="0xcc" name="FMT6_ASTC_10x10"/>
+       <value value="0xcd" name="FMT6_ASTC_12x10"/>
+       <value value="0xce" name="FMT6_ASTC_12x12"/>
+
+       <!-- same as X8Z24_UNORM but for sampling stencil (integer, 2nd channel) -->
+       <value value="0xea" name="FMT6_S8Z24_UINT"/>
 </enum>
 
 <enum name="a6xx_tex_fetchsize">
@@ -1011,14 +922,47 @@ to upconvert to 32b float internally?
        <reg32 offset="0x0806" name="CP_RB_RPTR"/>
        <reg32 offset="0x0807" name="CP_RB_WPTR"/>
        <reg32 offset="0x0808" name="CP_SQE_CNTL"/>
+       <reg32 offset="0x0812" name="CP_CP2GMU_STATUS">
+               <bitfield name="IFPC" pos="0" type="boolean"/>
+       </reg32>
        <reg32 offset="0x0821" name="CP_HW_FAULT"/>
        <reg32 offset="0x0823" name="CP_INTERRUPT_STATUS"/>
        <reg32 offset="0x0824" name="CP_PROTECT_STATUS"/>
        <reg32 offset="0x0830" name="CP_SQE_INSTR_BASE_LO"/>
        <reg32 offset="0x0831" name="CP_SQE_INSTR_BASE_HI"/>
        <reg32 offset="0x0840" name="CP_MISC_CNTL"/>
-       <reg32 offset="0x08C1" name="CP_ROQ_THRESHOLDS_1"/>
-       <reg32 offset="0x08C2" name="CP_ROQ_THRESHOLDS_2"/>
+       <!-- all the threshold values seem to be in units of quad-dwords: -->
+       <reg32 offset="0x08C1" name="CP_ROQ_THRESHOLDS_1">
+               <doc>
+                       b0..7 seems to contain the size of buffered by not yet processed
+                       RB level cmdstream.. it's possible that it is a low threshold
+                       and b8..15 is a high threshold?
+
+                       b16..23 identifies where IB1 data starts (and RB data ends?)
+
+                       b24..31 identifies where IB2 data starts (and IB1 data ends)
+               </doc>
+               <bitfield name="RB_LO" low="0" high="7" shr="2"/>
+               <bitfield name="RB_HI" low="8" high="15" shr="2"/>
+               <bitfield name="IB1_START" low="16" high="23" shr="2"/>
+               <bitfield name="IB2_START" low="24" high="31" shr="2"/>
+       </reg32>
+       <reg32 offset="0x08C2" name="CP_ROQ_THRESHOLDS_2">
+               <doc>
+                       low bits identify where CP_SET_DRAW_STATE stateobj
+                       processing starts (and IB2 data ends). I'm guessing
+                       b8 is part of this since (from downstream kgsl):
+
+                               /* ROQ sizes are twice as big on a640/a680 than on a630 */
+                               if (adreno_is_a640(adreno_dev) || adreno_is_a680(adreno_dev)) {
+                                       kgsl_regwrite(device, A6XX_CP_ROQ_THRESHOLDS_2, 0x02000140);
+                                       kgsl_regwrite(device, A6XX_CP_ROQ_THRESHOLDS_1, 0x8040362C);
+                               } ...
+               </doc>
+               <bitfield name="SDS_START" low="0" high="8" shr="2"/>
+               <!-- total ROQ size: -->
+               <bitfield name="ROQ_SIZE" low="16" high="31" shr="2"/>
+       </reg32>
        <reg32 offset="0x08C3" name="CP_MEM_POOL_SIZE"/>
        <reg32 offset="0x0841" name="CP_CHICKEN_DBG"/>
        <reg32 offset="0x0842" name="CP_ADDR_MODE_CNTL"/>
@@ -1075,6 +1019,26 @@ to upconvert to 32b float internally?
        <reg32 offset="0x092B" name="CP_IB2_BASE"/>
        <reg32 offset="0x092C" name="CP_IB2_BASE_HI"/>
        <reg32 offset="0x092D" name="CP_IB2_REM_SIZE"/>
+       <!-- SDS == CP_SET_DRAW_STATE: -->
+       <reg32 offset="0x092e" name="CP_SDS_BASE"/>
+       <reg32 offset="0x092f" name="CP_SDS_BASE_HI"/>
+       <reg32 offset="0x092e" name="CP_SDS_REM_SIZE"/>
+       <reg32 offset="0x0931" name="CP_BIN_SIZE_ADDRESS"/>
+       <reg32 offset="0x0932" name="CP_BIN_SIZE_ADDRESS_HI"/>
+       <reg32 offset="0x0934" name="CP_BIN_DATA_ADDR"/>
+       <reg32 offset="0x0935" name="CP_BIN_DATA_ADDR_HI"/>
+       <!--
+       There are probably similar registers for RB and SDS, teasing out SDS will
+       take a slightly better test case..
+        -->
+       <reg32 offset="0x0949" name="CP_CSQ_IB1_STAT">
+               <doc>number of remaining dwords incl current dword being consumed?</doc>
+               <bitfield name="REM" low="16" high="31"/>
+       </reg32>
+       <reg32 offset="0x094a" name="CP_CSQ_IB2_STAT">
+               <doc>number of remaining dwords incl current dword being consumed?</doc>
+               <bitfield name="REM" low="16" high="31"/>
+       </reg32>
        <reg32 offset="0x0980" name="CP_ALWAYS_ON_COUNTER_LO"/>
        <reg32 offset="0x0981" name="CP_ALWAYS_ON_COUNTER_HI"/>
        <reg32 offset="0x098D" name="CP_AHB_CNTL"/>
@@ -1108,7 +1072,9 @@ to upconvert to 32b float internally?
                <bitfield high="1" low="1" name="CP_AHB_BUSY_CP_MASTER" />
                <bitfield high="0" low="0" name="CP_AHB_BUSY_CX_MASTER"/>
        </reg32>
-       <reg32 offset="0x0213" name="RBBM_STATUS3"/>
+       <reg32 offset="0x0213" name="RBBM_STATUS3">
+               <bitfield pos="24" name="SMMU_STALLED_ON_FAULT" type="boolean"/>
+       </reg32>
        <reg32 offset="0x0215" name="RBBM_VBIF_GX_RESET_STATUS"/>
        <reg32 offset="0x0400" name="RBBM_PERFCTR_CP_0_LO"/>
        <reg32 offset="0x0401" name="RBBM_PERFCTR_CP_0_HI"/>
@@ -1777,6 +1743,7 @@ to upconvert to 32b float internally?
        </reg32>
        <reg32 offset="0x0c03" name="VSC_SIZE_ADDRESS_LO"/>
        <reg32 offset="0x0c04" name="VSC_SIZE_ADDRESS_HI"/>
+       <reg64 offset="0x0c03" name="VSC_SIZE_ADDRESS" type="waddress"/>
        <reg32 offset="0x0c06" name="VSC_BIN_COUNT">
                <bitfield name="NX" low="1" high="10" type="uint"/>
                <bitfield name="NY" low="11" high="20" type="uint"/>
@@ -1817,10 +1784,12 @@ to upconvert to 32b float internally?
         -->
        <reg32 offset="0x0c30" name="VSC_PIPE_DATA2_ADDRESS_LO"/>
        <reg32 offset="0x0c31" name="VSC_PIPE_DATA2_ADDRESS_HI"/>
+       <reg64 offset="0x0c30" name="VSC_PIPE_DATA2_ADDRESS" type="waddress"/>
        <reg32 offset="0x0c32" name="VSC_PIPE_DATA2_PITCH"/>
        <reg32 offset="0x0c33" name="VSC_PIPE_DATA2_ARRAY_PITCH" shr="4" type="uint"/>
        <reg32 offset="0x0c34" name="VSC_PIPE_DATA_ADDRESS_LO"/>
        <reg32 offset="0x0c35" name="VSC_PIPE_DATA_ADDRESS_HI"/>
+       <reg64 offset="0x0c34" name="VSC_PIPE_DATA_ADDRESS" type="waddress"/>
        <reg32 offset="0x0c36" name="VSC_PIPE_DATA_PITCH"/>
        <reg32 offset="0x0c37" name="VSC_PIPE_DATA_ARRAY_PITCH" shr="4" type="uint"/>
 
@@ -1854,7 +1823,20 @@ to upconvert to 32b float internally?
        <!-- always 0x03200000 ? -->
        <reg32 offset="0x0e12" name="UCHE_UNKNOWN_0E12"/>
 
-       <reg32 offset="0x8000" name="GRAS_UNKNOWN_8000"/>
+       <reg32 offset="0x8000" name="GRAS_CL_CNTL">
+               <bitfield name="CLIP_DISABLE" pos="0" type="boolean"/>
+               <bitfield name="ZNEAR_CLIP_DISABLE" pos="1" type="boolean"/>
+               <bitfield name="ZFAR_CLIP_DISABLE" pos="2" type="boolean"/>
+               <!-- set with depthClampEnable, not clear what it does -->
+               <bitfield name="UNK5" pos="5" type="boolean"/>
+               <!-- controls near z clip behavior (set for vulkan) -->
+               <bitfield name="ZERO_GB_SCALE_Z" pos="6" type="boolean"/>
+               <!-- guess based on a3xx and meaning of bits 8 and 9
+                    if the guess is right then this is related to point sprite clipping -->
+               <bitfield name="VP_CLIP_CODE_IGNORE" pos="7" type="boolean"/>
+               <bitfield name="VP_XFORM_DISABLE" pos="8" type="boolean"/>
+               <bitfield name="PERSP_DIVISION_DISABLE" pos="9" type="boolean"/>
+       </reg32>
        <reg32 offset="0x8001" name="GRAS_UNKNOWN_8001"/>
        <reg32 offset="0x8002" name="GRAS_UNKNOWN_8002"/>
        <reg32 offset="0x8003" name="GRAS_UNKNOWN_8003"/>
@@ -1866,10 +1848,8 @@ to upconvert to 32b float internally?
          <value value="0x3" name="LAYER_2D_ARRAY"/>
        </enum>
 
-       <reg32 offset="0x8004" name="GRAS_LAYER_CNTL">
-               <bitfield name="LAYERED" pos="0" type="boolean"/>
-               <bitfield name="TYPE" low="1" high="2" type="a6xx_layer_type"/>
-       </reg32>
+       <!-- index of highest layer that can be written to via gl_Layer -->
+       <reg32 offset="0x8004" name="GRAS_MAX_LAYER_INDEX" type="uint"/>
 
        <reg32 offset="0x8005" name="GRAS_CNTL">
                <!-- see also RB_RENDER_CONTROL0 -->
@@ -1901,6 +1881,10 @@ to upconvert to 32b float internally?
        <reg32 offset="0x8014" name="GRAS_CL_VPORT_ZOFFSET_0" type="float"/>
        <reg32 offset="0x8015" name="GRAS_CL_VPORT_ZSCALE_0" type="float"/>
 
+       <!-- not clear what it does, mirrors RB_Z_CLAMP_MIN -->
+       <reg32 offset="0x8070" name="GRAS_CL_Z_CLAMP_MIN" type="float"/>
+       <reg32 offset="0x8071" name="GRAS_CL_Z_CLAMP_MAX" type="float"/>
+
        <reg32 offset="0x8090" name="GRAS_SU_CNTL">
                <bitfield name="CULL_FRONT" pos="0" type="boolean"/>
                <bitfield name="CULL_BACK" pos="1" type="boolean"/>
@@ -1949,12 +1933,25 @@ to upconvert to 32b float internally?
                <bitfield name="MSAA_DISABLE" pos="2" type="boolean"/>
        </reg32>
 
-       <!-- always 0x0 -->
-       <reg32 offset="0x80a4" name="GRAS_UNKNOWN_80A4"/>
-       <!-- always 0x0 -->
-       <reg32 offset="0x80a5" name="GRAS_UNKNOWN_80A5"/>
-       <!-- always 0x0 -->
-       <reg32 offset="0x80a6" name="GRAS_UNKNOWN_80A6"/>
+       <bitset name="a6xx_sample_config" inline="yes">
+               <bitfield name="LOCATION_ENABLE" pos="1" type="boolean"/>
+       </bitset>
+
+       <bitset name="a6xx_sample_locations" inline="yes">
+               <bitfield name="SAMPLE_0_X" low="0" high="3" radix="4" type="fixed"/>
+               <bitfield name="SAMPLE_0_Y" low="4" high="7" radix="4" type="fixed"/>
+               <bitfield name="SAMPLE_1_X" low="8" high="11" radix="4" type="fixed"/>
+               <bitfield name="SAMPLE_1_Y" low="12" high="15" radix="4" type="fixed"/>
+               <bitfield name="SAMPLE_2_X" low="16" high="19" radix="4" type="fixed"/>
+               <bitfield name="SAMPLE_2_Y" low="20" high="23" radix="4" type="fixed"/>
+               <bitfield name="SAMPLE_3_X" low="24" high="27" radix="4" type="fixed"/>
+               <bitfield name="SAMPLE_3_Y" low="28" high="31" radix="4" type="fixed"/>
+       </bitset>
+
+       <reg32 offset="0x80a4" name="GRAS_SAMPLE_CONFIG" type="a6xx_sample_config"/>
+       <reg32 offset="0x80a5" name="GRAS_SAMPLE_LOCATION_0" type="a6xx_sample_locations"/>
+       <reg32 offset="0x80a6" name="GRAS_SAMPLE_LOCATION_1" type="a6xx_sample_locations"/>
+
        <!-- always 0x0 -->
        <reg32 offset="0x80af" name="GRAS_UNKNOWN_80AF"/>
 
@@ -1981,16 +1978,18 @@ to upconvert to 32b float internally?
        </reg32>
        <reg32 offset="0x8101" name="GRAS_UNKNOWN_8101"/>
        <reg32 offset="0x8102" name="GRAS_2D_BLIT_INFO">
-               <bitfield name="COLOR_FORMAT" low="0" high="7" type="a6xx_color_fmt"/>
+               <bitfield name="COLOR_FORMAT" low="0" high="7" type="a6xx_format"/>
        </reg32>
        <reg32 offset="0x8103" name="GRAS_LRZ_BUFFER_BASE_LO"/>
        <reg32 offset="0x8104" name="GRAS_LRZ_BUFFER_BASE_HI"/>
+       <reg64 offset="0x8103" name="GRAS_LRZ_BUFFER_BASE" type="waddress"/>
        <reg32 offset="0x8105" name="GRAS_LRZ_BUFFER_PITCH">
                <bitfield name="PITCH" low="0" high="10" shr="5" type="uint"/>
                <bitfield name="ARRAY_PITCH" low="11" high="21" shr="5" type="uint"/> <!-- ??? -->
        </reg32>
        <reg32 offset="0x8106" name="GRAS_LRZ_FAST_CLEAR_BUFFER_BASE_LO"/>
        <reg32 offset="0x8107" name="GRAS_LRZ_FAST_CLEAR_BUFFER_BASE_HI"/>
+       <reg32 offset="0x8106" name="GRAS_LRZ_FAST_CLEAR_BUFFER_BASE" type="waddress"/>
 
        <reg32 offset="0x8109" name="GRAS_SAMPLE_CNTL">
                <bitfield name="PER_SAMP_MODE" pos="0" type="boolean"/>
@@ -2010,7 +2009,7 @@ to upconvert to 32b float internally?
        <bitset name="a6xx_2d_blit_cntl" inline="yes">
                <bitfield name="ROTATE" low="0" high="2" type="a6xx_rotation"/>
                <bitfield name="SOLID_COLOR" pos="7" type="boolean"/>
-               <bitfield name="COLOR_FORMAT" low="8" high="15" type="a6xx_color_fmt"/>
+               <bitfield name="COLOR_FORMAT" low="8" high="15" type="a6xx_format"/>
                <bitfield name="SCISSOR" pos="16" type="boolean"/>
 
                <bitfield name="UNK" low="17" high="18" type="uint"/>
@@ -2072,12 +2071,9 @@ to upconvert to 32b float internally?
                <bitfield name="MSAA_DISABLE" pos="2" type="boolean"/>
        </reg32>
 
-       <!-- always 0x0 ? -->
-       <reg32 offset="0x8804" name="RB_UNKNOWN_8804"/>
-       <!-- always 0x0 ? -->
-       <reg32 offset="0x8805" name="RB_UNKNOWN_8805"/>
-       <!-- always 0x0 ? -->
-       <reg32 offset="0x8806" name="RB_UNKNOWN_8806"/>
+       <reg32 offset="0x8804" name="RB_SAMPLE_CONFIG" type="a6xx_sample_config"/>
+       <reg32 offset="0x8805" name="RB_SAMPLE_LOCATION_0" type="a6xx_sample_locations"/>
+       <reg32 offset="0x8806" name="RB_SAMPLE_LOCATION_1" type="a6xx_sample_locations"/>
 
        <!--
        note: maybe not actually called RB_RENDER_CONTROLn (since RB_RENDER_CNTL
@@ -2184,7 +2180,7 @@ to upconvert to 32b float internally?
                        <bitfield name="ALPHA_DEST_FACTOR" low="24" high="28" type="adreno_rb_blend_factor"/>
                </reg32>
                <reg32 offset="0x2" name="BUF_INFO">
-                       <bitfield name="COLOR_FORMAT" low="0" high="7" type="a6xx_color_fmt"/>
+                       <bitfield name="COLOR_FORMAT" low="0" high="7" type="a6xx_format"/>
                        <bitfield name="COLOR_TILE_MODE" low="8" high="9" type="a6xx_tile_mode"/>
                        <bitfield name="COLOR_SWAP" low="13" high="14" type="a3xx_color_swap"/>
                </reg32>
@@ -2202,6 +2198,9 @@ to upconvert to 32b float internally?
                 -->
                <reg32 offset="0x5" name="BASE_LO"/>
                <reg32 offset="0x6" name="BASE_HI"/>
+
+               <reg64 offset="0x5" name="BASE" type="waddress"/>
+
                <reg32 offset="0x7" name="BASE_GMEM"/>
        </array>
 
@@ -2229,6 +2228,7 @@ to upconvert to 32b float internally?
                <bitfield name="Z_ENABLE" pos="0" type="boolean"/>
                <bitfield name="Z_WRITE_ENABLE" pos="1" type="boolean"/>
                <bitfield name="ZFUNC" low="2" high="4" type="adreno_compare_func"/>
+               <bitfield name="Z_CLAMP_ENABLE" pos="5" type="boolean"/>
                <doc>Z_TEST_ENABLE bit is set for zfunc other than GL_ALWAYS or GL_NEVER</doc>
                <bitfield name="Z_TEST_ENABLE" pos="6" type="boolean"/>
        </reg32>
@@ -2245,6 +2245,7 @@ to upconvert to 32b float internally?
        </reg32>
        <reg32 offset="0x8875" name="RB_DEPTH_BUFFER_BASE_LO"/>
        <reg32 offset="0x8876" name="RB_DEPTH_BUFFER_BASE_HI"/>
+       <reg64 offset="0x8875" name="RB_DEPTH_BUFFER_BASE" type="waddress"/>
        <reg32 offset="0x8877" name="RB_DEPTH_BUFFER_BASE_GMEM"/>
 
        <!-- always 0x0 ? -->
@@ -2282,6 +2283,7 @@ to upconvert to 32b float internally?
        </reg32>
        <reg32 offset="0x8884" name="RB_STENCIL_BUFFER_BASE_LO"/>
        <reg32 offset="0x8885" name="RB_STENCIL_BUFFER_BASE_HI"/>
+       <reg64 offset="0x8884" name="RB_STENCIL_BUFFER_BASE" type="waddress"/>
        <reg32 offset="0x8886" name="RB_STENCIL_BUFFER_BASE_GMEM"/>
        <reg32 offset="0x8887" name="RB_STENCILREF">
                <bitfield name="REF" low="0" high="7"/>
@@ -2304,6 +2306,10 @@ to upconvert to 32b float internally?
                <bitfield name="ENABLE" pos="0" type="boolean"/>
        </reg32>
 
+       <!-- clamps depth value for depth test/write -->
+       <reg32 offset="0x88c0" name="RB_Z_CLAMP_MIN" type="float"/>
+       <reg32 offset="0x88c1" name="RB_Z_CLAMP_MAX" type="float"/>
+
        <reg32 offset="0x88d0" name="RB_UNKNOWN_88D0"/>
        <reg32 offset="0x88d1" name="RB_BLIT_SCISSOR_TL" type="adreno_reg_xy"/>
        <reg32 offset="0x88d2" name="RB_BLIT_SCISSOR_BR" type="adreno_reg_xy"/>
@@ -2317,14 +2323,16 @@ to upconvert to 32b float internally?
                <bitfield name="TILE_MODE" low="0" high="1" type="a6xx_tile_mode"/>
                <bitfield name="FLAGS" pos="2" type="boolean"/>
                <bitfield name="SAMPLES" low="3" high="4" type="a3xx_msaa_samples"/>
-               <bitfield name="COLOR_FORMAT" low="7" high="14" type="a6xx_color_fmt"/>
+               <bitfield name="COLOR_FORMAT" low="7" high="14" type="a6xx_format"/>
                <bitfield name="COLOR_SWAP" low="5" high="6" type="a3xx_color_swap"/>
        </reg32>
+       <reg64 offset="0x88d8" name="RB_BLIT_DST" type="waddress"/>
        <reg32 offset="0x88d8" name="RB_BLIT_DST_LO"/>
        <reg32 offset="0x88d9" name="RB_BLIT_DST_HI"/>
        <reg32 offset="0x88da" name="RB_BLIT_DST_PITCH" shr="6" type="uint"/>
        <!-- array-pitch is size of layer -->
        <reg32 offset="0x88db" name="RB_BLIT_DST_ARRAY_PITCH" shr="6" type="uint"/>
+       <reg64 offset="0x88dc" name="RB_BLIT_FLAG_DST" type="waddress"/>
        <reg32 offset="0x88dc" name="RB_BLIT_FLAG_DST_LO"/>
        <reg32 offset="0x88dd" name="RB_BLIT_FLAG_DST_HI"/>
        <reg32 offset="0x88de" name="RB_BLIT_FLAG_DST_PITCH">
@@ -2359,6 +2367,7 @@ to upconvert to 32b float internally?
 
        <reg32 offset="0x8900" name="RB_DEPTH_FLAG_BUFFER_BASE_LO"/>
        <reg32 offset="0x8901" name="RB_DEPTH_FLAG_BUFFER_BASE_HI"/>
+       <reg64 offset="0x8900" name="RB_DEPTH_FLAG_BUFFER_BASE" type="waddress"/>
        <reg32 offset="0x8902" name="RB_DEPTH_FLAG_BUFFER_PITCH">
                <bitfield name="PITCH" low="0" high="10" shr="6" type="uint"/>
                <bitfield name="ARRAY_PITCH" low="11" high="21" shr="7" type="uint"/>
@@ -2366,6 +2375,7 @@ to upconvert to 32b float internally?
        <array offset="0x8903" name="RB_MRT_FLAG_BUFFER" stride="3" length="8">
                <reg32 offset="0" name="ADDR_LO"/>
                <reg32 offset="1" name="ADDR_HI"/>
+               <reg64 offset="0" name="ADDR" type="waddress"/>
                <reg32 offset="2" name="PITCH">
                        <bitfield name="PITCH" low="0" high="10" shr="6" type="uint"/>
                        <bitfield name="ARRAY_PITCH" low="11" high="21" shr="7" type="uint"/> <!-- ??? -->
@@ -2378,27 +2388,30 @@ to upconvert to 32b float internally?
        <reg32 offset="0x8c01" name="RB_UNKNOWN_8C01"/>
 
        <bitset name="a6xx_2d_surf_info" inline="yes">
-               <bitfield name="COLOR_FORMAT" low="0" high="7" type="a6xx_color_fmt"/>
+               <bitfield name="COLOR_FORMAT" low="0" high="7" type="a6xx_format"/>
                <bitfield name="TILE_MODE" low="8" high="9" type="a6xx_tile_mode"/>
                <bitfield name="COLOR_SWAP" low="10" high="11" type="a3xx_color_swap"/>
-               <!-- b12 seems to be set when UBWC "FLAGS" buffer enabled -->
                <bitfield name="FLAGS" pos="12" type="boolean"/>
                <bitfield name="SRGB" pos="13" type="boolean"/>
                <!-- the rest is only for src -->
                <bitfield name="SAMPLES" low="14" high="15" type="a3xx_msaa_samples"/>
                <bitfield name="FILTER" pos="16" type="boolean"/>
                <bitfield name="SAMPLES_AVERAGE" pos="18" type="boolean"/>
+               <bitfield name="UNK20" pos="20" type="boolean"/>
+               <bitfield name="UNK22" pos="22" type="boolean"/>
        </bitset>
 
        <reg32 offset="0x8c17" name="RB_2D_DST_INFO" type="a6xx_2d_surf_info"/>
        <reg32 offset="0x8c18" name="RB_2D_DST_LO"/>
        <reg32 offset="0x8c19" name="RB_2D_DST_HI"/>
+       <reg64 offset="0x8c18" name="RB_2D_DST" type="waddress"/>
        <reg32 offset="0x8c1a" name="RB_2D_DST_SIZE">
                <bitfield name="PITCH" low="0" high="15" shr="6" type="uint"/>
        </reg32>
 
        <reg32 offset="0x8c20" name="RB_2D_DST_FLAGS_LO"/>
        <reg32 offset="0x8c21" name="RB_2D_DST_FLAGS_HI"/>
+       <reg64 offset="0x8c20" name="RB_2D_DST_FLAGS" type="waddress"/>
        <reg32 offset="0x8c22" name="RB_2D_DST_FLAGS_PITCH">
                <bitfield name="PITCH" low="0" high="10" shr="6" type="uint"/>
                <bitfield name="ARRAY_PITCH" low="11" high="21" shr="7" type="uint"/>
@@ -2415,7 +2428,19 @@ to upconvert to 32b float internally?
 
        <reg32 offset="0x8e04" name="RB_UNKNOWN_8E04"/>
 
-       <reg32 offset="0x8e07" name="RB_CCU_CNTL"/>  <!-- always 7c400004 or 10000000 -->
+       <reg32 offset="0x8e07" name="RB_CCU_CNTL">
+               <!-- offset into GMEM for something.
+                       important for sysmem path
+                       BLIT_OP_SCALE also writes to GMEM at this offset for GMEM store
+                       blob values for GMEM path (note: close to GMEM size):
+                       a618: 0x7c000 a630/a640: 0xf8000 a650: 0x114000
+                       SYSMEM path values:
+                       a618: 0x10000 a630/a640: 0x20000 a650: 0x30000
+               -->
+               <bitfield name="OFFSET" low="23" high="31" shr="12" type="uint"/>
+               <bitfield name="GMEM" pos="22" type="boolean"/> <!-- set for GMEM path -->
+               <bitfield name="UNK2" pos="2" type="boolean"/> <!-- sometimes set with GMEM? -->
+       </reg32>
 
        <reg32 offset="0x9100" name="VPC_UNKNOWN_9100"/>
 
@@ -2467,11 +2492,13 @@ to upconvert to 32b float internally?
        <reg32 offset="0x9219" name="VPC_SO_STREAM_COUNTS_HI"/>
 
        <array offset="0x921a" name="VPC_SO" stride="7" length="4">
+               <reg64 offset="0" name="BUFFER_BASE" type="waddress"/>
                <reg32 offset="0" name="BUFFER_BASE_LO"/>
                <reg32 offset="1" name="BUFFER_BASE_HI"/>
                <reg32 offset="2" name="BUFFER_SIZE"/>
                <reg32 offset="3" name="NCOMP"/>  <!-- component count -->
                <reg32 offset="4" name="BUFFER_OFFSET"/>
+               <reg64 offset="5" name="FLUSH_BASE" type="waddress"/>
                <reg32 offset="5" name="FLUSH_BASE_LO"/>
                <reg32 offset="6" name="FLUSH_BASE_HI"/>
        </array>
@@ -2600,7 +2627,7 @@ to upconvert to 32b float internally?
                        plus # of transform-feedback (streamout) varyings if using the
                        hw streamout (rather than stg instructions in shader)
                </doc>
-               <bitfield name="STRIDE_IN_VPC" low="0" high="6" type="uint"/>
+               <bitfield name="STRIDE_IN_VPC" low="0" high="7" type="uint"/>
                <bitfield name="PSIZE" pos="8" type="boolean"/>
        </reg32>
 
@@ -2608,7 +2635,7 @@ to upconvert to 32b float internally?
                <doc>
                  geometry shader
                </doc>
-               <bitfield name="STRIDE_IN_VPC" low="0" high="6" type="uint"/>
+               <bitfield name="STRIDE_IN_VPC" low="0" high="7" type="uint"/>
                <bitfield name="PSIZE" pos="8" type="boolean"/>
                <bitfield name="LAYER" pos="9" type="boolean"/>
                <bitfield name="PRIMITIVE_ID" pos="11" type="boolean"/>
@@ -2622,7 +2649,7 @@ to upconvert to 32b float internally?
                        plus # of transform-feedback (streamout) varyings if using the
                        hw streamout (rather than stg instructions in shader)
                </doc>
-               <bitfield name="STRIDE_IN_VPC" low="0" high="6" type="uint"/>
+               <bitfield name="STRIDE_IN_VPC" low="0" high="7" type="uint"/>
                <bitfield name="PSIZE" pos="8" type="boolean"/>
        </reg32>
        <reg32 offset="0x9b04" name="PC_PRIMITIVE_CNTL_4">
@@ -2632,7 +2659,7 @@ to upconvert to 32b float internally?
                        plus # of transform-feedback (streamout) varyings if using the
                        hw streamout (rather than stg instructions in shader)
                </doc>
-               <bitfield name="STRIDE_IN_VPC" low="0" high="6" type="uint"/>
+               <bitfield name="STRIDE_IN_VPC" low="0" high="7" type="uint"/>
                <bitfield name="PSIZE" pos="8" type="boolean"/>
        </reg32>
 
@@ -2657,11 +2684,22 @@ to upconvert to 32b float internally?
        <reg32 offset="0x9e08" name="PC_TESSFACTOR_ADDR_LO"/>
        <reg32 offset="0x9e09" name="PC_TESSFACTOR_ADDR_HI"/>
 
+       <!-- These match the contents of CP_SET_BIN_DATA (not written directly) -->
+       <reg32 offset="0x9e11" name="PC_VSTREAM_CONTROL">
+               <bitfield name="VSC_SIZE" low="16" high="21" type="uint"/>
+               <bitfield name="VSC_N" low="22" high="26" type="uint"/>
+       </reg32>
+       <reg32 offset="0x9e12" name="PC_BIN_DATA_ADDR2_LO"/>
+       <reg32 offset="0x9e13" name="PC_BIN_DATA_ADDR2_HI"/>
+       <reg32 offset="0x9e14" name="PC_BIN_DATA_ADDR_LO"/>
+       <reg32 offset="0x9e15" name="PC_BIN_DATA_ADDR_HI"/>
+
        <!-- always 0x0 -->
        <reg32 offset="0x9e72" name="PC_UNKNOWN_9E72"/>
 
        <reg32 offset="0xa000" name="VFD_CONTROL_0">
-               <bitfield name="VTXCNT" low="0" high="5" type="uint"/>
+               <bitfield name="FETCH_CNT" low="0" high="5" type="uint"/>
+               <bitfield name="DECODE_CNT" low="8" high="13" type="uint"/>
        </reg32>
        <reg32 offset="0xa001" name="VFD_CONTROL_1">
                <bitfield name="REGID4VTX" low="0" high="7" type="a3xx_regid"/>
@@ -2691,11 +2729,17 @@ to upconvert to 32b float internally?
 
        <!-- always 0x0 ? -->
        <reg32 offset="0xa008" name="VFD_UNKNOWN_A008"/>
-       <reg32 offset="0xa009" name="VFD_UNKNOWN_A009"/>
+       <reg32 offset="0xa009" name="VFD_ADD_OFFSET">
+               <!-- add VFD_INDEX_OFFSET to REGID4VTX -->
+               <bitfield name="VERTEX" pos="0" type="boolean"/>
+               <!-- add VFD_INSTANCE_START_OFFSET to REGID4INST -->
+               <bitfield name="INSTANCE" pos="1" type="boolean"/>
+       </reg32>
 
        <reg32 offset="0xa00e" name="VFD_INDEX_OFFSET"/>
        <reg32 offset="0xa00f" name="VFD_INSTANCE_START_OFFSET"/>
        <array offset="0xa010" name="VFD_FETCH" stride="4" length="32">
+               <reg64 offset="0x0" name="BASE" type="address"/>
                <reg32 offset="0x0" name="BASE_LO"/>
                <reg32 offset="0x1" name="BASE_HI"/>
                <reg32 offset="0x2" name="SIZE" type="uint"/>
@@ -2703,10 +2747,11 @@ to upconvert to 32b float internally?
        </array>
        <array offset="0xa090" name="VFD_DECODE" stride="2" length="32">
                <reg32 offset="0x0" name="INSTR">
-                       <!-- IDX appears to index into VFD_FETCH[] -->
+                       <!-- IDX and byte OFFSET into VFD_FETCH -->
                        <bitfield name="IDX" low="0" high="4" type="uint"/>
+                       <bitfield name="OFFSET" low="5" high="16"/>
                        <bitfield name="INSTANCED" pos="17" type="boolean"/>
-                       <bitfield name="FORMAT" low="20" high="27" type="a6xx_vtx_fmt"/>
+                       <bitfield name="FORMAT" low="20" high="27" type="a6xx_format"/>
                        <bitfield name="SWAP" low="28" high="29" type="a3xx_color_swap"/>
                        <bitfield name="UNK30" pos="30" type="boolean"/>
                        <bitfield name="FLOAT" pos="31" type="boolean"/>
@@ -2751,6 +2796,15 @@ to upconvert to 32b float internally?
        </bitset>
 
        <bitset name="a6xx_sp_xs_config" inline="yes">
+               <!--
+               Each of these are set if the given resource type is used
+               with the Vulkan/bindless binding model.
+               -->
+               <bitfield name="BINDLESS_TEX" pos="0" type="boolean"/>
+               <bitfield name="BINDLESS_SAMP" pos="1" type="boolean"/>
+               <bitfield name="BINDLESS_IBO" pos="2" type="boolean"/>
+               <bitfield name="BINDLESS_UBO" pos="3" type="boolean"/>
+
                <bitfield name="ENABLED" pos="8" type="boolean"/>
                <!--
                number of textures and samplers.. these might be swapped, with GL I
@@ -2764,7 +2818,7 @@ to upconvert to 32b float internally?
        <reg32 offset="0xa800" name="SP_VS_CTRL_REG0" type="a6xx_sp_xs_ctrl_reg0"/>
        <reg32 offset="0xa802" name="SP_PRIMITIVE_CNTL">
                <!-- # of VS outputs including pos/psize -->
-               <bitfield name="VSOUT" low="0" high="4" type="uint"/>
+               <bitfield name="VSOUT" low="0" high="5" type="uint"/>
        </reg32>
        <array offset="0xa803" name="SP_VS_OUT" stride="1" length="16">
                <reg32 offset="0x0" name="REG">
@@ -2840,7 +2894,7 @@ to upconvert to 32b float internally?
 
        <reg32 offset="0xa873" name="SP_PRIMITIVE_CNTL_GS">
                <!-- # of VS outputs including pos/psize -->
-               <bitfield name="GSOUT" low="0" high="4" type="uint"/>
+               <bitfield name="GSOUT" low="0" high="5" type="uint"/>
                <bitfield name="FLAGS_REGID" low="6" high="13" type="a3xx_regid"/>
        </reg32>
 
@@ -2934,7 +2988,7 @@ to upconvert to 32b float internally?
 
        <array offset="0xa996" name="SP_FS_MRT" stride="1" length="8">
                <reg32 offset="0" name="REG">
-                       <bitfield name="COLOR_FORMAT" low="0" high="7" type="a6xx_color_fmt"/>
+                       <bitfield name="COLOR_FORMAT" low="0" high="7" type="a6xx_format"/>
                        <bitfield name="COLOR_SINT" pos="8" type="boolean"/>
                        <bitfield name="COLOR_UINT" pos="9" type="boolean"/>
                </reg32>
@@ -2958,11 +3012,22 @@ to upconvert to 32b float internally?
                        <!--
                        CMD seems always 0x4??  3d, textureProj, textureLod seem to
                        skip pre-fetch.. TODO test texelFetch
+                        CMD is 0x6 when the Vulkan mode is enabled, and
+                        TEX_ID/SAMP_ID refer to the descriptor sets while the
+                        indices come from SP_FS_BINDLESS_PREFETCH[n]
                         -->
                        <bitfield name="CMD" low="27" high="31"/>
                </reg32>
        </array>
 
+       <!-- TODO confirm that this is actually an array -->
+       <array offset="0xa9a3" name="SP_FS_BINDLESS_PREFETCH" stride="1" length="4">
+               <reg32 offset="0" name="CMD">
+                       <bitfield name="SAMP_ID" low="0" high="7" type="uint"/>
+                       <bitfield name="TEX_ID" low="16" high="23" type="uint"/>
+               </reg32>
+       </array>
+
        <reg32 offset="0xa9a7" name="SP_FS_TEX_COUNT" type="uint"/>
 
        <!-- always 0x0 ? -->
@@ -2985,6 +3050,10 @@ to upconvert to 32b float internally?
        <reg32 offset="0xa9e6" name="SP_CS_TEX_CONST_LO"/>
        <reg32 offset="0xa9e7" name="SP_CS_TEX_CONST_HI"/>
 
+       <array offset="0xa9e8" name="SP_CS_BINDLESS_BASE" stride="2" length="5">
+               <reg64 offset="0" name="ADDR" type="waddress"/>
+       </array>
+
        <array offset="0xa98e" name="SP_FS_OUTPUT" stride="1" length="8">
                <doc>per MRT</doc>
                <reg32 offset="0x0" name="REG">
@@ -3012,6 +3081,10 @@ to upconvert to 32b float internally?
        <reg32 offset="0xab04" name="SP_FS_CONFIG" type="a6xx_sp_xs_config"/>
        <reg32 offset="0xab05" name="SP_FS_INSTRLEN" type="uint"/>
 
+       <array offset="0xab10" name="SP_BINDLESS_BASE" stride="2" length="5">
+               <reg64 offset="0" name="ADDR" type="waddress"/>
+       </array>
+
        <!--
        Combined IBO state for 3d pipe, used for Image and SSBO write/atomic
        instructions VS/HS/DS/GS/FS.  See SP_CS_IBO_* for compute shaders.
@@ -3029,7 +3102,7 @@ to upconvert to 32b float internally?
                <bitfield name="UINT" pos="2" type="boolean"/>
                <!-- looks like HW only cares about the base type of this format,
                     which matches the ifmt? -->
-               <bitfield name="COLOR_FORMAT" low="3" high="10" type="a6xx_color_fmt"/>
+               <bitfield name="COLOR_FORMAT" low="3" high="10" type="a6xx_format"/>
                <!-- set when ifmt is R2D_UNORM8_SRGB -->
                <bitfield name="SRGB" pos="11" type="boolean"/>
                <!-- some sort of channel mask, not sure what it is for -->
@@ -3045,6 +3118,12 @@ to upconvert to 32b float internally?
        <!-- always 0x3f ? -->
        <reg32 offset="0xae0f" name="SP_UNKNOWN_AE0F"/>
 
+       <!--
+       The downstream kernel calls the debug cluster of registers
+       "a6xx_sp_ps_tp_cluster" but this actually specifies the border
+       color base for compute shaders.
+       -->
+       <reg64 offset="0xb180" name="SP_PS_TP_BORDER_COLOR_BASE_ADDR" type="address"/>
        <!-- always 0x0 ? -->
        <reg32 offset="0xb182" name="SP_UNKNOWN_B182"/>
        <reg32 offset="0xb183" name="SP_UNKNOWN_B183"/>
@@ -3060,10 +3139,12 @@ to upconvert to 32b float internally?
        </reg32>
 
        <!-- looks to work in the same way as a5xx: -->
+       <reg64 offset="0xb302" name="SP_TP_BORDER_COLOR_BASE_ADDR" type="address"/>
        <reg32 offset="0xb302" name="SP_TP_BORDER_COLOR_BASE_ADDR_LO"/>
        <reg32 offset="0xb303" name="SP_TP_BORDER_COLOR_BASE_ADDR_HI"/>
-       <!-- always 0x0 ? -->
-       <reg32 offset="0xb304" name="SP_TP_UNKNOWN_B304"/>
+       <reg32 offset="0xb304" name="SP_TP_SAMPLE_CONFIG" type="a6xx_sample_config"/>
+       <reg32 offset="0xb305" name="SP_TP_SAMPLE_LOCATION_0" type="a6xx_sample_locations"/>
+       <reg32 offset="0xb306" name="SP_TP_SAMPLE_LOCATION_1" type="a6xx_sample_locations"/>
 
        <reg32 offset="0xb309" name="SP_TP_UNKNOWN_B309"/>
 
@@ -3079,12 +3160,14 @@ to upconvert to 32b float internally?
        </reg32>
        <reg32 offset="0xb4c2" name="SP_PS_2D_SRC_LO"/>
        <reg32 offset="0xb4c3" name="SP_PS_2D_SRC_HI"/>
+       <reg64 offset="0xb4c2" name="SP_PS_2D_SRC" type="waddress"/>
        <reg32 offset="0xb4c4" name="SP_PS_2D_SRC_PITCH">
           <bitfield name="PITCH" low="9" high="24" shr="6" type="uint"/>
        </reg32>
 
        <reg32 offset="0xb4ca" name="SP_PS_2D_SRC_FLAGS_LO"/>
        <reg32 offset="0xb4cb" name="SP_PS_2D_SRC_FLAGS_HI"/>
+       <reg64 offset="0xb4ca" name="SP_PS_2D_SRC_FLAGS" type="waddress"/>
        <reg32 offset="0xb4cc" name="SP_PS_2D_SRC_FLAGS_PITCH">
                <bitfield name="PITCH" low="0" high="10" shr="6" type="uint"/>
                <bitfield name="ARRAY_PITCH" low="11" high="21" shr="7" type="uint"/>
@@ -3174,6 +3257,11 @@ to upconvert to 32b float internally?
        <reg32 offset="0xb99a" name="HLSQ_CS_KERNEL_GROUP_Y"/>
        <reg32 offset="0xb99b" name="HLSQ_CS_KERNEL_GROUP_Z"/>
 
+       <!-- mirror of SP_CS_BINDLESS_BASE -->
+       <array offset="0xb9c0" name="HLSQ_CS_BINDLESS_BASE" stride="2" length="5">
+               <reg64 offset="0" name="ADDR" type="waddress"/>
+       </array>
+
        <!-- probably: -->
        <reg32 offset="0xbb08" name="HLSQ_UPDATE_CNTL"/>
 
@@ -3182,6 +3270,11 @@ to upconvert to 32b float internally?
        <!-- always 0x0 ? -->
        <reg32 offset="0xbb11" name="HLSQ_UNKNOWN_BB11"/>
 
+       <!-- mirror of SP_BINDLESS_BASE -->
+       <array offset="0xbb20" name="HLSQ_BINDLESS_BASE" stride="2" length="5">
+               <reg64 offset="0" name="ADDR" type="waddress"/>
+       </array>
+
        <!-- always 0x80 ? -->
        <reg32 offset="0xbe00" name="HLSQ_UNKNOWN_BE00"/>
        <!-- always 0x0 ? -->
@@ -3198,6 +3291,7 @@ to upconvert to 32b float internally?
                <value name="A6XX_TEX_NEAREST" value="0"/>
                <value name="A6XX_TEX_LINEAR" value="1"/>
                <value name="A6XX_TEX_ANISO" value="2"/>
+               <value name="A6XX_TEX_CUBIC" value="3"/> <!-- a650 only -->
        </enum>
        <enum name="a6xx_tex_clamp"> <!-- same as a4xx? -->
                <value name="A6XX_TEX_REPEAT" value="0"/>
@@ -3262,7 +3356,7 @@ to upconvert to 32b float internally?
                <bitfield name="SWIZ_W" low="13" high="15" type="a6xx_tex_swiz"/>
                <bitfield name="MIPLVLS" low="16" high="19" type="uint"/>
                <bitfield name="SAMPLES" low="20" high="21" type="a3xx_msaa_samples"/>
-               <bitfield name="FMT" low="22" high="29" type="a6xx_tex_fmt"/>
+               <bitfield name="FMT" low="22" high="29" type="a6xx_format"/>
                <bitfield name="SWAP" low="30" high="31" type="a3xx_color_swap"/>
        </reg32>
        <reg32 offset="1" name="1">
@@ -3343,7 +3437,7 @@ with a better name.
                used but if they are good chance position is same as TEX_CONST
                 -->
                <bitfield name="TILE_MODE" low="0" high="1" type="a6xx_tile_mode"/>
-               <bitfield name="FMT" low="22" high="29" type="a6xx_tex_fmt"/>
+               <bitfield name="FMT" low="22" high="29" type="a6xx_format"/>
        </reg32>
        <reg32 offset="1" name="1">
                <bitfield name="WIDTH" low="0" high="14" type="uint"/>
@@ -3402,19 +3496,7 @@ with a better name.
        </reg32>
        <reg32 offset="1" name="1">
                <bitfield name="BASE_HI" low="0" high="16"/>
-               <!-- size probably in high bits -->
-       </reg32>
-</domain>
-
-<domain name="CP_UNK_A6XX_55" width="32">
-       <reg32 offset="0" name="0">
-               <bitfield name="BASE_LO" low="0" high="31"/>
-       </reg32>
-       <reg32 offset="1" name="1">
-               <bitfield name="BASE_HI" low="0" high="16"/>
-       </reg32>
-       <reg32 offset="2" name="2">
-               <bitfield name="SIZE" low="0" high="15"/>
+               <bitfield name="SIZE" low="17" high="31"/> <!-- size in vec4 (4xDWORD) units -->
        </reg32>
 </domain>