util/format: Add some NEON intrinsics-based u_format_unpack.

In looking at the profile of dEQP, GLES3 was spending 5-10% of its time in
ReadPixels, and almost all of that is b8g8r8a8_unorm8.  It's really slow
because we're getting about 47MB/s by doing uncached reads 32 bits at a
time in the code-generated unpack.  If we use NEON to generate larger bus
transactions, we can speed things up to 136MB/s.  In comparison, raw
ldr/str read/writes with no byte swapping can hit a max of 216MB/sec.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10014>

src/util/format/u_format_table.py

@@ -166,8 +166,11 @@ def write_format_table(formats):
         print("   },")
 
     def generate_table_getter(type):
+        suffix = ""
+        if type == "unpack_":
+            suffix = "_generic"
         print("const struct util_format_%sdescription *" % type)
-        print("util_format_%sdescription(enum pipe_format format)" % type)
+        print("util_format_%sdescription%s(enum pipe_format format)" % (type, suffix))
         print("{")
         print("   if (format >= ARRAY_SIZE(util_format_%sdescriptions))" % (type))
         print("      return NULL;")
@@ -242,7 +245,6 @@ def write_format_table(formats):
     print("};")
     print()
     generate_table_getter("pack_")
-
     print('static const struct util_format_unpack_description')
     print('util_format_unpack_descriptions[] = {')
     for format in formats: