gallium: Use unpack_rgba() instead of fetch_rgba in translate_generic

This is the only user of fetch_rgba outside of llvmpipe, and it's in the
fallback path of this fallback path.  Looking at an example of these two
functions, b8g8r8a8's unpack_rgba is 2.7x as long as fetch_rgba.  It feels
reasonable to sacrifice some perf in this already slow (VBO readback, and
a function pointer call per attribute per vertex) path to reduce our
binary size.  And, if I ever finish getting unpack codegen to switch to
rows instead of rects, that factor will go back down.

Saves 40kb of binary on non-llvmpipe gallium drivers.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6305>
This commit is contained in:
Eric Anholt
2020-08-13 10:14:11 -07:00
committed by Marge Bot
parent 5b8d67cb64
commit 09196355b2
@@ -50,7 +50,9 @@ struct translate_generic {
struct {
enum translate_element_type type;
util_format_fetch_rgba_func_ptr fetch;
void (*fetch)(void *dst, unsigned dst_stride,
const uint8_t *src, unsigned src_stride,
unsigned width, unsigned height);
unsigned buffer;
unsigned input_offset;
unsigned instance_divisor;
@@ -623,7 +625,7 @@ generic_run_one(struct translate_generic *tg,
if (likely(copy_size >= 0)) {
memcpy(dst, src, copy_size);
} else {
tg->attrib[attr].fetch(data, src, 0, 0);
tg->attrib[attr].fetch(data, 0, src, 0, 1, 1);
if (0)
debug_printf("Fetch linear attr %d from %p stride %d index %d: "
@@ -796,6 +798,8 @@ translate_generic_create(const struct translate_key *key)
for (i = 0; i < key->nr_elements; i++) {
const struct util_format_description *format_desc =
util_format_description(key->element[i].input_format);
const struct util_format_unpack_description *unpack =
util_format_unpack_description(key->element[i].input_format);
assert(format_desc);
@@ -811,8 +815,7 @@ translate_generic_create(const struct translate_key *key)
}
}
tg->attrib[i].fetch =
util_format_fetch_rgba_func(key->element[i].input_format);
tg->attrib[i].fetch = unpack->unpack_rgba;
tg->attrib[i].buffer = key->element[i].input_buffer;
tg->attrib[i].input_offset = key->element[i].input_offset;
tg->attrib[i].instance_divisor = key->element[i].instance_divisor;