Almost always, the MOV will get copy propagated out. Even if it doesn't,
it's probably better to just reload the uniform at next use (to reduce
register pressure) rather than try to save instruction count.
I was looking at this because in the presence of texturing (which calls
add_uniform() directly to get the uniform load forced into the
instruction) the c->uniform_contents indices don't match 1:1 with the
temporary qregs.
get_temp_for_uniform(struct vc4_compile *c, enum quniform_contents contents,
uint32_t data)
{
- for (int i = 0; i < c->num_uniforms; i++) {
- if (c->uniform_contents[i] == contents &&
- c->uniform_data[i] == data)
- return c->uniforms[i];
- }
-
struct qreg u = add_uniform(c, contents, data);
struct qreg t = qir_MOV(c, u);
-
- resize_qreg_array(c, &c->uniforms, &c->uniforms_array_size,
- u.index + 1);
-
- c->uniforms[u.index] = t;
return t;
}
struct qreg *temps;
struct qreg *inputs;
struct qreg *outputs;
- struct qreg *uniforms;
struct qreg *consts;
uint32_t temps_array_size;
uint32_t inputs_array_size;