i965/fs: Patch the instruction generating discards; don't use CMP.Z.

CMP.Z doesn't work on Gen4-5 because the boolean isn't guaranteed to be
0 or 0xFFFFFFFF - only the low bit is defined.

We can call emit_bool_to_cond_code to generate the condition in f0.0;
the last instruction will generate the flag value.  We can patch it to
use f0.1, and negate the condition.

Fixes discard tests on Gen4-5.

Haswell shader-db stats:
total instructions in shared programs: 5770279 -> 5769112 (-0.02%)
instructions in affected programs:     64342 -> 63175 (-1.81%)
helped:                                1069

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
This commit is contained in:
Kenneth Graunke 2015-02-25 16:08:14 -08:00
parent 4ebacf8aa6
commit ea696be5ac
1 changed files with 3 additions and 2 deletions

View File

@ -2422,8 +2422,9 @@ fs_visitor::visit(ir_discard *ir)
*/
fs_inst *cmp;
if (ir->condition) {
ir->condition->accept(this);
cmp = emit(CMP(reg_null_f, this->result, fs_reg(0), BRW_CONDITIONAL_Z));
emit_bool_to_cond_code(ir->condition);
cmp = (fs_inst *) this->instructions.get_tail();
cmp->conditional_mod = brw_negate_cmod(cmp->conditional_mod);
} else {
fs_reg some_reg = fs_reg(retype(brw_vec8_grf(0, 0),
BRW_REGISTER_TYPE_UW));