brw: Strategically place flags initialization to help cmod prop

v2: Rebase on ac2b072312 ("brw: Add more specific brw_builder
helpers"), and fix a bug that caused the new instruction to possibly be
put in the wrong place.

No shader-db changes on any Intel platform.

fossil-db:

All Intel platforms had similar results. (Lunar Lake shown)
Totals:
Instrs: 233675305 -> 233641585 (-0.01%)
Cycle count: 32593658094 -> 32591467794 (-0.01%); split: -0.01%, +0.00%

Totals from 33513 (4.25% of 789264) affected shaders:
Instrs: 5200332 -> 5166612 (-0.65%)
Cycle count: 1499831128 -> 1497640828 (-0.15%); split: -0.15%, +0.00%

Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35444>

src/intel/compiler/brw_lower_subgroup_ops.cpp

@@ -389,6 +389,26 @@ brw_fill_flag(const brw_builder &bld, unsigned v, brw_inst *inst)
       }
    }
 
+   /* Performing the initialization before an instruction that writes to
+    * inst->src[0] might help cmod propagation. The expectation is that
+    * inst->src[0] will be used to generate flags. If scan_inst could be used
+    * to generate the flags instead, an instruction and temporary register
+    * might be saved.
+    */
+   foreach_inst_in_block_reverse_starting_from(brw_inst, scan_inst, inst) {
+      if (scan_inst->flags_read(devinfo))
+         break;
+
+      if (regions_overlap(scan_inst->dst, scan_inst->size_written,
+                          inst->src[0], inst->size_read(devinfo, 0))) {
+         ubld1.before(scan_inst).MOV(flag, value);
+         return flag;
+      }
+
+      if (scan_inst->flags_written(devinfo))
+         break;
+   }
+
    ubld1.MOV(flag, value);
 
    return flag;