break;
case SHADER_OPCODE_UNTYPED_ATOMIC:
+ case SHADER_OPCODE_UNTYPED_ATOMIC_FLOAT:
case SHADER_OPCODE_TYPED_ATOMIC:
/* Test code:
* mov(8) g112<1>ud 0x00000000ud { align1 WE_all 1Q };
latency = is_haswell ? 300 : 600;
break;
+ case SHADER_OPCODE_SEND:
+ switch (inst->sfid) {
+ default:
+ unreachable("Unknown SFID");
+ }
+ break;
+
default:
/* 2 cycles:
* mul(8) g4<1>F g2<0,1,0>F 0.5F { align1 WE_normal 1Q };
class instruction_scheduler {
public:
instruction_scheduler(backend_shader *s, int grf_count,
- int hw_reg_count, int block_count,
+ unsigned hw_reg_count, int block_count,
instruction_scheduler_mode mode)
{
this->bs = s;
bool post_reg_alloc;
int instructions_to_schedule;
int grf_count;
- int hw_reg_count;
+ unsigned hw_reg_count;
int reg_pressure;
int block_idx;
exec_list instructions;
int payload_last_use_ip[hw_reg_count];
v->calculate_payload_ranges(hw_reg_count, payload_last_use_ip);
- for (int i = 0; i < hw_reg_count; i++) {
+ for (unsigned i = 0; i < hw_reg_count; i++) {
if (payload_last_use_ip[i] == -1)
continue;
* After register allocation, reg_offsets are gone and we track individual
* GRF registers.
*/
- schedule_node *last_grf_write[grf_count * 16];
+ schedule_node **last_grf_write;
schedule_node *last_mrf_write[BRW_MAX_MRF(v->devinfo->gen)];
schedule_node *last_conditional_mod[8] = {};
schedule_node *last_accumulator_write = NULL;
*/
schedule_node *last_fixed_grf_write = NULL;
- memset(last_grf_write, 0, sizeof(last_grf_write));
+ last_grf_write = (schedule_node **)calloc(sizeof(schedule_node *), grf_count * 16);
memset(last_mrf_write, 0, sizeof(last_mrf_write));
/* top-to-bottom dependencies: RAW and WAW. */
}
/* bottom-to-top dependencies: WAR */
- memset(last_grf_write, 0, sizeof(last_grf_write));
+ memset(last_grf_write, 0, sizeof(schedule_node *) * grf_count * 16);
memset(last_mrf_write, 0, sizeof(last_mrf_write));
memset(last_conditional_mod, 0, sizeof(last_conditional_mod));
last_accumulator_write = NULL;
last_accumulator_write = n;
}
}
+
+ free(last_grf_write);
}
void