On 01/17/2017 01:07 AM, Kirill Batuzov wrote:
+ for (i = 0; i < 16; i++) {
+ overlap_temps[i][0] = GET_TCGV_V128(cpu_Q[i]);
+ overlap_temps[i][1] = (TCGArg)-1;
+ sub_temps[i][0] = GET_TCGV_V64(cpu_D[i * 2]);
+ sub_temps[i][1] = GET_TCGV_V64(cpu_D[i * 2 + 1]);
+ sub_temps[i][2] = (TCGArg)-1;
+ tcg_temp_set_overlap_temps(GET_TCGV_V64(cpu_D[i * 2]),
+ overlap_temps[i]);
+ tcg_temp_set_overlap_temps(GET_TCGV_V64(cpu_D[i * 2 + 1]),
+ overlap_temps[i]);
+ tcg_temp_set_sub_temps(GET_TCGV_V128(cpu_Q[i]), sub_temps[i]);
}
Should we simply detect this generically as the registers are declared? This
seems tedious to do for each target.
r~
