On Thu, Aug 31, 2017 at 2:25 PM, Jason Ekstrand <ja...@jlekstrand.net> wrote: > There doesn't seem to be a good reason for using EXECUTE_2 for JMPI and > brw_fs_generator manually stomps it to 1. This is a functional change > because it changes all of the clip and SF programs on gen4-5 to use > EXECUTE_1 for JMPI. Piglit says that this is ok.
The PRM says "The execution size must be 1." > --- > src/intel/compiler/brw_eu_emit.c | 2 +- > src/intel/compiler/brw_fs_generator.cpp | 1 - > 2 files changed, 1 insertion(+), 2 deletions(-) > > diff --git a/src/intel/compiler/brw_eu_emit.c > b/src/intel/compiler/brw_eu_emit.c > index 8c952e7..ada550c 100644 > --- a/src/intel/compiler/brw_eu_emit.c > +++ b/src/intel/compiler/brw_eu_emit.c > @@ -1183,7 +1183,7 @@ brw_JMPI(struct brw_codegen *p, struct brw_reg index, > struct brw_reg ip = brw_ip_reg(); > brw_inst *inst = brw_alu2(p, BRW_OPCODE_JMPI, ip, ip, index); > > - brw_inst_set_exec_size(devinfo, inst, BRW_EXECUTE_2); > + brw_inst_set_exec_size(devinfo, inst, BRW_EXECUTE_1); This is fixing a bug added in 2009 in commit 73137997e23ff6c1145d036315d1a9ad96651281 | insn->header.execution_size = 1; But the execution size encodings are BRW_EXECUTE_1 = 0, BRW_EXECUTE_2 = 1, BRW_EXECUTE_4 = 2, etc. With an updated commit message that contains this information: Reviewed-by: Matt Turner <matts...@gmail.com> _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
