On Thursday, July 6, 2017 4:48:23 PM PDT Matt Turner wrote:
> The implementation of ballotARB() will start by zeroing the flags
> register. So, a doing something like
>
> if (gl_SubGroupInvocationARB % 2u == 0u) {
> ... = ballotARB(true);
> [...]
> } else {
> ... = ballotARB(true);
> [...]
> }
>
> (like fs-ballot-if-else.shader_test does) would generate identical MOVs
> to the same destination (the flag register!), and we definitely do not
> want to pull that out of the control flow.
> ---
> src/intel/compiler/brw_fs_sel_peephole.cpp | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/src/intel/compiler/brw_fs_sel_peephole.cpp
> b/src/intel/compiler/brw_fs_sel_peephole.cpp
> index 8cd897f72e..fd02792beb 100644
> --- a/src/intel/compiler/brw_fs_sel_peephole.cpp
> +++ b/src/intel/compiler/brw_fs_sel_peephole.cpp
> @@ -68,7 +68,8 @@ count_movs_from_if(fs_inst *then_mov[MAX_MOVS], fs_inst
> *else_mov[MAX_MOVS],
> {
> int then_movs = 0;
> foreach_inst_in_block(fs_inst, inst, then_block) {
> - if (then_movs == MAX_MOVS || inst->opcode != BRW_OPCODE_MOV)
> + if (then_movs == MAX_MOVS || inst->opcode != BRW_OPCODE_MOV ||
> + inst->flags_written())
> break;
>
> then_mov[then_movs] = inst;
> @@ -77,7 +78,8 @@ count_movs_from_if(fs_inst *then_mov[MAX_MOVS], fs_inst
> *else_mov[MAX_MOVS],
>
> int else_movs = 0;
> foreach_inst_in_block(fs_inst, inst, else_block) {
> - if (else_movs == MAX_MOVS || inst->opcode != BRW_OPCODE_MOV)
> + if (else_movs == MAX_MOVS || inst->opcode != BRW_OPCODE_MOV ||
> + inst->flags_written())
> break;
>
> else_mov[else_movs] = inst;
> Reviewed-by: Kenneth Graunke <[email protected]>
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ mesa-dev mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/mesa-dev
