fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-07 04:58:05 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

broadcom/compiler: MULTOP in branch delay slots doesn't generate RTOP hazard

Browse source 
On unconditional branches qpu_set_branch_targets() can fill the delay slots
with a copy of the first instructions of the successor block.

As the qpu validator is sequential it would detect an incorrect hazard
when the MULTOP was copied but the UMUL24 wasn't.

This was identified in debug build when running gfxbench5.aztec_ruins_vk.

Assisted-by: Claude Opus 4.6
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40923>
This commit is contained in:
Jose Maria Casanova Crespo 2026-04-10 19:27:32 +02:00 • committed by Marge Bot
parent dd6e7c8ef0
commit 31c8e14df3
1 changed files with 11 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

13
src/broadcom/compiler/qpu_validate.c
View file

@ -139,8 +139,17 @@ qpu_validate_inst(struct v3d_qpu_validate_state *state, struct qinst *qinst)
assert(inst->type == V3D_QPU_INSTR_TYPE_ALU);
if (inst->alu.mul.op == V3D_QPU_M_MULTOP)
state->rtop_valid = true;
if (inst->alu.mul.op == V3D_QPU_M_MULTOP) {
/* On unconditional branches qpu_set_branch_targets() can fill the
* delay slots with a copy of the first instructions of the
* successor block. As the qpu validator is sequential it would
* detect a non real hazard when the MULTOP was copied but the
* UMUL24 wasn't. So we disable the hazard detection mechanism in
* this case.
*/
if (!in_branch_delay_slots(state))
state->rtop_valid = true;
}
if (inst->alu.mul.op == V3D_QPU_M_UMUL24) {
if (state->rtop_hazard)