From c3c9a8c85758796a26b48e484286e6b6f5a5299a Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Tue, 22 Oct 2013 12:27:06 -0700 Subject: [PATCH] i965: Try not to reverse-schedule things when doing LIFO scheduling. The LIFO plan was simple: Take the most recently made available instructions, and pick those first. But because of the order we were pushing things onto our list of available-to-schedule instructions, it meant that when a set of instructions was made available at the same time (for example, everything at the start of the program that didn't depend on other instructions) we'd schedule them in reverse order. If you had 10 texture calls in a row in your program, each with independent argument setup, we'd set up the last texture call's args and execute it first, even though we wouldn't be able to consume its results until we'd finished the other 9 texture calls (assuming consumption of texture results happens near each texture call, and combines it with another texture result, which is normal for a convolution shader). To fix this, walk the list for doing LIFO in the order that instructions were originally generated in the program, but choose to push newly-made-available instructions to the other end of the list instead. total instructions in shared programs: 1587242 -> 1586290 (-0.06%) instructions in affected programs: 7801 -> 6849 (-12.20%) GAINED: 76 LOST: 67 Thanks to Chia-I Wu for pointing out the bug in my first version of the patch that made it a huge loss. Reviewed-by: Matt Turner --- src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp b/src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp index 84b74ff2946..99538bddb47 100644 --- a/src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp +++ b/src/mesa/drivers/dri/i965/brw_schedule_instructions.cpp @@ -964,9 +964,7 @@ fs_instruction_scheduler::choose_instruction_to_schedule() * but also the MRF setup for the next sampler message, which in turn * unblocks the next sampler message). */ - for (schedule_node *node = (schedule_node *)instructions.get_tail(); - node != instructions.get_head()->prev; - node = (schedule_node *)node->prev) { + foreach_list(node, &instructions) { schedule_node *n = (schedule_node *)node; fs_inst *inst = (fs_inst *)n->inst; @@ -1059,7 +1057,7 @@ instruction_scheduler::schedule_instructions(backend_instruction *next_block_hea * be scheduled. Update the children's unblocked time for this * DAG edge as we do so. */ - for (int i = 0; i < chosen->child_count; i++) { + for (int i = chosen->child_count - 1; i >= 0; i--) { schedule_node *child = chosen->children[i]; child->unblocked_time = MAX2(child->unblocked_time, @@ -1075,7 +1073,7 @@ instruction_scheduler::schedule_instructions(backend_instruction *next_block_hea if (debug) { printf("\t\tnow available\n"); } - instructions.push_tail(child); + instructions.push_head(child); } } -- 2.30.2