On 5/27/24 11:52 AM, Hans-Peter Nilsson wrote:
Regtested cris-elf. Ok to commit?
-- >8 --
...and call compute_bb_for_insn in init_resource_info and
free_bb_for_insn in free_resource_info.
I put a gcc_unreachable in that else-clause for a failing
find_basic_block in mark_target_live_regs after the comment that says:
/* We didn't find the start of a basic block. Assume everything
in use. This should happen only extremely rarely. */
SET_HARD_REG_SET (res->regs);
and found that it fails not extremely rarely but extremely early in
the build (compiling libgcc).
That kind of pessimization leads to suboptimal delay-slot-filling.
Instead, do like many machine_dependent_reorg passes and call
compute_bb_for_insn as part of resource.cc initialization.
After this patch, there's a whole "if (b != -1)" conditional that's
dominated by a gcc_assert (b != -1). I separated that, as it's a NFC
whitespace patch that hampers patch readability.
Altogether this improved coremark performance for CRIS at -O2
-march=v10 by 0.36%.
* resource.cc: Include cfgrtl.h. Use BLOCK_FOR_INSN (insn)->index
instead of calling find_basic_block (insn). Assert for not -1.
(find_basic_block): Remove function.
(init_resource_info): Call compute_bb_for_insn.
(free_resource_info): Call free_bb_for_insn.
I'm pretty sure that code as part of the overall problem -- namely that
we didn't have good basic block info so we resorted to insn scanning.
Presumably we set BLOCK_FOR_INSN when we generate a wrapper SEQUENCE
insns for a filled delay slot? Assuming we do create the right mapping
for those new insns, then this is OK.
jeff
