Hi!
On Wed, Aug 11, 2021 at 11:02:25AM -0500, Pat Haugen wrote:
> * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add new flag.
> (POWERPC_MASKS): Likewise.
"Add OPTION_MASK_P10_FUSION_2STORE", instead.
> +static bool
> +is_fusable_store (rtx_insn *insn, rtx *str_mem)
> +{
> + /* Insn must be a non-prefixed base+disp form store. */
> + if (is_store_insn (insn, str_mem)
> + && get_attr_prefixed (insn) == PREFIXED_NO
> + && get_attr_update (insn) == UPDATE_NO
> + && get_attr_indexed (insn) == INDEXED_NO)
> + {
> + /* Further restictions by mode and size. */
(typo, "restrictions")
> + machine_mode mode = GET_MODE (*str_mem);
> + HOST_WIDE_INT size;
> + if (MEM_SIZE_KNOWN_P (*str_mem))
> + size = MEM_SIZE (*str_mem);
> + else
> + return false;
if (!MEM_SIZE_KNOWN_P (*str_mem))
return false;
machine_mode mode = GET_MODE (*str_mem);
HOST_WIDE_INT size = MEM_SIZE (*str_mem);
> +static int
> +power10_sched_reorder (rtx_insn **ready, int lastpos)
> +{
> + rtx mem1;
> +
> + /* Do store fusion during sched2 only. */
> + if (!reload_completed)
> + return cached_can_issue_more;
Should that just be "return false"?
> + {
> + int pos;
> + rtx mem2;
> +
> + /* A fusable store was just scheduled. Scan the ready list for another
> + store that it can fuse with. */
> + pos = lastpos;
Declare the var not before here?
> + while (pos >= 0)
> + {
And mem2 not before in this block?
> +/* { dg-final { scan-assembler-times {stw 4,4\(3\)\n\tstw 6,8\(3\)} 1 {
> target lp64 } } } */
> +/* { dg-final { scan-assembler-times {stw 4,4\(3\)\n\tstw 6,8\(3\)} 2 {
> target ilp32 } } } */
> +/* { dg-final { scan-assembler-times {std 4,8\(3\)\n\tstd 6,16\(3\)} 1 {
> target lp64 } } } */
> +/* { dg-final { scan-assembler-times {stfd 1,8\(3\)\n\tstfd 3,16\(3\)} 1 } }
> */
This is probably okay because p10 is not supported on any config that
uses register names.
> +/* { dg-final { scan-assembler-not {stw 4,4\(3\)\n\tstw 6,8\(3\)} } } */
> +/* { dg-final { scan-assembler-not {std 4,8\(3\)\n\tstd 6,16\(3\)} { target
> lp64 } } } */
> +/* { dg-final { scan-assembler-not {stfd 1,8\(3\)\n\tstfd 3,16\(3\)} } } */
Heh. A little fragile, the compiler could reorder the stores for other
reasons, but the best we can do here I guess.
Okay for trunk with the trivial cleanups. Thanks!
Segher
