在 2023/11/14 下午4:34, Xi Ruoyao 写道:
On Tue, 2023-11-14 at 10:26 +0800, chenglulu wrote:
Hi,
* Before calling this template, the function get_memmodel is called to
process memmodel, which has a piece of code:
/* Workaround for Bugzilla 59448. GCC doesn't track consume properly, so
be conservative and promote consume to acquire. */
if (val == MEMMODEL_CONSUME)
val = MEMMODEL_ACQUIRE;
* So I think MEMMODEL_CONSUME don't need to be processed here either.
Otherwise is OK.
Thanks, I've removed case MEMMODEL_CONSUME and there seems no issue.
RISC-V mem_thread_fence expansion also does not handle MEMMODEL_CONSUME.
Pushed r14-5432 with case MEMMODEL_CONSUME removed and comment adjusted,
as attached.
But curiously there are various references to MEMMODEL_CONSUME in
gcc/config:
$ grep -lr MEMMODEL_CONSUME gcc/config
gcc/config/aarch64/aarch64.cc
gcc/config/riscv/riscv.cc
gcc/config/ia64/ia64.cc
gcc/config/ia64/sync.md
gcc/config/gcn/gcn.md
gcc/config/loongarch/loongarch.cc
gcc/config/rs6000/rs6000.cc
gcc/config/rs6000/sync.md
gcc/config/nvptx/nvptx.cc
Maybe all of them are redundant?
I think so.:-)
*
