Closed davidharrishmc closed 3 months ago
Note that Sail does not yet support cbo.zero, so Sail throws an illegal instruction exception when attempting to simulate it during riscof compilation. Consequently I have been unable to fully test this. However, it should fix the problem Tim Hutt raised about the commit.
@jamesbeyond please have a look at it, this one is good to go, so you may merge this one
Regarding " It appears that the cbo tests don't use CSRs and that the Zicsr would likely be unnecessary. " While CBO tsts don't explicitly use CSRs, if there is any possibility they would trap, they can implicitly use them (at least, for arch-tests), So Zicsr can be removed if there is no possibility of a trap.
On Wed, Jun 12, 2024 at 3:34 AM David Harris @.***> wrote:
@.**** commented on this pull request.
On riscv-test-suite/rv64i_m/CMO/src/cbo.zero-01.S https://github.com/riscv-non-isa/riscv-arch-test/pull/473#discussion_r1636222460 :
I added Zicsr to the RV64 cbo.zero RVTEST_ISA string for consistency with the other tests per @UmerShahidengr https://github.com/UmerShahidengr 's comment.
I'm not the author of cbo. It appears that the cbo tests don't use CSRs and that the Zicsr would likely be unnecessary. I believe somebody from my team had experimented with that months ago and presumably removed the Zicsr in just this case, but I agree that it is better to be consistent. It's hard to envision a system with caches but no CSRs, so Zicsr is likely harmless, and if the tests are ever generalized to non-machine mode, Zicsr will be nessary to support menvcfg and the other modes.
@liweiwei90 https://github.com/liweiwei90 was the original author of PR
226 https://github.com/riscv-non-isa/riscv-arch-test/pull/226 with
cbo.zero tests. I will defer to the author if they decide to remove Zicsr from cbo.zero at a later point. If so, I think it should come out of RVTEST_ISA and RVTEST_CASE for both rv32i_m and rv64i_m.
— Reply to this email directly, view it on GitHub https://github.com/riscv-non-isa/riscv-arch-test/pull/473#pullrequestreview-2112621829, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHPXVJWSMQ4MK7YZGX6ONA3ZHAP5DAVCNFSM6AAAAABJE7T3W6VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDCMJSGYZDCOBSHE . You are receiving this because you are subscribed to this thread.Message ID: @.***>