qemu-arm
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH v2 2/5] target/arm: Ensure icount is enabled when emulating I


From: Philippe Mathieu-Daudé
Subject: Re: [PATCH v2 2/5] target/arm: Ensure icount is enabled when emulating INST_RETIRED
Date: Fri, 8 Dec 2023 12:23:41 +0100
User-agent: Mozilla Thunderbird

Hi Peter,

On 8/12/23 11:59, Peter Maydell wrote:
On Fri, 8 Dec 2023 at 10:36, Philippe Mathieu-Daudé <philmd@linaro.org> wrote:

On 7/12/23 23:12, Richard Henderson wrote:
On 12/7/23 07:45, Philippe Mathieu-Daudé wrote:
pmu_init() register its event checking the pm_event::supported()
handler. For INST_RETIRED, the event is only registered and the
bit enabled in the PMU Common Event Identification register when
icount is enabled as ICOUNT_PRECISE.

Assert the pm_event::get_count() and pm_event::ns_per_count()
handler will only be called under this icount mode.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
   target/arm/helper.c | 2 ++
   1 file changed, 2 insertions(+)

diff --git a/target/arm/helper.c b/target/arm/helper.c
index adb0960bba..333fd5f4bf 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -940,11 +940,13 @@ static bool instructions_supported(CPUARMState
*env)
   static uint64_t instructions_get_count(CPUARMState *env)
   {
+    assert(icount_enabled() == ICOUNT_PRECISE);
       return (uint64_t)icount_get_raw();
   }
   static int64_t instructions_ns_per(uint64_t icount)
   {
+    assert(icount_enabled() == ICOUNT_PRECISE);
       return icount_to_ns((int64_t)icount);
   }
   #endif

I don't think an assert is required -- that's exactly what the
.supported field is for. If you think this needs additional
clarification, a comment is sufficient.

Without this I'm getting this link failure with TCG disabled:

ld: Undefined symbols:
    _icount_to_ns, referenced from:
        _instructions_ns_per in target_arm_helper.c.o
clang: error: linker command failed with exit code 1 (use -v to see
invocation)

I think we should fix this earlier by not trying to enable
these TCG-only PMU event types in a non-TCG config.

I agree... but (as discussed yesterday on IRC), this is a bigger rework.

This icount cleanup blocks 60+ patches on top :/ Since I have a v3
addressing Richard's comments already done, I'll post it, mentioning the
PMU issue in the cover; then see if cleaning it isn't too invasive.
If I end in another rabbit hole, I'll suggest to accept this current
patch and clean the technical debt later, again.

Thanks,

Phil.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]