Re: [tip:perfcounters/core] perf_counter, x86: Implementgeneralized cache event types, add AMD support

[Ingo Molnar]

* Peter Zijlstra <a.p.zijlstra@xxxxxxxxx> wrote:

> On Mon, 2009-06-08 at 20:36 +0000, tip-bot for Thomas Gleixner wrote:
> > Commit-ID:  d3ad70660a0bf6d1a3692093939d238fe8add498
> > Gitweb:     http://git.kernel.org/tip/d3ad70660a0bf6d1a3692093939d238fe8add498
> > Author:     Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> > AuthorDate: Mon, 8 Jun 2009 22:33:10 +0200
> > Committer:  Ingo Molnar <mingo@xxxxxxx>
> > CommitDate: Mon, 8 Jun 2009 22:33:10 +0200
> > 
> > perf_counter, x86: Implement generalized cache event types, add AMD support
> > 
> > Fill in amd_hw_cache_event_id[] with the AMD CPU specific events,
> > for family 0x0f, 0x10 and 0x11.
> > 
> > There's apparently no distinction between load and store events, so
> > we only fill in the load events.
> 
> Ah, that's what you did.. :-)
> 
> > +static const u64 amd_0f_hw_cache_event_ids
> > +				[PERF_COUNT_HW_CACHE_MAX]
> > +				[PERF_COUNT_HW_CACHE_OP_MAX]
> > +				[PERF_COUNT_HW_CACHE_RESULT_MAX] =
> > +{
> > + [ C(L1D) ] = {
> > +	[ C(OP_READ) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0,
> > +		[ C(RESULT_MISS)   ] = 0,
> 
> 0x40 - Data Cache Access
> 0x41 - Data Cache Misses
> 
> > +	},
> > +	[ C(OP_WRITE) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0,
> > +		[ C(RESULT_MISS)   ] = 0,
> > +	},
> > +	[ C(OP_PREFETCH) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0,
> > +		[ C(RESULT_MISS)   ] = 0,
> 
> 0x4B - unit 0x3 (load+store) Prefetch Dispatched
> 
> > +	},
> > + },
> > + [ C(L1I ) ] = {
> > +	[ C(OP_READ) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0x0080, /* Instruction cache fetches  */
> > +		[ C(RESULT_MISS)   ] = 0x0081, /* Instruction cache misses   */
> > +	},
> > +	[ C(OP_WRITE) ] = {
> > +		[ C(RESULT_ACCESS) ] = -1,
> > +		[ C(RESULT_MISS)   ] = -1,
> > +	},
> > +	[ C(OP_PREFETCH) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0,
> > +		[ C(RESULT_MISS)   ] = 0,
> > +	},
> 
> 0x81 L1I Miss
> 0x82 L1I Miss, L2 Hit
> 0x83 L1I Miss, L2 Miss
> 
> > + },
> > + [ C(L2  ) ] = {
> > +	[ C(OP_READ) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0,
> > +		[ C(RESULT_MISS)   ] = 0,
> > +	},
> > +	[ C(OP_WRITE) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0,
> > +		[ C(RESULT_MISS)   ] = 0,
> > +	},
> > +	[ C(OP_PREFETCH) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0,
> > +		[ C(RESULT_MISS)   ] = 0,
> > +	},
> 
> 0x42 unit 0x1e (shared|exclusive|owned|mod) L1 Miss L2 Hit
> 0x43 unit 0x1e                              L2 Miss
> 
> 0x7d unit 0x3 (IC|DC) L2 Hit
> 0x7e unit 0x3 (IC|DC) L2 Miss
> 
> > + },
> > + [ C(DTLB) ] = {
> > +	[ C(OP_READ) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0,
> > +		[ C(RESULT_MISS)   ] = 0,
> > +	},
> 
> 0x4D - unit 0x1 4k DTLB L1 Hit
> 0x45 - unit 0x1 4k DTLB L1 Miss L2 Hit
> 0x46 - unit 0x1 4k DTLB L1 Miss L2 Miss
> 
> > +	[ C(OP_WRITE) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0,
> > +		[ C(RESULT_MISS)   ] = 0,
> > +	},
> > +	[ C(OP_PREFETCH) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0,
> > +		[ C(RESULT_MISS)   ] = 0,
> > +	},
> > + },
> > + [ C(ITLB) ] = {
> > +	[ C(OP_READ) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0x0080, /* Instruction fecthes        */
> > +		[ C(RESULT_MISS)   ] = 0x0085, /* Instr. fetch ITLB misses   */
> > +	},
> > +	[ C(OP_WRITE) ] = {
> > +		[ C(RESULT_ACCESS) ] = -1,
> > +		[ C(RESULT_MISS)   ] = -1,
> > +	},
> > +	[ C(OP_PREFETCH) ] = {
> > +		[ C(RESULT_ACCESS) ] = -1,
> > +		[ C(RESULT_MISS)   ] = -1,
> > +	},
> 
> 0x84               L1 ITLB Miss, L2 ITLB Hit
> 0x85 Unit 0x1 (4k) L1 ITLB Miss, L2 Miss
> 
> > + },
> > + [ C(BPU ) ] = {
> > +	[ C(OP_READ) ] = {
> > +		[ C(RESULT_ACCESS) ] = 0x00c2, /* Retired Branch Instr.      */
> > +		[ C(RESULT_MISS)   ] = 0x00c3, /* Retired Mispredicted BI    */
> 
> There's also Retired Taken in C4 and C5

Mind turning this into a patch? You did the hard work already :-)

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/