Skip to content

Support vector and more extended registers in perf

Dapeng Mi linux-perf-users lkml 1 version In review

Get series with b4
b4 am [email protected]

Versions newest first

  1. v4 Latest17 patches 4 repliesIn review

Changes in v4 from the cover letter

- Drop the SIMD registers if an NMI hits kernel mode for REGS_USER.
- Only dump the available regs, rather than zero and dump the unavailable regs. It's possible that the dumped registers are a subset of the requested registers.
- Some minor updates to address Dapeng's comments in V3.

Patches in v4 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/17perf/x86: Use x86_perf_regs in the x86 nmi handlerarch/x86/events/core.c+3 -10 replies
2/17perf/x86: Setup the regs dataarch/x86/events/core.c arch/x86/events/intel/ds.c arch/x86/events/perf_event.h+39 -10 replies
3/17x86/fpu/xstate: Add xsaves_nmiarch/x86/include/asm/fpu/xstate.h arch/x86/kernel/fpu/xstate.c+31 -04 replies
4/17perf: Move has_extended_regs() to header fileinclude/linux/perf_event.h kernel/events/core.c+9 -70 replies
5/17perf/x86: Support XMM register for non-PEBS and REGS_USERarch/x86/events/core.c arch/x86/events/intel/core.c arch/x86/events/intel/ds.c +4 more+172 -250 replies
6/17perf: Support SIMD registersinclude/linux/perf_event.h include/linux/perf_regs.h include/uapi/linux/perf_event.h +1 more+175 -70 replies
7/17perf/x86: Move XMM to sample_simd_vec_regsarch/x86/events/core.c arch/x86/events/intel/ds.c arch/x86/events/perf_event.h +3 more+195 -100 replies
8/17perf/x86: Add YMM into sample_simd_vec_regsarch/x86/events/core.c arch/x86/events/perf_event.h arch/x86/include/asm/perf_event.h +2 more+41 -30 replies
9/17perf/x86: Add ZMM into sample_simd_vec_regsarch/x86/events/core.c arch/x86/events/perf_event.h arch/x86/include/asm/perf_event.h +2 more+75 -30 replies
10/17perf/x86: Add OPMASK into sample_simd_pred_regarch/x86/events/core.c arch/x86/events/perf_event.h arch/x86/include/asm/perf_event.h +2 more+44 -40 replies
11/17perf/x86: Add eGPRs into sample_regsarch/x86/events/core.c arch/x86/events/perf_event.h arch/x86/include/asm/perf_event.h +2 more+83 -290 replies
12/17perf/x86: Add SSP into sample_regsarch/x86/events/core.c arch/x86/events/perf_event.h arch/x86/include/asm/perf_event.h +2 more+36 -40 replies
13/17perf/x86/intel: Enable PERF_PMU_CAP_SIMD_REGSarch/x86/events/intel/core.c+46 -20 replies
14/17perf tools: Only support legacy regs for the PT and PERF_REGS_MASKtools/perf/arch/x86/include/perf_regs.h tools/perf/util/intel-pt.c+2 -20 replies
15/17perf tools: headers: Sync with the kernel sourcestools/arch/x86/include/uapi/asm/perf_regs.h tools/include/uapi/linux/perf_event.h+105 -50 replies
16/17perf tools: parse-regs: Support the new SIMD formattools/perf/arch/x86/util/perf_regs.c tools/perf/util/evsel.c tools/perf/util/parse-regs-options.c +4 more+663 -140 replies
17/17perf tools: regs: Support to dump regs for PERF_SAMPLE_REGS_ABI_SIMDtools/perf/util/evsel.c tools/perf/util/perf-regs-arch/perf_regs_x86.c tools/perf/util/sample.h +1 more+143 -80 replies

Reviews trailers per patch

1 of 17 patches has a Reviewed-by or Acked-by.

PatchAdrian Hunter
1/17
2/17
3/17
4/17
5/17
6/17
7/17
8/17
9/17
10/17
11/17
12/17
13/17
14/17A-b
15/17
16/17
17/17

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.

Read the v4 thread

Cover letter [Patch v4 00/17] Support vector and more extended registers in perf

Changes since V3:
- Drop the SIMD registers if an NMI hits kernel mode for REGS_USER.
- Only dump the available regs, rather than zero and dump the unavailable regs. It's possible that the dumped registers are a subset of the requested registers.
- Some minor updates to address Dapeng's comments in V3.

Changes since V2:
- Use the FPU format for the x86_pmu.ext_regs_mask as well
- Add a check before invoking xsaves_nmi()
- Add perf_simd_reg_check() to retrieve the number of available registers. If the kernel fails to get the requested registers, e.g., XSAVES fails, nothing dumps to the userspace (the V2 dumps all 0s).
- Add POC perf tool patches

Changes since V1:
- Apply the new interfaces to configure and dump the SIMD registers
- Utilize the existing FPU functions, e.g., xstate_calculate_size, get_xsave_addr().

Starting from the Intel Ice Lake, the XMM registers can be collected in a PEBS record. More registers, e.g., YMM, ZMM, OPMASK, SPP and APX, will be added in the upcoming Architecture PEBS as well. But it requires the hardware support.

The patch set provides a software solution to mitigate the hardware requirement. It utilizes the XSAVES command to retrieve the requested registers in the overflow handler. The feature isn't limited to the PEBS event or specific platforms anymore.
The hardware solution (if available) is still preferred, since it has low overhead (especially with the large PEBS) and is more accurate.

In theory, the solution should work for all X86 platforms. But I only have newer Inter platforms to test. The patch set only enable the feature for Intel Ice Lake and later platforms.

The new registers include YMM, ZMM, OPMASK, SSP, and APX. The sample_regs_user/intr has run out. A new field in the struct perf_event_attr is required for the registers.

After a long discussion in V1,
https://lore.kernel.org/lkml/[email protected]/

The new field looks like as below.

@@ -543,6 +545,25 @@ struct perf_event_attr {
        __u64   sig_data;

        __u64   config3; /* extension of config2 */
+
+
+       /*
+        * Defines set of SIMD registers to dump on samples.
+        * The sample_simd_regs_enabled !=0 implies the
+        * set of SIMD registers is used to config all SIMD registers.
+        * If !sample_simd_regs_enabled, sample_regs_XXX may be used to
+        * config some SIMD registers on X86.
+        */
+       union {
+               __u16 sample_simd_regs_enabled;
+               __u16 sample_simd_pred_reg_qwords;
+       };
+       __u32 sample_simd_pred_reg_intr;
+       __u32 sample_simd_pred_reg_user;
+       __u16 sample_simd_vec_reg_qwords;
+       __u64 sample_simd_vec_reg_intr;
+       __u64 sample_simd_vec_reg_user;
+       __u32 __reserved_4;

};
@@ -1016,7 +1037,15 @@ enum perf_event_type {
* } && PERF_SAMPLE_BRANCH_STACK
*
* { u64 abi; # enum perf_sample_regs_abi

-        *        u64                   regs[weight(mask)]; } && PERF_SAMPLE_REGS_USER
+        *        u64                   regs[weight(mask)];
+        *        struct {
+        *              u16 nr_vectors;
+        *              u16 vector_qwords;
+        *              u16 nr_pred;
+        *              u16 pred_qwords;
+        *              u64 data[nr_vectors * vector_qwords + nr_pred * pred_qwords];
+        *        } && (abi & PERF_SAMPLE_REGS_ABI_SIMD)
+        *      } && PERF_SAMPLE_REGS_USER

*
* { u64 size;
* char data[size];
@@ -1043,7 +1072,15 @@ enum perf_event_type {

         *      { u64                   data_src; } && PERF_SAMPLE_DATA_SRC
         *      { u64                   transaction; } && PERF_SAMPLE_TRANSACTION
         *      { u64                   abi; # enum perf_sample_regs_abi
-        *        u64                   regs[weight(mask)]; } && PERF_SAMPLE_REGS_INTR
+        *        u64                   regs[weight(mask)];
+        *        struct {
+        *              u16 nr_vectors;
+        *              u16 vector_qwords;
+        *              u16 nr_pred;
+        *              u16 pred_qwords;
+        *              u64 data[nr_vectors * vector_qwords + nr_pred * pred_qwords];
+        *        } && (abi & PERF_SAMPLE_REGS_ABI_SIMD)
+        *      } && PERF_SAMPLE_REGS_INTR
         *      { u64                   phys_addr;} && PERF_SAMPLE_PHYS_ADDR
         *      { u64                   cgroup;} && PERF_SAMPLE_CGROUP
         *      { u64                   data_page_size;} && PERF_SAMPLE_DATA_PAGE_SIZE

Since there is only one vector qwords field, the qwords for the newest vector should be set by the tools. For example, if the end user wants XMM0 and YMM1, the vector qwords should be 4. The vector mask should be 0x3. The YMM0 and YMM1 will be dumped to the userspace. It's the tool's responsibility to output the XMM0 and YMM1 to the end user.

The POC perf tool patches for testing purposes is also attached.

Examples:
$perf record -I?
available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10 R11 R12 R13 R14 R15 SSP XMM0-31 YMM0-31 ZMM0-31 OPMASK0-7

$perf record --user-regs=?
available registers: AX BX CX DX SI DI BP SP IP FLAGS CS SS R8 R9 R10 R11 R12 R13 R14 R15 SSP XMM0-31 YMM0-31 ZMM0-31 OPMASK0-7

$perf record -e cycles:p -IXMM,YMM,OPMASK,SSP ./test $perf report -D
... ...
237538985992962 0x454d0 [0x480]: PERF_RECORD_SAMPLE(IP, 0x1):
179370/179370: 0xffffffff969627fc period: 124999 addr: 0 ... intr regs: mask 0x20000000000 ABI 64-bit
.... SSP 0x0000000000000000
... SIMD ABI nr_vectors 32 vector_qwords 4 nr_pred 8 pred_qwords 1

 .... YMM  [0] 0x0000000000004000
 .... YMM  [0] 0x000055e828695270
 .... YMM  [0] 0x0000000000000000
 .... YMM  [0] 0x0000000000000000
 .... YMM  [1] 0x000055e8286990e0
 .... YMM  [1] 0x000055e828698dd0
 .... YMM  [1] 0x0000000000000000
 .... YMM  [1] 0x0000000000000000
 ... ...
 .... YMM  [31] 0x0000000000000000
 .... YMM  [31] 0x0000000000000000
 .... YMM  [31] 0x0000000000000000
 .... YMM  [31] 0x0000000000000000

.... OPMASK[0] 0x0000000000100221
.... OPMASK[1] 0x0000000000000020
.... OPMASK[2] 0x000000007fffffff
.... OPMASK[3] 0x0000000000000000
.... OPMASK[4] 0x0000000000000000
.... OPMASK[5] 0x0000000000000000
.... OPMASK[6] 0x0000000000000000
.... OPMASK[7] 0x0000000000000000
... ...

History:
v3: https://lore.kernel.org/lkml/[email protected]/
v2: https://lore.kernel.org/lkml/[email protected]/
v1: https://lore.kernel.org/lkml/[email protected]/

Kan Liang (17):
perf/x86: Use x86_perf_regs in the x86 nmi handler perf/x86: Setup the regs data
x86/fpu/xstate: Add xsaves_nmi
perf: Move has_extended_regs() to header file
perf/x86: Support XMM register for non-PEBS and REGS_USER
perf: Support SIMD registers
perf/x86: Move XMM to sample_simd_vec_regs
perf/x86: Add YMM into sample_simd_vec_regs
perf/x86: Add ZMM into sample_simd_vec_regs
perf/x86: Add OPMASK into sample_simd_pred_reg
perf/x86: Add eGPRs into sample_regs
perf/x86: Add SSP into sample_regs
perf/x86/intel: Enable PERF_PMU_CAP_SIMD_REGS
perf tools: Only support legacy regs for the PT and PERF_REGS_MASK perf tools: headers: Sync with the kernel sources
perf tools: parse-regs: Support the new SIMD format
perf tools: regs: Support to dump regs for PERF_SAMPLE_REGS_ABI_SIMD

 arch/x86/events/core.c                        | 315 ++++++++++++-
 arch/x86/events/intel/core.c                  |  75 ++-
 arch/x86/events/intel/ds.c                    |  12 +-
 arch/x86/events/perf_event.h                  |  80 ++++
 arch/x86/include/asm/fpu/xstate.h             |   3 +
 arch/x86/include/asm/perf_event.h             |  30 +-
 arch/x86/include/uapi/asm/perf_regs.h         |  65 ++-
 arch/x86/kernel/fpu/xstate.c                  |  32 +-
 arch/x86/kernel/perf_regs.c                   | 139 +++++-
 include/linux/perf_event.h                    |  16 +
 include/linux/perf_regs.h                     |  26 +
 include/uapi/linux/perf_event.h               |  45 +-
 kernel/events/core.c                          | 111 ++++-
 tools/arch/x86/include/uapi/asm/perf_regs.h   |  65 ++-
 tools/include/uapi/linux/perf_event.h         |  45 +-
 tools/perf/arch/x86/include/perf_regs.h       |   2 +-
 tools/perf/arch/x86/util/perf_regs.c          | 443 +++++++++++++++++-
 tools/perf/util/evsel.c                       |  45 ++
 tools/perf/util/intel-pt.c                    |   2 +-
 tools/perf/util/parse-regs-options.c          | 133 +++++-
 .../perf/util/perf-regs-arch/perf_regs_x86.c  |  43 ++
 tools/perf/util/perf_event_attr_fprintf.c     |   6 +
 tools/perf/util/perf_regs.c                   |  54 +++
 tools/perf/util/perf_regs.h                   |  10 +
 tools/perf/util/record.h                      |   6 +
 tools/perf/util/sample.h                      |  10 +
 tools/perf/util/session.c                     |  78 ++-
 27 files changed, 1814 insertions(+), 77 deletions(-)

base-commit: 6d48436560e91be858158e227f21aab71698814e

Signature · 1 line
2.34.1