Skip to contentSkip to conversation

eal/ppc64: improve rte_rdtsc with ppc_get_timebase

PATCH Superseded Thinh Tran 3 participants 9 messages Jan 28, 2020 – Mar 11, 2020

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Thinh Tran

author
#0 lore raw

PATCH 1/1 eal/ppc64: improve rte_rdtsc with ppc_get_timebase

__ppc_get_timebase() is GNU extention and is more efficient

---
 .../common/include/arch/ppc_64/rte_cycles.h   | 28 ++-----------------
 1 file changed, 2 insertions(+), 26 deletions(-)
lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h +2 -26
index 8f2e98642..871f9b6e4 100644
@@ -14,6 +14,7 @@ extern "C" {
 
 #include <rte_byteorder.h>
 #include <rte_common.h>
+#include <sys/platform/ppc.h>
 
 /**
  * Read the time base register.
@@ -24,32 +25,7 @@ extern "C" {
 static inline uint64_t
 rte_rdtsc(void)
 {
-	union {
-		uint64_t tsc_64;
-		RTE_STD_C11
-		struct {
-#if RTE_BYTE_ORDER == RTE_BIG_ENDIAN
-			uint32_t hi_32;
-			uint32_t lo_32;
-#else
-			uint32_t lo_32;
-			uint32_t hi_32;
-#endif
-		};
-	} tsc;
-	uint32_t tmp;
-
-	asm volatile(
-			"0:\n"
-			"mftbu   %[hi32]\n"
-			"mftb    %[lo32]\n"
-			"mftbu   %[tmp]\n"
-			"cmpw    %[tmp],%[hi32]\n"
-			"bne     0b\n"
-			: [hi32] "=r"(tsc.hi_32), [lo32] "=r"(tsc.lo_32),
-			[tmp] "=r"(tmp)
-		    );
-	return tsc.tsc_64;
+	return __ppc_get_timebase();
 }
 
 static inline uint64_t
Signature · 1 line
2.17.1
author
#2 lore raw
In reply to Thinh Tran

Need to resubmit a new patch, since the same code may be used by other ppc platform such as FreeBSD that may not have the sys/platform/ppc.h Thinh Tran

Quoted 54 lines · Thinh Tran

__ppc_get_timebase() is GNU extention and is more efficient

---
  .../common/include/arch/ppc_64/rte_cycles.h   | 28 ++-----------------
  1 file changed, 2 insertions(+), 26 deletions(-)
lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h +2 -26
index 8f2e98642..871f9b6e4 100644
@@ -14,6 +14,7 @@ extern "C" {
 
  #include <rte_byteorder.h>
  #include <rte_common.h>
+#include <sys/platform/ppc.h>
 
  /**
   * Read the time base register.
@@ -24,32 +25,7 @@ extern "C" {
  static inline uint64_t
  rte_rdtsc(void)
  {
-	union {
-		uint64_t tsc_64;
-		RTE_STD_C11
-		struct {
-#if RTE_BYTE_ORDER == RTE_BIG_ENDIAN
-			uint32_t hi_32;
-			uint32_t lo_32;
-#else
-			uint32_t lo_32;
-			uint32_t hi_32;
-#endif
-		};
-	} tsc;
-	uint32_t tmp;
-
-	asm volatile(
-			"0:\n"
-			"mftbu   %[hi32]\n"
-			"mftb    %[lo32]\n"
-			"mftbu   %[tmp]\n"
-			"cmpw    %[tmp],%[hi32]\n"
-			"bne     0b\n"
-			: [hi32] "=r"(tsc.hi_32), [lo32] "=r"(tsc.lo_32),
-			[tmp] "=r"(tmp)
-		    );
-	return tsc.tsc_64;
+	return __ppc_get_timebase();
  }
 
  static inline uint64_t
author
#3 lore raw

PATCH v2 1/1 eal/ppc64: improve rte_rdtsc with ppc_get_timebase

In reply to Thinh Tran

__ppc_get_timebase() is GNU extension and is more efficient

v2: Advoid breaking other ppc_64 flatforms. The __ppc_get_timebase()
seems to be specific to powerpc platform and with GLIBC.

---
 lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h | 7 +++++++
 1 file changed, 7 insertions(+)
lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h +7 -0
index 8f2e98642..1c3fd556e 100644
@@ -14,6 +14,9 @@ extern "C" {
 
 #include <rte_byteorder.h>
 #include <rte_common.h>
+#if defined(__powerpc__) && defined(__GLIBC__)
+#include <sys/platform/ppc.h>
+#endif
 
 /**
  * Read the time base register.
@@ -24,6 +27,9 @@ extern "C" {
 static inline uint64_t
 rte_rdtsc(void)
 {
+#if defined(__powerpc__) && defined(__GLIBC__)
+	return __ppc_get_timebase();
+#else
 	union {
 		uint64_t tsc_64;
 		RTE_STD_C11
@@ -50,6 +56,7 @@ rte_rdtsc(void)
 			[tmp] "=r"(tmp)
 		    );
 	return tsc.tsc_64;
+#endif /* __powerpc__ && __GLIBC__ */
 }
 
 static inline uint64_t
Signature · 1 line
2.17.1
#5 lore raw
In reply to Thinh Tran
Quoted 2 lines · Thinh Tran

__ppc_get_timebase() is GNU extension and is more efficient

The commit title and log are quite short and give little idea on what this is about.

I had a look at this glibc helper:

/* Read the Time Base Register. */
static __inline__ uint64_t
__ppc_get_timebase (void)
{
#if __GNUC_PREREQ (4, 8)
return __builtin_ppc_get_timebase ();
#else

# ifdef __powerpc64__
  uint64_t __tb;
  /* "volatile" is necessary here, because the user expects this assembly
     isn't moved after an optimization.  */
  __asm__ volatile ("mfspr %0, 268" : "=r" (__tb));
  return __tb;
# else  /* not __powerpc64__ */
  uint32_t __tbu, __tbl, __tmp; \
  __asm__ volatile ("0:\n\t"
                    "mftbu %0\n\t"
                    "mftbl %1\n\t"
                    "mftbu %2\n\t"
                    "cmpw %0, %2\n\t"
                    "bne- 0b"
                    : "=r" (__tbu), "=r" (__tbl), "=r" (__tmp));
  return (((uint64_t) __tbu << 32) | __tbl);
# endif  /* not __powerpc64__ */
#endif
}

The last block is exactly the code we had in dpdk.
So I suppose we are trying to use mfspr for register 268 which seems linked to timebase (looking at the linux kernel sources).

Please, confirm this is an enhancement (and how this improves current ppc support).
Thanks.

Signature · 1 line
David Marchand
author
#6 lore raw

Hi, Sorry for late response.
Yes this is the enhancement for powerpc. Observations on our power8/9
the __ppc_get_timebase calls __builtin_ppc_get_timebase () which is result in calling the mftb instruction
__ppc_get_timebase():
mftb rA
this instruction on a 64-bit implementation copies the entire time base (TBU||TBL) into rA, which also reduces number of cycles significantly comparing to the current code (same as last block)

Take the simple reciprocal division perf test on power9 that heavily calls rte_rdtsc() to demonstrate:
- without this batch:
Validating unsigned 32bit division.
32bit Division results:
Total number of cycles normal division : 73744549935
Total number of cycles reciprocal division : 76954877143
Cycles per division(normal) : 17.17
Cycles per division(reciprocal) : 17.92
Validating unsigned 64bit division.
64bit Division results:
Total number of cycles normal division : 73932937051
Total number of cycles reciprocal division : 74598584339
Cycles per division(normal) : 17.21
Cycles per division(reciprocal) : 17.37
Validating unsigned 64bit division with 32bit divisor.
64bit Division results:
Total number of cycles normal division : 78660556171
Total number of cycles reciprocal division : 74566630579
Cycles per division(normal) : 18.31
Cycles per division(reciprocal) : 17.36
Validating division by power of 2.
64bit Division results:
Total number of cycles normal division : 1097
Total number of cycles reciprocal division : 1201
Cycles per division(normal) : 17.14
Cycles per division(reciprocal) : 18.77
Test OK
RTE>>
- with the patch:
Validating unsigned 32bit division.
32bit Division results:
Total number of cycles normal division : 41690214596
Total number of cycles reciprocal division : 44446377795
Cycles per division(normal) : 9.71
Cycles per division(reciprocal) : 10.35
Validating unsigned 64bit division.
64bit Division results:
Total number of cycles normal division : 41687737031
Total number of cycles reciprocal division : 41666358052
Cycles per division(normal) : 9.71
Cycles per division(reciprocal) : 9.70
Validating unsigned 64bit division with 32bit divisor.
64bit Division results:
Total number of cycles normal division : 46386969228
Total number of cycles reciprocal division : 41663680498
Cycles per division(normal) : 10.80
Cycles per division(reciprocal) : 9.70
Validating division by power of 2.
64bit Division results:
Total number of cycles normal division : 618
Total number of cycles reciprocal division : 618
Cycles per division(normal) : 9.66
Cycles per division(reciprocal) : 9.66
Test OK
RTE>>

I hope this explains it.
Thanks,
Thinh Tran

Quoted 46 lines · David Marchand
Quoted 2 lines · Thinh Tran

__ppc_get_timebase() is GNU extension and is more efficient

The commit title and log are quite short and give little idea on what this is about.

I had a look at this glibc helper:

/* Read the Time Base Register. */
static __inline__ uint64_t
__ppc_get_timebase (void)
{

#if __GNUC_PREREQ (4, 8)
   return __builtin_ppc_get_timebase ();
#else
# ifdef __powerpc64__
   uint64_t __tb;
   /* "volatile" is necessary here, because the user expects this assembly
      isn't moved after an optimization.  */
   __asm__ volatile ("mfspr %0, 268" : "=r" (__tb));
   return __tb;
# else  /* not __powerpc64__ */
   uint32_t __tbu, __tbl, __tmp; \
   __asm__ volatile ("0:\n\t"
                     "mftbu %0\n\t"
                     "mftbl %1\n\t"
                     "mftbu %2\n\t"
                     "cmpw %0, %2\n\t"
                     "bne- 0b"
                     : "=r" (__tbu), "=r" (__tbl), "=r" (__tmp));
   return (((uint64_t) __tbu << 32) | __tbl);
# endif  /* not __powerpc64__ */
#endif
}

The last block is exactly the code we had in dpdk.
So I suppose we are trying to use mfspr for register 268 which seems linked to timebase (looking at the linux kernel sources).

Please, confirm this is an enhancement (and how this improves current ppc support).
Thanks.

#7 lore raw
In reply to Thinh Tran
Quoted 5 lines · Thinh Tran

__ppc_get_timebase() is GNU extension and is more efficient

v2: Advoid breaking other ppc_64 flatforms. The __ppc_get_timebase()
seems to be specific to powerpc platform and with GLIBC.

dpdk only supports glibc at the moment.
https://doc.dpdk.org/guides/linux_gsg/sys_reqs.html#system-software

Now the 1M$ question is which C library are you using and on which platform :-).

Quoted 11 lines
---
 lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h | 7 +++++++
 1 file changed, 7 insertions(+)
lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h +0 -0
index 8f2e98642..1c3fd556e 100644

lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h extern "C" { · line 19

Quoted 6 lines
lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h +3 -0
@@ -14,6 +14,9 @@ extern "C" {
 
 #include <rte_byteorder.h>
 #include <rte_common.h>
+#if defined(__powerpc__) && defined(__GLIBC__)
+#include <sys/platform/ppc.h>
+#endif

libc headers must come first:
https://doc.dpdk.org/guides/contributing/coding_style.html#header-includes

Why do you need the __powerpc__ guard?

Quoted 24 lines
lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h +4 -1
 /**
  * Read the time base register.
@@ -24,6 +27,9 @@ extern "C" {
 static inline uint64_t
 rte_rdtsc(void)
 {
+#if defined(__powerpc__) && defined(__GLIBC__)
+       return __ppc_get_timebase();
+#else
        union {
                uint64_t tsc_64;
                RTE_STD_C11
@@ -50,6 +56,7 @@ rte_rdtsc(void)
                        [tmp] "=r"(tmp)
                    );
        return tsc.tsc_64;
+#endif /* __powerpc__ && __GLIBC__ */
 }
 
 static inline uint64_t
--

2.17.1

Signature · 1 line
David Marchand
author
#8 lore raw

Hi David, Thanks for your response.

We were worry about breaking the DPDK/FreeBSD on power, may hit the similar bug:
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=241146

From my understanding, and also from the documentation
https://doc.dpdk.org/guides/freebsd_gsg/install_from_ports.html
DPDK is supported, glibc may not be required by default, and it is not clear what the supported architectures are.

So if FreeBSD does not support DPDK on powerpc or FreeBSD supports DPDK on powerpc and requires glibc installed, then the GLIBC guard can be removed

This is unknown to me.

Agreed. __powerpc__ guard is not necessary in this path and will be removed.

Thanks,
Thinh Tran

Quoted 61 lines · David Marchand
Quoted 5 lines · Thinh Tran

__ppc_get_timebase() is GNU extension and is more efficient

v2: Advoid breaking other ppc_64 flatforms. The __ppc_get_timebase()
seems to be specific to powerpc platform and with GLIBC.

dpdk only supports glibc at the moment.
https://doc.dpdk.org/guides/linux_gsg/sys_reqs.html#system-software

Now the 1M$ question is which C library are you using and on which platform :-).

Quoted 11 lines
---
  lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h | 7 +++++++
  1 file changed, 7 insertions(+)
lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h +0 -0
index 8f2e98642..1c3fd556e 100644

lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h extern "C" { · line 19

Quoted 6 lines
lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h +3 -0
@@ -14,6 +14,9 @@ extern "C" {
 
  #include <rte_byteorder.h>
  #include <rte_common.h>
+#if defined(__powerpc__) && defined(__GLIBC__)
+#include <sys/platform/ppc.h>
+#endif

libc headers must come first:
https://doc.dpdk.org/guides/contributing/coding_style.html#header-includes

Why do you need the __powerpc__ guard?

Quoted 24 lines
lib/librte_eal/common/include/arch/ppc_64/rte_cycles.h +4 -1
  /**
   * Read the time base register.
@@ -24,6 +27,9 @@ extern "C" {
  static inline uint64_t
  rte_rdtsc(void)
  {
+#if defined(__powerpc__) && defined(__GLIBC__)
+       return __ppc_get_timebase();
+#else
         union {
                 uint64_t tsc_64;
                 RTE_STD_C11
@@ -50,6 +56,7 @@ rte_rdtsc(void)
                         [tmp] "=r"(tmp)
                     );
         return tsc.tsc_64;
+#endif /* __powerpc__ && __GLIBC__ */
  }
 
  static inline uint64_t
--

2.17.1