sync/atomic, runtime/internal/atomic: improve ppc64x atomics

The following performance improvements have been made to the
low-level atomic functions for ppc64le & ppc64:

- For those cases containing a lwarx and stwcx (or other sizes):
sync, lwarx, maybe something, stwcx, loop to sync, sync, isync
The sync is moved before (outside) the lwarx/stwcx loop, and the
 sync after is removed, so it becomes:
sync, lwarx, maybe something, stwcx, loop to lwarx, isync

- For the Or8 and And8, the shifting and manipulation of the
address to the word aligned version were removed and the
instructions were changed to use lbarx, stbcx instead of
register shifting, xor, then lwarx, stwcx.

- New instructions LWSYNC, LBAR, STBCC were tested and added.
runtime/atomic_ppc64x.s was changed to use the LWSYNC opcode
instead of the WORD encoding.

Fixes #15469

Ran some of the benchmarks in the runtime and sync directories.
Some results varied from run to run but the trend was improvement
based on best times for base and new:

runtime.test:
BenchmarkChanNonblocking-128         0.88          0.89          +1.14%
BenchmarkChanUncontended-128         569           511           -10.19%
BenchmarkChanContended-128           63110         53231         -15.65%
BenchmarkChanSync-128                691           598           -13.46%
BenchmarkChanSyncWork-128            11355         11649         +2.59%
BenchmarkChanProdCons0-128           2402          2090          -12.99%
BenchmarkChanProdCons10-128          1348          1363          +1.11%
BenchmarkChanProdCons100-128         1002          746           -25.55%
BenchmarkChanProdConsWork0-128       2554          2720          +6.50%
BenchmarkChanProdConsWork10-128      1909          1804          -5.50%
BenchmarkChanProdConsWork100-128     1624          1580          -2.71%
BenchmarkChanCreation-128            237           212           -10.55%
BenchmarkChanSem-128                 705           667           -5.39%
BenchmarkChanPopular-128             5081190       4497566       -11.49%

BenchmarkCreateGoroutines-128             532           473           -11.09%
BenchmarkCreateGoroutinesParallel-128     35.0          34.7          -0.86%
BenchmarkCreateGoroutinesCapture-128      4923          4200          -14.69%

sync.test:
BenchmarkUncontendedSemaphore-128      112           94.2          -15.89%
BenchmarkContendedSemaphore-128        133           128           -3.76%
BenchmarkMutexUncontended-128          1.90          1.67          -12.11%
BenchmarkMutex-128                     353           310           -12.18%
BenchmarkMutexSlack-128                304           283           -6.91%
BenchmarkMutexWork-128                 554           541           -2.35%
BenchmarkMutexWorkSlack-128            567           556           -1.94%
BenchmarkMutexNoSpin-128               275           242           -12.00%
BenchmarkMutexSpin-128                 1129          1030          -8.77%
BenchmarkOnce-128                      1.08          0.96          -11.11%
BenchmarkPool-128                      29.8          27.4          -8.05%
BenchmarkPoolOverflow-128              40564         36583         -9.81%
BenchmarkSemaUncontended-128           3.14          2.63          -16.24%
BenchmarkSemaSyntNonblock-128          1087          1069          -1.66%
BenchmarkSemaSyntBlock-128             897           893           -0.45%
BenchmarkSemaWorkNonblock-128          1034          1028          -0.58%
BenchmarkSemaWorkBlock-128             949           886           -6.64%

Change-Id: I4403fb29d3cd5254b7b1ce87a216bd11b391079e
Reviewed-on: https://go-review.googlesource.com/22549
Reviewed-by: Michael Munday <munday@ca.ibm.com>
Reviewed-by: Minux Ma <minux@golang.org>

1 files changed, 16 insertions, 53 deletions

diff --git a/src/runtime/internal/atomic/asm_ppc64x.s b/src/runtime/internal/atomic/asm_ppc64x.s
index 4a776787a2..a928e400d2 100644
--- a/src/runtime/internal/atomic/asm_ppc64x.s
+++ b/src/runtime/internal/atomic/asm_ppc64x.s
@@ -17,21 +17,20 @@ TEXT runtime∕internal∕atomic·Cas(SB), NOSPLIT, $0-17
 	MOVD	ptr+0(FP), R3
 	MOVWZ	old+8(FP), R4
 	MOVWZ	new+12(FP), R5
-cas_again:
 	SYNC
+cas_again:
 	LWAR	(R3), R6
 	CMPW	R6, R4
 	BNE	cas_fail
 	STWCCC	R5, (R3)
 	BNE	cas_again
 	MOVD	$1, R3
-	SYNC
 	ISYNC
 	MOVB	R3, ret+16(FP)
 	RET
 cas_fail:
-	MOVD	$0, R3
-	BR	-5(PC)
+	MOVB	R0, ret+16(FP)
+	RET
 
 // bool	runtime∕internal∕atomic·Cas64(uint64 *ptr, uint64 old, uint64 new)
 // Atomically:
@@ -45,21 +44,20 @@ TEXT runtime∕internal∕atomic·Cas64(SB), NOSPLIT, $0-25
 	MOVD	ptr+0(FP), R3
 	MOVD	old+8(FP), R4
 	MOVD	new+16(FP), R5
-cas64_again:
 	SYNC
+cas64_again:
 	LDAR	(R3), R6
 	CMP	R6, R4
 	BNE	cas64_fail
 	STDCCC	R5, (R3)
 	BNE	cas64_again
 	MOVD	$1, R3
-	SYNC
 	ISYNC
 	MOVB	R3, ret+24(FP)
 	RET
 cas64_fail:
-	MOVD	$0, R3
-	BR	-5(PC)
+	MOVB	R0, ret+24(FP)
+	RET
 
 TEXT runtime∕internal∕atomic·Casuintptr(SB), NOSPLIT, $0-25
 	BR	runtime∕internal∕atomic·Cas64(SB)
@@ -103,8 +101,7 @@ TEXT runtime∕internal∕atomic·Xadd(SB), NOSPLIT, $0-20
 	LWAR	(R4), R3
 	ADD	R5, R3
 	STWCCC	R3, (R4)
-	BNE	-4(PC)
-	SYNC
+	BNE	-3(PC)
 	ISYNC
 	MOVW	R3, ret+16(FP)
 	RET
@@ -116,8 +113,7 @@ TEXT runtime∕internal∕atomic·Xadd64(SB), NOSPLIT, $0-24
 	LDAR	(R4), R3
 	ADD	R5, R3
 	STDCCC	R3, (R4)
-	BNE	-4(PC)
-	SYNC
+	BNE	-3(PC)
 	ISYNC
 	MOVD	R3, ret+16(FP)
 	RET
@@ -128,8 +124,7 @@ TEXT runtime∕internal∕atomic·Xchg(SB), NOSPLIT, $0-20
 	SYNC
 	LWAR	(R4), R3
 	STWCCC	R5, (R4)
-	BNE	-3(PC)
-	SYNC
+	BNE	-2(PC)
 	ISYNC
 	MOVW	R3, ret+16(FP)
 	RET
@@ -140,8 +135,7 @@ TEXT runtime∕internal∕atomic·Xchg64(SB), NOSPLIT, $0-24
 	SYNC
 	LDAR	(R4), R3
 	STDCCC	R5, (R4)
-	BNE	-3(PC)
-	SYNC
+	BNE	-2(PC)
 	ISYNC
 	MOVD	R3, ret+16(FP)
 	RET
@@ -171,26 +165,12 @@ TEXT runtime∕internal∕atomic·Store64(SB), NOSPLIT, $0-16
 TEXT runtime∕internal∕atomic·Or8(SB), NOSPLIT, $0-9
 	MOVD	ptr+0(FP), R3
 	MOVBZ	val+8(FP), R4
-	// Align ptr down to 4 bytes so we can use 32-bit load/store.
-	// R5 = (R3 << 0) & ~3
-	RLDCR	$0, R3, $~3, R5
-	// Compute val shift.
-#ifdef GOARCH_ppc64
-	// Big endian.  ptr = ptr ^ 3
-	XOR	$3, R3
-#endif
-	// R6 = ((ptr & 3) * 8) = (ptr << 3) & (3*8)
-	RLDC	$3, R3, $(3*8), R6
-	// Shift val for aligned ptr. R4 = val << R6
-	SLD	R6, R4, R4
-
-again:
 	SYNC
-	LWAR	(R5), R6
+again:
+	LBAR	(R3), R6
 	OR	R4, R6
-	STWCCC	R6, (R5)
+	STBCCC	R6, (R3)
 	BNE	again
-	SYNC
 	ISYNC
 	RET
 
@@ -198,28 +178,11 @@ again:
 TEXT runtime∕internal∕atomic·And8(SB), NOSPLIT, $0-9
 	MOVD	ptr+0(FP), R3
 	MOVBZ	val+8(FP), R4
-	// Align ptr down to 4 bytes so we can use 32-bit load/store.
-	// R5 = (R3 << 0) & ~3
-	RLDCR	$0, R3, $~3, R5
-	// Compute val shift.
-#ifdef GOARCH_ppc64
-	// Big endian.  ptr = ptr ^ 3
-	XOR	$3, R3
-#endif
-	// R6 = ((ptr & 3) * 8) = (ptr << 3) & (3*8)
-	RLDC	$3, R3, $(3*8), R6
-	// Shift val for aligned ptr. R4 = val << R6 | ^(0xFF << R6)
-	MOVD	$0xFF, R7
-	SLD	R6, R4
-	SLD	R6, R7
-	XOR $-1, R7
-	OR	R7, R4
-again:
 	SYNC
-	LWAR	(R5), R6
+again:
+	LBAR	(R3), R6
 	AND	R4, R6
-	STWCCC	R6, (R5)
+	STBCCC	R6, (R3)
 	BNE	again
-	SYNC
 	ISYNC
 	RET