cmd/compile: don't fold address of global into load/store on PPC64

On PPC64 (and a few other architectures), accessing global
requires multiple instructions and use of temp register.
The compiler emits a single MOV prog, and the assembler
expands it to multiple instructions. If globals are accessed
multiple times, each time it generates a reload of the temp
register. As this is done by the assembler, the compiler
cannot optimize it.

This CL makes the compiler not fold address of global into load
and store. If a global is accessed multiple times, or multiple
fields of a struct are accessed, the compiler can CSE the
address. Currently, this doesn't help the case where different
globals are accessed, even though they may be close to each
other in the address space (which we don't know at compile time).

It helps a little bit in go1 benchmark:

name                     old time/op    new time/op    delta
BinaryTree17-2              4.84s ± 1%     4.84s ± 1%    ~     (p=0.796 n=10+10)
Fannkuch11-2                4.10s ± 0%     4.08s ± 0%  -0.58%  (p=0.000 n=9+8)
FmtFprintfEmpty-2          97.9ns ± 1%    96.8ns ± 1%  -1.08%  (p=0.000 n=10+10)
FmtFprintfString-2          147ns ± 0%     147ns ± 1%    ~     (p=0.129 n=9+10)
FmtFprintfInt-2             152ns ± 0%     152ns ± 0%    ~     (p=0.294 n=10+8)
FmtFprintfIntInt-2          218ns ± 1%     217ns ± 0%  -0.64%  (p=0.000 n=10+8)
FmtFprintfPrefixedInt-2     263ns ± 1%     256ns ± 0%  -2.77%  (p=0.000 n=10+8)
FmtFprintfFloat-2           375ns ± 1%     368ns ± 0%  -1.95%  (p=0.000 n=10+7)
FmtManyArgs-2               849ns ± 0%     850ns ± 0%    ~     (p=0.621 n=8+9)
GobDecode-2                12.3ms ± 1%    12.2ms ± 1%  -0.94%  (p=0.003 n=10+10)
GobEncode-2                10.3ms ± 1%    10.5ms ± 1%  +2.03%  (p=0.000 n=10+10)
Gzip-2                      414ms ± 1%     414ms ± 0%    ~     (p=0.842 n=9+10)
Gunzip-2                   66.3ms ± 0%    66.4ms ± 0%    ~     (p=0.077 n=9+9)
HTTPClientServer-2         66.3µs ± 5%    66.4µs ± 1%    ~     (p=0.661 n=10+9)
JSONEncode-2               23.9ms ± 1%    23.9ms ± 1%    ~     (p=0.905 n=10+9)
JSONDecode-2                119ms ± 1%     116ms ± 0%  -2.65%  (p=0.000 n=10+10)
Mandelbrot200-2            5.11ms ± 0%    4.92ms ± 0%  -3.71%  (p=0.000 n=10+10)
GoParse-2                  5.81ms ± 1%    5.84ms ± 1%    ~     (p=0.052 n=10+10)
RegexpMatchEasy0_32-2       315ns ± 0%     317ns ± 0%  +0.67%  (p=0.000 n=10+10)
RegexpMatchEasy0_1K-2       658ns ± 0%     638ns ± 0%  -3.01%  (p=0.000 n=9+9)
RegexpMatchEasy1_32-2       315ns ± 1%     317ns ± 0%  +0.56%  (p=0.000 n=9+9)
RegexpMatchEasy1_1K-2       935ns ± 0%     926ns ± 0%  -0.96%  (p=0.000 n=9+9)
RegexpMatchMedium_32-2      394ns ± 0%     396ns ± 1%  +0.46%  (p=0.001 n=10+10)
RegexpMatchMedium_1K-2     65.1µs ± 0%    64.5µs ± 0%  -0.90%  (p=0.000 n=9+9)
RegexpMatchHard_32-2       3.16µs ± 0%    3.17µs ± 0%  +0.35%  (p=0.000 n=10+9)
RegexpMatchHard_1K-2       89.4µs ± 0%    89.3µs ± 0%    ~     (p=0.136 n=9+9)
Revcomp-2                   703ms ± 2%     694ms ± 2%  -1.41%  (p=0.009 n=10+10)
Template-2                  107ms ± 1%     107ms ± 1%    ~     (p=0.053 n=9+10)
TimeParse-2                 526ns ± 0%     524ns ± 0%  -0.34%  (p=0.002 n=9+9)
TimeFormat-2                534ns ± 0%     504ns ± 1%  -5.51%  (p=0.000 n=10+10)
[Geo mean]                 93.8µs         93.1µs       -0.70%

It also helps in the case mentioned in issue #17110, main.main
in package math's test. Now it generates 4 loads of R31 instead
of 10, for the same piece of code.

This causes a slight increase of binary size: cmd/go increases
0.66%.

If this is a good idea, we should do it on other architectures
where accessing global is expensive.

Updates #17110.

Change-Id: I2687af6eafc04f2a57c19781ec300c33567094b6
Reviewed-on: https://go-review.googlesource.com/68250
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>

src/cmd/compile/internal/ssa/gen/PPC64.rules

@@ -760,35 +760,54 @@
 (FMOVSstore [off1] {sym} (ADDconst [off2] ptr) val mem) && is16Bit(off1+off2) -> (FMOVSstore [off1+off2] {sym} ptr val mem)
 (FMOVDstore [off1] {sym} (ADDconst [off2] ptr) val mem) && is16Bit(off1+off2) -> (FMOVDstore [off1+off2] {sym} ptr val mem)
 
-(MOVBstore [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2) ->
+// Fold address into load/store.
+// The assembler needs to generate several instructions and use
+// temp register for accessing global, and each time it will reload
+// the temp register. So don't fold address of global, unless there
+// is only one use.
+(MOVBstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (MOVBstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)
-(MOVHstore [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2) ->
+(MOVHstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (MOVHstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)
-(MOVWstore [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2) ->
+(MOVWstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (MOVWstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)
-(MOVDstore [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2) ->
+(MOVDstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (MOVDstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)
 
-(FMOVSstore [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2) ->
+(FMOVSstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (FMOVSstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)
-(FMOVDstore [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2) ->
+(FMOVDstore [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) val mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (FMOVDstore [off1+off2] {mergeSym(sym1,sym2)} ptr val mem)
 
-(MOVBZload [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2) ->
+(MOVBZload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (MOVBZload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
-(MOVHload [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2) ->
+(MOVHload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (MOVHload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
-(MOVHZload [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2) ->
+(MOVHZload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (MOVHZload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
-(MOVWload [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2) ->
+(MOVWload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (MOVWload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
-(MOVWZload [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2) ->
+(MOVWZload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (MOVWZload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
-(MOVDload [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2) ->
+(MOVDload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (MOVDload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
-(FMOVSload [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2) ->
+(FMOVSload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (FMOVSload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
-(FMOVDload [off1] {sym1} (MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2) ->
+(FMOVDload [off1] {sym1} p:(MOVDaddr [off2] {sym2} ptr) mem) && canMergeSym(sym1,sym2)
+	&& (ptr.Op != OpSB || p.Uses == 1) ->
         (FMOVDload [off1+off2] {mergeSym(sym1,sym2)} ptr mem)
 
 // Fold offsets for loads.
@@ -819,13 +838,17 @@
     (MOVBstorezero [off1+off2] {sym} x mem)
 
 // Fold symbols into storezero
-(MOVDstorezero [off1] {sym1} (MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2) ->
+(MOVDstorezero [off1] {sym1} p:(MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2)
+	&& (x.Op != OpSB || p.Uses == 1) ->
     (MOVDstorezero [off1+off2] {mergeSym(sym1,sym2)} x mem)
-(MOVWstorezero [off1] {sym1} (MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2) ->
+(MOVWstorezero [off1] {sym1} p:(MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2)
+	&& (x.Op != OpSB || p.Uses == 1) ->
     (MOVWstorezero [off1+off2] {mergeSym(sym1,sym2)} x mem)
-(MOVHstorezero [off1] {sym1} (MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2) ->
+(MOVHstorezero [off1] {sym1} p:(MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2)
+	&& (x.Op != OpSB || p.Uses == 1) ->
     (MOVHstorezero [off1+off2] {mergeSym(sym1,sym2)} x mem)
-(MOVBstorezero [off1] {sym1} (MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2) ->
+(MOVBstorezero [off1] {sym1} p:(MOVDaddr [off2] {sym2} x) mem) && canMergeSym(sym1,sym2)
+	&& (x.Op != OpSB || p.Uses == 1) ->
     (MOVBstorezero [off1+off2] {mergeSym(sym1,sym2)} x mem)
 
 // atomic intrinsics