cmd/compile/internal/ssa: expand runtime.memequal for length {3,5,6,7}

This CL slightly speeds up strings.HasPrefix when testing constant
prefixes of length {3,5,6,7}.

goos: linux
goarch: amd64
cpu: Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz
                │      old     │                 new                 │
                │    sec/op    │   sec/op     vs base                │
StringPrefix3-8   11.125n ± 2%   8.539n ± 1%  -23.25% (p=0.000 n=20)
StringPrefix5-8   11.170n ± 2%   8.700n ± 1%  -22.11% (p=0.000 n=20)
StringPrefix6-8   11.190n ± 2%   8.655n ± 1%  -22.65% (p=0.000 n=20)
StringPrefix7-8   11.095n ± 1%   8.878n ± 1%  -19.98% (p=0.000 n=20)

Change-Id: I510a80d59cf78680b57d68780d35d212d24030e2
Reviewed-on: https://go-review.googlesource.com/c/go/+/700816
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Auto-Submit: Keith Randall <khr@golang.org>

1 files changed, 113 insertions, 1 deletions

diff --git a/src/cmd/compile/internal/ssa/_gen/generic.rules b/src/cmd/compile/internal/ssa/_gen/generic.rules
index c0806c2243..58872ca85a 100644
--- a/src/cmd/compile/internal/ssa/_gen/generic.rules
+++ b/src/cmd/compile/internal/ssa/_gen/generic.rules
@@ -2084,7 +2084,7 @@
 (NilCheck ptr:(NilCheck _ _) _ ) => ptr
 
 // for late-expanded calls, recognize memequal applied to a single constant byte
-// Support is limited by 1, 2, 4, 8 byte sizes
+// Support is limited by [1-8] byte sizes
 (StaticLECall {callAux} sptr (Addr {scon} (SB)) (Const64 [1]) mem)
   && isSameCall(callAux, "runtime.memequal")
   && symIsRO(scon)
@@ -2131,6 +2131,118 @@
   && canLoadUnaligned(config) && config.PtrSize == 8
   => (MakeResult (Eq64 (Load <typ.Int64> sptr mem) (Const64 <typ.Int64> [int64(read64(scon,0,config.ctxt.Arch.ByteOrder))])) mem)
 
+(StaticLECall {callAux} sptr (Addr {scon} (SB)) (Const64 [3]) mem)
+  && isSameCall(callAux, "runtime.memequal")
+  && symIsRO(scon)
+  && canLoadUnaligned(config) =>
+  (MakeResult
+    (Eq32
+      (Or32 <typ.Int32>
+        (ZeroExt16to32 <typ.Int32> (Load <typ.Int16> sptr mem))
+        (Lsh32x32 <typ.Int32>
+          (ZeroExt8to32 <typ.Int32> (Load <typ.Int8> (OffPtr <typ.BytePtr> [2] sptr) mem))
+          (Const32 <typ.Int32> [16])))
+      (Const32 <typ.Int32> [int32(uint32(read16(scon,0,config.ctxt.Arch.ByteOrder))|(uint32(read8(scon,2))<<16))]))
+    mem)
+
+(StaticLECall {callAux} (Addr {scon} (SB)) sptr (Const64 [3]) mem)
+  && isSameCall(callAux, "runtime.memequal")
+  && symIsRO(scon)
+  && canLoadUnaligned(config) =>
+  (MakeResult
+    (Eq32
+      (Or32 <typ.Int32>
+        (ZeroExt16to32 <typ.Int32> (Load <typ.Int16> sptr mem))
+        (Lsh32x32 <typ.Int32>
+          (ZeroExt8to32 <typ.Int32> (Load <typ.Int8> (OffPtr <typ.BytePtr> [2] sptr) mem))
+          (Const32 <typ.Int32> [16])))
+      (Const32 <typ.Int32> [int32(uint32(read16(scon,0,config.ctxt.Arch.ByteOrder))|(uint32(read8(scon,2))<<16))]))
+    mem)
+
+(StaticLECall {callAux} sptr (Addr {scon} (SB)) (Const64 [5]) mem)
+  && isSameCall(callAux, "runtime.memequal")
+  && symIsRO(scon)
+  && canLoadUnaligned(config) && config.PtrSize == 8 =>
+  (MakeResult
+    (Eq64
+      (Or64 <typ.Int64>
+        (ZeroExt32to64 <typ.Int64> (Load <typ.Int32> sptr mem))
+        (Lsh64x64 <typ.Int64>
+          (ZeroExt8to64 <typ.Int64> (Load <typ.Int8> (OffPtr <typ.BytePtr> [4] sptr) mem))
+          (Const64 <typ.Int64> [32])))
+      (Const64 <typ.Int64> [int64(uint64(read32(scon,0,config.ctxt.Arch.ByteOrder))|(uint64(read8(scon,4))<<32))]))
+    mem)
+
+(StaticLECall {callAux} (Addr {scon} (SB)) sptr (Const64 [5]) mem)
+  && isSameCall(callAux, "runtime.memequal")
+  && symIsRO(scon)
+  && canLoadUnaligned(config) && config.PtrSize == 8 =>
+  (MakeResult
+    (Eq64
+      (Or64 <typ.Int64>
+        (ZeroExt32to64 <typ.Int64> (Load <typ.Int32> sptr mem))
+        (Lsh64x64 <typ.Int64>
+          (ZeroExt8to64 <typ.Int64> (Load <typ.Int8> (OffPtr <typ.BytePtr> [4] sptr) mem))
+          (Const64 <typ.Int64> [32])))
+      (Const64 <typ.Int64> [int64(uint64(read32(scon,0,config.ctxt.Arch.ByteOrder))|(uint64(read8(scon,4))<<32))]))
+    mem)
+
+(StaticLECall {callAux} sptr (Addr {scon} (SB)) (Const64 [6]) mem)
+  && isSameCall(callAux, "runtime.memequal")
+  && symIsRO(scon)
+  && canLoadUnaligned(config) && config.PtrSize == 8 =>
+  (MakeResult
+    (Eq64
+      (Or64 <typ.Int64>
+        (ZeroExt32to64 <typ.Int64> (Load <typ.Int32> sptr mem))
+        (Lsh64x64 <typ.Int64>
+          (ZeroExt16to64 <typ.Int64> (Load <typ.Int16> (OffPtr <typ.BytePtr> [4] sptr) mem))
+          (Const64 <typ.Int64> [32])))
+      (Const64 <typ.Int64> [int64(uint64(read32(scon,0,config.ctxt.Arch.ByteOrder))|(uint64(read16(scon,4,config.ctxt.Arch.ByteOrder))<<32))]))
+    mem)
+
+(StaticLECall {callAux} (Addr {scon} (SB)) sptr (Const64 [6]) mem)
+  && isSameCall(callAux, "runtime.memequal")
+  && symIsRO(scon)
+  && canLoadUnaligned(config) && config.PtrSize == 8 =>
+  (MakeResult
+    (Eq64
+      (Or64 <typ.Int64>
+        (ZeroExt32to64 <typ.Int64> (Load <typ.Int32> sptr mem))
+        (Lsh64x64 <typ.Int64>
+          (ZeroExt16to64 <typ.Int64> (Load <typ.Int16> (OffPtr <typ.BytePtr> [4] sptr) mem))
+          (Const64 <typ.Int64> [32])))
+      (Const64 <typ.Int64> [int64(uint64(read32(scon,0,config.ctxt.Arch.ByteOrder))|(uint64(read16(scon,4,config.ctxt.Arch.ByteOrder))<<32))]))
+    mem)
+
+(StaticLECall {callAux} sptr (Addr {scon} (SB)) (Const64 [7]) mem)
+  && isSameCall(callAux, "runtime.memequal")
+  && symIsRO(scon)
+  && canLoadUnaligned(config) && config.PtrSize == 8 =>
+  (MakeResult
+    (Eq64
+      (Or64 <typ.Int64>
+        (ZeroExt32to64 <typ.Int64> (Load <typ.Int32> sptr mem))
+        (Lsh64x64 <typ.Int64>
+          (ZeroExt32to64 <typ.Int64> (Load <typ.Int32> (OffPtr <typ.BytePtr> [3] sptr) mem))
+          (Const64 <typ.Int64> [32])))
+      (Const64 <typ.Int64> [int64(uint64(read32(scon,0,config.ctxt.Arch.ByteOrder))|(uint64(read32(scon,3,config.ctxt.Arch.ByteOrder))<<32))]))
+    mem)
+
+(StaticLECall {callAux} (Addr {scon} (SB)) sptr (Const64 [7]) mem)
+  && isSameCall(callAux, "runtime.memequal")
+  && symIsRO(scon)
+  && canLoadUnaligned(config) && config.PtrSize == 8 =>
+  (MakeResult
+    (Eq64
+      (Or64 <typ.Int64>
+        (ZeroExt32to64 <typ.Int64> (Load <typ.Int32> sptr mem))
+        (Lsh64x64 <typ.Int64>
+          (ZeroExt32to64 <typ.Int64> (Load <typ.Int32> (OffPtr <typ.BytePtr> [3] sptr) mem))
+          (Const64 <typ.Int64> [32])))
+      (Const64 <typ.Int64> [int64(uint64(read32(scon,0,config.ctxt.Arch.ByteOrder))|(uint64(read32(scon,3,config.ctxt.Arch.ByteOrder))<<32))]))
+    mem)
+
 (StaticLECall {callAux} _ _ (Const64 [0]) mem)
   && isSameCall(callAux, "runtime.memequal")
   => (MakeResult (ConstBool <typ.Bool> [true]) mem)