Stowage/go
2
0
Fork 0
mirror of https://github.com/golang/go.git synced 2025-12-08 06:10:04 +00:00
Code Issues Projects Releases Packages Wiki Activity

cmd/compile: use generated loops instead of DUFFCOPY on riscv64

Browse source 
MemmoveKnownSize112-4            632.1Mi ± 1%   1288.5Mi ± 0%  +103.85% (p=0.000 n=10)
MemmoveKnownSize128-4            636.1Mi ± 0%   1280.9Mi ± 1%  +101.36% (p=0.000 n=10)
MemmoveKnownSize192-4            645.3Mi ± 0%   1306.9Mi ± 1%  +102.53% (p=0.000 n=10)
MemmoveKnownSize248-4            650.2Mi ± 2%   1312.5Mi ± 1%  +101.87% (p=0.000 n=10)
MemmoveKnownSize256-4            650.7Mi ± 0%   1303.6Mi ± 1%  +100.33% (p=0.000 n=10)
MemmoveKnownSize512-4            658.2Mi ± 1%   1293.9Mi ± 0%   +96.60% (p=0.000 n=10)
MemmoveKnownSize1024-4           662.1Mi ± 0%   1312.6Mi ± 0%   +98.26% (p=0.000 n=10)

Change-Id: I43681ca029880025558b33ddc4295da3947c9b28
Reviewed-on: https://go-review.googlesource.com/c/go/+/700537
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Mark Freeman <markfreeman@google.com>
This commit is contained in:
Meng Zhuo 2025-09-03 09:55:56 +08:00
parent 879ff736d3
commit 4dac9e093f
5 changed files with 186 additions and 222 deletions
Download patch file Download diff file Expand all files Collapse all files

57
src/cmd/compile/internal/ssa/_gen/RISCV64Ops.go
View file

@ -117,6 +117,7 @@ func init() {
regCtxt := regNamed["X26"]
callerSave := gpMask | fpMask | regNamed["g"]
r5toR6 := regNamed["X5"] | regNamed["X6"]
var (
gpstore = regInfo{inputs: []regMask{gpspsbMask, gpspMask, 0}} // SB in first input so we can load from a global, but not in second to avoid using SB as a temporary register
@ -354,27 +355,51 @@ func init() {
},
// general unaligned move
// arg0 = address of dst memory (in X5, changed as side effect)
// arg1 = address of src memory (in X6, changed as side effect)
// arg2 = address of the last element of src (can't be X7 as we clobber it before using arg2)
// arg3 = mem
// auxint = alignment
// clobbers X7 as a tmp register.
// arg0 = address of dst memory (clobber)
// arg1 = address of src memory (clobber)
// arg2 = mem
// auxint = size and type alignment
// returns mem
// mov (X6), X7
// mov X7, (X5)
// ADD $sz, X5
// ADD $sz, X6
// BGEU Rarg2, X5, -4(PC)
// mov (offset)(Rarg1), TMP
// mov TMP, (offset)(Rarg0)
{
name: "LoweredMove",
aux: "Int64",
argLength: 4,
aux: "SymValAndOff",
symEffect: "Write",
argLength: 3,
reg: regInfo{
inputs: []regMask{regNamed["X5"], regNamed["X6"], gpMask &^ regNamed["X7"]},
clobbers: regNamed["X5"] | regNamed["X6"] | regNamed["X7"],
inputs: []regMask{gpMask &^ regNamed["X5"], gpMask &^ regNamed["X5"]},
clobbers: regNamed["X5"],
},
faultOnNilArg0: true,
faultOnNilArg1: true,
},
// general unaligned move
// arg0 = address of dst memory (clobber)
// arg1 = address of src memory (clobber)
// arg3 = mem
// auxint = alignment
// returns mem
// ADD $sz, X6
//loop:
// mov (Rarg1), X5
// mov X5, (Rarg0)
// ...rest 7 mov...
// ADD $sz, Rarg0
// ADD $sz, Rarg1
// BNE X6, Rarg1, loop
{
name: "LoweredMoveLoop",
aux: "SymValAndOff",
argLength: 3,
symEffect: "Write",
reg: regInfo{
inputs: []regMask{gpMask &^ r5toR6, gpMask &^ r5toR6},
clobbers: r5toR6,
clobbersArg0: true,
clobbersArg1: true,
},
typ: "Mem",
faultOnNilArg0: true,
faultOnNilArg1: true,
},