1
0
mirror of https://github.com/golang/go synced 2024-09-30 18:18:32 -06:00

cmd/compile: improve Slicemask on riscv64

Implement Slicemask the same way every other architecture does - negate
then arithmetic right shift. This sets or clears the sign bit, before
extending it to the entire register.

Removes around 2,500 instructions from the Go binary on linux/riscv64.

Change-Id: I4d675b826e7eb23fe2b1e6e46b95dcd49ab49733
Reviewed-on: https://go-review.googlesource.com/c/go/+/426354
Reviewed-by: Meng Zhuo <mzh@golangcn.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: Joel Sing <joel@sing.id.au>
Reviewed-by: Wayne Zuo <wdvxdr@golangcn.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
This commit is contained in:
Joel Sing 2022-08-29 20:36:51 +10:00
parent ba1ef54c1e
commit c011270fa5
2 changed files with 6 additions and 16 deletions

View File

@ -142,14 +142,7 @@
(Round32F ...) => (Copy ...)
(Round64F ...) => (Copy ...)
// From genericOps.go:
// "0 if arg0 == 0, -1 if arg0 > 0, undef if arg0<0"
//
// Like other arches, we compute ~((x-1) >> 63), with arithmetic right shift.
// For positive x, bit 63 of x-1 is always 0, so the result is -1.
// For zero x, bit 63 of x-1 is 1, so the result is 0.
//
(Slicemask <t> x) => (NOT (SRAI <t> [63] (ADDI <t> [-1] x)))
(Slicemask <t> x) => (SRAI [63] (NEG <t> x))
// Truncations
// We ignore the unused high parts of registers, so truncates are just copies.

View File

@ -7329,17 +7329,14 @@ func rewriteValueRISCV64_OpSlicemask(v *Value) bool {
v_0 := v.Args[0]
b := v.Block
// match: (Slicemask <t> x)
// result: (NOT (SRAI <t> [63] (ADDI <t> [-1] x)))
// result: (SRAI [63] (NEG <t> x))
for {
t := v.Type
x := v_0
v.reset(OpRISCV64NOT)
v0 := b.NewValue0(v.Pos, OpRISCV64SRAI, t)
v0.AuxInt = int64ToAuxInt(63)
v1 := b.NewValue0(v.Pos, OpRISCV64ADDI, t)
v1.AuxInt = int64ToAuxInt(-1)
v1.AddArg(x)
v0.AddArg(v1)
v.reset(OpRISCV64SRAI)
v.AuxInt = int64ToAuxInt(63)
v0 := b.NewValue0(v.Pos, OpRISCV64NEG, t)
v0.AddArg(x)
v.AddArg(v0)
return true
}