go/test/codegen/noextend.go

287 lines
5.4 KiB
Go
Raw Normal View History

cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// asmcheck
// Copyright 2018 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package codegen
import "math/bits"
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
var sval64 [8]int64
var sval32 [8]int32
var sval16 [8]int16
var sval8 [8]int8
var val64 [8]uint64
var val32 [8]uint32
var val16 [8]uint16
var val8 [8]uint8
// Avoid zero/sign extensions following a load
// which has extended the value correctly.
// Note: No tests are done for int8 since
// an extra extension is usually needed due to
// no signed byte load.
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
func set16(x8 int8, u8 *uint8, y8 int8, z8 uint8) {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// Truncate not needed, load does sign/zero extend
// ppc64x:-"MOVBZ\tR\\d+,\\sR\\d+"
val16[0] = uint16(*u8)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// AND not needed due to size
// ppc64x:-"ANDCC"
sval16[1] = 255 & int16(x8+y8)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"ANDCC"
val16[1] = 255 & uint16(*u8+z8)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
}
func shiftidx(u8 *uint8, x16 *int16, u16 *uint16) {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVBZ\tR\\d+,\\sR\\d+"
val16[0] = uint16(sval16[*u8>>2])
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVH\tR\\d+,\\sR\\d+"
sval16[1] = int16(val16[*x16>>1])
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVHZ\tR\\d+,\\sR\\d+"
val16[1] = uint16(sval16[*u16>>2])
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
}
func setnox(x8 int8, u8 *uint8, y8 *int8, z8 *uint8, x16 *int16, u16 *uint16, x32 *int32, u32 *uint32) {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVBZ\tR\\d+,\\sR\\d+"
val16[0] = uint16(*u8)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// AND not needed due to size
// ppc64x:-"ANDCC"
sval16[1] = 255 & int16(x8+*y8)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"ANDCC"
val16[1] = 255 & uint16(*u8+*z8)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVH\tR\\d+,\\sR\\d+"
sval32[1] = int32(*x16)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVBZ\tR\\d+,\\sR\\d+"
val32[0] = uint32(*u8)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVHZ\tR\\d+,\\sR\\d+"
val32[1] = uint32(*u16)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVH\tR\\d+,\\sR\\d+"
sval64[1] = int64(*x16)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVW\tR\\d+,\\sR\\d+"
sval64[2] = int64(*x32)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVBZ\tR\\d+,\\sR\\d+"
val64[0] = uint64(*u8)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVHZ\tR\\d+,\\sR\\d+"
val64[1] = uint64(*u16)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVWZ\tR\\d+,\\sR\\d+"
val64[2] = uint64(*u32)
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
}
func cmp16(u8 *uint8, x32 *int32, u32 *uint32, x64 *int64, u64 *uint64) bool {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVBZ\tR\\d+,\\sR\\d+"
if uint16(*u8) == val16[0] {
return true
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
}
// ppc64x:-"MOVHZ\tR\\d+,\\sR\\d+"
if uint16(*u32>>16) == val16[0] {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
return true
}
// ppc64x:-"MOVHZ\tR\\d+,\\sR\\d+"
if uint16(*u64>>48) == val16[0] {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
return true
}
// Verify the truncates are using the correct sign.
// ppc64x:-"MOVHZ\tR\\d+,\\sR\\d+"
if int16(*x32) == sval16[0] {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
return true
}
// ppc64x:-"MOVH\tR\\d+,\\sR\\d+"
if uint16(*u32) == val16[0] {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
return true
}
// ppc64x:-"MOVHZ\tR\\d+,\\sR\\d+"
if int16(*x64) == sval16[0] {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
return true
}
// ppc64x:-"MOVH\tR\\d+,\\sR\\d+"
if uint16(*u64) == val16[0] {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
return true
}
return false
}
func cmp32(u8 *uint8, x16 *int16, u16 *uint16, x64 *int64, u64 *uint64) bool {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVBZ\tR\\d+,\\sR\\d+"
if uint32(*u8) == val32[0] {
return true
}
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVH\tR\\d+,\\sR\\d+"
if int32(*x16) == sval32[0] {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
return true
}
// ppc64x:-"MOVHZ\tR\\d+,\\sR\\d+"
if uint32(*u16) == val32[0] {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
return true
}
// Verify the truncates are using the correct sign.
// ppc64x:-"MOVWZ\tR\\d+,\\sR\\d+"
if int32(*x64) == sval32[0] {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
return true
}
// ppc64x:-"MOVW\tR\\d+,\\sR\\d+"
if uint32(*u64) == val32[0] {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
return true
}
return false
}
func cmp64(u8 *uint8, x16 *int16, u16 *uint16, x32 *int32, u32 *uint32) bool {
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVBZ\tR\\d+,\\sR\\d+"
if uint64(*u8) == val64[0] {
return true
}
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVH\tR\\d+,\\sR\\d+"
if int64(*x16) == sval64[0] {
return true
}
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVHZ\tR\\d+,\\sR\\d+"
if uint64(*u16) == val64[0] {
return true
}
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVW\tR\\d+,\\sR\\d+"
if int64(*x32) == sval64[0] {
return true
}
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
// ppc64x:-"MOVWZ\tR\\d+,\\sR\\d+"
if uint64(*u32) == val64[0] {
return true
}
return false
cmd/compile: improve rules for PPC64.rules This adds some improvements to the rules for PPC64 to eliminate unnecessary zero or sign extends, and fix some rule for truncates which were not always using the correct sign instruction. This reduces of size of many functions by 1 or 2 instructions and can improve performance in cases where the execution time depends on small loops where at least 1 instruction was removed and where that loop contributes a significant amount of the total execution time. Included is a testcase for codegen to verify the sign/zero extend instructions are omitted. An example of the improvement (strings): IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10) IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9) IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6) IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10) IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10) IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10) IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8) IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10) IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10) IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10) Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500 Reviewed-on: https://go-review.googlesource.com/129875 Reviewed-by: David Chase <drchase@google.com>
2018-08-15 21:34:06 +00:00
}
// no unsign extension following 32 bits ops
func noUnsignEXT(t1, t2, t3, t4 uint32, k int64) uint64 {
var ret uint64
// arm64:"RORW",-"MOVWU"
ret += uint64(bits.RotateLeft32(t1, 7))
// arm64:"MULW",-"MOVWU"
ret *= uint64(t1 * t2)
// arm64:"MNEGW",-"MOVWU"
ret += uint64(-t1 * t3)
// arm64:"UDIVW",-"MOVWU"
ret += uint64(t1 / t4)
// arm64:-"MOVWU"
ret += uint64(t2 % t3)
// arm64:"MSUBW",-"MOVWU"
ret += uint64(t1 - t2*t3)
// arm64:"MADDW",-"MOVWU"
ret += uint64(t3*t4 + t2)
// arm64:"REVW",-"MOVWU"
ret += uint64(bits.ReverseBytes32(t1))
// arm64:"RBITW",-"MOVWU"
ret += uint64(bits.Reverse32(t1))
// arm64:"CLZW",-"MOVWU"
ret += uint64(bits.LeadingZeros32(t1))
// arm64:"REV16W",-"MOVWU"
ret += uint64(((t1 & 0xff00ff00) >> 8) | ((t1 & 0x00ff00ff) << 8))
// arm64:"EXTRW",-"MOVWU"
ret += uint64((t1 << 25) | (t2 >> 7))
return ret
}
// no sign extension when the upper bits of the result are zero
func noSignEXT(x int) int64 {
t1 := int32(x)
var ret int64
// arm64:-"MOVW"
ret += int64(t1 & 1)
// arm64:-"MOVW"
ret += int64(int32(x & 0x7fffffff))
// arm64:-"MOVH"
ret += int64(int16(x & 0x7fff))
// arm64:-"MOVB"
ret += int64(int8(x & 0x7f))
return ret
}
// corner cases that sign extension must not be omitted
func shouldSignEXT(x int) int64 {
t1 := int32(x)
var ret int64
// arm64:"MOVW"
ret += int64(t1 & (-1))
// arm64:"MOVW"
ret += int64(int32(x & 0x80000000))
// arm64:"MOVW"
ret += int64(int32(x & 0x1100000011111111))
// arm64:"MOVH"
ret += int64(int16(x & 0x1100000000001111))
// arm64:"MOVB"
ret += int64(int8(x & 0x1100000000000011))
return ret
}
func noIntermediateExtension(a, b, c uint32) uint32 {
// arm64:-"MOVWU"
return a*b*9 + c
}