</p>

<p>
For instance, some architectures provide a "fused multiply and add" (FMA) instruction
that computes <code>x*y + z</code> without rounding the intermediate result <code>x*y</code>.
These examples show when a Go implementation can use that instruction:
</p>

<pre>
// FMA allowed for computing r, because x*y is not explicitly rounded:
r  = x*y + z
r  = z;   r += x*y

// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

// This file encapsulates some of the odd characteristics of the
// s390x instruction set, to minimize its interaction
// with the core of the assembler.

package arch

import (
	"cmd/internal/obj/s390x"
)

func jumpS390x(word string) bool {
	switch word {
	case "BRC",
		"BC",

// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

// This file encapsulates some of the odd characteristics of the
// 64-bit PowerPC (PPC64) instruction set, to minimize its interaction
// with the core of the assembler.

package arch

import (
	"cmd/internal/obj"
	"cmd/internal/obj/ppc64"
)

func jumpPPC64(word string) bool {
	switch word {

		for _, test := range cat.tests {
			parser.allowABI = cat.allowABI
			parser.errorCount = 0
			parser.lineNum++
			if !parser.pseudo(test.pseudo, tokenize(test.operands)) {
				t.Fatalf("Wrong pseudo-instruction: %s", test.pseudo)
			}
			errorLine := buf.String()
			if test.expected != errorLine {
				t.Errorf("Unexpected error %q; expected %q", errorLine, test.expected)
			}
			buf.Reset()
		}
	}

	MOVB	$255, 4096(R4)        // ebff40000152
	MOVB	$-128, -524288(R5)    // eb8050008052
	MOVB	$1, -524289(R6)       // c0a1fff7ffff41aa60009201a000

	// RX (12-bit displacement) and RXY (20-bit displacement) instruction encoding (e.g: ST vs STY)
	MOVW	R1, 4095(R2)(R3)       // 50132fff
	MOVW	R1, 4096(R2)(R3)       // e31320000150
	MOVWZ	R1, 4095(R2)(R3)       // 50132fff
	MOVWZ	R1, 4096(R2)(R3)       // e31320000150

	// limit (8KB), we stop growing the source string once the limit
	// is reached and keep reusing the same source string - that
	// should therefore be always resident in the L1 cache - until we
	// have completed the construction of the result.
	// This yields significant speedups (up to +100%) in cases where
	// the result length is large (roughly, over L2 cache size).
	const chunkLimit = 8 * 1024
	chunkMax := n
	if chunkMax > chunkLimit {

	MOVWU	X5, (X6)			// ERROR "unsupported unsigned store"
	MOVF	F0, F1, F2			// ERROR "illegal MOV instruction"
	MOVD	F0, F1, F2			// ERROR "illegal MOV instruction"
	MOV	X10, X11, X12			// ERROR "illegal MOV instruction"
	MOVW	X10, X11, X12			// ERROR "illegal MOV instruction"
	RORI	$64, X5, X6			// ERROR "immediate out of range 0 to 63"
	SLLI	$64, X5, X6			// ERROR "immediate out of range 0 to 63"

// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

// This file encapsulates some of the odd characteristics of the ARM
// instruction set, to minimize its interaction with the core of the
// assembler.

package arch

import (
	"strings"

	"cmd/internal/obj"
	"cmd/internal/obj/arm"
)

var armLS = map[string]uint8{

// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

// This file encapsulates some of the odd characteristics of the
// Loong64 (LoongArch64) instruction set, to minimize its interaction
// with the core of the assembler.

package arch

import (
	"cmd/internal/obj"
	"cmd/internal/obj/loong64"
)

func jumpLoong64(word string) bool {
	switch word {

Instead, the compiler operates on a kind of semi-abstract instruction set,
and instruction selection occurs partly after code generation.
The assembler works on the semi-abstract form, so
when you see an instruction like <code>MOV</code>
what the toolchain actually generates for that operation might
not be a move instruction at all, perhaps a clear or load.
Or it might correspond exactly to the machine instruction with that name.