diff --git a/huff0/_generate/gen.go b/huff0/_generate/gen.go
index 485eed6472..6eab53ff4f 100644
--- a/huff0/_generate/gen.go
+++ b/huff0/_generate/gen.go
@@ -49,7 +49,7 @@ func (d decompress4x) generateProcedure(name string) {
 	exhausted := GP64()
 	XORQ(exhausted.As64(), exhausted.As64()) // exhausted = false
 
-	limitPtr := AllocLocal(8)
+	limit := GP64()
 
 	bufferOrigin := GP64()
 	peekBits := GP64()
@@ -57,25 +57,17 @@ func (d decompress4x) generateProcedure(name string) {
 	dstEvery := GP64()
 	table := GP64()
 
-	br0 := GP64()
-	br1 := GP64()
-	br2 := GP64()
-	br3 := GP64()
+	br := GP64()
 
 	Comment("Preload values")
 	{
 		ctx := Dereference(Param("ctx"))
 		Load(ctx.Field("peekBits"), peekBits)
-		Load(ctx.Field("out"), buffer)
-		MOVQ(buffer, bufferOrigin)
-		limit := Load(ctx.Field("limit"), GP64())
-		MOVQ(limit, limitPtr)
+		Load(ctx.Field("out"), bufferOrigin)
+		Load(ctx.Field("limit"), limit)
 		Load(ctx.Field("dstEvery"), dstEvery)
 		Load(ctx.Field("tbl"), table)
-		Load(ctx.Field("pbr0"), br0)
-		Load(ctx.Field("pbr1"), br1)
-		Load(ctx.Field("pbr2"), br2)
-		Load(ctx.Field("pbr3"), br3)
+		Load(ctx.Field("pbr"), br)
 	}
 
 	Comment("Main loop")
@@ -83,15 +75,15 @@ func (d decompress4x) generateProcedure(name string) {
 
 	MOVQ(bufferOrigin, buffer)
 	// Check if we have space
-	CMPQ(buffer, limitPtr)
+	CMPQ(buffer, limit)
 	SETGE(exhausted.As8())
-	d.decodeTwoValues(0, br0, peekBits, table, buffer, exhausted)
+	d.decodeTwoValues(0, br, peekBits, table, buffer, exhausted)
 	ADDQ(dstEvery, buffer)
-	d.decodeTwoValues(1, br1, peekBits, table, buffer, exhausted)
+	d.decodeTwoValues(1, br, peekBits, table, buffer, exhausted)
 	ADDQ(dstEvery, buffer)
-	d.decodeTwoValues(2, br2, peekBits, table, buffer, exhausted)
+	d.decodeTwoValues(2, br, peekBits, table, buffer, exhausted)
 	ADDQ(dstEvery, buffer)
-	d.decodeTwoValues(3, br3, peekBits, table, buffer, exhausted)
+	d.decodeTwoValues(3, br, peekBits, table, buffer, exhausted)
 
 	ADDQ(U8(2), bufferOrigin) // off += 2
 
@@ -100,10 +92,9 @@ func (d decompress4x) generateProcedure(name string) {
 
 	{
 		ctx := Dereference(Param("ctx"))
-		tmp := Load(ctx.Field("out"), GP64())
-		decoded := GP64()
-		MOVQ(bufferOrigin, decoded)
-		SUBQ(tmp, decoded)
+		ctxout, _ := ctx.Field("out").Resolve()
+		decoded := bufferOrigin
+		SUBQ(ctxout.Addr, decoded)
 		SHLQ(U8(2), decoded) // decoded *= 4
 
 		Store(decoded, ctx.Field("decoded"))
@@ -118,6 +109,7 @@ const bitReader_in = 0
 const bitReader_off = bitReader_in + 3*8 // {ptr, len, cap}
 const bitReader_value = bitReader_off + 8
 const bitReader_bitsRead = bitReader_value + 8
+const bitReader__size = bitReader_bitsRead + 8
 
 func (d decompress4x) decodeTwoValues(id int, br, peekBits, table, buffer, exhausted reg.GPVirtual) {
 	brValue, brBitsRead := d.fillFast32(id, 32, br, exhausted)
@@ -157,9 +149,10 @@ func (d decompress4x) decodeTwoValues(id int, br, peekBits, table, buffer, exhau
 	Comment("out[id * dstEvery + 1] = uint8(v1.entry >> 8)")
 	MOVW(out.As16(), Mem{Base: buffer})
 
-	Comment("update the bitrader reader structure")
-	MOVQ(brValue, Mem{Base: br, Disp: bitReader_value})
-	MOVB(brBitsRead.As8(), Mem{Base: br, Disp: bitReader_bitsRead})
+	Comment("update the bitreader structure")
+	offset := id * bitReader__size
+	MOVQ(brValue, Mem{Base: br, Disp: offset + bitReader_value})
+	MOVB(brBitsRead.As8(), Mem{Base: br, Disp: offset + bitReader_bitsRead})
 }
 
 func (d decompress4x) generateProcedure4x8bit(name string) {
@@ -171,33 +164,25 @@ func (d decompress4x) generateProcedure4x8bit(name string) {
 	exhausted := GP64()                      // Fixed since we need 8H
 	XORQ(exhausted.As64(), exhausted.As64()) // exhausted = false
 
-	bufferOrigin := AllocLocal(8)
-	limitPtr := AllocLocal(8)
+	bufferOrigin := GP64()
+	limit := GP64()
 
 	peekBits := GP64()
 	buffer := GP64()
 	dstEvery := GP64()
 	table := GP64()
 
-	br0 := GP64()
-	br1 := GP64()
-	br2 := GP64()
-	br3 := GP64()
+	br := GP64()
 
 	Comment("Preload values")
 	{
 		ctx := Dereference(Param("ctx"))
 		Load(ctx.Field("peekBits"), peekBits)
-		Load(ctx.Field("out"), buffer)
-		MOVQ(buffer, bufferOrigin)
-		limit := Load(ctx.Field("limit"), GP64())
-		MOVQ(limit, limitPtr)
+		Load(ctx.Field("out"), bufferOrigin)
+		Load(ctx.Field("limit"), limit)
 		Load(ctx.Field("dstEvery"), dstEvery)
 		Load(ctx.Field("tbl"), table)
-		Load(ctx.Field("pbr0"), br0)
-		Load(ctx.Field("pbr1"), br1)
-		Load(ctx.Field("pbr2"), br2)
-		Load(ctx.Field("pbr3"), br3)
+		Load(ctx.Field("pbr"), br)
 	}
 
 	Comment("Main loop")
@@ -205,15 +190,15 @@ func (d decompress4x) generateProcedure4x8bit(name string) {
 
 	MOVQ(bufferOrigin, buffer)
 	// Check if we have space
-	CMPQ(buffer, limitPtr)
+	CMPQ(buffer, limit)
 	SETGE(exhausted.As8())
-	d.decodeFourValues(0, br0, peekBits, table, buffer, exhausted)
+	d.decodeFourValues(0, br, peekBits, table, buffer, exhausted)
 	ADDQ(dstEvery, buffer)
-	d.decodeFourValues(1, br1, peekBits, table, buffer, exhausted)
+	d.decodeFourValues(1, br, peekBits, table, buffer, exhausted)
 	ADDQ(dstEvery, buffer)
-	d.decodeFourValues(2, br2, peekBits, table, buffer, exhausted)
+	d.decodeFourValues(2, br, peekBits, table, buffer, exhausted)
 	ADDQ(dstEvery, buffer)
-	d.decodeFourValues(3, br3, peekBits, table, buffer, exhausted)
+	d.decodeFourValues(3, br, peekBits, table, buffer, exhausted)
 
 	ADDQ(U8(4), bufferOrigin) // off += 4
 
@@ -222,10 +207,9 @@ func (d decompress4x) generateProcedure4x8bit(name string) {
 
 	{
 		ctx := Dereference(Param("ctx"))
-		tmp := Load(ctx.Field("out"), GP64())
-		decoded := GP64()
-		MOVQ(bufferOrigin, decoded)
-		SUBQ(tmp, decoded)
+		ctxout, _ := ctx.Field("out").Resolve()
+		decoded := bufferOrigin
+		SUBQ(ctxout.Addr, decoded)
 		SHLQ(U8(2), decoded) // decoded *= 4
 
 		Store(decoded, ctx.Field("decoded"))
@@ -234,7 +218,7 @@ func (d decompress4x) generateProcedure4x8bit(name string) {
 }
 
 func (d decompress4x) decodeFourValues(id int, br, peekBits, table, buffer, exhausted reg.GPVirtual) {
-	brValue, brBitsRead := d.fillFast32(id+1000, 32, br, exhausted)
+	brValue, brBitsRead := d.fillFast32(id, 32, br, exhausted)
 
 	decompress := func(valID int, outByte reg.Register) {
 		CX := reg.CL
@@ -269,9 +253,10 @@ func (d decompress4x) decodeFourValues(id int, br, peekBits, table, buffer, exha
 	Comment("out[id * dstEvery + 4] = uint8(v3.entry >> 8)")
 	MOVL(out.As32(), Mem{Base: buffer})
 
-	Comment("update the bitreader reader structure")
-	MOVQ(brValue, Mem{Base: br, Disp: bitReader_value})
-	MOVB(brBitsRead.As8(), Mem{Base: br, Disp: bitReader_bitsRead})
+	Comment("update the bitreader structure")
+	offset := id * bitReader__size
+	MOVQ(brValue, Mem{Base: br, Disp: offset + bitReader_value})
+	MOVB(brBitsRead.As8(), Mem{Base: br, Disp: offset + bitReader_bitsRead})
 }
 
 func (d decompress4x) fillFast32(id, atLeast int, br, exhausted reg.GPVirtual) (brValue, brBitsRead reg.GPVirtual) {
@@ -281,14 +266,15 @@ func (d decompress4x) fillFast32(id, atLeast int, br, exhausted reg.GPVirtual) (
 	Commentf("br%d.fillFast32()", id)
 	brValue = GP64()
 	brBitsRead = GP64()
-	MOVQ(Mem{Base: br, Disp: bitReader_value}, brValue)
-	MOVBQZX(Mem{Base: br, Disp: bitReader_bitsRead}, brBitsRead)
+	offset := bitReader__size * id
+	MOVQ(Mem{Base: br, Disp: offset + bitReader_value}, brValue)
+	MOVBQZX(Mem{Base: br, Disp: offset + bitReader_bitsRead}, brBitsRead)
 
 	// We must have at least 2 * max tablelog left
 	CMPQ(brBitsRead, U8(64-atLeast))
 	JBE(LabelRef("skip_fill" + strconv.Itoa(id)))
 	brOffset := GP64()
-	MOVQ(Mem{Base: br, Disp: bitReader_off}, brOffset)
+	MOVQ(Mem{Base: br, Disp: offset + bitReader_off}, brOffset)
 
 	SUBQ(U8(32), brBitsRead) // b.bitsRead -= 32
 	SUBQ(U8(4), brOffset)    // b.off -= 4
@@ -297,7 +283,7 @@ func (d decompress4x) fillFast32(id, atLeast int, br, exhausted reg.GPVirtual) (
 	// v = v[:4]
 	// low := (uint32(v[0])) | (uint32(v[1]) << 8) | (uint32(v[2]) << 16) | (uint32(v[3]) << 24)
 	tmp := GP64()
-	MOVQ(Mem{Base: br, Disp: bitReader_in}, tmp)
+	MOVQ(Mem{Base: br, Disp: offset + bitReader_in}, tmp)
 
 	Comment("b.value |= uint64(low) << (b.bitsRead & 63)")
 	addr := Mem{Base: brOffset, Index: tmp.As64(), Scale: 1}
@@ -306,7 +292,7 @@ func (d decompress4x) fillFast32(id, atLeast int, br, exhausted reg.GPVirtual) (
 	MOVQ(brBitsRead, CX.As64())
 	SHLQ(CX, tmp.As64())
 
-	MOVQ(brOffset, Mem{Base: br, Disp: bitReader_off})
+	MOVQ(brOffset, Mem{Base: br, Disp: offset + bitReader_off})
 	ORQ(tmp.As64(), brValue)
 	{
 		Commentf("exhausted = exhausted || (br%d.off < 4)", id)
@@ -474,11 +460,9 @@ func (d decompress1x) generateProcedure(name string) {
 	{
 		// calculate decoded as current `out` - initial `out`
 		ctx := Dereference(Param("ctx"))
-		decoded := GP64()
-		tmp := GP64()
-		MOVQ(buffer, decoded)
-		Load(ctx.Field("out"), tmp)
-		SUBQ(tmp, decoded)
+		ctxout, _ := ctx.Field("out").Resolve()
+		decoded := buffer
+		SUBQ(ctxout.Addr, decoded)
 		Store(decoded, ctx.Field("decoded"))
 
 		pbr := Dereference(ctx.Field("pbr"))
diff --git a/huff0/decompress_amd64.go b/huff0/decompress_amd64.go
index 671e630a84..9f3e9f79e2 100644
--- a/huff0/decompress_amd64.go
+++ b/huff0/decompress_amd64.go
@@ -27,10 +27,7 @@ func decompress4x_8b_main_loop_amd64(ctx *decompress4xContext)
 const fallback8BitSize = 800
 
 type decompress4xContext struct {
-	pbr0     *bitReaderShifted
-	pbr1     *bitReaderShifted
-	pbr2     *bitReaderShifted
-	pbr3     *bitReaderShifted
+	pbr      *[4]bitReaderShifted
 	peekBits uint8
 	out      *byte
 	dstEvery int
@@ -89,10 +86,7 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 
 	if len(out) > 4*4 && !(br[0].off < 4 || br[1].off < 4 || br[2].off < 4 || br[3].off < 4) {
 		ctx := decompress4xContext{
-			pbr0:     &br[0],
-			pbr1:     &br[1],
-			pbr2:     &br[2],
-			pbr3:     &br[3],
+			pbr:      &br,
 			peekBits: uint8((64 - d.actualTableLog) & 63), // see: bitReaderShifted.peekBitsFast()
 			out:      &out[0],
 			dstEvery: dstEvery,
diff --git a/huff0/decompress_amd64.s b/huff0/decompress_amd64.s
index 6c65c6e2b2..dd1a5aecd6 100644
--- a/huff0/decompress_amd64.s
+++ b/huff0/decompress_amd64.s
@@ -4,45 +4,40 @@
 // +build amd64,!appengine,!noasm,gc
 
 // func decompress4x_main_loop_amd64(ctx *decompress4xContext)
-TEXT ·decompress4x_main_loop_amd64(SB), $8-8
+TEXT ·decompress4x_main_loop_amd64(SB), $0-8
 	XORQ DX, DX
 
 	// Preload values
 	MOVQ    ctx+0(FP), AX
-	MOVBQZX 32(AX), SI
-	MOVQ    40(AX), DI
-	MOVQ    DI, BX
-	MOVQ    72(AX), CX
-	MOVQ    CX, (SP)
-	MOVQ    48(AX), R8
-	MOVQ    56(AX), R9
-	MOVQ    (AX), R10
-	MOVQ    8(AX), R11
-	MOVQ    16(AX), R12
-	MOVQ    24(AX), R13
+	MOVBQZX 8(AX), DI
+	MOVQ    16(AX), SI
+	MOVQ    48(AX), BX
+	MOVQ    24(AX), R9
+	MOVQ    32(AX), R10
+	MOVQ    (AX), R11
 
 	// Main loop
 main_loop:
-	MOVQ  BX, DI
-	CMPQ  DI, (SP)
+	MOVQ  SI, R8
+	CMPQ  R8, BX
 	SETGE DL
 
 	// br0.fillFast32()
-	MOVQ    32(R10), R14
-	MOVBQZX 40(R10), R15
-	CMPQ    R15, $0x20
+	MOVQ    32(R11), R12
+	MOVBQZX 40(R11), R13
+	CMPQ    R13, $0x20
 	JBE     skip_fill0
-	MOVQ    24(R10), AX
-	SUBQ    $0x20, R15
+	MOVQ    24(R11), AX
+	SUBQ    $0x20, R13
 	SUBQ    $0x04, AX
-	MOVQ    (R10), BP
+	MOVQ    (R11), R14
 
 	// b.value |= uint64(low) << (b.bitsRead & 63)
-	MOVL (AX)(BP*1), BP
-	MOVQ R15, CX
-	SHLQ CL, BP
-	MOVQ AX, 24(R10)
-	ORQ  BP, R14
+	MOVL (AX)(R14*1), R14
+	MOVQ R13, CX
+	SHLQ CL, R14
+	MOVQ AX, 24(R11)
+	ORQ  R14, R12
 
 	// exhausted = exhausted || (br0.off < 4)
 	CMPQ  AX, $0x04
@@ -51,57 +46,57 @@ main_loop:
 
 skip_fill0:
 	// val0 := br0.peekTopBits(peekBits)
-	MOVQ R14, BP
-	MOVQ SI, CX
-	SHRQ CL, BP
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v0 := table[val0&mask]
-	MOVW (R9)(BP*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br0.advance(uint8(v0.entry)
 	MOVB CH, AL
-	SHLQ CL, R14
-	ADDB CL, R15
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val1 := br0.peekTopBits(peekBits)
-	MOVQ SI, CX
-	MOVQ R14, BP
-	SHRQ CL, BP
+	MOVQ DI, CX
+	MOVQ R12, R14
+	SHRQ CL, R14
 
 	// v1 := table[val1&mask]
-	MOVW (R9)(BP*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br0.advance(uint8(v1.entry))
 	MOVB CH, AH
-	SHLQ CL, R14
-	ADDB CL, R15
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// these two writes get coalesced
 	// out[id * dstEvery + 0] = uint8(v0.entry >> 8)
 	// out[id * dstEvery + 1] = uint8(v1.entry >> 8)
-	MOVW AX, (DI)
+	MOVW AX, (R8)
 
-	// update the bitrader reader structure
-	MOVQ R14, 32(R10)
-	MOVB R15, 40(R10)
-	ADDQ R8, DI
+	// update the bitreader structure
+	MOVQ R12, 32(R11)
+	MOVB R13, 40(R11)
+	ADDQ R9, R8
 
 	// br1.fillFast32()
-	MOVQ    32(R11), R14
-	MOVBQZX 40(R11), R15
-	CMPQ    R15, $0x20
+	MOVQ    80(R11), R12
+	MOVBQZX 88(R11), R13
+	CMPQ    R13, $0x20
 	JBE     skip_fill1
-	MOVQ    24(R11), AX
-	SUBQ    $0x20, R15
+	MOVQ    72(R11), AX
+	SUBQ    $0x20, R13
 	SUBQ    $0x04, AX
-	MOVQ    (R11), BP
+	MOVQ    48(R11), R14
 
 	// b.value |= uint64(low) << (b.bitsRead & 63)
-	MOVL (AX)(BP*1), BP
-	MOVQ R15, CX
-	SHLQ CL, BP
-	MOVQ AX, 24(R11)
-	ORQ  BP, R14
+	MOVL (AX)(R14*1), R14
+	MOVQ R13, CX
+	SHLQ CL, R14
+	MOVQ AX, 72(R11)
+	ORQ  R14, R12
 
 	// exhausted = exhausted || (br1.off < 4)
 	CMPQ  AX, $0x04
@@ -110,57 +105,57 @@ skip_fill0:
 
 skip_fill1:
 	// val0 := br1.peekTopBits(peekBits)
-	MOVQ R14, BP
-	MOVQ SI, CX
-	SHRQ CL, BP
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v0 := table[val0&mask]
-	MOVW (R9)(BP*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br1.advance(uint8(v0.entry)
 	MOVB CH, AL
-	SHLQ CL, R14
-	ADDB CL, R15
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val1 := br1.peekTopBits(peekBits)
-	MOVQ SI, CX
-	MOVQ R14, BP
-	SHRQ CL, BP
+	MOVQ DI, CX
+	MOVQ R12, R14
+	SHRQ CL, R14
 
 	// v1 := table[val1&mask]
-	MOVW (R9)(BP*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br1.advance(uint8(v1.entry))
 	MOVB CH, AH
-	SHLQ CL, R14
-	ADDB CL, R15
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// these two writes get coalesced
 	// out[id * dstEvery + 0] = uint8(v0.entry >> 8)
 	// out[id * dstEvery + 1] = uint8(v1.entry >> 8)
-	MOVW AX, (DI)
+	MOVW AX, (R8)
 
-	// update the bitrader reader structure
-	MOVQ R14, 32(R11)
-	MOVB R15, 40(R11)
-	ADDQ R8, DI
+	// update the bitreader structure
+	MOVQ R12, 80(R11)
+	MOVB R13, 88(R11)
+	ADDQ R9, R8
 
 	// br2.fillFast32()
-	MOVQ    32(R12), R14
-	MOVBQZX 40(R12), R15
-	CMPQ    R15, $0x20
+	MOVQ    128(R11), R12
+	MOVBQZX 136(R11), R13
+	CMPQ    R13, $0x20
 	JBE     skip_fill2
-	MOVQ    24(R12), AX
-	SUBQ    $0x20, R15
+	MOVQ    120(R11), AX
+	SUBQ    $0x20, R13
 	SUBQ    $0x04, AX
-	MOVQ    (R12), BP
+	MOVQ    96(R11), R14
 
 	// b.value |= uint64(low) << (b.bitsRead & 63)
-	MOVL (AX)(BP*1), BP
-	MOVQ R15, CX
-	SHLQ CL, BP
-	MOVQ AX, 24(R12)
-	ORQ  BP, R14
+	MOVL (AX)(R14*1), R14
+	MOVQ R13, CX
+	SHLQ CL, R14
+	MOVQ AX, 120(R11)
+	ORQ  R14, R12
 
 	// exhausted = exhausted || (br2.off < 4)
 	CMPQ  AX, $0x04
@@ -169,57 +164,57 @@ skip_fill1:
 
 skip_fill2:
 	// val0 := br2.peekTopBits(peekBits)
-	MOVQ R14, BP
-	MOVQ SI, CX
-	SHRQ CL, BP
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v0 := table[val0&mask]
-	MOVW (R9)(BP*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br2.advance(uint8(v0.entry)
 	MOVB CH, AL
-	SHLQ CL, R14
-	ADDB CL, R15
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val1 := br2.peekTopBits(peekBits)
-	MOVQ SI, CX
-	MOVQ R14, BP
-	SHRQ CL, BP
+	MOVQ DI, CX
+	MOVQ R12, R14
+	SHRQ CL, R14
 
 	// v1 := table[val1&mask]
-	MOVW (R9)(BP*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br2.advance(uint8(v1.entry))
 	MOVB CH, AH
-	SHLQ CL, R14
-	ADDB CL, R15
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// these two writes get coalesced
 	// out[id * dstEvery + 0] = uint8(v0.entry >> 8)
 	// out[id * dstEvery + 1] = uint8(v1.entry >> 8)
-	MOVW AX, (DI)
+	MOVW AX, (R8)
 
-	// update the bitrader reader structure
-	MOVQ R14, 32(R12)
-	MOVB R15, 40(R12)
-	ADDQ R8, DI
+	// update the bitreader structure
+	MOVQ R12, 128(R11)
+	MOVB R13, 136(R11)
+	ADDQ R9, R8
 
 	// br3.fillFast32()
-	MOVQ    32(R13), R14
-	MOVBQZX 40(R13), R15
-	CMPQ    R15, $0x20
+	MOVQ    176(R11), R12
+	MOVBQZX 184(R11), R13
+	CMPQ    R13, $0x20
 	JBE     skip_fill3
-	MOVQ    24(R13), AX
-	SUBQ    $0x20, R15
+	MOVQ    168(R11), AX
+	SUBQ    $0x20, R13
 	SUBQ    $0x04, AX
-	MOVQ    (R13), BP
+	MOVQ    144(R11), R14
 
 	// b.value |= uint64(low) << (b.bitsRead & 63)
-	MOVL (AX)(BP*1), BP
-	MOVQ R15, CX
-	SHLQ CL, BP
-	MOVQ AX, 24(R13)
-	ORQ  BP, R14
+	MOVL (AX)(R14*1), R14
+	MOVQ R13, CX
+	SHLQ CL, R14
+	MOVQ AX, 168(R11)
+	ORQ  R14, R12
 
 	// exhausted = exhausted || (br3.off < 4)
 	CMPQ  AX, $0x04
@@ -228,149 +223,142 @@ skip_fill2:
 
 skip_fill3:
 	// val0 := br3.peekTopBits(peekBits)
-	MOVQ R14, BP
-	MOVQ SI, CX
-	SHRQ CL, BP
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v0 := table[val0&mask]
-	MOVW (R9)(BP*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br3.advance(uint8(v0.entry)
 	MOVB CH, AL
-	SHLQ CL, R14
-	ADDB CL, R15
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val1 := br3.peekTopBits(peekBits)
-	MOVQ SI, CX
-	MOVQ R14, BP
-	SHRQ CL, BP
+	MOVQ DI, CX
+	MOVQ R12, R14
+	SHRQ CL, R14
 
 	// v1 := table[val1&mask]
-	MOVW (R9)(BP*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br3.advance(uint8(v1.entry))
 	MOVB CH, AH
-	SHLQ CL, R14
-	ADDB CL, R15
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// these two writes get coalesced
 	// out[id * dstEvery + 0] = uint8(v0.entry >> 8)
 	// out[id * dstEvery + 1] = uint8(v1.entry >> 8)
-	MOVW AX, (DI)
+	MOVW AX, (R8)
 
-	// update the bitrader reader structure
-	MOVQ  R14, 32(R13)
-	MOVB  R15, 40(R13)
-	ADDQ  $0x02, BX
+	// update the bitreader structure
+	MOVQ  R12, 176(R11)
+	MOVB  R13, 184(R11)
+	ADDQ  $0x02, SI
 	TESTB DL, DL
 	JZ    main_loop
 	MOVQ  ctx+0(FP), AX
-	MOVQ  40(AX), CX
-	MOVQ  BX, DX
-	SUBQ  CX, DX
-	SHLQ  $0x02, DX
-	MOVQ  DX, 64(AX)
+	SUBQ  16(AX), SI
+	SHLQ  $0x02, SI
+	MOVQ  SI, 40(AX)
 	RET
 
 // func decompress4x_8b_main_loop_amd64(ctx *decompress4xContext)
-TEXT ·decompress4x_8b_main_loop_amd64(SB), $16-8
+TEXT ·decompress4x_8b_main_loop_amd64(SB), $0-8
 	XORQ DX, DX
 
 	// Preload values
 	MOVQ    ctx+0(FP), CX
-	MOVBQZX 32(CX), BX
-	MOVQ    40(CX), SI
-	MOVQ    SI, (SP)
-	MOVQ    72(CX), DX
-	MOVQ    DX, 8(SP)
-	MOVQ    48(CX), DI
-	MOVQ    56(CX), R8
-	MOVQ    (CX), R9
-	MOVQ    8(CX), R10
-	MOVQ    16(CX), R11
-	MOVQ    24(CX), R12
+	MOVBQZX 8(CX), DI
+	MOVQ    16(CX), BX
+	MOVQ    48(CX), SI
+	MOVQ    24(CX), R9
+	MOVQ    32(CX), R10
+	MOVQ    (CX), R11
 
 	// Main loop
 main_loop:
-	MOVQ  (SP), SI
-	CMPQ  SI, 8(SP)
+	MOVQ  BX, R8
+	CMPQ  R8, SI
 	SETGE DL
 
-	// br1000.fillFast32()
-	MOVQ    32(R9), R13
-	MOVBQZX 40(R9), R14
-	CMPQ    R14, $0x20
-	JBE     skip_fill1000
-	MOVQ    24(R9), R15
-	SUBQ    $0x20, R14
-	SUBQ    $0x04, R15
-	MOVQ    (R9), BP
+	// br0.fillFast32()
+	MOVQ    32(R11), R12
+	MOVBQZX 40(R11), R13
+	CMPQ    R13, $0x20
+	JBE     skip_fill0
+	MOVQ    24(R11), R14
+	SUBQ    $0x20, R13
+	SUBQ    $0x04, R14
+	MOVQ    (R11), R15
 
 	// b.value |= uint64(low) << (b.bitsRead & 63)
-	MOVL (R15)(BP*1), BP
-	MOVQ R14, CX
-	SHLQ CL, BP
-	MOVQ R15, 24(R9)
-	ORQ  BP, R13
-
-	// exhausted = exhausted || (br1000.off < 4)
-	CMPQ  R15, $0x04
+	MOVL (R14)(R15*1), R15
+	MOVQ R13, CX
+	SHLQ CL, R15
+	MOVQ R14, 24(R11)
+	ORQ  R15, R12
+
+	// exhausted = exhausted || (br0.off < 4)
+	CMPQ  R14, $0x04
 	SETLT AL
 	ORB   AL, DL
 
-skip_fill1000:
+skip_fill0:
 	// val0 := br0.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v0 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br0.advance(uint8(v0.entry)
 	MOVB CH, AL
-	SHLQ CL, R13
-	ADDB CL, R14
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val1 := br0.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v1 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br0.advance(uint8(v1.entry)
 	MOVB   CH, AH
-	SHLQ   CL, R13
-	ADDB   CL, R14
+	SHLQ   CL, R12
+	ADDB   CL, R13
 	BSWAPL AX
 
 	// val2 := br0.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v2 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br0.advance(uint8(v2.entry)
 	MOVB CH, AH
-	SHLQ CL, R13
-	ADDB CL, R14
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val3 := br0.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v3 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br0.advance(uint8(v3.entry)
 	MOVB   CH, AL
-	SHLQ   CL, R13
-	ADDB   CL, R14
+	SHLQ   CL, R12
+	ADDB   CL, R13
 	BSWAPL AX
 
 	// these four writes get coalesced
@@ -378,88 +366,88 @@ skip_fill1000:
 	// out[id * dstEvery + 1] = uint8(v1.entry >> 8)
 	// out[id * dstEvery + 3] = uint8(v2.entry >> 8)
 	// out[id * dstEvery + 4] = uint8(v3.entry >> 8)
-	MOVL AX, (SI)
-
-	// update the bitreader reader structure
-	MOVQ R13, 32(R9)
-	MOVB R14, 40(R9)
-	ADDQ DI, SI
-
-	// br1001.fillFast32()
-	MOVQ    32(R10), R13
-	MOVBQZX 40(R10), R14
-	CMPQ    R14, $0x20
-	JBE     skip_fill1001
-	MOVQ    24(R10), R15
-	SUBQ    $0x20, R14
-	SUBQ    $0x04, R15
-	MOVQ    (R10), BP
+	MOVL AX, (R8)
+
+	// update the bitreader structure
+	MOVQ R12, 32(R11)
+	MOVB R13, 40(R11)
+	ADDQ R9, R8
+
+	// br1.fillFast32()
+	MOVQ    80(R11), R12
+	MOVBQZX 88(R11), R13
+	CMPQ    R13, $0x20
+	JBE     skip_fill1
+	MOVQ    72(R11), R14
+	SUBQ    $0x20, R13
+	SUBQ    $0x04, R14
+	MOVQ    48(R11), R15
 
 	// b.value |= uint64(low) << (b.bitsRead & 63)
-	MOVL (R15)(BP*1), BP
-	MOVQ R14, CX
-	SHLQ CL, BP
-	MOVQ R15, 24(R10)
-	ORQ  BP, R13
-
-	// exhausted = exhausted || (br1001.off < 4)
-	CMPQ  R15, $0x04
+	MOVL (R14)(R15*1), R15
+	MOVQ R13, CX
+	SHLQ CL, R15
+	MOVQ R14, 72(R11)
+	ORQ  R15, R12
+
+	// exhausted = exhausted || (br1.off < 4)
+	CMPQ  R14, $0x04
 	SETLT AL
 	ORB   AL, DL
 
-skip_fill1001:
+skip_fill1:
 	// val0 := br1.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v0 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br1.advance(uint8(v0.entry)
 	MOVB CH, AL
-	SHLQ CL, R13
-	ADDB CL, R14
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val1 := br1.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v1 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br1.advance(uint8(v1.entry)
 	MOVB   CH, AH
-	SHLQ   CL, R13
-	ADDB   CL, R14
+	SHLQ   CL, R12
+	ADDB   CL, R13
 	BSWAPL AX
 
 	// val2 := br1.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v2 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br1.advance(uint8(v2.entry)
 	MOVB CH, AH
-	SHLQ CL, R13
-	ADDB CL, R14
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val3 := br1.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v3 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br1.advance(uint8(v3.entry)
 	MOVB   CH, AL
-	SHLQ   CL, R13
-	ADDB   CL, R14
+	SHLQ   CL, R12
+	ADDB   CL, R13
 	BSWAPL AX
 
 	// these four writes get coalesced
@@ -467,88 +455,88 @@ skip_fill1001:
 	// out[id * dstEvery + 1] = uint8(v1.entry >> 8)
 	// out[id * dstEvery + 3] = uint8(v2.entry >> 8)
 	// out[id * dstEvery + 4] = uint8(v3.entry >> 8)
-	MOVL AX, (SI)
-
-	// update the bitreader reader structure
-	MOVQ R13, 32(R10)
-	MOVB R14, 40(R10)
-	ADDQ DI, SI
-
-	// br1002.fillFast32()
-	MOVQ    32(R11), R13
-	MOVBQZX 40(R11), R14
-	CMPQ    R14, $0x20
-	JBE     skip_fill1002
-	MOVQ    24(R11), R15
-	SUBQ    $0x20, R14
-	SUBQ    $0x04, R15
-	MOVQ    (R11), BP
+	MOVL AX, (R8)
+
+	// update the bitreader structure
+	MOVQ R12, 80(R11)
+	MOVB R13, 88(R11)
+	ADDQ R9, R8
+
+	// br2.fillFast32()
+	MOVQ    128(R11), R12
+	MOVBQZX 136(R11), R13
+	CMPQ    R13, $0x20
+	JBE     skip_fill2
+	MOVQ    120(R11), R14
+	SUBQ    $0x20, R13
+	SUBQ    $0x04, R14
+	MOVQ    96(R11), R15
 
 	// b.value |= uint64(low) << (b.bitsRead & 63)
-	MOVL (R15)(BP*1), BP
-	MOVQ R14, CX
-	SHLQ CL, BP
-	MOVQ R15, 24(R11)
-	ORQ  BP, R13
-
-	// exhausted = exhausted || (br1002.off < 4)
-	CMPQ  R15, $0x04
+	MOVL (R14)(R15*1), R15
+	MOVQ R13, CX
+	SHLQ CL, R15
+	MOVQ R14, 120(R11)
+	ORQ  R15, R12
+
+	// exhausted = exhausted || (br2.off < 4)
+	CMPQ  R14, $0x04
 	SETLT AL
 	ORB   AL, DL
 
-skip_fill1002:
+skip_fill2:
 	// val0 := br2.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v0 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br2.advance(uint8(v0.entry)
 	MOVB CH, AL
-	SHLQ CL, R13
-	ADDB CL, R14
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val1 := br2.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v1 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br2.advance(uint8(v1.entry)
 	MOVB   CH, AH
-	SHLQ   CL, R13
-	ADDB   CL, R14
+	SHLQ   CL, R12
+	ADDB   CL, R13
 	BSWAPL AX
 
 	// val2 := br2.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v2 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br2.advance(uint8(v2.entry)
 	MOVB CH, AH
-	SHLQ CL, R13
-	ADDB CL, R14
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val3 := br2.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v3 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br2.advance(uint8(v3.entry)
 	MOVB   CH, AL
-	SHLQ   CL, R13
-	ADDB   CL, R14
+	SHLQ   CL, R12
+	ADDB   CL, R13
 	BSWAPL AX
 
 	// these four writes get coalesced
@@ -556,88 +544,88 @@ skip_fill1002:
 	// out[id * dstEvery + 1] = uint8(v1.entry >> 8)
 	// out[id * dstEvery + 3] = uint8(v2.entry >> 8)
 	// out[id * dstEvery + 4] = uint8(v3.entry >> 8)
-	MOVL AX, (SI)
-
-	// update the bitreader reader structure
-	MOVQ R13, 32(R11)
-	MOVB R14, 40(R11)
-	ADDQ DI, SI
-
-	// br1003.fillFast32()
-	MOVQ    32(R12), R13
-	MOVBQZX 40(R12), R14
-	CMPQ    R14, $0x20
-	JBE     skip_fill1003
-	MOVQ    24(R12), R15
-	SUBQ    $0x20, R14
-	SUBQ    $0x04, R15
-	MOVQ    (R12), BP
+	MOVL AX, (R8)
+
+	// update the bitreader structure
+	MOVQ R12, 128(R11)
+	MOVB R13, 136(R11)
+	ADDQ R9, R8
+
+	// br3.fillFast32()
+	MOVQ    176(R11), R12
+	MOVBQZX 184(R11), R13
+	CMPQ    R13, $0x20
+	JBE     skip_fill3
+	MOVQ    168(R11), R14
+	SUBQ    $0x20, R13
+	SUBQ    $0x04, R14
+	MOVQ    144(R11), R15
 
 	// b.value |= uint64(low) << (b.bitsRead & 63)
-	MOVL (R15)(BP*1), BP
-	MOVQ R14, CX
-	SHLQ CL, BP
-	MOVQ R15, 24(R12)
-	ORQ  BP, R13
-
-	// exhausted = exhausted || (br1003.off < 4)
-	CMPQ  R15, $0x04
+	MOVL (R14)(R15*1), R15
+	MOVQ R13, CX
+	SHLQ CL, R15
+	MOVQ R14, 168(R11)
+	ORQ  R15, R12
+
+	// exhausted = exhausted || (br3.off < 4)
+	CMPQ  R14, $0x04
 	SETLT AL
 	ORB   AL, DL
 
-skip_fill1003:
+skip_fill3:
 	// val0 := br3.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v0 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br3.advance(uint8(v0.entry)
 	MOVB CH, AL
-	SHLQ CL, R13
-	ADDB CL, R14
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val1 := br3.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v1 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br3.advance(uint8(v1.entry)
 	MOVB   CH, AH
-	SHLQ   CL, R13
-	ADDB   CL, R14
+	SHLQ   CL, R12
+	ADDB   CL, R13
 	BSWAPL AX
 
 	// val2 := br3.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v2 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br3.advance(uint8(v2.entry)
 	MOVB CH, AH
-	SHLQ CL, R13
-	ADDB CL, R14
+	SHLQ CL, R12
+	ADDB CL, R13
 
 	// val3 := br3.peekTopBits(peekBits)
-	MOVQ R13, R15
-	MOVQ BX, CX
-	SHRQ CL, R15
+	MOVQ R12, R14
+	MOVQ DI, CX
+	SHRQ CL, R14
 
 	// v3 := table[val0&mask]
-	MOVW (R8)(R15*2), CX
+	MOVW (R10)(R14*2), CX
 
 	// br3.advance(uint8(v3.entry)
 	MOVB   CH, AL
-	SHLQ   CL, R13
-	ADDB   CL, R14
+	SHLQ   CL, R12
+	ADDB   CL, R13
 	BSWAPL AX
 
 	// these four writes get coalesced
@@ -645,20 +633,18 @@ skip_fill1003:
 	// out[id * dstEvery + 1] = uint8(v1.entry >> 8)
 	// out[id * dstEvery + 3] = uint8(v2.entry >> 8)
 	// out[id * dstEvery + 4] = uint8(v3.entry >> 8)
-	MOVL AX, (SI)
+	MOVL AX, (R8)
 
-	// update the bitreader reader structure
-	MOVQ  R13, 32(R12)
-	MOVB  R14, 40(R12)
-	ADDQ  $0x04, (SP)
+	// update the bitreader structure
+	MOVQ  R12, 176(R11)
+	MOVB  R13, 184(R11)
+	ADDQ  $0x04, BX
 	TESTB DL, DL
 	JZ    main_loop
 	MOVQ  ctx+0(FP), AX
-	MOVQ  40(AX), CX
-	MOVQ  (SP), DX
-	SUBQ  CX, DX
-	SHLQ  $0x02, DX
-	MOVQ  DX, 64(AX)
+	SUBQ  16(AX), BX
+	SHLQ  $0x02, BX
+	MOVQ  BX, 40(AX)
 	RET
 
 // func decompress1x_main_loop_amd64(ctx *decompress1xContext)
@@ -750,10 +736,8 @@ loop_condition:
 
 	// Update ctx structure
 	MOVQ ctx+0(FP), AX
-	MOVQ DX, CX
-	MOVQ 16(AX), DX
-	SUBQ DX, CX
-	MOVQ CX, 40(AX)
+	SUBQ 16(AX), DX
+	MOVQ DX, 40(AX)
 	MOVQ (AX), AX
 	MOVQ R9, 24(AX)
 	MOVQ R10, 32(AX)
@@ -847,10 +831,8 @@ loop_condition:
 
 	// Update ctx structure
 	MOVQ ctx+0(FP), AX
-	MOVQ DX, CX
-	MOVQ 16(AX), DX
-	SUBQ DX, CX
-	MOVQ CX, 40(AX)
+	SUBQ 16(AX), DX
+	MOVQ DX, 40(AX)
 	MOVQ (AX), AX
 	MOVQ R9, 24(AX)
 	MOVQ R10, 32(AX)