Write LZW output 8 bytes at a time, not 1

The benchmarks are mixed, but I think it's still a net win.

----
name                                    old speed     new speed     delta

wuffs_gif_decode_1k_bw/clang            258MB/s ± 0%  253MB/s ± 1%    -2.04%  (p=0.008 n=5+5)
wuffs_gif_decode_1k_color/clang         159MB/s ± 0%  116MB/s ± 0%   -26.76%  (p=0.008 n=5+5)
wuffs_gif_decode_10k/clang              167MB/s ± 0%  138MB/s ± 0%   -17.42%  (p=0.008 n=5+5)
wuffs_gif_decode_100k_artificial/clang  205MB/s ± 0%  439MB/s ± 0%  +114.53%  (p=0.008 n=5+5)
wuffs_gif_decode_100k_realistic/clang   151MB/s ± 0%  164MB/s ± 0%    +8.41%  (p=0.008 n=5+5)
wuffs_gif_decode_1000k/clang            155MB/s ± 0%  165MB/s ± 0%    +6.50%  (p=0.008 n=5+5)
wuffs_gif_decode_anim_screencap/clang   255MB/s ± 0%  921MB/s ± 0%  +261.50%  (p=0.008 n=5+5)

wuffs_gif_decode_1k_bw/gcc              316MB/s ± 2%  267MB/s ± 1%   -15.78%  (p=0.008 n=5+5)
wuffs_gif_decode_1k_color/gcc           184MB/s ± 1%  134MB/s ± 0%   -27.57%  (p=0.008 n=5+5)
wuffs_gif_decode_10k/gcc                196MB/s ± 0%  161MB/s ± 0%   -17.50%  (p=0.008 n=5+5)
wuffs_gif_decode_100k_artificial/gcc    287MB/s ± 0%  449MB/s ± 1%   +56.22%  (p=0.008 n=5+5)
wuffs_gif_decode_100k_realistic/gcc     182MB/s ± 0%  182MB/s ± 1%      ~     (p=0.690 n=5+5)
wuffs_gif_decode_1000k/gcc              187MB/s ± 1%  185MB/s ± 1%      ~     (p=0.056 n=5+5)
wuffs_gif_decode_anim_screencap/gcc     397MB/s ± 0%  973MB/s ± 0%  +145.42%  (p=0.008 n=5+5)

wuffs_lzw_decode_20k/clang              170MB/s ± 0%  191MB/s ± 1%   +12.52%  (p=0.008 n=5+5)
wuffs_lzw_decode_100k/clang             242MB/s ± 0%  312MB/s ± 2%   +29.00%  (p=0.008 n=5+5)

wuffs_lzw_decode_20k/gcc                220MB/s ± 0%  227MB/s ± 0%    +3.31%  (p=0.008 n=5+5)
wuffs_lzw_decode_100k/gcc               317MB/s ± 0%  369MB/s ± 1%   +16.53%  (p=0.008 n=5+5)
----

LZW is the compression method used by GIF images. Roughly speaking, LZW
is a sequence of codes, and each code emits either 1 byte (a literal
code) or multiple bytes (a copy code). The speed gain in this commit
comes from, when emitting an N-byte copy code, this now requires N/8
iterations instead of N iterations, as each iteration now writes 8 bytes
instead of 1. Literal codes, both before and after this commit, requires
a single iteration.

Decoding should therefore get faster or slower depending on whether it's
copy-code-heavy or literal-code-heavy respectively. For what it's worth,
copy-code-heavy GIF images also usually compress better.

Running script/extract-giflzw.go (or examining the cached output at
test/data/analysis/giflzw-emission-lengths.txt) on the anim_screencap
(which got faster) and 1k_color (which got slower) cases (also known as
gifplayer-muybridge.gif and hippopotamus.regular.gif) shows that the
proportion of literal codes agrees with that explanation.

----
gifplayer-muybridge.gif (all frames)
 Percent   Count  Emission_Length_Bucket
   12.78   34658           1 (Literal Code)
   14.05   38106           2
   13.25   35937           3
   10.59   28707           4
    7.85   21286           5
    6.08   16481           6
    4.85   13156           7
    3.60    9764           8
    9.34   25318    9 ..  16
    4.89   13257   17 ..  32
    4.78   12954   33 ..  64
    4.72   12806   65 .. 128
    3.02    8190  127 .. 256
    0.20     550  256+

hippopotamus.regular.gif (all frames)
 Percent   Count  Emission_Length_Bucket
   82.66     672           1 (Literal Code)
   13.04     106           2
    2.83      23           3
    0.86       7           4
    0.37       3           5
    0.25       2           6
    0.00       0           7
    0.00       0           8
    0.00       0    9 ..  16
    0.00       0   17 ..  32
    0.00       0   33 ..  64
    0.00       0   65 .. 128
    0.00       0  127 .. 256
    0.00       0  256+
----

The new algorithm also adds more CPU and RAM overhead, regardless of
whether the input GIF image's LZW data is copy-code-heavy or
literal-code-heavy. For example, the suffix table grows by 8x, from 4KiB
to 32 KiB.

The gifplayer-muybridge.gif image is also larger (472 x 298) and the
hippopotamus.regular.gif image is also smaller (36 x 28). Higher fixed
costs may have larger (negative) relative effects on decoding smaller
images.

To repeat, I think it's still a net win.

An upcoming commit will update the std/lzw/README.md file to provide
more detail on the algorithm.

Thanks to Tushar Khandelwal for the idea.
This commit is contained in:
Nigel Tao committed 2018-10-21 14:48:40 +11:00
1 parent 8d7bd36179
commit 3338d6d1bc
7 files changed
+292 -67

No files matched your search

+59 -1
View File
@@ -22,7 +22,10 @@ import (
t "github.com/google/wuffs/lang/token"
)
var errNoSuchBuiltin = errors.New("cgen: internal error: no such built-in")
var (
errNoSuchBuiltin = errors.New("cgen: internal error: no such built-in")
errOptimizationNotApplicable = errors.New("cgen: internal error: optimization not applicable")
)
func (g *gen) writeBuiltinCall(b *buffer, n *a.Expr, rp replacementPolicy, depth uint32) error {
if n.Operator() != t.IDOpenParen {
@@ -323,6 +326,10 @@ func (g *gen) writeBuiltinNumType(b *buffer, recv *a.Expr, method t.ID, args []*
func (g *gen) writeBuiltinSlice(b *buffer, recv *a.Expr, method t.ID, args []*a.Node, rp replacementPolicy, depth uint32) error {
switch method {
case t.IDCopyFromSlice:
if err := g.writeBuiltinSliceCopyFromSlice8(b, recv, method, args, rp, depth); err != errOptimizationNotApplicable {
return err
}
// TODO: don't assume that the slice is a slice of base.u8.
b.writes("wuffs_base__slice_u8__copy_from_slice(")
if err := g.writeExpr(b, recv, rp, depth); err != nil {
@@ -351,6 +358,57 @@ func (g *gen) writeBuiltinSlice(b *buffer, recv *a.Expr, method t.ID, args []*a.
return errNoSuchBuiltin
}
// writeBuiltinSliceCopyFromSlice8 writes an optimized version of:
//
// foo[fIndex:fIndex + 8].copy_from_slice!(s:bar[bIndex:bIndex + 8])
func (g *gen) writeBuiltinSliceCopyFromSlice8(b *buffer, recv *a.Expr, method t.ID, args []*a.Node, rp replacementPolicy, depth uint32) error {
if method != t.IDCopyFromSlice || len(args) != 1 {
return errOptimizationNotApplicable
}
foo, fIndex := matchFooIndexIndexPlus8(recv)
bar, bIndex := matchFooIndexIndexPlus8(args[0].AsArg().Value())
if foo == nil || bar == nil {
return errOptimizationNotApplicable
}
b.writes("memcpy((")
if err := g.writeExpr(b, foo, rp, depth); err != nil {
return err
}
b.writes(")+(")
if err := g.writeExpr(b, fIndex, rp, depth); err != nil {
return err
}
b.writes("),(")
if err := g.writeExpr(b, bar, rp, depth); err != nil {
return err
}
b.writes(")+(")
if err := g.writeExpr(b, bIndex, rp, depth); err != nil {
return err
}
// TODO: don't assume that the slice is a slice of base.u8.
b.writes("), 8)")
return nil
}
// matchFooIndexIndexPlus8 matches n with "foo[index:index + 8]". It returns
// nil values if there isn't a match.
func matchFooIndexIndexPlus8(n *a.Expr) (foo *a.Expr, index *a.Expr) {
if n.Operator() != t.IDColon {
return nil, nil
}
foo = n.LHS().AsExpr()
index = n.MHS().AsExpr()
rhs := n.RHS().AsExpr()
if index == nil || rhs == nil || rhs.Operator() != t.IDXBinaryPlus || !rhs.LHS().AsExpr().Eq(index) {
return nil, nil
}
if cv := rhs.RHS().AsExpr().ConstValue(); cv == nil || cv.Cmp(eight) != 0 {
return nil, nil
}
return foo, index
}
func (g *gen) writeBuiltinTable(b *buffer, recv *a.Expr, method t.ID, args []*a.Node, rp replacementPolicy, depth uint32) error {
field := ""
+5 -2
View File
@@ -40,8 +40,11 @@ import (
)
var (
zero = big.NewInt(0)
one = big.NewInt(1)
zero = big.NewInt(0)
one = big.NewInt(1)
eight = big.NewInt(8)
mibi = big.NewInt(1 << 20)
)
// Prefixes are prepended to names to form a namespace and to avoid e.g.
+36 -21
View File
@@ -98,26 +98,43 @@ func (g *gen) writeStatementAssign(b *buffer, n *a.Assign, depth uint32) error {
if err := g.writeSuspendibles(b, n.RHS(), depth); err != nil {
return err
}
opName, tilde := "", false
opName, closer := "", ""
op := n.Operator()
switch op {
case t.IDTildeSatPlusEq, t.IDTildeSatMinusEq:
uBits := uintBits(n.LHS().MType().QID())
if uBits == 0 {
return fmt.Errorf("unsupported tilde-operator type %q", n.LHS().MType().Str(g.tm))
}
uOp := "add"
if op != t.IDTildeSatPlusEq {
uOp = "sub"
}
b.printf("wuffs_base__u%d__sat_%s_indirect(&", uBits, uOp)
opName, tilde = ",", true
if lTyp := n.LHS().MType(); lTyp.IsNumType() || lTyp.IsBool() || lTyp.IsSliceType() {
op := n.Operator()
switch op {
case t.IDTildeSatPlusEq, t.IDTildeSatMinusEq:
uBits := uintBits(n.LHS().MType().QID())
if uBits == 0 {
return fmt.Errorf("unsupported tilde-operator type %q", lTyp.Str(g.tm))
}
uOp := "add"
if op != t.IDTildeSatPlusEq {
uOp = "sub"
}
b.printf("wuffs_base__u%d__sat_%s_indirect(&", uBits, uOp)
opName, closer = ",", ")"
default:
opName = cOpName(op)
if opName == "" {
return fmt.Errorf("unrecognized operator %q", op.AmbiguousForm().Str(g.tm))
default:
opName = cOpName(op)
if opName == "" {
return fmt.Errorf("unrecognized operator %q", op.AmbiguousForm().Str(g.tm))
}
}
} else {
if lTyp.IsArrayType() {
if nElem := lTyp.ArrayLength().ConstValue(); nElem.Sign() > 0 && nElem.Cmp(mibi) <= 0 {
if elemTyp := lTyp.Inner(); elemTyp.IsNumType() {
if nBits := uintBits(elemTyp.QID()); nBits != 0 {
b.writes("memcpy(")
opName, closer = ",", fmt.Sprintf(", %d)", nElem.Int64()*int64(nBits/8))
}
}
}
}
if closer == "" {
return fmt.Errorf("unsupported assignment type %q", lTyp.Str(g.tm))
}
}
@@ -128,9 +145,7 @@ func (g *gen) writeStatementAssign(b *buffer, n *a.Assign, depth uint32) error {
if err := g.writeExpr(b, n.RHS(), replaceCallSuspendibles, depth); err != nil {
return err
}
if tilde {
b.writeb(')')
}
b.writes(closer)
b.writes(";\n")
return nil
}
+61
View File
@@ -35,6 +35,8 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryGreaterThan, xb, xa); err != nil {
return err
}
_ = xa
_ = xb
return nil
}},
@@ -55,6 +57,9 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryLessThan, xc, xb); err != nil {
return err
}
_ = xa
_ = xb
_ = xc
return nil
}},
@@ -75,6 +80,9 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryEqEq, xc, xb); err != nil {
return err
}
_ = xa
_ = xb
_ = xc
return nil
}},
@@ -95,6 +103,9 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryLessThan, xc, xb); err != nil {
return err
}
_ = xa
_ = xb
_ = xc
return nil
}},
@@ -115,6 +126,9 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryLessEq, xc, xb); err != nil {
return err
}
_ = xa
_ = xb
_ = xc
return nil
}},
@@ -135,6 +149,9 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryLessThan, xc, xb); err != nil {
return err
}
_ = xa
_ = xb
_ = xc
return nil
}},
@@ -147,6 +164,8 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryLessThan, xb, xa); err != nil {
return err
}
_ = xa
_ = xb
return nil
}},
@@ -159,6 +178,8 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryGreaterEq, xb, xa); err != nil {
return err
}
_ = xa
_ = xb
return nil
}},
@@ -179,6 +200,9 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryLessEq, xc, xb); err != nil {
return err
}
_ = xa
_ = xb
_ = xc
return nil
}},
@@ -199,6 +223,9 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryEqEq, xc, xb); err != nil {
return err
}
_ = xa
_ = xb
_ = xc
return nil
}},
@@ -219,6 +246,9 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryLessEq, xc, xb); err != nil {
return err
}
_ = xa
_ = xb
_ = xc
return nil
}},
@@ -231,6 +261,8 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryLessEq, xb, xa); err != nil {
return err
}
_ = xa
_ = xb
return nil
}},
@@ -251,6 +283,9 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryLessEq, zeroExpr, xb); err != nil {
return err
}
_ = xa
_ = xb
_ = xc
return nil
}},
@@ -284,6 +319,29 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryLessEq, xc0, xc); err != nil {
return err
}
_ = xa
_ = xb
_ = xb0
_ = xc
_ = xc0
return nil
}},
{`"a <= (a + b): 0 <= b"`, func(q *checker, n *a.Assert) error {
op, xa, t0 := parseBinaryOp(n.Condition())
if op != t.IDXBinaryLessEq {
return errFailed
}
op, xa, xb := parseBinaryOp(t0)
if op != t.IDXBinaryPlus {
return errFailed
}
// 0 <= b
if err := proveReasonRequirement(q, t.IDXBinaryLessEq, zeroExpr, xb); err != nil {
return err
}
_ = xa
_ = xb
return nil
}},
@@ -301,6 +359,9 @@ var reasons = [...]struct {
if err := proveReasonRequirement(q, t.IDXBinaryLessEq, xa, t1); err != nil {
return err
}
_ = xa
_ = xb
_ = xc
return nil
}},
}
+14
View File
@@ -26,6 +26,7 @@ import (
"go/format"
"io/ioutil"
"os"
"sort"
"strings"
)
@@ -49,6 +50,8 @@ var reasons = [...]string{
"a < (b + c): a < c; 0 <= b",
"a < (b + c): a < (b0 + c0); b0 <= b; c0 <= c",
"a <= (a + b): 0 <= b",
"(a + b) <= c: a <= (c - b)",
}
@@ -109,6 +112,17 @@ func gen(reason string) error {
}
}
{
namesAsList := []string(nil)
for name := range names {
namesAsList = append(namesAsList, name)
}
sort.Strings(namesAsList)
for _, name := range namesAsList {
fmt.Fprintf(&out, "_ = x%s\n", name)
}
}
fmt.Fprintf(&out, "return nil }},\n")
return nil
}
+68 -20
View File
@@ -2569,10 +2569,10 @@ struct wuffs_lzw__decoder__struct {
uint32_t f_literal_width;
uint32_t f_flush_j;
uint8_t f_suffixes[4096];
uint8_t f_suffixes[4096][8];
uint16_t f_prefixes[4096];
uint16_t f_prefix_lengths[4096];
uint8_t f_output[8192];
uint8_t f_output[8199];
struct {
uint32_t coro_susp_point;
@@ -2586,8 +2586,12 @@ struct wuffs_lzw__decoder__struct {
uint32_t v_bits;
uint32_t v_n_bits;
uint32_t v_code;
uint16_t v_lm1_a;
uint32_t v_c;
uint32_t v_o;
uint32_t v_steps;
uint8_t v_first_byte;
uint16_t v_lm1_b;
} c_decode[1];
struct {
uint32_t coro_susp_point;
@@ -9507,8 +9511,12 @@ wuffs_lzw__decoder__decode(wuffs_lzw__decoder* self,
uint32_t v_bits;
uint32_t v_n_bits;
uint32_t v_code;
uint16_t v_lm1_a;
uint32_t v_c;
uint32_t v_o;
uint32_t v_steps;
uint8_t v_first_byte;
uint16_t v_lm1_b;
uint8_t* iop_a_src = NULL;
uint8_t* io0_a_src = NULL;
@@ -9539,8 +9547,12 @@ wuffs_lzw__decoder__decode(wuffs_lzw__decoder* self,
v_bits = self->private_impl.c_decode[0].v_bits;
v_n_bits = self->private_impl.c_decode[0].v_n_bits;
v_code = self->private_impl.c_decode[0].v_code;
v_lm1_a = self->private_impl.c_decode[0].v_lm1_a;
v_c = self->private_impl.c_decode[0].v_c;
v_o = self->private_impl.c_decode[0].v_o;
v_steps = self->private_impl.c_decode[0].v_steps;
v_first_byte = self->private_impl.c_decode[0].v_first_byte;
v_lm1_b = self->private_impl.c_decode[0].v_lm1_b;
} else {
}
switch (coro_susp_point) {
@@ -9553,11 +9565,12 @@ wuffs_lzw__decoder__decode(wuffs_lzw__decoder* self,
v_clear_code = (((uint32_t)(1)) << v_literal_width);
v_end_code = (v_clear_code + 1);
v_save_code = v_end_code;
v_prev_code = 0;
v_prev_code = v_end_code;
v_width = (v_literal_width + 1);
v_j = 0;
while (v_j < v_clear_code) {
self->private_impl.f_prefix_lengths[v_j] = 0;
self->private_impl.f_suffixes[v_j][0] = ((uint8_t)(v_j));
v_j += 1;
}
v_j = 0;
@@ -9604,11 +9617,21 @@ wuffs_lzw__decoder__decode(wuffs_lzw__decoder* self,
self->private_impl.f_output[v_j] = ((uint8_t)(v_code));
v_j = ((v_j + 1) & 8191);
if (v_save_code <= 4095) {
self->private_impl.f_suffixes[v_save_code] = ((uint8_t)(v_code));
self->private_impl.f_prefixes[v_save_code] =
((uint16_t)(v_prev_code));
self->private_impl.f_prefix_lengths[v_save_code] =
v_lm1_a =
((self->private_impl.f_prefix_lengths[v_prev_code] + 1) & 4095);
self->private_impl.f_prefix_lengths[v_save_code] = v_lm1_a;
if ((v_lm1_a % 8) != 0) {
self->private_impl.f_prefixes[v_save_code] =
self->private_impl.f_prefixes[v_prev_code];
memcpy(self->private_impl.f_suffixes[v_save_code],
self->private_impl.f_suffixes[v_prev_code], 8);
self->private_impl.f_suffixes[v_save_code][(v_lm1_a % 8)] =
((uint8_t)(v_code));
} else {
self->private_impl.f_prefixes[v_save_code] =
((uint16_t)(v_prev_code));
self->private_impl.f_suffixes[v_save_code][0] = ((uint8_t)(v_code));
}
v_save_code += 1;
if ((v_save_code == (((uint32_t)(1)) << v_width)) && (v_width < 12)) {
v_width += 1;
@@ -9620,32 +9643,53 @@ wuffs_lzw__decoder__decode(wuffs_lzw__decoder* self,
goto label_1_break;
}
v_save_code = v_end_code;
v_prev_code = 0;
v_prev_code = v_end_code;
v_width = (v_literal_width + 1);
} else if (v_code <= v_save_code) {
v_c = v_code;
if (v_code == v_save_code) {
v_c = v_prev_code;
}
v_o = ((v_j + ((uint32_t)(self->private_impl.f_prefix_lengths[v_c]))) &
v_o = ((v_j + (((uint32_t)(self->private_impl.f_prefix_lengths[v_c])) &
4294967288)) &
8191);
v_j = ((v_o + 1) & 8191);
while (v_c >= v_clear_code) {
self->private_impl.f_output[v_o] = self->private_impl.f_suffixes[v_c];
v_o = ((v_o - 1) & 8191);
v_j = ((v_j + 1 +
((uint32_t)(self->private_impl.f_prefix_lengths[v_c]))) &
8191);
v_steps = (((uint32_t)(self->private_impl.f_prefix_lengths[v_c])) >> 3);
while (true) {
memcpy((self->private_impl.f_output) + (v_o),
(self->private_impl.f_suffixes[v_c]) + (0), 8);
if (v_steps <= 0) {
goto label_2_break;
}
v_steps -= 1;
v_o = ((v_o - 8) & 8191);
v_c = ((uint32_t)(self->private_impl.f_prefixes[v_c]));
}
self->private_impl.f_output[v_o] = ((uint8_t)(v_c));
label_2_break:;
v_first_byte = self->private_impl.f_suffixes[v_c][0];
if (v_code == v_save_code) {
self->private_impl.f_output[v_j] = ((uint8_t)(v_c));
self->private_impl.f_output[v_j] = v_first_byte;
v_j = ((v_j + 1) & 8191);
}
if (v_save_code <= 4095) {
self->private_impl.f_suffixes[v_save_code] = ((uint8_t)(v_c));
self->private_impl.f_prefixes[v_save_code] =
((uint16_t)(v_prev_code));
self->private_impl.f_prefix_lengths[v_save_code] =
v_lm1_b =
((self->private_impl.f_prefix_lengths[v_prev_code] + 1) & 4095);
self->private_impl.f_prefix_lengths[v_save_code] = v_lm1_b;
if ((v_lm1_b % 8) != 0) {
self->private_impl.f_prefixes[v_save_code] =
self->private_impl.f_prefixes[v_prev_code];
memcpy(self->private_impl.f_suffixes[v_save_code],
self->private_impl.f_suffixes[v_prev_code], 8);
self->private_impl.f_suffixes[v_save_code][(v_lm1_b % 8)] =
v_first_byte;
} else {
self->private_impl.f_prefixes[v_save_code] =
((uint16_t)(v_prev_code));
self->private_impl.f_suffixes[v_save_code][0] =
((uint8_t)(v_first_byte));
}
v_save_code += 1;
if ((v_save_code == (((uint32_t)(1)) << v_width)) && (v_width < 12)) {
v_width += 1;
@@ -9715,8 +9759,12 @@ suspend:
self->private_impl.c_decode[0].v_bits = v_bits;
self->private_impl.c_decode[0].v_n_bits = v_n_bits;
self->private_impl.c_decode[0].v_code = v_code;
self->private_impl.c_decode[0].v_lm1_a = v_lm1_a;
self->private_impl.c_decode[0].v_c = v_c;
self->private_impl.c_decode[0].v_o = v_o;
self->private_impl.c_decode[0].v_steps = v_steps;
self->private_impl.c_decode[0].v_first_byte = v_first_byte;
self->private_impl.c_decode[0].v_lm1_b = v_lm1_b;
goto exit;
exit:
@@ -9785,7 +9833,7 @@ wuffs_lzw__decoder__flush(wuffs_lzw__decoder* self,
v_s = wuffs_base__slice_u8__subslice_ij(
((wuffs_base__slice_u8){
.ptr = self->private_impl.f_output,
.len = 8192,
.len = 8199,
}),
v_i, v_j);
v_n = wuffs_base__io_writer__copy_from_slice(&iop_a_dst, io1_a_dst, v_s);
+49 -23
View File
@@ -19,10 +19,10 @@ pri status "?internal error: inconsistent I/O"
pub struct decoder?(
literal_width base.u32[..8],
flush_j base.u32[..8191],
suffixes array[4096] base.u8,
suffixes array[4096] array[8] base.u8,
prefixes array[4096] base.u16[..4095],
prefix_lengths array[4096] base.u16[..4095],
output array[8192] base.u8,
prefix_lengths array[4096] base.u16[..4095], // TODO: rename this field.
output array[8192 + 7] base.u8,
)
pub func decoder.set_literal_width!(lw base.u32[2..8]) {
@@ -44,7 +44,7 @@ pub func decoder.decode!??(dst base.io_writer, src base.io_reader) {
// next back-reference expands to. The README.md file also calls this value
// `max`. 4096 means do not save.
var save_code base.u32[..4096] = end_code
var prev_code base.u32[..4095]
var prev_code base.u32[..4095] = end_code
var width base.u32[..12] = literal_width + 1
// First, use the j variable to iterate over prefix_lengths.
@@ -52,6 +52,7 @@ pub func decoder.decode!??(dst base.io_writer, src base.io_reader) {
while j < clear_code {
assert j < 256 via "a < b: a < c; c <= b"(c:clear_code)
this.prefix_lengths[j] = 0
this.suffixes[j][0] = j as base.u8
j += 1
}
// For the rest of this function, j is such that output[:j] accumulates the
@@ -109,9 +110,18 @@ pub func decoder.decode!??(dst base.io_writer, src base.io_reader) {
this.output[j] = code as base.u8
j = (j + 1) & 8191
if save_code <= 4095 {
this.suffixes[save_code] = code as base.u8
this.prefixes[save_code] = prev_code as base.u16
this.prefix_lengths[save_code] = (this.prefix_lengths[prev_code] + 1) & 4095
var lm1_a base.u16[..4095] = (this.prefix_lengths[prev_code] + 1) & 4095
this.prefix_lengths[save_code] = lm1_a
if (lm1_a % 8) != 0 {
this.prefixes[save_code] = this.prefixes[prev_code]
this.suffixes[save_code] = this.suffixes[prev_code]
this.suffixes[save_code][lm1_a % 8] = code as base.u8
} else {
this.prefixes[save_code] = prev_code as base.u16
this.suffixes[save_code][0] = code as base.u8
}
save_code += 1
if (save_code == ((1 as base.u32) << width)) and (width < 12) {
width += 1
@@ -124,8 +134,9 @@ pub func decoder.decode!??(dst base.io_writer, src base.io_reader) {
break
}
save_code = end_code
prev_code = 0
prev_code = end_code
width = literal_width + 1
} else if code <= save_code {
var c base.u32[..4095] = code
if code == save_code {
@@ -134,35 +145,50 @@ pub func decoder.decode!??(dst base.io_writer, src base.io_reader) {
// Letting old_j and new_j denote the values of j before and after
// these two lines of code, the decoded bytes will be written to
// output[old_j:new_j]. They will be written back-to-front,
// starting by writing output[o], also known as output[new_j - 1].
// output[old_j:new_j]. They will be written back-to-front, 8 bytes
// at a time, starting by writing output[o:o + 8], which will
// contain output[new_j - 1].
//
// In the special case that code == save_code, the decoded bytes
// contain an extra copy (at the end) of the first byte, and will
// be written to output[old_j:new_j + 1].
var o base.u32[..8191] = (j + (this.prefix_lengths[c] as base.u32)) & 8191
j = (o + 1) & 8191
var o base.u32[..8191] = (j + ((this.prefix_lengths[c] as base.u32) & 0xFFFFFFF8)) & 8191
j = (j + 1 + (this.prefix_lengths[c] as base.u32)) & 8191
while c >= clear_code,
post c < 256 via "a < b: a < c; c <= b"(c:clear_code),
{
this.output[o] = this.suffixes[c]
// This line is essentially "o -= 1". The "& 8191" is a no-op
var steps base.u32 = (this.prefix_lengths[c] as base.u32) >> 3
while true {
assert o <= (o + 8) via "a <= (a + b): 0 <= b"(b:8)
this.output[o:o + 8].copy_from_slice!(s:this.suffixes[c][0:0 + 8])
if steps <= 0 {
break
}
steps -= 1
// This line is essentially "o -= 8". The "& 8191" is a no-op
// in practice, but is necessary for the overflow checker.
o = (o ~mod- 1) & 8191
o = (o ~mod- 8) & 8191
c = this.prefixes[c] as base.u32
}
this.output[o] = c as base.u8
var first_byte base.u8 = this.suffixes[c][0]
if code == save_code {
this.output[j] = c as base.u8
this.output[j] = first_byte
j = (j + 1) & 8191
}
if save_code <= 4095 {
this.suffixes[save_code] = c as base.u8
this.prefixes[save_code] = prev_code as base.u16
this.prefix_lengths[save_code] = (this.prefix_lengths[prev_code] + 1) & 4095
var lm1_b base.u16[..4095] = (this.prefix_lengths[prev_code] + 1) & 4095
this.prefix_lengths[save_code] = lm1_b
if (lm1_b % 8) != 0 {
this.prefixes[save_code] = this.prefixes[prev_code]
this.suffixes[save_code] = this.suffixes[prev_code]
this.suffixes[save_code][lm1_b % 8] = first_byte
} else {
this.prefixes[save_code] = prev_code as base.u16
this.suffixes[save_code][0] = first_byte as base.u8
}
save_code += 1
if (save_code == ((1 as base.u32) << width)) and (width < 12) {
width += 1