Remove more branches in std/png filters

wuffs_png_decode_19k_8bpp/clang9          95.5MB/s ± 0%  91.8MB/s ± 1%   -3.90%  (p=0.008 n=5+5)
wuffs_png_decode_40k_24bpp/clang9          115MB/s ± 0%   115MB/s ± 1%     ~     (p=0.151 n=5+5)
wuffs_png_decode_77k_8bpp/clang9           358MB/s ± 0%   345MB/s ± 0%   -3.68%  (p=0.008 n=5+5)
wuffs_png_decode_552k_32bpp/clang9         124MB/s ± 0%   152MB/s ± 1%  +22.35%  (p=0.008 n=5+5)
wuffs_png_decode_4002k_24bpp/clang9        116MB/s ± 0%   116MB/s ± 0%     ~     (p=0.151 n=5+5)

wuffs_png_decode_filter_1_sub/clang9       924MB/s ± 0%   982MB/s ± 0%   +6.28%  (p=0.008 n=5+5)
wuffs_png_decode_filter_2_up/clang9       13.1GB/s ± 0%  12.7GB/s ± 0%   -2.73%  (p=0.008 n=5+5)
wuffs_png_decode_filter_3_average/clang9   640MB/s ± 0%   757MB/s ± 0%  +18.31%  (p=0.008 n=5+5)
wuffs_png_decode_filter_4_paeth/clang9    64.8MB/s ± 0%  82.1MB/s ± 0%  +26.72%  (p=0.008 n=5+5)

wuffs_png_decode_19k_8bpp/gcc10           99.0MB/s ± 0%  98.9MB/s ± 0%     ~     (p=0.222 n=5+5)
wuffs_png_decode_40k_24bpp/gcc10           114MB/s ± 1%   118MB/s ± 1%   +3.38%  (p=0.008 n=5+5)
wuffs_png_decode_77k_8bpp/gcc10            320MB/s ± 0%   321MB/s ± 0%   +0.38%  (p=0.016 n=5+4)
wuffs_png_decode_552k_32bpp/gcc10          141MB/s ± 0%   144MB/s ± 0%   +1.80%  (p=0.008 n=5+5)
wuffs_png_decode_4002k_24bpp/gcc10         118MB/s ± 0%   118MB/s ± 1%     ~     (p=0.151 n=5+5)

wuffs_png_decode_filter_1_sub/gcc10       1.11GB/s ± 0%  0.90GB/s ± 0%  -18.57%  (p=0.008 n=5+5)
wuffs_png_decode_filter_2_up/gcc10        11.1GB/s ± 0%  11.0GB/s ± 0%   -0.33%  (p=0.008 n=5+5)
wuffs_png_decode_filter_3_average/gcc10    751MB/s ± 0%   749MB/s ± 0%   -0.29%  (p=0.008 n=5+5)
wuffs_png_decode_filter_4_paeth/gcc10     76.9MB/s ± 0%  81.8MB/s ± 0%   +6.33%  (p=0.008 n=5+5)
This commit is contained in:
Nigel Tao committed 2021-01-02 23:07:04 +11:00
1 parent be00ccf61f
commit e403659c05
4 files changed
+156 -80

No files matched your search

+3
View File
@@ -25,6 +25,7 @@ editing this list.
---
- `"a <= b: a == b"`
- `"a <= b: b >= a"`
- `"a <= b: a <= c; c <= b"`
- `"a <= b: a <= c; c == b"`
@@ -32,7 +33,9 @@ editing this list.
---
- `"a >= b: a == b"`
- `"a >= b: b <= a"`
- `"a >= b: a >= (b + c); 0 <= c"`
---
+52
View File
@@ -169,6 +169,20 @@ var reasons = [...]struct {
return nil
}},
{`"a <= b: a == b"`, func(q *checker, n *a.Assert) error {
op, xa, xb := parseBinaryOp(n.Condition())
if op != t.IDXBinaryLessEq {
return errFailed
}
// a == b
if err := proveReasonRequirement(q, t.IDXBinaryEqEq, xa, xb); err != nil {
return err
}
_ = xa
_ = xb
return nil
}},
{`"a <= b: b >= a"`, func(q *checker, n *a.Assert) error {
op, xa, xb := parseBinaryOp(n.Condition())
if op != t.IDXBinaryLessEq {
@@ -252,6 +266,20 @@ var reasons = [...]struct {
return nil
}},
{`"a >= b: a == b"`, func(q *checker, n *a.Assert) error {
op, xa, xb := parseBinaryOp(n.Condition())
if op != t.IDXBinaryGreaterEq {
return errFailed
}
// a == b
if err := proveReasonRequirement(q, t.IDXBinaryEqEq, xa, xb); err != nil {
return err
}
_ = xa
_ = xb
return nil
}},
{`"a >= b: b <= a"`, func(q *checker, n *a.Assert) error {
op, xa, xb := parseBinaryOp(n.Condition())
if op != t.IDXBinaryGreaterEq {
@@ -266,6 +294,30 @@ var reasons = [...]struct {
return nil
}},
{`"a >= b: a >= (b + c); 0 <= c"`, func(q *checker, n *a.Assert) error {
op, xa, xb := parseBinaryOp(n.Condition())
if op != t.IDXBinaryGreaterEq {
return errFailed
}
// a >= (b + c)
xc := argValue(q.tm, n.Args(), "c")
if xc == nil {
return errFailed
}
t0 := a.NewExpr(0, t.IDXBinaryPlus, 0, xb.AsNode(), nil, xc.AsNode(), nil)
if err := proveReasonRequirement(q, t.IDXBinaryGreaterEq, xa, t0); err != nil {
return err
}
// 0 <= c
if err := proveReasonRequirement(q, t.IDXBinaryLessEq, zeroExpr, xc); err != nil {
return err
}
_ = xa
_ = xb
_ = xc
return nil
}},
{`"a < (b + c): a < c; 0 <= b"`, func(q *checker, n *a.Assert) error {
op, xa, t0 := parseBinaryOp(n.Condition())
if op != t.IDXBinaryLessThan {
+33 -35
View File
@@ -30040,22 +30040,19 @@ wuffs_png__decoder__filter_3(
if (((uint64_t)(a_prev.len)) == 0) {
v_i = v_filter_distance;
while (v_i < ((uint64_t)(a_curr.len))) {
if (v_i >= v_filter_distance) {
if ((v_i - v_filter_distance) < ((uint64_t)(a_curr.len))) {
a_curr.ptr[v_i] += (a_curr.ptr[(v_i - v_filter_distance)] / 2);
}
}
a_curr.ptr[v_i] += (a_curr.ptr[(v_i - v_filter_distance)] / 2);
v_i += 1;
}
} else {
v_n = wuffs_base__u64__min(((uint64_t)(a_curr.len)), ((uint64_t)(a_prev.len)));
v_i = 0;
while ((v_i < v_n) && (v_i < v_filter_distance)) {
a_curr.ptr[v_i] += (a_prev.ptr[v_i] / 2);
v_i += 1;
}
v_i = v_filter_distance;
while (v_i < v_n) {
if (v_i >= v_filter_distance) {
a_curr.ptr[v_i] += ((uint8_t)(((((uint32_t)(a_curr.ptr[(v_i - v_filter_distance)])) + ((uint32_t)(a_prev.ptr[v_i]))) / 2)));
} else {
a_curr.ptr[v_i] += (a_prev.ptr[v_i] / 2);
}
a_curr.ptr[v_i] += ((uint8_t)(((((uint32_t)(a_curr.ptr[(v_i - v_filter_distance)])) + ((uint32_t)(a_prev.ptr[v_i]))) / 2)));
v_i += 1;
}
}
@@ -30083,33 +30080,34 @@ wuffs_png__decoder__filter_4(
v_filter_distance = ((uint64_t)(self->private_impl.f_filter_distance));
v_n = wuffs_base__u64__min(((uint64_t)(a_curr.len)), ((uint64_t)(a_prev.len)));
v_i = 0;
while ((v_i < v_n) && (v_i < v_filter_distance)) {
a_curr.ptr[v_i] += a_prev.ptr[v_i];
v_i += 1;
}
v_i = v_filter_distance;
while (v_i < v_n) {
if (v_i < v_filter_distance) {
a_curr.ptr[v_i] += a_prev.ptr[v_i];
v_fa = ((uint32_t)(a_curr.ptr[(v_i - v_filter_distance)]));
v_fb = ((uint32_t)(a_prev.ptr[v_i]));
v_fc = ((uint32_t)(a_prev.ptr[(v_i - v_filter_distance)]));
v_pp = ((v_fa + v_fb) - v_fc);
v_pa = (v_pp - v_fa);
if (v_pa >= 2147483648) {
v_pa = (0 - v_pa);
}
v_pb = (v_pp - v_fb);
if (v_pb >= 2147483648) {
v_pb = (0 - v_pb);
}
v_pc = (v_pp - v_fc);
if (v_pc >= 2147483648) {
v_pc = (0 - v_pc);
}
if ((v_pa <= v_pb) && (v_pa <= v_pc)) {
a_curr.ptr[v_i] += ((uint8_t)((v_fa & 255)));
} else if (v_pb <= v_pc) {
a_curr.ptr[v_i] += ((uint8_t)((v_fb & 255)));
} else {
v_fa = ((uint32_t)(a_curr.ptr[(v_i - v_filter_distance)]));
v_fb = ((uint32_t)(a_prev.ptr[v_i]));
v_fc = ((uint32_t)(a_prev.ptr[(v_i - v_filter_distance)]));
v_pp = ((v_fa + v_fb) - v_fc);
v_pa = (v_pp - v_fa);
if (v_pa >= 2147483648) {
v_pa = (0 - v_pa);
}
v_pb = (v_pp - v_fb);
if (v_pb >= 2147483648) {
v_pb = (0 - v_pb);
}
v_pc = (v_pp - v_fc);
if (v_pc >= 2147483648) {
v_pc = (0 - v_pc);
}
if ((v_pa <= v_pb) && (v_pa <= v_pc)) {
a_curr.ptr[v_i] += ((uint8_t)((v_fa & 255)));
} else if (v_pb <= v_pc) {
a_curr.ptr[v_i] += ((uint8_t)((v_fb & 255)));
} else {
a_curr.ptr[v_i] += ((uint8_t)((v_fc & 255)));
}
a_curr.ptr[v_i] += ((uint8_t)((v_fc & 255)));
}
v_i += 1;
}
+68 -45
View File
@@ -61,35 +61,47 @@ pri func decoder.filter_3!(curr: slice base.u8, prev: slice base.u8) {
filter_distance = this.filter_distance as base.u64
if args.prev.length() == 0 {
i = filter_distance
while i < args.curr.length() {
assert i >= filter_distance via "a >= b: a == b"()
while i < args.curr.length(),
inv i >= filter_distance,
{
assert i < 0xFFFF_FFFF_FFFF_FFFF via "a < b: a < c; c <= b"(c: args.curr.length())
if i >= filter_distance {
if (i - filter_distance) < args.curr.length() {
args.curr[i] ~mod+= args.curr[i - filter_distance] / 2
}
}
assert (i - filter_distance) < args.curr.length() via "(a - b) < c: a < c; 0 <= b"()
args.curr[i] ~mod+= args.curr[i - filter_distance] / 2
i += 1
assert i >= filter_distance via "a >= b: a >= (b + c); 0 <= c"(c: 1)
} endwhile
} else {
n = args.curr.length().min(a: args.prev.length())
i = 0
while i < n,
while (i < n) and (i < filter_distance),
inv n <= args.curr.length(),
inv n <= args.prev.length(),
{
assert i < 0xFFFF_FFFF_FFFF_FFFF via "a < b: a < c; c <= b"(c: n)
assert i < args.curr.length() via "a < b: a < c; c <= b"(c: n)
assert i < args.prev.length() via "a < b: a < c; c <= b"(c: n)
if i >= filter_distance {
assert (i - filter_distance) < args.curr.length() via "(a - b) < c: a < c; 0 <= b"()
args.curr[i] ~mod+= ((
(args.curr[i - filter_distance] as base.u32) +
(args.prev[i] as base.u32)) / 2) as base.u8
} else {
args.curr[i] ~mod+= args.prev[i] / 2
}
args.curr[i] ~mod+= args.prev[i] / 2
i += 1
} endwhile
i = filter_distance
assert i >= filter_distance via "a >= b: a == b"()
while i < n,
inv i >= filter_distance,
inv n <= args.curr.length(),
inv n <= args.prev.length(),
{
assert i < 0xFFFF_FFFF_FFFF_FFFF via "a < b: a < c; c <= b"(c: n)
assert i < args.curr.length() via "a < b: a < c; c <= b"(c: n)
assert i < args.prev.length() via "a < b: a < c; c <= b"(c: n)
assert (i - filter_distance) < args.curr.length() via "(a - b) < c: a < c; 0 <= b"()
args.curr[i] ~mod+= ((
(args.curr[i - filter_distance] as base.u32) +
(args.prev[i] as base.u32)) / 2) as base.u8
i += 1
assert i >= filter_distance via "a >= b: a >= (b + c); 0 <= c"(c: 1)
} endwhile
}
}
@@ -109,42 +121,53 @@ pri func decoder.filter_4!(curr: slice base.u8, prev: slice base.u8) {
filter_distance = this.filter_distance as base.u64
n = args.curr.length().min(a: args.prev.length())
i = 0
while i < n,
while (i < n) and (i < filter_distance),
inv n <= args.curr.length(),
inv n <= args.prev.length(),
{
assert i < 0xFFFF_FFFF_FFFF_FFFF via "a < b: a < c; c <= b"(c: n)
assert i < args.curr.length() via "a < b: a < c; c <= b"(c: n)
assert i < args.prev.length() via "a < b: a < c; c <= b"(c: n)
if i < filter_distance {
args.curr[i] ~mod+= args.prev[i]
} else {
assert (i - filter_distance) < args.curr.length() via "(a - b) < c: a < c; 0 <= b"()
assert (i - filter_distance) < args.prev.length() via "(a - b) < c: a < c; 0 <= b"()
fa = args.curr[i - filter_distance] as base.u32
fb = args.prev[i] as base.u32
fc = args.prev[i - filter_distance] as base.u32
pp = (fa ~mod+ fb) ~mod- fc
pa = pp ~mod- fa
if pa >= 0x8000_0000 {
pa = 0 ~mod- pa
}
pb = pp ~mod- fb
if pb >= 0x8000_0000 {
pb = 0 ~mod- pb
}
pc = pp ~mod- fc
if pc >= 0x8000_0000 {
pc = 0 ~mod- pc
}
if (pa <= pb) and (pa <= pc) {
args.curr[i] ~mod+= (fa & 0xFF) as base.u8
} else if pb <= pc {
args.curr[i] ~mod+= (fb & 0xFF) as base.u8
} else {
args.curr[i] ~mod+= (fc & 0xFF) as base.u8
}
}
args.curr[i] ~mod+= args.prev[i]
i += 1
} endwhile
i = filter_distance
assert i >= filter_distance via "a >= b: a == b"()
while i < n,
inv i >= filter_distance,
inv n <= args.curr.length(),
inv n <= args.prev.length(),
{
assert i < 0xFFFF_FFFF_FFFF_FFFF via "a < b: a < c; c <= b"(c: n)
assert i < args.curr.length() via "a < b: a < c; c <= b"(c: n)
assert i < args.prev.length() via "a < b: a < c; c <= b"(c: n)
assert (i - filter_distance) < args.curr.length() via "(a - b) < c: a < c; 0 <= b"()
assert (i - filter_distance) < args.prev.length() via "(a - b) < c: a < c; 0 <= b"()
fa = args.curr[i - filter_distance] as base.u32
fb = args.prev[i] as base.u32
fc = args.prev[i - filter_distance] as base.u32
pp = (fa ~mod+ fb) ~mod- fc
pa = pp ~mod- fa
if pa >= 0x8000_0000 {
pa = 0 ~mod- pa
}
pb = pp ~mod- fb
if pb >= 0x8000_0000 {
pb = 0 ~mod- pb
}
pc = pp ~mod- fc
if pc >= 0x8000_0000 {
pc = 0 ~mod- pc
}
if (pa <= pb) and (pa <= pc) {
args.curr[i] ~mod+= (fa & 0xFF) as base.u8
} else if pb <= pc {
args.curr[i] ~mod+= (fb & 0xFF) as base.u8
} else {
args.curr[i] ~mod+= (fc & 0xFF) as base.u8
}
i += 1
assert i >= filter_distance via "a >= b: a >= (b + c); 0 <= c"(c: 1)
} endwhile
}