mleku/wazero
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
1,842 Commits 1 Branch 30 Tags
65650d399d105f339f77f4d4b256268a842e5cf5
Commit Graph

18 Commits

Author SHA1 Message Date
Takeshi Yoneda
65650d399d ssa: reuses slices for basicBlock.params (#2247) 
This replaces the basicBlock.params field with the reusable
VarLength[Value] type. As a result, the compilation starts
using less memory and allocations.

```
goos: darwin
goarch: arm64
pkg: github.com/tetratelabs/wazero
                      │  old.txt   │             new.txt              │
                      │   sec/op   │   sec/op    vs base              │
Compilation/wazero-10   2.004 ± 2%   2.001 ± 0%       ~ (p=0.620 n=7)
Compilation/zig-10      4.164 ± 1%   4.174 ± 3%       ~ (p=0.097 n=7)
geomean                 2.888        2.890       +0.06%

                      │   old.txt    │              new.txt               │
                      │     B/op     │     B/op      vs base              │
Compilation/wazero-10   297.7Mi ± 0%   297.5Mi ± 0%  -0.06% (p=0.001 n=7)
Compilation/zig-10      594.0Mi ± 0%   593.9Mi ± 0%  -0.01% (p=0.001 n=7)
geomean                 420.5Mi        420.3Mi       -0.03%

                      │   old.txt   │              new.txt              │
                      │  allocs/op  │  allocs/op   vs base              │
Compilation/wazero-10   472.5k ± 0%   457.1k ± 0%  -3.25% (p=0.001 n=7)
Compilation/zig-10      277.2k ± 0%   275.7k ± 0%  -0.53% (p=0.001 n=7)
geomean                 361.9k        355.0k       -1.90%
```

Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2024-06-13 13:01:58 -07:00
Takeshi Yoneda
5c8366f8d5 ssa: optimizes slice allocations (#2242) 
This makes the compilation faster and use less memory:

```
goos: darwin
goarch: arm64
pkg: github.com/tetratelabs/wazero
                      │  old.txt   │             new.txt              │
                      │   sec/op   │   sec/op    vs base              │
Compilation/wazero-10   2.184 ± 0%   2.110 ± 0%  -3.40% (p=0.001 n=7)
Compilation/zig-10      4.331 ± 1%   4.187 ± 1%  -3.31% (p=0.001 n=7)
geomean                 3.075        2.972       -3.36%

                      │   old.txt    │               new.txt               │
                      │     B/op     │     B/op      vs base               │
Compilation/wazero-10   337.3Mi ± 0%   301.9Mi ± 0%  -10.49% (p=0.001 n=7)
Compilation/zig-10      599.3Mi ± 0%   594.3Mi ± 0%   -0.84% (p=0.001 n=7)
geomean                 449.6Mi        423.6Mi        -5.79%

                      │   old.txt   │              new.txt               │
                      │  allocs/op  │  allocs/op   vs base               │
Compilation/wazero-10   592.9k ± 0%   527.9k ± 0%  -10.97% (p=0.001 n=7)
Compilation/zig-10      287.8k ± 0%   278.6k ± 0%   -3.20% (p=0.001 n=7)
geomean                 413.1k        383.5k        -7.17%
```

Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2024-06-10 08:25:29 -07:00
Takeshi Yoneda
747609b0f5 ssa: removes map use for block traversals (#2235) 
This removes the use of map in basic block traversals.
As a result, overall compilation perf improves like the below:

### Zig
```
goos: darwin
goarch: arm64
pkg: github.com/tetratelabs/wazero/internal/integration_test/stdlibs
                             │ old_zig.txt │            new_zig.txt            │
                             │   sec/op    │   sec/op    vs base               │
Zig/Compile/test-opt.wasm-10    4.438 ± 1%   3.778 ± 0%  -14.87% (p=0.002 n=6)
Zig/Run/test-opt.wasm-10        18.77 ± 1%   18.76 ± 0%        ~ (p=0.818 n=6)
Zig/Compile/test.wasm-10        5.083 ± 0%   4.673 ± 0%   -8.07% (p=0.002 n=6)
Zig/Run/test.wasm-10            19.27 ± 1%   19.30 ± 1%        ~ (p=0.699 n=6)
geomean                         9.504        8.941        -5.92%

                             │ old_zig.txt  │            new_zig.txt             │
                             │     B/op     │     B/op      vs base              │
Zig/Compile/test-opt.wasm-10   396.7Mi ± 0%   394.7Mi ± 0%  -0.51% (p=0.002 n=6)
Zig/Run/test-opt.wasm-10       741.7Mi ± 0%   741.7Mi ± 0%       ~ (p=0.900 n=6)
Zig/Compile/test.wasm-10       660.0Mi ± 0%   659.5Mi ± 0%  -0.08% (p=0.002 n=6)
Zig/Run/test.wasm-10           1.296Gi ± 0%   1.296Gi ± 0%       ~ (p=0.892 n=6)
geomean                        712.6Mi        711.5Mi       -0.15%

                             │ old_zig.txt │            new_zig.txt            │
                             │  allocs/op  │  allocs/op   vs base              │
Zig/Compile/test-opt.wasm-10   363.2k ± 0%   362.6k ± 0%  -0.17% (p=0.002 n=6)
Zig/Run/test-opt.wasm-10       51.58k ± 0%   51.58k ± 0%       ~ (p=0.933 n=6)
Zig/Compile/test.wasm-10       515.2k ± 0%   515.4k ± 0%       ~ (p=0.485 n=6)
Zig/Run/test.wasm-10           2.156M ± 0%   2.156M ± 0%       ~ (p=0.998 n=6)
geomean                        379.8k        379.7k       -0.03%
```

### wasip1

```
goos: darwin
goarch: arm64
pkg: github.com/tetratelabs/wazero/internal/integration_test/stdlibs
                                            │ old_wasip1.txt │            new_wasip1.txt             │
                                            │     sec/op     │    sec/op      vs base                │
Wasip1/Compile/src_archive_tar.test-10            2.198 ± 1%    2.067 ± 1%    -5.96% (p=0.001 n=7)
Wasip1/Run/src_archive_tar.test-10               398.8m ± 0%   398.8m ± 0%         ~ (p=0.902 n=7)
Wasip1/Compile/src_bufio.test-10                  1.492 ± 0%    1.409 ± 1%    -5.57% (p=0.001 n=7)
Wasip1/Run/src_bufio.test-10                     120.5m ± 1%   121.0m ± 1%    +0.44% (p=0.017 n=7)
Wasip1/Compile/src_bytes.test-10                  1.543 ± 0%    1.454 ± 0%    -5.72% (p=0.001 n=7)
Wasip1/Run/src_bytes.test-10                     469.0m ± 1%   467.4m ± 1%         ~ (p=0.209 n=7)
Wasip1/Compile/src_context.test-10                1.664 ± 0%    1.564 ± 1%    -6.00% (p=0.001 n=7)
Wasip1/Run/src_context.test-10                   31.54m ± 1%   31.57m ± 0%         ~ (p=0.445 n=6+7)
Wasip1/Compile/src_encoding_ascii85.test-10                     1.261 ±  ∞ ¹
geomean                                          527.3m        565.9m         -2.92%
¹ need >= 6 samples for confidence interval at level 0.95

                                            │ old_wasip1.txt │             new_wasip1.txt             │
                                            │      B/op      │      B/op       vs base                │
Wasip1/Compile/src_archive_tar.test-10          93.44Mi ± 0%   93.17Mi ± 0%    -0.30% (p=0.001 n=7)
Wasip1/Run/src_archive_tar.test-10              286.0Mi ± 0%   286.0Mi ± 0%         ~ (p=0.593 n=7)
Wasip1/Compile/src_bufio.test-10                74.38Mi ± 0%   74.13Mi ± 0%    -0.35% (p=0.001 n=7)
Wasip1/Run/src_bufio.test-10                    105.3Mi ± 0%   105.3Mi ± 0%         ~ (p=0.780 n=7)
Wasip1/Compile/src_bytes.test-10                75.58Mi ± 0%   75.32Mi ± 0%    -0.35% (p=0.001 n=7)
Wasip1/Run/src_bytes.test-10                    605.0Mi ± 0%   605.0Mi ± 0%         ~ (p=0.331 n=7)
Wasip1/Compile/src_context.test-10              78.33Mi ± 0%   78.07Mi ± 0%    -0.33% (p=0.001 n=7)
Wasip1/Run/src_context.test-10                  71.52Mi ± 0%   71.52Mi ± 0%         ~ (p=1.000 n=6+7)
Wasip1/Compile/src_encoding_ascii85.test-10                    70.38Mi ±  ∞ ¹
geomean                                         123.4Mi        115.7Mi         -0.17%
¹ need >= 6 samples for confidence interval at level 0.95

                                            │ old_wasip1.txt │             new_wasip1.txt              │
                                            │   allocs/op    │   allocs/op    vs base                  │
Wasip1/Compile/src_archive_tar.test-10           265.4k ± 0%   265.0k ± 0%    -0.16% (p=0.001 n=7)
Wasip1/Run/src_archive_tar.test-10               7.831k ± 0%   7.830k ± 0%         ~ (p=1.000 n=7)
Wasip1/Compile/src_bufio.test-10                 195.6k ± 0%   195.4k ± 0%    -0.12% (p=0.001 n=7)
Wasip1/Run/src_bufio.test-10                     3.728k ± 0%   3.728k ± 0%         ~ (p=1.000 n=7)   ¹
Wasip1/Compile/src_bytes.test-10                 204.1k ± 0%   203.7k ± 0%    -0.20% (p=0.001 n=7)
Wasip1/Run/src_bytes.test-10                     6.377k ± 0%   6.377k ± 0%         ~ (p=1.000 n=7)
Wasip1/Compile/src_context.test-10               221.7k ± 0%   221.6k ± 0%    -0.06% (p=0.001 n=7)
Wasip1/Run/src_context.test-10                   3.814k ± 0%   3.814k ± 1%         ~ (p=0.140 n=6+7)
Wasip1/Compile/src_encoding_ascii85.test-10                    182.3k ±  ∞ ²
geomean                                          33.71k        40.64k         -0.07%
¹ all samples are equal
² need >= 6 samples for confidence interval at level 0.95
```


### TinyGo
```
goos: darwin
goarch: arm64
pkg: github.com/tetratelabs/wazero/internal/integration_test/stdlibs
                                      │ old_tinygo.txt │            new_tinygo.txt             │
                                      │     sec/op     │    sec/op      vs base                │
TinyGo/Compile/container_heap.test-10    410.8m ± 1%     399.8m ± 0%    -2.69% (p=0.001 n=7)
TinyGo/Run/container_heap.test-10        14.41m ± 0%     14.29m ± 2%    -0.77% (p=0.026 n=7)
TinyGo/Compile/container_list.test-10    410.5m ± 1%     398.1m ± 0%    -3.02% (p=0.001 n=7)
TinyGo/Run/container_list.test-10        14.27m ± 2%     14.16m ± 1%         ~ (p=0.073 n=7)
TinyGo/Compile/container_ring.test-10    403.7m ± 1%     392.5m ± 2%    -2.77% (p=0.001 n=7)
TinyGo/Run/container_ring.test-10        14.24m ± 0%     14.27m ± 1%         ~ (p=0.259 n=7)
TinyGo/Compile/crypto_des.test-10        418.8m ± 0%     408.1m ± 0%    -2.56% (p=0.001 n=7)
TinyGo/Run/crypto_des.test-10            18.23m ± 0%     18.17m ± 1%         ~ (p=0.456 n=7)
TinyGo/Compile/crypto_md5.test-10        417.3m ± 2%     406.1m ± 1%    -2.68% (p=0.001 n=7)
TinyGo/Run/crypto_md5.test-10            20.50m ± 0%     20.45m ± 1%         ~ (p=0.128 n=7)
TinyGo/Compile/crypto_rc4.test-10        402.2m ± 1%     390.5m ± 0%    -2.90% (p=0.001 n=7)
TinyGo/Run/crypto_rc4.test-10            160.8m ± 0%     161.0m ± 1%         ~ (p=1.000 n=7)
TinyGo/Compile/crypto_sha1.test-10       417.2m ± 1%     404.5m ± 1%    -3.04% (p=0.001 n=7)
TinyGo/Run/crypto_sha1.test-10           15.93m ± 1%     15.90m ± 1%         ~ (p=0.710 n=7)
TinyGo/Compile/crypto_sha256.test-10     423.4m ± 1%     412.4m ± 1%    -2.60% (p=0.001 n=7)
TinyGo/Run/crypto_sha256.test-10         16.16m ±  ∞ ¹   16.05m ±  ∞ ¹       ~ (p=0.381 n=2+5)
geomean                                  94.17m          92.70m         -1.56%
¹ need >= 6 samples for confidence interval at level 0.95

                                      │ old_tinygo.txt │             new_tinygo.txt             │
                                      │      B/op      │      B/op       vs base                │
TinyGo/Compile/container_heap.test-10   48.55Mi ± 0%     48.30Mi ± 0%    -0.52% (p=0.001 n=7)
TinyGo/Run/container_heap.test-10       16.63Mi ± 0%     16.63Mi ± 0%         ~ (p=0.557 n=7)
TinyGo/Compile/container_list.test-10   48.53Mi ± 0%     48.29Mi ± 0%    -0.51% (p=0.001 n=7)
TinyGo/Run/container_list.test-10       16.40Mi ± 0%     16.40Mi ± 0%         ~ (p=0.364 n=7)
TinyGo/Compile/container_ring.test-10   47.78Mi ± 0%     47.53Mi ± 0%    -0.52% (p=0.001 n=7)
TinyGo/Run/container_ring.test-10       16.30Mi ± 0%     16.30Mi ± 0%         ~ (p=0.128 n=7)
TinyGo/Compile/crypto_des.test-10       48.67Mi ± 0%     48.42Mi ± 0%    -0.51% (p=0.001 n=7)
TinyGo/Run/crypto_des.test-10           16.76Mi ± 0%     16.76Mi ± 0%         ~ (p=0.902 n=7)
TinyGo/Compile/crypto_md5.test-10       48.73Mi ± 0%     48.48Mi ± 0%    -0.51% (p=0.001 n=7)
TinyGo/Run/crypto_md5.test-10           44.97Mi ± 0%     44.97Mi ± 0%         ~ (p=0.402 n=7)
TinyGo/Compile/crypto_rc4.test-10       47.76Mi ± 0%     47.52Mi ± 0%    -0.51% (p=0.001 n=7)
TinyGo/Run/crypto_rc4.test-10           29.28Mi ± 0%     29.28Mi ± 0%         ~ (p=0.104 n=7)
TinyGo/Compile/crypto_sha1.test-10      48.97Mi ± 0%     48.72Mi ± 0%    -0.52% (p=0.001 n=7)
TinyGo/Run/crypto_sha1.test-10          17.44Mi ± 0%     17.44Mi ± 0%         ~ (p=1.000 n=7)
TinyGo/Compile/crypto_sha256.test-10    48.81Mi ± 0%     48.56Mi ± 0%    -0.51% (p=0.001 n=7)
TinyGo/Run/crypto_sha256.test-10        17.53Mi ±  ∞ ¹   17.53Mi ±  ∞ ¹       ~ (p=0.381 n=2+5)
geomean                                 31.45Mi          31.37Mi         -0.26%
¹ need >= 6 samples for confidence interval at level 0.95

                                      │ old_tinygo.txt │            new_tinygo.txt             │
                                      │   allocs/op    │   allocs/op    vs base                │
TinyGo/Compile/container_heap.test-10    83.67k ± 0%     83.46k ± 0%    -0.25% (p=0.011 n=7)
TinyGo/Run/container_heap.test-10        374.9k ± 0%     374.9k ± 0%         ~ (p=1.000 n=7)
TinyGo/Compile/container_list.test-10    83.34k ± 0%     83.19k ± 0%    -0.19% (p=0.002 n=7)
TinyGo/Run/container_list.test-10        370.0k ± 0%     370.0k ± 0%         ~ (p=0.674 n=7)
TinyGo/Compile/container_ring.test-10    83.26k ± 0%     83.08k ± 0%    -0.22% (p=0.004 n=7)
TinyGo/Run/container_ring.test-10        367.6k ± 0%     367.6k ± 0%         ~ (p=0.249 n=7)
TinyGo/Compile/crypto_des.test-10        83.68k ± 0%     83.53k ± 0%    -0.18% (p=0.004 n=7)
TinyGo/Run/crypto_des.test-10            378.1k ± 0%     378.1k ± 0%         ~ (p=0.437 n=7)
TinyGo/Compile/crypto_md5.test-10        83.86k ± 0%     83.67k ± 0%    -0.23% (p=0.001 n=7)
TinyGo/Run/crypto_md5.test-10            393.3k ± 0%     393.3k ± 0%         ~ (p=0.592 n=7)
TinyGo/Compile/crypto_rc4.test-10        83.32k ± 0%     83.20k ± 0%    -0.14% (p=0.011 n=7)
TinyGo/Run/crypto_rc4.test-10            367.1k ± 0%     367.1k ± 0%         ~ (p=0.102 n=7)
TinyGo/Compile/crypto_sha1.test-10       84.05k ± 0%     83.87k ± 0%    -0.21% (p=0.002 n=7)
TinyGo/Run/crypto_sha1.test-10           392.7k ± 0%     392.7k ± 0%         ~ (p=1.000 n=7)
TinyGo/Compile/crypto_sha256.test-10     83.86k ± 0%     83.67k ± 0%    -0.24% (p=0.001 n=7)
TinyGo/Run/crypto_sha256.test-10         394.5k ±  ∞ ¹   394.5k ±  ∞ ¹       ~ (p=0.952 n=2+5)
geomean                                  178.2k          178.0k         -0.10%
```

### wazero compiled as wasip1 binary

```
goos: darwin
goarch: arm64
pkg: github.com/tetratelabs/wazero
               │  old.txt   │             new.txt              │
               │   sec/op   │   sec/op    vs base              │
Compilation-10   2.413 ± 0%   2.258 ± 1%  -6.42% (p=0.001 n=7)

               │   old.txt    │              new.txt               │
               │     B/op     │     B/op      vs base              │
Compilation-10   339.9Mi ± 0%   337.7Mi ± 0%  -0.63% (p=0.001 n=7)

               │   old.txt   │              new.txt              │
               │  allocs/op  │  allocs/op   vs base              │
Compilation-10   603.9k ± 0%   602.4k ± 0%  -0.25% (p=0.001 n=7)
```


Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2024-06-07 11:42:45 -07:00
Takeshi Yoneda
eb24363d34 ssa: empirically faster passRedundantPhiEliminationOpt (#2214) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2024-05-23 16:12:40 +09:00
Takeshi Yoneda
b9633563c0 ssa: simplifies basicBlock.params (#2212) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2024-05-21 15:51:41 +09:00
Takeshi Yoneda
9e4a5544bd wazevo(ssa): recursive constant phi eliminations (#2109) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2024-03-04 14:58:09 +09:00
Takeshi Yoneda
52212ad018 wazevo(ssa): refactors pass execution (#2089) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2024-02-24 10:44:08 +09:00
Takeshi Yoneda
7d1818c227 wazevo(ssa): reuse slices (#2041) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2024-02-11 10:27:31 -08:00
Takeshi Yoneda
8cf0fc3f4b wazevo(ssa): avoids allocation during natural order sort (#2024) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2024-02-07 17:21:15 -08:00
Takeshi Yoneda
6f16354ec7 wazevo: refactors liveness analysis (#1821) 
This refactors the liveness analysis and starts using the different algorithm,
which results in 30s -> 12s for Python binary compilation.

Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2023-10-31 08:12:17 +09:00
Takeshi Yoneda
ee91780e3d wazevo(ssa): eliminates no-op shifts by constant zeros (#1763) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2023-10-06 11:27:35 +09:00
Takeshi Yoneda
c592c8e5c1 wazevo: removes unused SSA insts, centralizes debug toggles (#1671) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2023-08-30 08:16:51 +09:00
Takeshi Yoneda
b6ace485b6 wazevo: passes call_indirect, linking spectest (#1670) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2023-08-29 12:09:18 +09:00
Takeshi Yoneda
828498dd89 wazevo(ssa): introduce trappable side effect (#1668) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2023-08-29 08:04:10 +09:00
Takeshi Yoneda
757c4a6436 wazevo: passes br_if, switch and other spectests (#1659) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2023-08-25 10:07:49 +09:00
Takeshi Yoneda
f96f097b39 wazevo: pass memory_grow.wast (#1656) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2023-08-24 10:51:28 +09:00
Takeshi Yoneda
1223d11d6a wazevo: pass memory_size.wast (#1653) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2023-08-23 14:59:29 +09:00
Takeshi Yoneda
02900879bf wazevo: initial impl of the new optimizing backend (#1615) 
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
 2023-08-09 10:45:23 +09:00