You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: bench.md
+3-5Lines changed: 3 additions & 5 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -9,9 +9,7 @@ This file documents the coremark bench results to keep track of performance impr
9
9
- repr(C) for the SideTableEntry struct caused mysterious improvements, not sure if it is a fluke
10
10
- cc02503: avg = 855.71814, n = 20 (remove defensive malformed check in main loop)
11
11
- since the module is already validated at run time, there is no reason for the check to exist, it was a remnant of early development phase that lacked proper handling for some malformed modules
12
-
- current: no significant difference
13
-
14
-
On nightly, the performance is slightly better (sometimes reaching 900)
12
+
- current: avg = 948.285031, n = 20 (use peek for stack access, reorganization and simplifications)
15
13
16
14
Next step: use direct threading to improve branch prediction
17
15
@@ -28,6 +26,6 @@ wasmi: ~1700
28
26
tinywasm: ~630
29
27
30
28
Goal:
31
-
We expect/hope to reach ~1200 after threaded dispatch implementation. It seems like Ben Titzer only reached performance comparable to production-ready, optimizing interpreters through manually crafted assembly code for hot paths.
29
+
I expect/hope to reach ~1200 after threaded dispatch implementation. It seems like Ben Titzer only reached performance comparable to production-ready, optimizing interpreters through manually crafted assembly code for hot paths.
32
30
33
-
Higher performance may not be pursued after the point and instead I might focus on adding more instructions to achieve Wasm 2.0 spec parity (should be easy with AI).
31
+
Higher performance may not be pursued after the point and instead I might focus on adding more instructions to achieve Wasm 2.0 spec parity.
0 commit comments