no cache hit · compiler development

Stream: compiler development

Topic: no cache hit

Anton (Jun 24 2026 at 14:53):

This should have been faster on the second run right? As in, it should have been cached.

❯ ./zig-out/bin/roc test/echo/all_syntax_test.roc
roc
  ✓ Resolving Dependencies  52ms
  ✓ Parsing                 2ms
  ✓ Name Resolution         5ms
  ✓ Type Inference          84ms
  ✓ Specializing            192ms
[dbg] 42.0

roc on  add-to-all-syntax [?] via ↯ v0.16.0
❯ ./zig-out/bin/roc test/echo/all_syntax_test.roc
roc
  ✓ Resolving Dependencies  57ms
  ✓ Parsing                 3ms
  ✓ Name Resolution         5ms
  ✓ Type Inference          89ms
  ✓ Specializing            195ms

Richard Feldman (Jun 24 2026 at 15:05):

for sure

Richard Feldman (Jun 24 2026 at 15:05):

we should update those stats to include cache info

Richard Feldman (Jun 24 2026 at 15:07):

e.g. 5ms (5 cached, 6 uncached modules)

Richard Feldman (Jun 24 2026 at 15:09):

we could also add a separate step for like "Source File I/O" at the very start

Richard Feldman (Jun 24 2026 at 15:09):

and another for "Hash Source Files" (the BLAKE3 hash for cache lookup) and one more for "Cache Reads" (and actually maybe even a "Cache Writes" since that takes time when we're populating the cache for the first time)