Multicore OCaml includes an event tracing system that act as a GC profiling tool for the development effort on the multicore runtime.
Inspired by GHC's ThreadScope, the idea is provide a low-overhead framework to collect runtime events and statistics, and process them offline in order to profile and debug performance problems.
Multicore OCaml emit such event traces in a format compatible with the
| .L111: | |
| .loc 3 562 0 | |
| movq -16(%rbp), %rax | |
| andb $252, %ah | |
| movq %rax, %rdx | |
| movq -8(%rbp), %rax | |
| movq %rdx, (%rax) | |
| .loc 3 563 0 | |
| nop | |
| jmp .L102 |
| # The main loop condition | |
| .L108: | |
| .loc 3 542 0 | |
| testq %r13, %r13 | |
| jg .L111 | |
| [...] | |
| # the if condition checking to load next memory chunk on chunk boundary | |
| .L103: | |
| .loc 3 567 0 |
| @@ -464,6 +464,7 @@ | |
| Removing basic block 19 | |
| Removing basic block 20 | |
| Removing basic block 21 | |
| +__attribute__((optimize (1, "tree-vrp", "no-strict-aliasing", "wrapv"))) | |
| sweep_slice (intnat work) | |
| { | |
| void (*<T7fc>) (value) final_fun; | |
| @@ -494,7 +495,7 @@ |
| while (work > 0){ | |
| if (caml_gc_sweep_hp < limit){ | |
| hp = caml_gc_sweep_hp; | |
| hd = Hd_hp (hp); | |
| work -= Whsize_hd (hd); | |
| caml_gc_sweep_hp += Bhsize_hd (hd); | |
| switch (Color_hd (hd)){ | |
| case Caml_white: | |
| if (Tag_hd (hd) == Custom_tag){ |
| module I1 = struct type i2 = int;; let p3 = Extprot.Pretty_print.pp_int;; end;; | |
| module I4 = struct type i5 = string;; let p6 = Extprot.Pretty_print.pp_string;; end;; | |
| module L7 = struct type l8 = string;; let p9 = Extprot.Pretty_print.pp_string;; end;; | |
| module S10 = struct type s11 = string;; let p12 = Extprot.Pretty_print.pp_string;; end;; | |
| module S13 = | |
| struct | |
| type s14 = S10.s11 list;; | |
| let p15 = Extprot.Pretty_print.pp_list S10.p12;; | |
| end;; |
I hereby claim:
To claim this, I am signing this object:
