Open jaykrell opened 6 years ago
-pb/-parallelback/M3_PARALLEL_BACK?
Just intuitively, I haven't been too unhappy, comparing the build speed of m3-libs & m3-front to m3cc in C.
On 05/12/2018 07:07 AM, Jay Krell wrote:
We need to figure out the bottlenecks and speed it up.
multi-thread/proc? batch up cm3cg/as invocations / .ms file content? Too much autoconf-within-quake (I see fgrep).
m3core takes 18 seconds to compile. Let's try to get it to 10? Best bet might be entire library at once in one .ms file. And similar for the .c files.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/modula3/cm3/issues/26, or mute the thread https://github.com/notifications/unsubscribe-auth/AGVctBwQqE0M9b9WzW9kIIBaCf4FNmKzks5txtB7gaJpZM4T8Y3U.
This looks like done work? In HEAD I see M3_PARALLEL_BACK, but cm3-commands shown cm3cg invocations... I have been absent from cm3 for some time... Can someone enlighten me here...? Does this mean current cm3 builds in parallel and defaults to cm3cg (no-C, no-LLVM) backend?
I think I checked in the code to do up to 10 parallel back-end builds at the same time, for AMD64.
But yes it also defaults to cm3cg.
Mika
On Sun, Jul 22, 2018 at 1:00 AM, Dragiša Durić notifications@github.com wrote:
This looks like done work? In HEAD I see M3_PARALLEL_BACK, but cm3-commands shown cm3cg invocations... I have been absent from cm3 for some time... Can someone enlighten me here...? Does this mean current cm3 builds in parallel and defaults to cm3cg (no-C, no-LLVM) backend?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/modula3/cm3/issues/26#issuecomment-406848501, or mute the thread https://github.com/notifications/unsubscribe-auth/AFWGpYmh1PBNSbko8yj1-gc3tpqOzP2vks5uJDEjgaJpZM4T8Y3U .
Ok, This was more of a self-check. Not being in for some time, I am trying to catch up to speed :). Thanks!
I like this parallel build. It was fast enough for me before, but now it looks unreal :). On Sun, 2018-07-22 at 01:23 -0700, Mika Nyström wrote:
I think I checked in the code to do up to 10 parallel back-end builds at the same time, for AMD64.
But yes it also defaults to cm3cg.
Mika
On Sun, Jul 22, 2018 at 1:00 AM, Dragiša Durić <notifications@github. com> wrote:
This looks like done work? In HEAD I see M3_PARALLEL_BACK, but cm3-commands shown cm3cg invocations... I have been absent from cm3 for some time... Can someone enlighten me here...? Does this mean current cm3 builds in parallel and defaults to cm3cg (no-C, no-LLVM) backend?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/modula3/cm3/issues/26#issuecomment-406848501, or mute the thread https://github.com/notifications/unsubscribe-auth/AFWGpYmh1PBNSbko 8yj1-gc3tpqOzP2vks5uJDEjgaJpZM4T8Y3U .
— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.
We need to figure out the bottlenecks and speed it up.
multi-thread/proc? batch up cm3cg/as invocations / .ms file content? Too much autoconf-within-quake (I see fgrep).
m3core takes 18 seconds to compile. Let's try to get it to 10? Best bet might be entire library at once in one .ms file. And similar for the .c files.