build-bench-env.sh fails to install alt allocators (and no bench.sh?)

[daanx]

build-bench-env.sh fails to install alt allocators (and no bench.sh?)

[daanx]

Thanks for the feedback -- so bad bench.sh was missing! Just added it and will test with it.
Otherwise:

• hoard: hmm, what does that error mean? perhaps running with sudo fixes it? I think we can work around it by not installing hoard, just compiling it. I'll look into it.
• supermalloc: I'll try to run it with warnings as warnings . I guess this is new behaviour for GCC 9?
• rpmalloc: I think we need to disable -Werror,-Watomic-implicit-seq-cst
  I'll try to fix it and perhaps you can try again.
  Thanks!!

[SteveBronder]

Thanks for including it! I'll downgrade to gcc 8 and try again.

hoard: hmm, what does that error mean? perhaps running with sudo fixes it? I think we can work around it by not installing hoard, just compiling it. I'll look into it.

I've gotten hoard to compile before on this machine, I think it's just that calling make plain doesn't call the correct config for hoard. I can give it another go later this week.

https://github.com/emeryberger/Hoard/blob/master/src/GNUmakefile#L39

For the other two I've never touched them so I have no clue

[daanx]

I just pushed an update to the build scripts that build all external dependencies in an extern folder local to mimalloc-bench. Also, I think i fixed the supermalloc and hoard builds.

[SteveBronder]

Nice ty! I was also seeing an error in rpmalloc

[1/40] CC test/thread.c
FAILED: build/ninja/linux/debug/b'x86_64'/test-57ec084/thread-35aa063.o 
clang -MMD -MT 'build/ninja/linux/debug/b'\''x86_64'\''/test-57ec084/thread-35aa063.o' -MF 'build/ninja/linux/debug/b'\''x86_64'\''/test-57ec084/thread-35aa063.o'.d -I. -Irpmalloc -Itest -DRPMALLOC_COMPILE=1 -funit-at-a-time -fstrict-aliasing -fno-math-errno -ffinite-math-only -funsafe-math-optimizations -fno-trapping-math -ffast-math -D_GNU_SOURCE=1 -W -Werror -pedantic -Wall -Weverything -Wno-padded -Wno-documentation-unknown-command -std=c11 -Werror -Watomic-implicit-seq-cst  -DBUILD_DEBUG=1 -g -DENABLE_ASSERTS=1 -DENABLE_STATISTICS=1 -c test/thread.c -o 'build/ninja/linux/debug/b'\''x86_64'\''/test-57ec084/thread-35aa063.o'
test/thread.c:95:2: error: implicit use of sequentially-consistent atomic may incur stronger memory barriers than necessary [-Werror,-Watomic-implicit-seq-cst]
        __sync_synchronize();
        ^~~~~~~~~~~~~~~~~~
test/thread.c:104:2: error: implicit use of sequentially-consistent atomic may incur stronger memory barriers than necessary [-Werror,-Watomic-implicit-seq-cst]
        __sync_synchronize();
        ^~~~~~~~~~~~~~~~~~
2 errors generated.
[2/40] CC test/thread.c

But that's an error in their stuff and they have a fix for it on their develop branch

mjansson/rpmalloc#86

You should be able to fix it by changing the build script to use

git clone https://github.com/mjansson/rpmalloc/issues/86 -b develop

Seems like after that all the benchmarks are running tho!

[SteveBronder]

This puppy purrs though!

Below is a graph of the tests on my machine with (Ryzen 7 CPU)

Some of those are within a pretty reasonable margin of error. tbb is v v v slow which I'm surprised by