FYI: SoftIron Overdrive 1000 vs. Pine64+ 2GB buildworld buildkernel time examples (fairly modern head versions)

[Mark Millard]

[I forgot to list the builds as -j4 builds in both examples.]

On 2018-Oct-5, at 11:30 AM, Mark Millard <marklmi at yahoo.com> wrote:

> Self hosted (re-)build from scratch (empty build tree at start):
> 
> SoftIron Overdrive 1000: a little under  3 hrs and 20 min
> Pine64+ 2GB:             a little under 13 hrs and 45 min
> 
> So somewhat more than a 10 hour difference and a factor of
> somewhat over 4 for total time. (I was testing the Overdrive
> and was curious. The Overdrive 1000 seems to be working fine.)
> 
> 
> Supporting context details (including the unusual storage) if
> you care . . .
> 
> Both had:

-j4 used to do the build.

> make[1]: "/usr/src/Makefile.inc1" line 341: SYSTEM_COMPILER: Determined that CC=cc matches the source tree.  Not bootstrapping a cross-compiler.
> make[1]: "/usr/src/Makefile.inc1" line 346: SYSTEM_LINKER: Determined that LD=ld matches the source tree.  Not bootstrapping a cross-linker.
> 
> and:
> 
> #WITH_CROSS_COMPILER=
> WITH_SYSTEM_COMPILER=
> WITH_SYSTEM_LINKER=
> #
> #CPUTYPE=soft
> WITH_LIBCPLUSPLUS=
> #WITH_LLD_BOOTSTRAP=
> WITHOUT_BINUTILS_BOOTSTRAP=
> WITH_ELFTOOLCHAIN_BOOTSTRAP=
> WITH_LLVM_TARGET_ALL=
> #WITH_CLANG_BOOTSTRAP=
> WITH_CLANG=
> WITH_CLANG_IS_CC=
> WITH_CLANG_FULL=
> WITH_CLANG_EXTRAS=
> WITH_LLD=
> WITH_LLD_IS_LD=
> WITHOUT_BINUTILS=
> WITH_LLDB=
> #
> WITH_BOOT=
> WITHOUT_LIB32=
> #
> WITHOUT_GCC_BOOTSTRAP=
> WITHOUT_GCC=
> WITHOUT_GCC_IS_CC=
> WITHOUT_GNUCXX=
> #
> NO_WERROR=
> #WERROR=
> MALLOC_PRODUCTION=
> 
> Both were using: LDFLAGS.lld+= -Wl,--no-threads
> (via make.conf).
> 
> Both were running non-debug builds (with symbols), including
> for the kernel. Both were building non-debug builds (with
> symbols).
> 
> The file system is UFS with soft updates (no journaling)
> in both cases. File systems each is one partition. Trim
> enabled, vfs.ffs.dotrimcons=1 . fsck_ffs -E had been used
> in a prior boot in both cases.
> 
> vm.pageout_oom_seq=120 was in use for both. (The default of
> 12 is not sufficient for the Pine64+ 2GB to complete
> buildworld buildkernel.)
> 
> 
> The Overdrive context was based on:
> 
> CFLAGS.clang+= -mcpu=cortex-a57
> CXXFLAGS.clang+= -mcpu=cortex-a57
> CPPFLAGS.clang+= -mcpu=cortex-a57
> ACFLAGS.arm64cpuid.S+=  -mcpu=cortex-a57+crypto
> ACFLAGS.aesv8-armx.S+=  -mcpu=cortex-a57+crypto
> ACFLAGS.ghashv8-armx.S+=        -mcpu=cortex-a57+crypto
> 
> and was a -r339076 context building itself.
> 
> 
> The Pine64+ 2GB context was based on:
> 
> CFLAGS.clang+= -mcpu=cortex-a53
> CXXFLAGS.clang+= -mcpu=cortex-a53
> CPPFLAGS.clang+= -mcpu=cortex-a53
> ACFLAGS.arm64cpuid.S+=  -mcpu=cortex-a53+crypto
> ACFLAGS.aesv8-armx.S+=  -mcpu=cortex-a53+crypto
> ACFLAGS.ghashv8-armx.S+=        -mcpu=cortex-a53+crypto
> 
> and was a -r338860 context building itself. (Some
> notes are from memory for this older build.)
> 
> 
> The Overdrive had the hard disk replaced by a SanDisk
> 860 Pro SSD (1TB). Around 95% of the space being free.
> It still had the original 8 GiByte RAM. 28 GiByte swap
> space on the same media.
> 
> My modified version of top reported for the Overdrive:
> no observed use of swap.
> 
> My modified version of top reported:
> Maximum Observed Active Mem 1627M.
> No observed use of Swap.
> 
> 
> The Pine64+ 2GB had a 128 GB e.MMC on a microsd adapter
> (used in DDR52 mode via local changes). Around 50% of
> the space being free. 3584M swap space on the same media.
> 
> My modified version of top reported:
> Maximum Observed Active Mem 1501M.
> Maximum Observed Used Swap 198M.
> 
> The Pine64+ 2GB does have a case, heat sinks, and a fan,
> but by no means equivalent to the Overdrive configuration.


===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)