From: Sergey Bronnikov via Tarantool-patches <tarantool-patches@dev.tarantool.org> To: Sergey Kaplun <skaplun@tarantool.org>, Maxim Kokryashkin <m.kokryashkin@tarantool.org> Cc: tarantool-patches@dev.tarantool.org Subject: Re: [Tarantool-patches] [PATCH luajit 4/5] Fix pow() optimization inconsistencies. Date: Fri, 18 Aug 2023 15:45:57 +0300 [thread overview] Message-ID: <dbd256a4-c150-4f9c-f0a9-4352435452bc@tarantool.org> (raw) In-Reply-To: <d10307e2a3f603d058bbc6969468116ccf191c9e.1692089299.git.skaplun@tarantool.org> Hi, Sergey! thanks for the patch! LGTM with two minor comments inline. On 8/15/23 12:36, Sergey Kaplun wrote: > From: Mike Pall <mike> > > (cherry-picked from commit 9512d5c1aced61e13e7be2d3208ec7ae3516b458) > > This patch fixes different misbehaviour between JIT-compiled code and misbehaviour -> misbehaviours > the interpreter for power operator with the following ways: > * Drop folding optimizations for base ^ 0.5 => sqrt(base), as far as > pow(base, 0.5) isn't interchangeable and depends on the <math.h> > implementation. > * Drop folding optimizations for 2 ^ int_pow => ldexp(1.0, int_pow), to > avoid dependcy on the <math.h> implementation. dependcy -> dependency > * Now `asm_pow()` always assemble a call to the `lj_vm_powi()` function, > that is general now for all CPU architectures. Using this internal > function instead of toolchain-provided `pow()` guarantees consistency > between interpreter and JIT results. Also, it drops custom > implementation for the `vm_powi_sse()` on x86_64. > * `math_extern2` macro in the VM may take the second argument, that is > used as the target function to call. The first argument is still the > name for `func_nnsse` macro. > * Narrowing for power operation avoids range guard for non-constant base > IR. This leads to invalid result if value on trace is out of range. > Now it is done unconditionally. > > Be aware, that [220/502] lib/string/format/num.lua test [1] from > LuaJIT-test suite fails after this commit. > > [1]: https://www.exploringbinary.com/incorrect-floating-point-to-decimal-conversions/ > > Sergey Kaplun: > * added the description and the test for the problem <snipped>
next prev parent reply other threads:[~2023-08-18 12:45 UTC|newest] Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top 2023-08-15 9:36 [Tarantool-patches] [PATCH luajit 0/5] Fix pow inconsistencies and improve asserts Sergey Kaplun via Tarantool-patches 2023-08-15 9:36 ` [Tarantool-patches] [PATCH luajit 1/5] test: introduce `samevalues()` TAP checker Sergey Kaplun via Tarantool-patches 2023-08-17 14:03 ` Maxim Kokryashkin via Tarantool-patches 2023-08-17 15:03 ` Sergey Kaplun via Tarantool-patches 2023-08-18 10:43 ` Sergey Bronnikov via Tarantool-patches 2023-08-18 10:58 ` Sergey Kaplun via Tarantool-patches 2023-08-18 11:12 ` Sergey Bronnikov via Tarantool-patches 2023-08-21 10:47 ` Igor Munkin via Tarantool-patches 2023-08-24 7:44 ` Sergey Bronnikov via Tarantool-patches 2023-08-15 9:36 ` [Tarantool-patches] [PATCH luajit 2/5] Remove pow() splitting and cleanup backends Sergey Kaplun via Tarantool-patches 2023-08-17 14:52 ` Maxim Kokryashkin via Tarantool-patches 2023-08-17 15:33 ` Sergey Kaplun via Tarantool-patches 2023-08-20 9:48 ` Maxim Kokryashkin via Tarantool-patches 2023-08-18 11:08 ` Sergey Bronnikov via Tarantool-patches 2023-08-15 9:36 ` [Tarantool-patches] [PATCH luajit 3/5] Improve assertions Sergey Kaplun via Tarantool-patches 2023-08-17 14:58 ` Maxim Kokryashkin via Tarantool-patches 2023-08-18 7:56 ` Sergey Kaplun via Tarantool-patches 2023-08-18 11:20 ` Sergey Bronnikov via Tarantool-patches 2023-08-15 9:36 ` [Tarantool-patches] [PATCH luajit 4/5] Fix pow() optimization inconsistencies Sergey Kaplun via Tarantool-patches 2023-08-18 12:45 ` Sergey Bronnikov via Tarantool-patches [this message] 2023-08-21 8:07 ` Sergey Kaplun via Tarantool-patches 2023-08-20 9:26 ` Maxim Kokryashkin via Tarantool-patches 2023-08-21 8:06 ` Sergey Kaplun via Tarantool-patches 2023-08-21 9:00 ` Maxim Kokryashkin via Tarantool-patches 2023-08-21 9:31 ` Sergey Kaplun via Tarantool-patches 2023-08-15 9:36 ` [Tarantool-patches] [PATCH luajit 5/5] Revert to trival pow() optimizations to prevent inaccuracies Sergey Kaplun via Tarantool-patches 2023-08-18 12:49 ` Sergey Bronnikov via Tarantool-patches 2023-08-21 8:16 ` Sergey Kaplun via Tarantool-patches 2023-08-20 9:37 ` Maxim Kokryashkin via Tarantool-patches 2023-08-21 8:15 ` Sergey Kaplun via Tarantool-patches 2023-08-21 9:06 ` Maxim Kokryashkin via Tarantool-patches 2023-08-21 9:36 ` Sergey Kaplun via Tarantool-patches 2023-08-24 7:47 ` [Tarantool-patches] [PATCH luajit 0/5] Fix pow inconsistencies and improve asserts Sergey Bronnikov via Tarantool-patches 2023-08-31 15:18 ` Igor Munkin via Tarantool-patches
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=dbd256a4-c150-4f9c-f0a9-4352435452bc@tarantool.org \ --to=tarantool-patches@dev.tarantool.org \ --cc=m.kokryashkin@tarantool.org \ --cc=sergeyb@tarantool.org \ --cc=skaplun@tarantool.org \ --subject='Re: [Tarantool-patches] [PATCH luajit 4/5] Fix pow() optimization inconsistencies.' \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox