PostgreSQL 16 Reaches Beta With Performance Improvements - Including SIMD for x86 & ARM

[phoronix]

Phoronix: PostgreSQL 16 Reaches Beta With Performance Improvements - Including SIMD for x86 & ARM

PostgreSQL 16 is up to its first beta milestone today with new performance optimizations and continued security enhancements...

https://www.phoronix.com/news/PostgreSQL-16-Beta-1

[lejeczek]

faster performance.. right, really?

[uid313]

PostgreSQL doesn't support RISC-V. 😥️

What do you like about PostgreSQL? What is your favorite thing about PostgreSQL? What makes PostgreSQL good?

To me it seems like a database that is stable because it has been around for such a long time, and that it has a lot of features, and much of the SQL standards implemented. On the other hand, it has a vanity license and a messy user space.

[wswartzendruber]

I haven't worked with it in a while, but when I did, I liked how functions were separated into frontend and backend. Ergo, I could write a function in C if I wanted to, and the called wouldn't have to know or care. I could also use SQL or pgPLSQL.

[Vistaus]

PostgreSQL doesn't support RISC-V. 😥️

The 5 users of RISC-V will be sad.

Not to hate or anything, I think RISC-V has a lot of potential and I hope it overturns other architectures, but at the moment, usage outside of embedded is limited. And AFAIK, PostgreSQL isn't something you want to use on embedded.

[andrei_me]

Stability was what made us change back in 2007, MySQL couldn't handle in the hardware that we had.

On the dev quality of life, I liked the schema to separate objects, many different data types, functions, explain detailing the query plan, the extensions available like fuzzy match, citus, the foreign data wrappers

I think it was postgres 8.3 or 8.4 back then, nowadays we parallel queries, jsonb, GIN indexes, FILTER in aggregation functions, and many new features

From this release, I like simple things like underscore for big number (100_000_000), zstd and lz4 in pg_dump, more queries supporting parallel execution

The elephant keeps learning new tricks 🙂

[ryao]

The 5 users of RISC-V will be sad.

Not to hate or anything, I think RISC-V has a lot of potential and I hope it overturns other architectures, but at the moment, usage outside of embedded is limited. And AFAIK, PostgreSQL isn't something you want to use on embedded.

That should change in the near future with higher end designs being made:

https://www.prnewswire.com/news-releases/ventana-introduces-veyron-worlds-first-data-center-class-risc-v-cpu-product-family-301700985.html

https://www.tomshardware.com/news/tenstorrent-shares-roadmap-of-ultra-high-performance-risc-v-cpus-and-ai-accelerators

There are others, but those two caught my attention. At the low end, you have things like these:

https://www.tomshardware.com/news/milk-v-offers-a-trio-of-risc-v-raspberry-pi-alternatives

RISC-V is basically where ARM was 10 to 15 years ago, and it is catching up fast. ARM’s recent moves to change its business model are accelerating this trend. Suing Qualcomm and then talking about moving royalties to device makers based on the cost of the device to be paid several times more for the same thing are moves pushing companies to RISC-V.

That said, there is the chicken and egg problem to address in that people have little incentive to adopt RISC-V hardware without software support and there is little incentive to adopt it without hardware adoption. The open source community really should be working on RISC-V support to help break this chicken and egg situation, lest we be stuck with lousy options from ISA IP rights holders demanding ever increasing sums of money for the same thing. I recently have been increasingly looking at RISC-V.

Unfortunately, hardware that I consider good enough for me to get started with RISC-V was only recently announced and hardware suitable for a company I know that has almost finished its ARM migration is not on the market yet. The closest thing to something suitable for them is this:

https://www.cnx-software.com/2022/05/31/aries-embedded-osm-modules-renesas-arm-or-risc-v-microprocessors/

If it had a GPU for hardware accelerated electron on X11 on a touch screen, it would be perfect, provided electron begins to support RISC-V.

[jacob]

That should change in the near future with higher end designs being made:

https://www.prnewswire.com/news-releases/ventana-introduces-veyron-worlds-first-data-center-class-risc-v-cpu-product-family-301700985.html

https://www.tomshardware.com/news/tenstorrent-shares-roadmap-of-ultra-high-performance-risc-v-cpus-and-ai-accelerators

There are others, but those two caught my attention. At the low end, you have things like these:

https://www.tomshardware.com/news/milk-v-offers-a-trio-of-risc-v-raspberry-pi-alternatives

RISC-V is basically where ARM was 10 to 15 years ago, and it is catching up fast. ARM’s recent moves to change its business model are accelerating this trend. Suing Qualcomm and then talking about moving royalties to device makers based on the cost of the device to be paid several times more for the same thing are moves pushing companies to RISC-V.

That said, there is the chicken and egg problem to address in that people have little incentive to adopt RISC-V hardware without software support and there is little incentive to adopt it without hardware adoption. The open source community really should be working on RISC-V support to help break this chicken and egg situation, lest we be stuck with lousy options from ISA IP rights holders demanding ever increasing sums of money for the same thing. I recently have been increasingly looking at RISC-V.

Unfortunately, hardware that I consider good enough for me to get started with RISC-V was only recently announced and hardware suitable for a company I know that has almost finished its ARM migration is not on the market yet. The closest thing to something suitable for them is this:

https://www.cnx-software.com/2022/05/31/aries-embedded-osm-modules-renesas-arm-or-risc-v-microprocessors/

If it had a GPU for hardware accelerated electron on X11 on a touch screen, it would be perfect, provided electron begins to support RISC-V.

Yeah yeah, powerful high end RISC-V hardware is just around the corner, right. We've been listening to that for how many years now? It has only ever been vaporware, nothing more.