LibreOffice Enables RISC-V 64-bit Support

[kvuj]

Why does libreoffice need to modify its source for risc-v?

"bridges/source/cpp_uno/gcc3_linux_riscv64/abi.cxx"
huh?

"bridges/source/cpp_uno/gcc3_linux_x86-64/abi.cxx"
"This is an implementation of the x86-64 ABI as described in 'System V Application Binary Interface, AMD64 Architecture Processor Supplement' (http://www.x86-64.org/documentation/abi-0.95.pdf)"
Why?

[kpedersen]

Why does libreoffice need to modify its source for risc-v?

"bridges/source/cpp_uno/gcc3_linux_riscv64/abi.cxx"

Why?

Hmm, It looks like the UNO bridge: https://help.libreoffice.org/latest/...o_objects.html

So serializing / sending objects across a wire; I suppose needs the endianness to match or be consistent.

So the patch / bridge needs to kind of achieve similar to a htons(3) on steroids?

(I imagine 99.9% of users don't even use this. Plus it probably isn't available on the WebAssembly port that they seem to be focusing on so much).

[ayumu]

Firefox is available too.
Now if only there could be a cheap relatively high performance SoC that would be great. Something like the Raspberry Pi at the price of the Raspberry Pi.

Look into the VisionFive 2 (kickstarter over... shipping soon, alongside general availability).

Less than $100, and it gets you 80% there CPU performance wise, but in exchange has 4x rpi4 GPU performance, 2x GbE, 1x M_2, cryptography acceleration and a better video encode/decode block.

[Quackdoc]

Look into the VisionFive 2 (kickstarter over... shipping soon, alongside general availability).

Less than $100, and it gets you 80% there CPU performance wise, but in exchange has 4x rpi4 GPU performance, 2x GbE, 1x M_2, cryptography acceleration and a better video encode/decode block.

im concerned about the GPU, IMO mesa is necessary, and being IT gpu ""BXE-4-32 GPU"" I haven't been able to find out if this will eventually be supported in PVR. now I know PVR supports the ""BXS-4-64"", I'm just not sure if they are the same enough to make it a simple port or not.

[ayumu]

im concerned about the GPU, IMO mesa is necessary, and being IT gpu ""BXE-4-32 GPU"" I haven't been able to find out if this will eventually be supported in PVR. now I know PVR supports the ""BXS-4-64"", I'm just not sure if they are the same enough to make it a simple port or not.

While I agree Mesa3d is necessary, I don't think there's reason to doubt their word that an open driver is in the works. If the board was more expensive, maybe I would feel more cautious.

Note that you can also look at the Star64, which is based on the same SoC, but offers a PCIe port, with which you can use a GPU that's already supported by Mesa3d. I would assume whatever works on the previous SiFive boards also works here, as the CPU's the same (U74, although a newer revision in VisionFive 2).

[ll1025]

I would love if, instead of doing the fun geeky technical bits in LibreOffice (which have nothing to do with it's also-ran status), there would be anything resembling an effort to improve its interface.

I did a usability study for some undergrad class some years ago, nothing terribly rigorous, but even in designing the tests and doing trial runs it was incredible how convoluted and painful LibreOffice makes very common tasks compared with how easy they are in Excel. There's no real equivalent to Excel's "Format as Table" which creates a named range, adds sort / filter tabs, and adds alternate-row formatting. You can sort of recreate some of it in LO, with about 2-dozen clicks, but its painful. Formatting for print, or printing to PDF, is a horrorshow. And there's nothing at all like the keyboard accessibility shortcuts that let you perform complex analysis without ever touching a mouse.

It's kind of incredible that something like LibreOffice doesn't even add the usual FOSS goodies like regex support in calc or commandline / API access; even on that front, Office wins.

I have tried for years to love LibreOffice because the idea of being tied into a proprietary system is hateful, but I'm not going to throw away most of the utility of a system just because it isn't FOSS.

[Quackdoc]

I would love if, instead of doing the fun geeky technical bits in LibreOffice (which have nothing to do with it's also-ran status), there would be anything resembling an effort to improve its interface.

I did a usability study for some undergrad class some years ago, nothing terribly rigorous, but even in designing the tests and doing trial runs it was incredible how convoluted and painful LibreOffice makes very common tasks compared with how easy they are in Excel. There's no real equivalent to Excel's "Format as Table" which creates a named range, adds sort / filter tabs, and adds alternate-row formatting. You can sort of recreate some of it in LO, with about 2-dozen clicks, but its painful. Formatting for print, or printing to PDF, is a horrorshow. And there's nothing at all like the keyboard accessibility shortcuts that let you perform complex analysis without ever touching a mouse.

It's kind of incredible that something like LibreOffice doesn't even add the usual FOSS goodies like regex support in calc or commandline / API access; even on that front, Office wins.

I have tried for years to love LibreOffice because the idea of being tied into a proprietary system is hateful, but I'm not going to throw away most of the utility of a system just because it isn't FOSS.

I use only office myself, its not all the way there, but at least it doesnt make my eyes bleed and for basic usage its fine, im not sure if that feature is supported though

[ll1025]

I use only office myself, its not all the way there, but at least it doesnt make my eyes bleed and for basic usage its fine, im not sure if that feature is supported though

From a 30 second demo it appears to have that feature and seems lightyears ahead of LibreOffice, though it does not add in the named references that are core for complex formulae. E.g, if you have columns "Firstname", "Surname", "DisplayName", format as table lets your formula for "DisplayName" become

=concat([@[Surname]],", ",[@[Firstname]])

When dealing with things like investing or debt calculation it is way easier to deal with references to [@[CostBasis]] than 23 references to "$A3".

It's really remarkable how many years the LO team has wasted on non-features while zero progress has been made on the UX/UI.

EDIT: It looks like formatting as a table in OnlyOffice also creates a generic named range that expands with the size of the table, and I'm just not using it right. Seems like a good contender for an Office replacement down the line.

[kvuj]

Hmm, It looks like the UNO bridge: https://help.libreoffice.org/latest/...o_objects.html

So serializing / sending objects across a wire; I suppose needs the endianness to match or be consistent.

So the patch / bridge needs to kind of achieve similar to a htons(3) on steroids?

(I imagine 99.9% of users don't even use this. Plus it probably isn't available on the WebAssembly port that they seem to be focusing on so much).

Cool, thanks a lot on the research!

On a side note, god I love reading OpenBSD source code. htonl() function