|
What is this visual meme thing called?
|
# ? Apr 24, 2020 18:24 |
|
|
# ? Jun 7, 2024 01:49 |
|
re: isa chat NOP? more like no, op
|
# ? Apr 24, 2020 18:29 |
|
Schadenboner posted:What is this visual meme thing called? I'm pretty sure the "cute embedded nonsense hacks" controversy predates Helvetica T-shirts
|
# ? Apr 24, 2020 18:30 |
|
echinopsis posted:maybe they just want to slowly control as much as possible and rely on intel as little as possible that, and perf/watt on intel chips is hitting a wall lately, which was also the primary reason for their jump from power for x86 years ago
|
# ? Apr 24, 2020 18:32 |
|
i hate adobe softare so, so very much
|
# ? Apr 24, 2020 18:33 |
|
illustrator, indesign and photoshop are still the absolute best at what they do, and indesign has remarkably been getting less janky over time (unlike the other two)
|
# ? Apr 24, 2020 19:30 |
|
The Management posted:the insane variable length instruction encoding that makes it impossible to traverse an instruction stream backwards, or even really forwards unless you want to implement a full instruction decoder in software i love how crazily complicated instruction decoding is on x86. like you can point at the decoder on an x86 chip, because it’s so goddamn difficult that it makes up an appreciable amount of the die space. x86 chips were groundbreakers in pipelining the decode stage because x86 makes it too hard to decode a fetched instruction in one cycle. that crazy 20-cycle pentium 4 pipeline was like one-quarter decode. x86 chips actually decode every instruction two ways in parallel because they can’t afford to wait for the normal decode to finish just to figure out what the next IP is going to be. iirc there are multiple of these secondary decoders on every chip
|
# ? Apr 24, 2020 19:30 |
|
wow its been so long, i can only defend like half of that. but yeah, all of those details jive
|
# ? Apr 24, 2020 19:36 |
|
rjmccall posted:i love how crazily complicated instruction decoding is on x86. like you can point at the decoder on an x86 chip, because it’s so goddamn difficult that it makes up an appreciable amount of the die space. x86 chips were groundbreakers in pipelining the decode stage because x86 makes it too hard to decode a fetched instruction in one cycle. that crazy 20-cycle pentium 4 pipeline was like one-quarter decode. x86 chips actually decode every instruction two ways in parallel because they can’t afford to wait for the normal decode to finish just to figure out what the next IP is going to be. iirc there are multiple of these secondary decoders on every chip spending half your power budget on figuring out what the instruction is supposed to be seems like a good use of energy
|
# ? Apr 24, 2020 19:46 |
|
someone who is good at chip design help me budget this. my instruction set is dying
|
# ? Apr 24, 2020 19:48 |
|
also you'd think all of this crazy decoding would at least save you a bunch of code size, and it sortof does, but it's hilariously sabotaged on x86-64 because some idiot at amd thought that people would want to run legacy code directly in 64-bit mode, so all the instructions default to a 32-bit operand width and need a prefix to be used with pointers, which is much more common. and then they hosed over the legacy code thing anyway by taking away some common instruction encodings in order to make room for that prefix, so you can only run carefully-specialized 32-bit code in that mode anyway. oh and by the way this means decoding is also processor-mode-sensitive, although that at least happens on some other architectures
|
# ? Apr 24, 2020 20:00 |
|
maybe we're all wrong about apple moving to arm maybe they're bringing transmeta back from the dead!!
|
# ? Apr 24, 2020 20:13 |
|
qirex posted:illustrator, indesign and photoshop are still the absolute best at what they do, and indesign has remarkably been getting less janky over time (unlike the other two) coming from node based image manipulation and into stack based like photoshop is so god drat lame. it feels very restrictive to have to constrain everything to a large serial set of transforms. maybe I just don’t know how to use it properly because it doesn’t really cause problems, but as an example, maybe I want to screen A with B but only B and not just everything under B. node editing feels so powerful and manly rjmccall posted:also you'd think all of this crazy decoding would at least save you a bunch of code size, and it sortof does, but it's hilariously sabotaged on x86-64 because some idiot at amd thought that people would want to run legacy code directly in 64-bit mode, so all the instructions default to a 32-bit operand width and need a prefix to be used with pointers, which is much more common. and then they hosed over the legacy code thing anyway by taking away some common instruction encodings in order to make room for that prefix, so you can only run carefully-specialized 32-bit code in that mode anyway. oh and by the way this means decoding is also processor-mode-sensitive, although that at least happens on some other architectures so what you’re telling me is that threads are getting double ripped and then we throw away halve? actually don’t answer that because I know it’s a nonsense statement but god drat do I love reading about this kind of poo poo despite not really knowing anything about it
|
# ? Apr 24, 2020 20:50 |
|
has anyone read brian greene’s universe books? wish someone would write a book like that about processors
|
# ? Apr 24, 2020 20:51 |
|
echinopsis posted:has anyone read brian greene’s universe books? wish someone would write a book like that about processors Jon Stokes' Inside The Machine is v.deece but kinda old? E: Also I seem to remember the guy's politics might be ? But v.worth it if you can pick it up cheap used?
|
# ? Apr 24, 2020 21:19 |
|
echinopsis posted:node editing feels so powerful and manly you are a weird dude
|
# ? Apr 24, 2020 21:32 |
|
but is he right
|
# ? Apr 24, 2020 21:36 |
|
infernal machines posted:you are a weird dude my sense of humour isn’t really funny I’m going for some kind of post-cringe
|
# ? Apr 24, 2020 21:46 |
|
echi: original cringepunk
|
# ? Apr 24, 2020 21:47 |
|
https://youtu.be/F_2EdnWHUNk i’d like to think if I ever did stand up it was like this
|
# ? Apr 24, 2020 21:48 |
|
as crazy as x86 decoding is, how much impact is it having on end performance? like the die area consumed by decoders isn’t that huge, especially since the cpu cores look small these days compared to cache/gpu/etc
|
# ? Apr 24, 2020 22:09 |
|
I mean obviously everything should just move back to POWER
|
# ? Apr 24, 2020 22:11 |
|
DuckConference posted:as crazy as x86 decoding is, how much impact is it having on end performance? like the die area consumed by decoders isn’t that huge, especially since the cpu cores look small these days compared to cache/gpu/etc depends. the complex decoding takes a lot of power. power means heat, heat means reduced performance. it probably doesn’t affect the pure instructions per clock throughput directly, unless you miss a branch prediction and you have a longer pipeline to restart.
|
# ? Apr 24, 2020 22:27 |
|
is there like a babies first instruction set or something? like a web tool for doing assembler but very simple? management did you ever play TIS-100?
|
# ? Apr 24, 2020 22:48 |
|
When I was forced to take a course on assembly it was all using MIPS for being simple for child. I can't remember what tool we used for progamming in it though
|
# ? Apr 24, 2020 22:51 |
|
echinopsis posted:is there like a babies first instruction set or something? like a web tool for doing assembler but very simple? babbys first instruction set is probably mips. I’m sure you can find a thing for that, it’s commonly used in education. though maybe it’s riscv these days. never played TIS-100.
|
# ? Apr 24, 2020 22:59 |
|
TIS-100 is cool and fun. but not *that* deeply related to any other matter discussed in this thread.
|
# ? Apr 24, 2020 23:01 |
|
No Mods No Masters posted:When I was forced to take a course on assembly it was all using MIPS for being simple for child. I can't remember what tool we used for progamming in it though do mips i. delay slots rule and are definitely an excellent idea
|
# ? Apr 24, 2020 23:42 |
|
oh yeah, if there's one thing that is perennially cool in an instruction set it is encoding implementation details of its first implementation.
|
# ? Apr 24, 2020 23:57 |
|
The best part about learning MIPS was the rockstar programmers in class — who kept talking up how badass learning assembly was going to be — completely failing at the first take home assignment beyond “set up the simulator”.
|
# ? Apr 25, 2020 00:00 |
|
i liked sparc
|
# ? Apr 25, 2020 00:03 |
|
The Management posted:the stack based argument passing, the fact that instructions can only target certain registers, the legacy of direct memory accessing accumulator architecture still making up the primary programming model, the existence of extra registers that the compiler can never take advantage of, the insane variable length instruction encoding that makes it impossible to traverse an instruction stream backwards, or even really forwards unless you want to implement a full instruction decoder in software. oh, and the loving lack of broadcasting TLB maintenance ops. will apple begin to deviate from the arm instruction set (or does it already)
|
# ? Apr 25, 2020 00:11 |
|
Cybernetic Vermin posted:oh yeah, if there's one thing that is perennially cool in an instruction set it is encoding implementation details of its first implementation. is this like coding a compiler? I need to do something like this one of these days
|
# ? Apr 25, 2020 00:22 |
|
theadder posted:will apple begin to deviate from the arm instruction set (or does it already) no. Apple paid arm to create an instruction set that matched its microarchitecture plans, that’s what arm64 is. as an architecture licensee, they are contractually required to produce ISA-compatible cores. there is a process for licensees to add their own instructions in the reserved instruction space, and there is a process to create instructions that will be rolled into the ISA. Apple has already done that. they added instructions for WKDM compression, which is used by the iOS kernel memory compressor. they’ve added the link register signing instructions, which performs a reversible transform on link register values to mitigate ROP attacks. and some others that I forget.
|
# ? Apr 25, 2020 03:20 |
|
No Mods No Masters posted:When I was forced to take a course on assembly it was all using MIPS for being simple for child. I can't remember what tool we used for progamming in it though we used SPIM
|
# ? Apr 25, 2020 07:36 |
|
whats the good ipad to get
|
# ? Apr 25, 2020 15:07 |
|
my sister just bought a iPad (7th gen) and then she got a pencil ($120) but I think they won’t work together? I think she needs the Apple Pencil ($100) instead? apple is a deeply troubled company e: the first thing on the site where you choose your Apple Pencil version is a link to online chat support, incredible Jenny Agutter fucked around with this message at 15:41 on Apr 25, 2020 |
# ? Apr 25, 2020 15:33 |
|
barkbell posted:whats the good ipad to get the most expensive one?? is this a joke
|
# ? Apr 25, 2020 15:37 |
|
barkbell posted:whats the good ipad to get don’t buy the iPad (regular). 2 year old chip is going to shorten the lifespan and the front camera is garbage. iPad Air is good. iPad Pro is better but way more expensive.
|
# ? Apr 25, 2020 15:47 |
|
|
# ? Jun 7, 2024 01:49 |
|
barkbell posted:whats the good ipad to get buy a refurb pro because the new one is exactly the same except for cameras and a bit more GPU but unless you’re planning on doing pre-viz for film/TV productions don’t bother
|
# ? Apr 25, 2020 15:56 |