News for hackers. Stuff that matters.
Re:Browser Bloat (1996) by derefr on Tuesday August 30, 2016 @02:20AM (#12387902)
I would assume that most of the logic for what Netscape did was embedded in the Netscape binary, while most modern browser features are 1% browser binary code, and 99% calls into the huge iceberg of OS-provided libraries.

Compare, for example, how VRML worked (entirely software rendering, all the code inside the browser) to how WebGL works (an API to talk to the OS graphics driver, which in turn is just a very fat API on top of the very CISC instruction set of a GPU.)

Re:Browser Bloat (1996) by djsumdog on Tuesday August 30, 2016 @04:00AM (#12388263)
Not true. To keep things compatible and minimize platform bugs, browsers embed a lot. Firefox has it's own jpeg, png, video codec and spelling libraries built-in. Try compiling it sometime. It takes forever.
Re:Browser Bloat (1996) by dietrichepp on Tuesday August 30, 2016 @02:50AM (#12388016)
Why do you say that the GPU has a "very CISC instruction set"? The RISC/CISC divide is a somewhat historical notion, and GPUs don't really seem to fall in either camp to me (separate load/store + tons of GPRs + VLIW = RISC-like, transcendental functions + texture fetch with interpolation = CISC-like).
Re:Browser Bloat (1996) by derefr on Tuesday August 30, 2016 @03:11PM (#12392994)
I was taking a bit of poetic license there; what I meant is that the ABI that the GPU presents to the CPU contains a number of high-level instructions that must be further decoded within the GPU, in the same way that modern "CISC-presenting" processors decode CISC ops into RISC μops.

That's only for graphics programming, mind you; the GPGPU interface is pretty directly representative of the GPU's internal RISC execution strategy.