Conversation
|
I think this is valuable work and having multiple frontends generating wasm Is Fannkuch long-running? If so, let's try to separate it from "unit" tests On Sat, Aug 29, 2015 at 3:24 AM, Katelyn Gadd notifications@github.com
|
With n=7 the C# version takes about 1ms (including startup overhead) and n=8 takes about 5ms. |
|
Added a working Sieve of Eratosthenes test (and implemented fp sqrt, since it needs it). |
This commit fixes WebAssembly#34 by specifying that the flags field (which indicates if a segment is passive) is a `varuint32` instead of a `uint8`. It was discovered in WebAssembly#34 that the memory index located at that position today is a `varuint32`, which can be validly encoded as `0x80 0x00` in addition to `0x00` (in addition to a number of other encodings). This means that if the first field were repurposed as a single byte of flags, it would break these existing modules that work today. It's not currently known how many modules in the wild actually take advantage of such an encoding, but it's probably better to be safe than sorry! Closes WebAssembly#34
… op proposal (WebAssembly#35) Adds table.size, table.grow, table.fill to overview, spec, interpreter, and tests, as decided at recent CG meeting. Also adds a few more tests for table.get and table.set. Also, change interpreter's segment encoding to match bulk ops proposal, addressing WebAssembly#18. (Not updated in spec yet, since corresponding spec text is still missing from bulk ops proposal.)
[spec] fix resulttype in soundness - admin-instr - label
Really we should be merging in all change from upstream but that is still in progress. Fixes: #35
Mention Wasmi in the list of engines supporting the proposal
2 participants
This isn't necessarily something we want to merge yet, but I wanted to make a PR so people can look at it and we can start a discussion.
I did some work to cross-compile simple test cases from C# to wasm. I have some basic stuff working. CountUp passes; Fannkuch starts, runs, then fails (which is progress, since it was hanging before!)
See https://gist.github.com/kg/3f66f50e2428d9d47a41 for the source to Fannkuch and https://gist.github.com/kg/4cd6e23a73ac232dcca5 for the source to CountUp.
My goal here was to start translating more-realistic functions and workflows from a source language to wasm, instead of just hand-writing wasm tests. This will be helpful in the interim until LLVM's wasm backend is working, and we might still want to have another source language for tests just to make sure we aren't overly focused on LLVM.
It would be great to get help stepping through fannkuch to figure out why it's broken. The current state of the interpreter is basically impossible to debug - if it hangs, it just hangs, and
-tdoesn't seem to trace the actual instructions being executed. I may take a shot at adding more instrumentation to the interpreter once I can figure out the code.Some related code quality/syntax questions came up while porting these tests, so I'll file issues about them separately.