Update: The code has been cleaned and tools/presubmit.py returns with no (significant) errors. So the only work left to do before another commit is to make the mips cctest tests work and check the simulator.
Alexandre On Jan 20, 3:22 pm, Alexandre Rames <[email protected]> wrote: > Hi, > > Could you give me some hint on how to run tests I implemented in test- > assembler-mips.cc ? > The file looks like this: > > #include ... > [...] > static void InitializeVM() { [...] } > #define __ assm. > > TEST(MIPS0) { > InitializeVM(); > v8::HandleScope scope; > Assembler assm(NULL, 0); > > // addition > __ addu(v0, a0, Operand(a1)); > __ jr(ra); > __ nop(); > > CodeDesc desc; > assm.GetCode(&desc); > [...] > CHECK_EQ(0xfffffffb, res); > > } > > [...] > > Thanks, > > Alexandre > > On Jan 20, 2:48 pm, Alexandre Rames <[email protected]> wrote: > > > > > Hi, > > > Thanks for reviewing my code. > > Here are some details and answers on several points. > > > About the contributor licence agreement: I have given it to my boss. > > It will be sent quite soon. > > > I have corrected everything you told, except one issue I detail below. > > I hope the fixes will be enough to have it work on a 64bit Linux. I > > don't have one so please let me know of other future errors. > > > I worked on the simulator, the disassembler, and the debugger. (The > > arm code was really helpful. I mainly adapted it.) They all work, but > > the simulator still needs a few tests. However most instructions seem > > to work fine. I am currently working on some tests like the ones in > > test-assembler-<arch>.cc and test-disasm-<arch>.cc . This should be > > done today. Should I commit the code when the tests are implemented? > > There may still be a few mistakes, but there would at least be a good > > basis to work on. > > > About the issue I did not change: > > The concerned review urls are > > http://codereview.chromium.org/549079/diff/1/11#newcode153 > > and http://codereview.chromium.org/549079/diff/1/11#newcode153 > > > I had to use functions with such template when it uses conditional > > execution. > > On MIPS there is 'cmp' instruction or conditional execution like on > > ARM. MIPS assembly has instead contional branch instructions like "beq > > r1 r2 offset". > > In MIPS CodeGenerator::Branch function s5 and s6 registers are loaded > > with the values to be compared and the condition is a property of the > > CodeGenerator. I do this because when generating code comparisons are > > done before their result is evaluated to branch. > > However using this kind of solution for other functions would require > > to setup s5 and s6 every time we want to compare something. That would > > be really unefficient. > > I can't see another simple way to pass the arguments to functions like > > DeferredCodeBranch(). > > We could get rid of the preprocessor #ifdef by using the mips template > > for other architectures, which would then not use the extra arguments, > > but that would be ugly... > > > Feel free to ask more details if you need. > > > Alexandre
-- v8-users mailing list [email protected] http://groups.google.com/group/v8-users
