Vm-dev June 2016

vm-dev@lists.squeakfoundation.org

50 participants
129 discussions

StackInterp/Cog>>transferTo: - assertValidExecutionPointe:r:s differences
by Ben Coman 06 May '18

06 May '18

Just curious about the difference in instructionPointer between these two... StackInterpreter>>transferTo: self assertValidExecutionPointe: instructionPointer + 1 r: framePointer s: stackPointer. CoIntepreter>>transferTo:from: self assertValidExecutionPointe: instructionPointer r: framePointer s: stackPointer. cheers -ben

3 8

Amazing ARM simulator experience
by Eliot Miranda 03 May '17

03 May '17

Hi All, just had to tell people about this morning's experience using the ARM simulator. I've been building Smalltalk VMs since 1983, so 33 years. My first on the Three Rivers PERQ was dog slow. My undergraduate project was done the the National Semiconductor 32016 based Whitechapel Computer Works workstation and its and 32032-based successor. This morning I was revamping register management in Cog's ARMv5/v6 back end, making more registers available by using two of the C argument registers for two of registers the Cogit assigns to various fixed tasks, and using store and load multiple instructions to save and restore concisely the registers around calls into the runtime. Remember the architecture here. The simulator generates ARM machine code into the ByteArray that re[resents the address space, holding all of generated machine code, a small C stack, and the Smalltalk heap. A plugin, derived from Gdb's ARM simulator written in C, interprets that machine code, running for a couple of milliseconds at a time in a loop, applying breakpoint calculations, and asserts on every call into the run-time, and that calls into the run-time and accesses of variables in the simulator is done by using illegal addresses in the generated machine code. Each illegal access causes the Gdb-derived machine code interpreter to return with an error, this error is turned into an exception, the handler for which maps the specific illegal address into a variable or message selector, accesses the variable or activates the message selector, providing the result, and allowing the execution to continue. In changing the register management I had a test case that worked, an image that prompted for an expression ands evaluated it, which worked both in the simulator and with the generated VM. But the real VM, crashed when used on a proper image. So to debug I started launching the real interactive image in the simulator. Well, the amazing experience is that that image, whose machine code is being _interpreted by a C program_ feels /faster/ than my 32016 based implementation back in 1984-ish. Quite amazing. I can open windows, type text, access source code (was playing with Message Names) etc. It's sluggish, but usable. Amazing how fast modern machines are. All on my 2012 vintage 2.2 Ghz Core i7 MacBook Pro. I'm blown away :-) _,,,^..^,,,_ best, Eliot

5 7

error "please insert disk" (windows vm)
by Nicolai Hess 14 Jan '17

14 Jan '17

we have the following bug report 16928 <https://pharo.fogbugz.com/f/cases/16928/no-disk-in-the-drive-reply-to-World…> "no disk in the drive" reply to World | Save As (and other places) "On a clean Pharo4.0 #40618 image, on Windows 7 World Menu | Save As brings up a modal alert window, "There is no disk in the drive. Please insert a disk into drive <etc>" On my system, it does this 4 times. I'm assuming that each instance refers to one of the card-reader drives which are attached, but without a card inserted. This is a pain to deal with every single time a File Browser or World | Save As dialog is opened. Could the alerts either be - only raised when that specific drive is requested by a user action or, less preferably - given by a non-modal window, and fading away after a period" I was able to reproduce this error on squeaks and pharos window vm. (insert a usb card reader, with card, wait some time , remove only the card, every access on this drive, for example open a FileList and scroll to the drive letter). The solution I found is to call SetErrorMode( *SEM_FAILCRITICALERRORS)* I would like to propose the following change to sqwin32directory.c wrap the call to FindFirstFileW/FindNextFileW with UINT prevMode = SetErrorMode(SEM_FAILCRITICALERRORS); call FindFirstFileW/FindNextFileW SetErrorMode(prevMode); ... and call this functions in at least dir_lookup() (the other file/directory methods are save). Alternative solution, we could call the SetErrorMode function once at program start but I don't know if there are other "useful" errors message that we don't want to disable. I tested both solution, both seems to work fine. What do you think?

1 3

New case to crash Spur VM
by Denis Kudriashov 09 Dec '16

09 Dec '16

Hi In debugger (Squeak or Pharo) try "step over" for expression "Object basicSize": image will crash. We have opened issue for this in Pharo tracker 17962 <https://pharo.fogbugz.com/f/cases/17962/Image-is-crashed-when-step-over-Obj…>

5 12

Script needed; git mavens can you help?
by Eliot Miranda 27 Sep '16

27 Sep '16

Hi git mavens, I've used a script called svnrevert in the svn version that reverts any changed file in its argument(s), which may be directories, /and/ sets the modification date to the date at which the file was last checked in. I'd like the same functionality in git. Her'es the subversion version: #!/bin/bash # Revert file(s) or all modified files in a directory # and touch them back to the checkin date for f in "$@" do if [ -d "$f" ]; then $0 `svn st "$f" | grep "^M" | sed 's/^M *//'` else if svn revert "$f"; then changed="`svn info \"$f\" | grep 'Last Changed Date:' | sed 's/ *(.*//'`" touch -t "`date -j -f 'Last Changed Date: %Y-%m-%d %H:%M:%S %z' \"$changed\" '+%Y%m%d%H%M.%S'`" "$f" fi fi done _,,,^..^,,,_ best, Eliot

5 5

Cog + Pi + OSProcess
by Henrik Johansen 19 Aug '16

19 Aug '16

On the Cog VM shipped with latest Raspbian (3553, plugins 3543) running on a Pi3, after loading OSProcess, evaluating (PipeableOSProcess command: '/bin/echo this is a test') output segfaults the VM, crash.dmp indicating UnixOSProcessAccessor >> #forkAndExec: <snip> The shipped interpreter vm (4.10.2 platform rev2614) works with OSProcess, so I can use that instead, but has anyone encountered this already and happen to have a plugin built for Cog/Pi that allows command: use?. Cheers, Henry PS. Doing the layman's hack (copying over and renaming so.AioPlugin/so.UnixOSProcessPlugin from interpreter vm) does not work very well, no more crashes, but the PipeableOSProcess created has pid nil, status notYetRunning and an empty string output, so I assume a Cog-specific variant is needed.

8 28

ffi-test-main.c fails on FreeBSD
by Gary Dunn 12 Aug '16

12 Aug '16

In conjunction with getting Open Cobalt running on FreeBSD I have been looking into the problems with SqueakFFIPrims on FreeBSD and have reached a point that is beyond my understanding. I am not sure if the fix should come from the VM side or the FreeBSD port maintainer. I needed to get source code back into the ports tree so I did a make in /usr/ports/lang/squeak. I already had the 3.9 tarball, so that just repopulated the source and did the initial build. In the "files" folder I found patch files typical for a FreeBSD port. These patch source files from a generic tarball to work with FreeBSD. The one of interest was patch-platforms__unix__plugins__SqueakFFIPrims__ffi-config which contains ----x-----x----- --- platforms/unix/plugins/SqueakFFIPrims/ffi-config.org Wed Apr 26 20:27:53 2006 +++ platforms/unix/plugins/SqueakFFIPrims/ffi-config Wed Apr 26 20:29:00 2006 @@ -39,6 +39,7 @@ case ${abi} in linux) abi=sysv;; + freebsd*) abi=sysv;; darwin*) abi=darwin;; *) abi=libffi; lib="-lffi";; esac -----x-----x----- I then had the folder work/Squeak-3.9-7/platforms/unix/plugins/SqueakFFIPrims In it is a file 00README which explains how to test FFI. It says to run ffi-test-config but I have no such file. It says to type "make" to build a test suite but the build failed. It says to try with make CPU=any ABI=libffi LIB=-lffi which also failed. I traced the problem to the gcc search paths for includes and libraries -- turns out FreeBSD only searches /usr/include and /usr/lib when the required files are in /usr/local/include and /usr/local/lib. After some fiddling I finally had a "main" to test. It failed: ffi-test-main.c failed at line 361. Here is a bit from that file, ending with line 361: ffiInitialize(); for (ul= 0; ul < 15; ++ul) ffiPushSingleFloat(fa[ul]); GO(FFITypeSingleFloat, many); f= ffiReturnFloatValue(); ffiCleanup(); CHECK(f - ff < FLT_EPSILON); I worked out the CHECK macro but could not find what FLT_EPSILON is. At that point I decided to stop and come to you for help. Where do I go from here? -- Gary Dunn, Honolulu osp(a)aloha.com http://openslate.net/ http://e9erust.blogspot.com/ Sent from Slate001

5 8

Tasks of the heartbeat code
by Holger Freyther 27 Jul '16

27 Jul '16

Dear Eliot, I tried to understand the basic load of the VM and I have two questions for the heartbeat code. Is it right that there are two concerns for the code? 1.) Call forceInterruptCheckFromHeartbeat periodically 2.) Update the clock as seen by the image 1st is useful to interrupt Processes like […] repeat with a pre-emptive scheduler? On a different note do backwards jumps generate interrupt checks too? 2nd is this an optimization to not have to generate the clock when the primitive is being used? thanks holger

4 12

New Cog VMs available
by Eliot Miranda 14 Jul '16

14 Jul '16

... at http://www.mirandabanda.org/files/Cog/VM/VM.r3684 CogVM binaries as per VMMaker.oscog-eem.1834/r3684 General: Correct undue sign extension wile promoting 32 to 64 bits int in fetchLong64:ofObject: This was causing trouble in non spur object memory (V3). Fix slowdown of update on 64-bit x64 by using int variables for the tides in sqExternalSemaphores.c and inline assembler for the sqCompareAndSwap in sqAtomicOps.h. Fix slips in offset time primitives. They need to pop argumentCount + 1, not argumentCount. Define an optional primitive as the fast primitive fail code so that unselected optional primitives are in the prmitiveTable as fast primitive fails. Spur: Have WideString>>at: fail to answer an out-of-range character in the interpreter primitive (as well as in the JIT). Fix bug in following state on primitive failure. The old code would always follow to depth 1, even if the accessor depth was 0. Hard-code primitiveSize's depth to 0 (accessing length in at:[put:] and size causes the stack depth computation to answer 1 instead of 0 for these primitives. Fix assert and dequeueMourner for case where mournQueue is nil. - fixed a bug in receiver accessing in immutability primitive for mirror primitive. - change primitiveSetOrHasIdentityHash to patch the class table if the new hash is set to a behavior (the primitive knows it's a behavior if the second (optional) argument is true) For example: FullBlockClosure tryPrimitive: 161 withArgs: {38.true} No longer follow the method and context fields in a closure in activateNewClosureMethod:numArgs:mayContextSwitch:; the caller will have failed if these are forwarders, so no need to check again. Cogit: Add a primitive that answers pc map data for methods which can be used to better decorate methods in the VM Profiler. Refactor the pc map enumeration facilities so that the Sista pic data primitive can share the same enumerator. Do this by collapsing the isBackwardBranch and annotation parameters into a single parameter. Follow selectors in the openPICList post Spur become. Fix the ARM's caller-saved register mask now that we can name all the actual registers. Reworked machine code generation of immutability so for common stores it uses a single trampoline for both store checks and immutability checks. - improved support for register allocation: branch merge successfully compiled with register moved instead of spilling. Sista Cogit: Don't bother to add counters to conditional jumps implementing and: and or:. Added the remote inst var access bytecode in sista V1 bytecode set without interfering with existing code. Plugins: Upgrade LargeIntegersPlugin to v2.0 LargeInteger primitives now deal with 32-bits digits. No change to image code. Memory is 8 bytes aligned on Spur. When storing 32/64 bits large integers values, allways fill the eight bytes whatever the effectivily used size, rather than bother with dissertion of size. Generate integer type checking as C macros rather than direct/indirect interpreterProxy function call in plugins. This, and 32-bit accessing mean singificantly faster large integer arithmetic. Fix primAlienReplace to use positiveMachineIntegerValueOf: instead of positive32BitValueOf:. BitBltPlugin operates on 32bit word units, therefore it's better to declare its operands as 'unsigned int' rather than sqInt. On 32bit VM, this doesn't change anything, but on 64bits spur, it makes this snippet work: | wideString source pos blt expectedWideString | source := #[1 64 255 14 1 64 48 251]. expectedWideString := WideString fromByteArray: source. wideString := WideString new: source size // 4. pos := 0. blt := (BitBlt toForm: (Form new hackBits: wideString)) sourceForm: (Form new hackBits: source). blt combinationRule: Form over; sourceX: 0; sourceY: pos // 4; height: wideString byteSize // 4; width: 4; destX: 0; destY: 0; copyBits. wideString restoreEndianness. self assert: wideString = expectedWideString Hence it fixes loading/diffing MCZ with wide character. Mac OS Builds: Don't link with -fvisibility=hidden; it breaks external plugins. Use -fvisibility=default instead. This fixes e.g. UnixOSProcessPlugin on Mac OS X. Windows Builds: The Windows VMs are no longer "dpiAware". If you want one-to-one pixel mapping check the README for ins ructions. It's a simple edit. _,,,^..^,,,_ best, Eliot

7 20

Re: [Vm-dev] VM Maker: CMakeVMMakerSqueak-tty.126.mcz
by Holger Freyther 11 Jul '16

11 Jul '16

> On 18 Jun 2016, at 23:28, commits(a)source.squeak.org wrote: > > > Timothy M uploaded a new version of CMakeVMMakerSqueak to project VM Maker: > http://source.squeak.org/VMMaker/CMakeVMMakerSqueak-tty.126.mcz Hi, I might be teared and feathered but what would it take to maintain the CMakeLists.txt directly in git and not of generate them from Smalltalk? From my limited point of view it seems cmake supports targeting multiple platforms/configurations from within the CMakeLists. And when trying to build the PharoVM on FreeBSD I had a lot of trouble[1] and waited[2] a bit too much. kind regards holger [1] Trouble in the sense that the generated paths were absolute, that some CFLAGS didn't apply to the compiler used. [2] I seem to have failed to just generate the CMakeLists.txt/* and decided to generate everything which took some time. E.g. the files not being truncated before they are written to

3 9

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Vm-dev June 2016