Vm-dev March 2022

vm-dev@lists.squeakfoundation.org

14 participants
35 discussions

[OpenSmalltalk/opensmalltalk-vm] b558ef: Attempt to fix https://github.com/OpenSmalltalk/op...
by Nicolas Cellier 09 Mar '22

09 Mar '22

Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: b558efdd677dee45458c6d176b642b2b36f46e28 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/b558efdd677dee4545… Author: Nicolas Cellier <nicolas.cellier.aka.nice(a)gmail.com> Date: 2022-03-09 (Wed, 09 Mar 2022) Changed paths: M platforms/Cross/plugins/BitBltPlugin/BitBltInternal.h Log Message: ----------- Attempt to fix https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/617 I think that using a uint32_t for representing a negative pitch does not work on 64 bits architecture. It will lead to large positive offset (2^32-pitch) instead of negative (-pitch). We must use either a signed int, or a uint64_t. Let's choose usqInt which will work for both 32 and 64bits arch. Using a signed int would be a good option too, but without a careful analysis of potential signed arithmetic overflow, keep it safe, stay unsigned...

1 0

[OpenSmalltalk/opensmalltalk-vm] a50a6a: Increase stack headroom on all 64-bit Macs. A rec...
by Eliot Miranda 08 Mar '22

08 Mar '22

Branch: refs/heads/Cog Home: https://github.com/OpenSmalltalk/opensmalltalk-vm Commit: a50a6a770d7354270840cfc48f3fd09c041f6630 https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/a50a6a770d73542708… Author: Eliot Miranda <eliot.miranda(a)gmail.com> Date: 2022-03-07 (Mon, 07 Mar 2022) Changed paths: M platforms/iOS/vm/Common/Classes/sqSqueakMainApp.m M platforms/iOS/vm/Common/Classes/sqSqueakMainApplication.m Log Message: ----------- Increase stack headroom on all 64-bit Macs. A recent Vrtend startup crash (sorry Tim!) shows that x86_64 needs more headroom too. Have the reportheadroom flag obeyed on MacOS.

1 0

VM Maker: VMMaker-dtl.433.mcz
by commits＠source.squeak.org 07 Mar '22

07 Mar '22

David T. Lewis uploaded a new version of VMMaker to project VM Maker: http://source.squeak.org/VMMaker/VMMaker-dtl.433.mcz ==================== Summary ==================== Name: VMMaker-dtl.433 Author: dtl Time: 6 March 2022, 5:44:02.279 pm UUID: 1a9d1fbe-9459-4275-84b8-b668eed57607 Ancestors: VMMaker-dtl.432 VMMaker 4.19.14 Add the new BitBlt rules from oscog (Nicolas Cellier) Name: VMMaker.oscog-nice.2909 Author: nice Time: 23 December 2020, 9:54:51.201058 pm UUID: 6869c4f1-9d80-40c3-8409-7c05e087edf5 Ancestors: VMMaker.oscog-eem.2908 WIP: implement new BitBlt ops (rules) for alpha compositing "42" alphaScale:with: "43" alphaUnscale:with: "44" alphaBlendUnscaled:with: This is a proposal implementation for https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/505 The generated C code compiles, but it's otherwise mostly yet unused and untested code. I commit it before I loose it. There's probably more to do for letting accelerated code work (PI). Name: VMMaker.oscog-nice.3169 Author: nice Time: 24 February 2022, 1:53:42.760293 am UUID: e2ca1296-dc0c-418e-b1ac-7def7ff9a33d Ancestors: VMMaker.oscog-eem.3168 Fix alphaBlendUnscaled some bits were leaking in the neighbour color. Name: VMMaker.oscog-nice.3170 Author: nice Time: 24 February 2022, 2:48:39.990237 pm UUID: 6b851cd8-b485-1047-acfd-8adee37e3660 Ancestors: VMMaker.oscog-nice.3169 Revise the 3 new BitBlt rules again (see https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/505) - use a proper division by 255 (factor alpha/255) in alphaScale:with: instead of approximate factor alpha/256 previously - use a proper rounding (by adding 16r80 before the division by 255) in alphaScale:with: - split the unscale ops in 3 separate r/g/b channels, otherwise, when grouping rb, the division by alpha will leak red bits not only on green components but also on blue - fixup the carry used for saturating the rbg components in case of 16rFF overflow in alphaUnscale:with: Note: division by 255 can be checked thru this snippet (should answer 0): (0 to: 16rFF squared) inject: 0 into: [:max :num | | rat | rat := num + 16r80. "+ 16r80 for rounding" rat := rat + (rat - 1 >> 8 bitAnd: 16rFF) >> 8 bitAnd: 16rFF. "divide by 255" ((num/255) rounded - rat) abs max: max]. =============== Diff against VMMaker-dtl.432 =============== Item was changed: SmartSyntaxInterpreterPlugin subclass: #BitBltSimulation + instanceVariableNames: 'destForm sourceForm halftoneForm combinationRule destX destY width height sourceX sourceY clipX clipY clipWidth clipHeight sourceWidth sourceHeight sourceDepth sourcePitch sourceBits sourcePPW sourceMSB destWidth destHeight destDepth destPitch destBits destPPW destMSB bitCount skew mask1 mask2 preload nWords destMask hDir vDir sourceIndex sourceDelta destIndex destDelta sx sy dx dy bbW bbH halftoneHeight noSource noHalftone halftoneBase sourceAlpha srcBitShift dstBitShift bitBltOop affectedL affectedR affectedT affectedB opTable maskTable ditherMatrix4x4 ditherThresholds16 ditherValues16 hasSurfaceLock warpSrcShift warpSrcMask warpAlignShift warpAlignMask warpBitShiftTable querySurfaceFn lockSurfaceFn unlockSurfaceFn isWarping cmFlags cmMask cmShiftTable cmMaskTable cmLookupTable cmBitsPerColor dither8Lookup componentAlphaModeColor componentAlphaModeAlpha ungammaLookupTable gammaLookupTable numGCsOnInvocation bitBltIsReceiver endOfDestination endOfSource' + classVariableNames: 'AllOnes AlphaIndex BBClipHeightIndex BBClipWidthIndex BBClipXIndex BBClipYIndex BBColorMapIndex BBDestFormIndex BBDestXIndex BBDestYIndex BBHalftoneFormIndex BBHeightIndex BBLastIndex BBRuleIndex BBSourceFormIndex BBSourceXIndex BBSourceYIndex BBWarpBase BBWidthIndex BBXTableIndex BEBitBltIndex BinaryPoint BlueIndex ColorMapFixedPart ColorMapIndexedPart ColorMapNewStyle ColorMapPresent CrossedX EndOfRun FixedPt1 FormBitsIndex FormDepthIndex FormHeightIndex FormWidthIndex GreenIndex OpTable OpTableSize RedIndex' - instanceVariableNames: 'destForm sourceForm halftoneForm combinationRule destX destY width height sourceX sourceY clipX clipY clipWidth clipHeight sourceWidth sourceHeight sourceDepth sourcePitch sourceBits sourcePPW sourceMSB destWidth destHeight destDepth destPitch destBits destPPW destMSB bitCount skew mask1 mask2 preload nWords destMask hDir vDir sourceIndex sourceDelta destIndex destDelta sx sy dx dy bbW bbH halftoneHeight noSource noHalftone halftoneBase sourceAlpha srcBitShift dstBitShift bitBltOop affectedL affectedR affectedT affectedB opTable maskTable ditherMatrix4x4 ditherThresholds16 ditherValues16 hasSurfaceLock warpSrcShift warpSrcMask warpAlignShift warpAlignMask warpBitShiftTable querySurfaceFn lockSurfaceFn unlockSurfaceFn isWarping cmFlags cmMask cmShiftTable cmMaskTable cmLookupTable cmBitsPerColor dither8Lookup componentAlphaModeColor componentAlphaModeAlpha ungammaLookupTable gammaLookupTable' - classVariableNames: 'AllOnes AlphaIndex BBClipHeightIndex BBClipWidthIndex BBClipXIndex BBClipYIndex BBColorMapIndex BBDestFormIndex BBDestXIndex BBDestYIndex BBHalftoneFormIndex BBHeightIndex BBLastIndex BBRuleIndex BBSourceFormIndex BBSourceXIndex BBSourceYIndex BBWarpBase BBWidthIndex BBXTableIndex BinaryPoint BlueIndex ColorMapFixedPart ColorMapIndexedPart ColorMapNewStyle ColorMapPresent CrossedX EndOfRun FixedPt1 FormBitsIndex FormDepthIndex FormHeightIndex FormWidthIndex GreenIndex JitBltHookSize OpTable OpTableSize RedIndex' poolDictionaries: '' category: 'VMMaker-Interpreter'! + !BitBltSimulation commentStamp: 'nice 10/31/2020 23:39' prior: 0! - !BitBltSimulation commentStamp: 'tpr 3/25/2013 16:50' prior: 0! This class implements BitBlt, much as specified in the Blue Book spec. Performance has been enhanced through the use of pointer variables such as sourceIndex and destIndex, and by separating several special cases of the inner loop. Operation has been extended to color, with support for 1, 2, 4, 8, 16, and 32-bit pixel sizes. Conversion between different pixel sizes is facilitated by accepting an optional color map. In addition to the original 16 combination rules, this BitBlt supports 16 fail (for old paint mode) 17 fail (for old mask mode) 18 sourceWord + destinationWord 19 sourceWord - destinationWord 20 rgbAdd: sourceWord with: destinationWord 21 rgbSub: sourceWord with: destinationWord 22 OLDrgbDiff: sourceWord with: destinationWord 23 OLDtallyIntoMap: destinationWord -- old vers doesn't clip to bit boundary 24 alphaBlend: sourceWord with: destinationWord 25 pixPaint: sourceWord with: destinationWord 26 pixMask: sourceWord with: destinationWord 27 rgbMax: sourceWord with: destinationWord 28 rgbMin: sourceWord with: destinationWord 29 rgbMin: sourceWord bitInvert32 with: destinationWord 30 alphaBlendConst: sourceWord with: destinationWord -- alpha passed as an arg 31 alphaPaintConst: sourceWord with: destinationWord -- alpha passed as an arg 32 rgbDiff: sourceWord with: destinationWord 33 tallyIntoMap: destinationWord 34 alphaBlendScaled: sourceWord with: destinationWord 35 alphaBlendScaled: sourceWord with: "unused here - only used by FXBlt" 36 alphaBlendScaled: sourceWord with: "unused here - only used by FXBlt" 37 rgbMul: sourceWord with: destinationWord 38 pixSwap: sourceWord with: destinationWord 39 pixClear: sourceWord with: destinationWord 40 fixAlpha: sourceWord with: destinationWord 41 rgbComponentAlpha: sourceWord with: destinationWord + 42 alphaScale: ignoredSourceWord with: destinationWord + 43 alphaUnscale: ignoredSourceWord with: destinationWord + 44 alphaBlendUnscaled: sourceWord with: destinationWord This implementation has also been fitted with an experimental "warp drive" that allows abritrary scaling and rotation (and even limited affine deformations) with all BitBlt storage modes supported. To add a new rule to BitBlt... 1. add the new rule method or methods in the category 'combination rules' of BBSim 2. describe it in the class comment of BBSim and in the class comment for BitBlt 3. add refs to initializeRuleTable in proper positions 4. add refs to initBBOpTable, following the pattern ! Item was changed: ----- Method: BitBltSimulation class>>initializeRuleTable (in category 'initialization') ----- initializeRuleTable "BitBltSimulation initializeRuleTable" "**WARNING** You MUST change initBBOpTable if you change this" OpTable := #( "0" clearWord:with: "1" bitAnd:with: "2" bitAndInvert:with: "3" sourceWord:with: "4" bitInvertAnd:with: "5" destinationWord:with: "6" bitXor:with: "7" bitOr:with: "8" bitInvertAndInvert:with: "9" bitInvertXor:with: "10" bitInvertDestination:with: "11" bitOrInvert:with: "12" bitInvertSource:with: "13" bitInvertOr:with: "14" bitInvertOrInvert:with: "15" destinationWord:with: "16" destinationWord:with: "unused - was old paint" "17" destinationWord:with: "unused - was old mask" "18" addWord:with: "19" subWord:with: "20" rgbAdd:with: "21" rgbSub:with: "22" OLDrgbDiff:with: "23" OLDtallyIntoMap:with: "24" alphaBlend:with: "25" pixPaint:with: "26" pixMask:with: "27" rgbMax:with: "28" rgbMin:with: "29" rgbMinInvert:with: "30" alphaBlendConst:with: "31" alphaPaintConst:with: "32" rgbDiff:with: "33" tallyIntoMap:with: "34" alphaBlendScaled:with: "35" alphaBlendScaled:with: "unused here - only used by FXBlt" "36" alphaBlendScaled:with: "unused here - only used by FXBlt" "37" rgbMul:with: "38" pixSwap:with: "39" pixClear:with: "40" fixAlpha:with: "41" rgbComponentAlpha:with: + "42" alphaScale:with: + "43" alphaUnscale:with: + "44" alphaBlendUnscaled:with: ). OpTableSize := OpTable size + 1. "0-origin indexing" ! Item was added: + ----- Method: BitBltSimulation>>alphaBlendUnscaled:with: (in category 'combination rules') ----- + alphaBlendUnscaled: sourceWord with: destinationWord + "Blend sourceWord with destinationWord using the alpha value from both sourceWord and destinationWord. + Alpha is encoded as 0 meaning 0.0, and 255 meaning 1.0. + The alpha channel and color produced are + + srcAlpha + (destAlpha*(1-srcAlpha)) + (srcAlpha*srcColor + (destAlpha*(1-srcAlpha)*dstColor)) / (srcAlpha + (destAlpha*(1-srcAlpha))) + + In contrast to alphaBlend:with: the method does not assume that destination form is opaque. + In contrast to alphaBlendScaled:with: the method does not assume that colors have been pre-scaled (muliplied) by alpha channel." + | alpha blendA result blendR blendB blendG | + <inline: false> + <returnTypeC: 'unsigned int'> + <var: 'sourceWord' type: #'unsigned int'> + <var: 'destinationWord' type: #'unsigned int'> + <var: 'blendA' type: #'unsigned int'> + <var: 'blendR' type: #'unsigned int'> + <var: 'blendG' type: #'unsigned int'> + <var: 'blendB' type: #'unsigned int'> + <var: 'result' type: #'unsigned int'> + <var: 'alpha' type: #'unsigned int'> + alpha := sourceWord >> 24. "High 8 bits of source pixel, assuming ARGB encoding" + alpha = 0 ifTrue: [ ^ destinationWord ]. + alpha = 255 ifTrue: [ ^ sourceWord ]. + + blendA := 16rFF * alpha + (16rFF - alpha * (destinationWord >> 24)) + 16r80. "blend alpha channels" + blendA := blendA + (blendA - 1 >> 8 bitAnd: 16rFF) >> 8 bitAnd: 16rFF. "divide by 255" + + blendR := ((sourceWord bitAnd: 16rFF0000) * alpha) + + ((destinationWord bitAnd: 16rFF0000) * (blendA-alpha)) + +(blendA<<15) + // blendA bitAnd: 16rFF0000. "blend red" + + blendG := ((sourceWord bitAnd: 16r00FF00) * alpha) + + ((destinationWord bitAnd: 16r00FF00) * (blendA-alpha)) + +(blendA<<7) + // blendA bitAnd: 16r00FF00. "blend green" + + blendB := ((sourceWord bitAnd: 16r0000FF) * alpha) + + ((destinationWord bitAnd: 16r0000FF) * (blendA-alpha)) + +(blendA>>1) + // blendA bitAnd: 16r0000FF. "blend blue" + + result := ((blendR bitOr: blendB) bitOr: blendG) bitOr: blendA << 24. + ^ result + ! Item was added: + ----- Method: BitBltSimulation>>alphaScale:with: (in category 'combination rules') ----- + alphaScale: sourceWord with: destinationWord + "Scale (premultiply) the destination with its alpha channel. + Note that sourceWord is ignored." + | alpha rb g | + <inline: false> "Do NOT inline this into optimized loops" + <returnTypeC: 'unsigned int'> + <var: 'sourceWord' type: #'unsigned int'> + <var: 'destinationWord' type: #'unsigned int'> + <var: 'rb' type: #'unsigned int'> + <var: 'g' type: #'unsigned int'> + <var: 'alpha' type: #'unsigned int'> + alpha := destinationWord >> 24. "High 8 bits is opacity (ARGB format)" + rb := (destinationWord bitAnd: 16rFF00FF) * alpha + 16r800080. "scale red and blue components" + rb := rb + (rb - 16r010001 >> 8 bitAnd: 16rFF00FF) >> 8 bitAnd: 16rFF00FF. "divide by 255" + g := (destinationWord bitAnd: 16r00FF00) * alpha + 16r008000. "scale green component" + g := g + (g - 16r000100 >> 8 bitAnd: 16r00FF00) >> 8 bitAnd: 16r00FF00. "divide by 255" + ^(g bitOr: rb) bitOr: (destinationWord bitAnd: 16rFF000000) "recompose"! Item was added: + ----- Method: BitBltSimulation>>alphaUnscale:with: (in category 'combination rules') ----- + alphaUnscale: sourceWord with: destinationWord + "Unscale (divide) the destination with its alpha channel. + Note that sourceWord is ignored." + | alpha r b g rgb carry | + <inline: false> "Do NOT inline this into optimized loops" + <returnTypeC: 'unsigned int'> + <var: 'sourceWord' type: #'unsigned int'> + <var: 'destinationWord' type: #'unsigned int'> + <var: 'r' type: #'unsigned int'> + <var: 'g' type: #'unsigned int'> + <var: 'b' type: #'unsigned int'> + <var: 'rgb' type: #'unsigned int'> + <var: 'alpha' type: #'unsigned int'> + <var: 'carry' type: #'unsigned int'> + alpha := destinationWord >> 24. "High 8 bits is opacity (ARGB format)" + alpha = 0 ifTrue: [^0]. + r := (destinationWord bitAnd: 16rFF0000) * 255 + (alpha+1<<15) // alpha. "unscale red components" + g := (destinationWord bitAnd: 16r00FF00) * 255 + (alpha+1<<7) // alpha. "unscale green component" + b := (destinationWord bitAnd: 16r0000FF) * 255 + (alpha+1>>1) // alpha. "unscale blue components" + carry := (((r bitAnd: 16rFF000000) bitOr: (g bitAnd: 16rFF0000)) bitOr: (b bitAnd: 16rFF00)) >> 8. + carry := carry bitOr: ((carry bitAnd: 16rAAAAAA) >> 1 bitOr: (carry bitAnd: 16r555555) << 1). + carry := carry bitOr: ((carry bitAnd: 16rCCCCCC) >> 2 bitOr: (carry bitAnd: 16r333333) << 2). + carry := carry bitOr: ((carry bitAnd: 16rF0F0F0) >> 4 bitOr: (carry bitAnd: 16r0F0F0F) << 4). + rgb := ((r bitAnd: 16rFF0000) bitOr: (g bitAnd: 16r00FF00)) bitOr: (b bitAnd: 16r0000FF). + rgb := rgb bitOr: carry. "saturate RGB components if division overflows" + ^rgb bitOr: (destinationWord bitAnd: 16rFF000000) "restore alpha"! Item was changed: ----- Method: BitBltSimulation>>initBBOpTable (in category 'initialize-release') ----- initBBOpTable self cCode: 'opTable[0+1] = (void *)clearWordwith'. self cCode: 'opTable[1+1] = (void *)bitAndwith'. self cCode: 'opTable[2+1] = (void *)bitAndInvertwith'. self cCode: 'opTable[3+1] = (void *)sourceWordwith'. self cCode: 'opTable[4+1] = (void *)bitInvertAndwith'. self cCode: 'opTable[5+1] = (void *)destinationWordwith'. self cCode: 'opTable[6+1] = (void *)bitXorwith'. self cCode: 'opTable[7+1] = (void *)bitOrwith'. self cCode: 'opTable[8+1] = (void *)bitInvertAndInvertwith'. self cCode: 'opTable[9+1] = (void *)bitInvertXorwith'. self cCode: 'opTable[10+1] = (void *)bitInvertDestinationwith'. self cCode: 'opTable[11+1] = (void *)bitOrInvertwith'. self cCode: 'opTable[12+1] = (void *)bitInvertSourcewith'. self cCode: 'opTable[13+1] = (void *)bitInvertOrwith'. self cCode: 'opTable[14+1] = (void *)bitInvertOrInvertwith'. self cCode: 'opTable[15+1] = (void *)destinationWordwith'. self cCode: 'opTable[16+1] = (void *)destinationWordwith'. self cCode: 'opTable[17+1] = (void *)destinationWordwith'. self cCode: 'opTable[18+1] = (void *)addWordwith'. self cCode: 'opTable[19+1] = (void *)subWordwith'. self cCode: 'opTable[20+1] = (void *)rgbAddwith'. self cCode: 'opTable[21+1] = (void *)rgbSubwith'. self cCode: 'opTable[22+1] = (void *)OLDrgbDiffwith'. self cCode: 'opTable[23+1] = (void *)OLDtallyIntoMapwith'. self cCode: 'opTable[24+1] = (void *)alphaBlendwith'. self cCode: 'opTable[25+1] = (void *)pixPaintwith'. self cCode: 'opTable[26+1] = (void *)pixMaskwith'. self cCode: 'opTable[27+1] = (void *)rgbMaxwith'. self cCode: 'opTable[28+1] = (void *)rgbMinwith'. self cCode: 'opTable[29+1] = (void *)rgbMinInvertwith'. self cCode: 'opTable[30+1] = (void *)alphaBlendConstwith'. self cCode: 'opTable[31+1] = (void *)alphaPaintConstwith'. self cCode: 'opTable[32+1] = (void *)rgbDiffwith'. self cCode: 'opTable[33+1] = (void *)tallyIntoMapwith'. self cCode: 'opTable[34+1] = (void *)alphaBlendScaledwith'. self cCode: 'opTable[35+1] = (void *)alphaBlendScaledwith'. self cCode: 'opTable[36+1] = (void *)alphaBlendScaledwith'. self cCode: 'opTable[37+1] = (void *)rgbMulwith'. self cCode: 'opTable[38+1] = (void *)pixSwapwith'. self cCode: 'opTable[39+1] = (void *)pixClearwith'. self cCode: 'opTable[40+1] = (void *)fixAlphawith'. + self cCode: 'opTable[41+1] = (void *)rgbComponentAlphawith'. + self cCode: 'opTable[42+1] = (void *)alphaScalewith'. + self cCode: 'opTable[43+1] = (void *)alphaUnscalewith'. + self cCode: 'opTable[44+1] = (void *)alphaBlendUnscaledwith'.! - self cCode: 'opTable[41+1] = (void *)rgbComponentAlphawith'.! Item was changed: ----- Method: VMMaker class>>versionString (in category 'version testing') ----- versionString "VMMaker versionString" + ^'4.19.14'! - ^'4.19.13'!

1 0

ArmV8 linux crash in MVC issue 617
by Bruce O'Neel 05 Mar '22

05 Mar '22

Hi all, So I spent a few hours last night and a few more today running the broken code in Arm32 and Arm64. I don't see the problem. I can not decide if it is in the the BitBlt plugin built by cog, though I think not, nor in the lower level code called to actually move bits because that code is called from so many places and runs well. The bug only happens on Armv8 linux because the bug is in code that is only built on Arm* linux. The code only happens in MVC with a 32 bit screen depth. But this is because the code path is taken only for 32 bit screen depths. So.... We would like to release a new version of Squeak soonish. I think we have a few choices: 1. Assume that the combo of ArmV8 and MVC will be rare enough that almost no one will stumble on it. If they do there is a workaround of using some other screen depth in MVC. A touch ugly but to most of today's youth I'll bet they already thought MVC was ugly and won't use it. 2. Put my patch in place for the moment. In both cases I want to keep the issue open so it is not forgotten. I would like to fix it I just don't see the problem. Thoughts? cheers bruce

2 1

64 Bit Raspian
by ken.dickey＠whidbey.com 03 Mar '22

03 Mar '22

I noted that Raspberry Pi finally runs a 64bit version of Raspian Linux. Both the Cog and Stack VMs build and seem to run fine w current. Thanks again to all! FYI, -KenD

3 4

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

Vm-dev March 2022