73530 2011-11-30 23:35:24 -0800 Optimize to not use pow() in the inner loop in AudioParamTimeline 2011-12-13 20:57:17 -0800 1 1 1 Unclassified WebKit Web Audio 528+ (Nightly build) Unspecified Unspecified RESOLVED FIXED P2 Normal --- 1 james.wei james.wei crogers james.wei kbr rtoy webkit.review.bot oldest_to_newest 512532 0 james.wei 2011-11-30 23:35:24 -0800 Optimize to not use pow() in the inner loop 512545 1 mrowe 2011-11-30 23:56:45 -0800 Can you be a little more specific as to what you’d like to optimize to not use pow() in the inner loop? 512559 2 117360 james.wei 2011-12-01 00:14:02 -0800 Created attachment 117360 optimize to not use pow in the inner loop 512561 3 james.wei 2011-12-01 00:17:02 -0800 in AudioParamTimeline::valuesForTimeRangeImpl to handle AudioEvent, it is mentioned to optimize to not use pow() in the inner loop as it is just a simple exponential ramp. this patch to implement this optimization. 512577 4 james.wei 2011-12-01 00:30:16 -0800 (In reply to comment #1) > Can you be a little more specific as to what you’d like to optimize to not use pow() in the inner loop? an error occurs when I use webkit-patch to create this bug entry and leaves just a simple description. Sorry for it. I have added a more detailed information and patch. thanks 512817 5 rtoy 2011-12-01 09:03:13 -0800 Thanks for the patch. I have not yet examined it, but note that there is some discussion in the audio w3c group (http://lists.w3.org/Archives/Public/public-audio/2011OctDec/0107.html) about what the ramp is supposed to do. The issue is that the starting value cannot currently be 0, because the ramp is, essentially, value1 * exp(t/T). The implementation of the ramp may change. 512946 6 117360 kbr 2011-12-01 11:39:04 -0800 Comment on attachment 117360 optimize to not use pow in the inner loop Raymond and I both reviewed the math and it looks correct. Despite the fact that the ramp function might be redefined soon, I think it's worth adding this optimization. I think the patch needs to be rebaselined to top of tree though. Could you take care of that? r- just for this issue. 512986 7 rtoy 2011-12-01 12:18:53 -0800 Also, can you add a comment on what the formula actually is and what the optimization is now doing to the formula? I think the original is v(t) = 2^((1-x(t))*v1 + x(t)*v2) with x(t) = (t-t1)*k You've changed this to be v(t+dt) = v(t)*2^((v2-v1)*k) roughly. This makes it easier to tell that the math is right. 513264 8 117522 james.wei 2011-12-01 16:41:05 -0800 Created attachment 117522 optimize to not use pow in the inner loop -- updated Fix patch top tree issue and add comments to explain the formula 513267 9 james.wei 2011-12-01 16:46:36 -0800 (In reply to comment #5) > Thanks for the patch. I have not yet examined it, but note that there is some discussion in the audio w3c group (http://lists.w3.org/Archives/Public/public-audio/2011OctDec/0107.html) about what the ramp is supposed to do. The issue is that the starting value cannot currently be 0, because the ramp is, essentially, value1 * exp(t/T). The implementation of the ramp may change. Raymond, thanks for the information. I noticed the notes Alistair sent out. If the final exponential ramp algorithm finalized, I have the interest to implement it then. thanks 513270 10 james.wei 2011-12-01 16:49:38 -0800 (In reply to comment #6) > (From update of attachment 117360 [details]) > Raymond and I both reviewed the math and it looks correct. Despite the fact that the ramp function might be redefined soon, I think it's worth adding this optimization. > > I think the patch needs to be rebaselined to top of tree though. Could you take care of that? r- just for this issue. Ken, thanks for the review. I have updated the patch based on the top of WebKit tree. Also, I added some comments to explain the algorithm of optimization. 513276 11 117522 kbr 2011-12-01 16:51:41 -0800 Comment on attachment 117522 optimize to not use pow in the inner loop -- updated Thanks for the update. Looks good. Please use "webkit-patch upload" in the future to upload patches as it will streamline some of the process. If you'd like this committed please mark it "cq?". r=me 513290 12 james.wei 2011-12-01 17:03:39 -0800 (In reply to comment #11) > (From update of attachment 117522 [details]) > Thanks for the update. Looks good. Please use "webkit-patch upload" in the future to upload patches as it will streamline some of the process. If you'd like this committed please mark it "cq?". r=me Ken, thanks for the review and the remind. I have marked it as cq?. could you please help to commit it? thanks 513298 13 kbr 2011-12-01 17:06:24 -0800 (In reply to comment #12) > (In reply to comment #11) > > (From update of attachment 117522 [details] [details]) > > Thanks for the update. Looks good. Please use "webkit-patch upload" in the future to upload patches as it will streamline some of the process. If you'd like this committed please mark it "cq?". r=me > > Ken, thanks for the review and the remind. I have marked it as cq?. could you please help to commit it? thanks I'll mark it r+ as soon as it passes the Chromium Linux EWS bot. 513299 14 kbr 2011-12-01 17:06:35 -0800 (In reply to comment #13) > (In reply to comment #12) > > (In reply to comment #11) > > > (From update of attachment 117522 [details] [details] [details]) > > > Thanks for the update. Looks good. Please use "webkit-patch upload" in the future to upload patches as it will streamline some of the process. If you'd like this committed please mark it "cq?". r=me > > > > Ken, thanks for the review and the remind. I have marked it as cq?. could you please help to commit it? thanks > > I'll mark it r+ as soon as it passes the Chromium Linux EWS bot. Er, I meant cq+. 513458 15 james.wei 2011-12-01 21:08:55 -0800 (In reply to comment #13) > (In reply to comment #12) > > (In reply to comment #11) > > > (From update of attachment 117522 [details] [details] [details]) > > > Thanks for the update. Looks good. Please use "webkit-patch upload" in the future to upload patches as it will streamline some of the process. If you'd like this committed please mark it "cq?". r=me > > > > Ken, thanks for the review and the remind. I have marked it as cq?. could you please help to commit it? thanks > > I'll mark it r+ as soon as it passes the Chromium Linux EWS bot. thanks. 513629 16 crogers 2011-12-02 02:18:42 -0800 Can we get a layout test to validate? The currentTime var is no longer getting updated - may not be correct... 513634 17 crogers 2011-12-02 02:30:44 -0800 Wei thanks for the patch! I'm on vacation right now, but would like refine this a bit more when I get back (hope you can wait). Btw the starting at zero issue is not a bug so I dont't expect the math to change. 513695 18 james.wei 2011-12-02 05:36:52 -0800 (In reply to comment #16) > Can we get a layout test to validate? The currentTime var is no longer getting updated - may not be correct... Roger, thanks for the comments. Is there any reference to write layout test for webaudio AudioParam? I have tried to write a layout test but failed to find a reference and don't know how to use the layout test to verify the change. So I just wrote some unit tests to verify the correctness of the optimization algorithm and test the performance of the function. for the var currentTime, it is a local variable and not used after the if statement for ExponentialRampToValue, so I just ignore it. you can double check it. thanks 513698 19 james.wei 2011-12-02 05:42:00 -0800 (In reply to comment #17) > Wei thanks for the patch! I'm on vacation right now, but would like refine this a bit more when I get back (hope you can wait). Btw the starting at zero issue is not a bug so I dont't expect the math to change. Roger, that's ok for me. It is my honor if I can contribute to webaudio. I am very interested in this area but still a newbie. Hope I can be an expert on this area and contribute more to it in future. thanks 514016 20 117522 kbr 2011-12-02 12:41:25 -0800 Comment on attachment 117522 optimize to not use pow in the inner loop -- updated Since Chris indicated this needs more work I'm changing the review state to r-. Let's iterate more next week. 514387 21 crogers 2011-12-02 19:59:50 -0800 (In reply to comment #19) > (In reply to comment #17) > > Wei thanks for the patch! I'm on vacation right now, but would like refine this a bit more when I get back (hope you can wait). Btw the starting at zero issue is not a bug so I dont't expect the math to change. > > Roger, that's ok for me. It is my honor if I can contribute to webaudio. I am very interested in this area but still a newbie. Hope I can be an expert on this area and contribute more to it in future. thanks James, thanks for you patience. I'm getting back around Dec. 11 and look forward to working more with you. Thanks for your help! 518953 22 james.wei 2011-12-11 17:36:40 -0800 (In reply to comment #21) > (In reply to comment #19) > > (In reply to comment #17) > > > Wei thanks for the patch! I'm on vacation right now, but would like refine this a bit more when I get back (hope you can wait). Btw the starting at zero issue is not a bug so I dont't expect the math to change. > > > > Roger, that's ok for me. It is my honor if I can contribute to webaudio. I am very interested in this area but still a newbie. Hope I can be an expert on this area and contribute more to it in future. thanks > > James, thanks for you patience. I'm getting back around Dec. 11 and look forward to working more with you. Thanks for your help! roger, I have marked the patch as r?, could you help to review it? thanks 519026 23 crogers 2011-12-11 22:56:24 -0800 I'll try to have a look tomorrow. 519644 24 117522 crogers 2011-12-12 16:10:13 -0800 Comment on attachment 117522 optimize to not use pow in the inner loop -- updated View in context: https://bugs.webkit.org/attachment.cgi?id=117522&action=review > Source/WebCore/webaudio/AudioParamTimeline.cpp:262 > + // So v(t+dt) = v(t) * 2 ^ (k * sampleFrameTimeIncr * (v2 - v1)) This can be made quite a bit simpler if you get rid of the log2() conversions on lines 249:250 and do the math directly in linear space. Then you can have something like: double numSampleFrames = deltaTime * sampleRate; double multiplier = pow(value2 / value1, 1 / numSampleFrames); Note: I really wish we still were using double-precision here, but to avoid warnings caused by 64bit -> 32bit conversion (double -> float) we switched a lot of this code over to use float. Later to avoid accumulated error we probably need to look at this a bit more closely and switch some code back to using doubles. but (at least for now), we should calculate "numSampleFrames" and "multiplier" as float > Source/WebCore/webaudio/AudioParamTimeline.cpp:269 > + writeIndex++; I would remove lines 264:269 > Source/WebCore/webaudio/AudioParamTimeline.cpp:273 > + double multiplier = powf(2.0f, k * sampleFrameTimeIncr * (value2 - value1)); and calculate multiplier as I suggest above (but using float) > Source/WebCore/webaudio/AudioParamTimeline.cpp:276 > + values[writeIndex] = temp; We need to maintain/update the two variables "value" and "currentTime" since they're propagated to other parts of the code within this loop. I would suggest this code for the inner loop: values[writeIndex] = value; value *= multiplier; currentTime += sampleFrameTimeIncr; *** actually the "currentTime" updating can be pulled outside of the loop here and in the linear case... 519732 25 117522 james.wei 2011-12-12 17:28:51 -0800 Comment on attachment 117522 optimize to not use pow in the inner loop -- updated View in context: https://bugs.webkit.org/attachment.cgi?id=117522&action=review >> Source/WebCore/webaudio/AudioParamTimeline.cpp:262 >> + // So v(t+dt) = v(t) * 2 ^ (k * sampleFrameTimeIncr * (v2 - v1)) > > This can be made quite a bit simpler if you get rid of the log2() conversions on lines 249:250 and do the math directly in linear space. > > Then you can have something like: > > double numSampleFrames = deltaTime * sampleRate; > double multiplier = pow(value2 / value1, 1 / numSampleFrames); > > > Note: I really wish we still were using double-precision here, but to avoid warnings caused by 64bit -> 32bit conversion (double -> float) > we switched a lot of this code over to use float. Later to avoid accumulated error we probably need to look at this a bit more closely and switch some code back to using doubles. > > but (at least for now), we should calculate "numSampleFrames" and "multiplier" as float thanks, I will follow this to update the patch. thanks >> Source/WebCore/webaudio/AudioParamTimeline.cpp:269 >> + writeIndex++; > > I would remove lines 264:269 A question: if we remove this block, the first value will be defaultValue. is it correct? >> Source/WebCore/webaudio/AudioParamTimeline.cpp:273 >> + double multiplier = powf(2.0f, k * sampleFrameTimeIncr * (value2 - value1)); > > and calculate multiplier as I suggest above (but using float) I will do it. thanks >> Source/WebCore/webaudio/AudioParamTimeline.cpp:276 >> + values[writeIndex] = temp; > > We need to maintain/update the two variables "value" and "currentTime" since they're propagated to other parts of the code within this loop. > > I would suggest this code for the inner loop: > > values[writeIndex] = value; > value *= multiplier; > currentTime += sampleFrameTimeIncr; > > > *** actually the "currentTime" updating can be pulled outside of the loop here and in the linear case... got it. thanks 519738 26 crogers 2011-12-12 17:40:06 -0800 > > I would remove lines 264:269 > > A question: if we remove this block, the first value will be defaultValue. is it correct? > value is initialized to defaultValue at the top of the function, but also its value can change multiple times through the loop as different segments are processed. So, it should be already at the correct value. 519822 27 118942 james.wei 2011-12-12 19:52:01 -0800 Created attachment 118942 Patch 519823 28 james.wei 2011-12-12 19:52:57 -0800 (In reply to comment #26) > > > I would remove lines 264:269 > > > > A question: if we remove this block, the first value will be defaultValue. is it correct? > > > > value is initialized to defaultValue at the top of the function, but also its value can change multiple times through the loop as different segments are processed. > So, it should be already at the correct value. thanks, I have updated the patch. please help to review. thanks 519832 29 118942 crogers 2011-12-12 20:10:59 -0800 Comment on attachment 118942 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=118942&action=review > Source/WebCore/webaudio/AudioParamTimeline.cpp:267 > + float multiplier = powf(value2 / value1, 1 / numSampleFrames); We should handle the case where value1 is zero to avoid divide-by-zero, and make multiplier zero in this case. I think we should remove the comments on lines 249:259 and 261:266 and have a very simple one or two line comment directly describing the math in 267, something like: // The value goes exponentially from value1 to value2 in a duration of deltaTime seconds (corresponding to numSampleFrames). // Compute the per-sample multiplier. 519846 30 118942 james.wei 2011-12-12 21:12:11 -0800 Comment on attachment 118942 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=118942&action=review >> Source/WebCore/webaudio/AudioParamTimeline.cpp:267 >> + float multiplier = powf(value2 / value1, 1 / numSampleFrames); > > We should handle the case where value1 is zero to avoid divide-by-zero, and make multiplier zero in this case. > > I think we should remove the comments on lines 249:259 and 261:266 and have a very simple one or two line comment directly describing the math in 267, something like: > > // The value goes exponentially from value1 to value2 in a duration of deltaTime seconds (corresponding to numSampleFrames). > // Compute the per-sample multiplier. roger, for the case that value1 == 0 case, it will not go to this block. as the code said: if (value1 <= 0 || value2 <= 0) { // Handle negative values error case by propagating previous value. for (; writeIndex < fillToFrame; ++writeIndex) values[writeIndex] = value; } else { ...... I will simplify the comments. thanks 519850 31 118949 james.wei 2011-12-12 21:29:39 -0800 Created attachment 118949 Patch 520435 32 crogers 2011-12-13 15:47:07 -0800 Looks fine to me. 520442 33 118949 kbr 2011-12-13 15:51:35 -0800 Comment on attachment 118949 Patch Sounds good. rs=me 520473 34 james.wei 2011-12-13 16:34:56 -0800 (In reply to comment #33) > (From update of attachment 118949 [details]) > Sounds good. rs=me Ken, I mark this patch as cq?, could you help to land it? thanks 520652 35 118949 webkit.review.bot 2011-12-13 20:57:11 -0800 Comment on attachment 118949 Patch Clearing flags on attachment: 118949 Committed r102733: <http://trac.webkit.org/changeset/102733> 520653 36 webkit.review.bot 2011-12-13 20:57:17 -0800 All reviewed patches have been landed. Closing bug. 117360 2011-12-01 00:14:02 -0800 2011-12-01 16:51:51 -0800 optimize to not use pow in the inner loop opt-pow.patch text/plain 2212 james.wei SW5kZXg6IFdlYkNvcmUvQ2hhbmdlTG9nCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUvQ2hhbmdlTG9n CShyZXZpc2lvbiAxMDE2MjgpCisrKyBXZWJDb3JlL0NoYW5nZUxvZwkod29ya2luZyBjb3B5KQpA QCAtMSwzICsxLDE2IEBACisyMDExLTEyLTAxICBKYW1lcyBXZWkgPGphbWVzLndlaUBpbnRlbC5j b20+IAorCisgICAgICAgIE9wdGltaXplIHRvIG5vdCB1c2UgcG93KCkgaW4gdGhlIGlubmVyIGxv b3AKKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTczNTMw CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkKKworICAgICAgICBObyBuZXcg dGVzdHMuCisKKyAgICAgICAgKiB3ZWJhdWRpby9BdWRpb1BhcmFtVGltZWxpbmUuY3BwOgorICAg ICAgICAoV2ViQ29yZTpBdWRpb1BhcmFtVGltZWxpbmU6dmFsdWVzRm9yVGltZVJhbmdlSW1wbCk6 CisKKwogMjAxMS0xMi0wMSAgS2VudGFybyBIYXJhICA8aGFyYWtlbkBjaHJvbWl1bS5vcmc+CiAK ICAgICAgICAgW0pTQ10gV2hlbiBYWFhYIGhhcyBhIE5hbWVkQ29uc3RydWN0b3IsIHdpbmRvdy5Y WFhYIHNob3VsZCBiZSBYWFhYQ29uc3RydWN0b3IKSW5kZXg6IFdlYkNvcmUvd2ViYXVkaW8vQXVk aW9QYXJhbVRpbWVsaW5lLmNwcAo9PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBXZWJDb3JlL3dlYmF1ZGlvL0F1ZGlv UGFyYW1UaW1lbGluZS5jcHAJKHJldmlzaW9uIDEwMTYyOCkKKysrIFdlYkNvcmUvd2ViYXVkaW8v QXVkaW9QYXJhbVRpbWVsaW5lLmNwcAkod29ya2luZyBjb3B5KQpAQCAtMjQ5LDE0ICsyNDksMjEg QEAKICAgICAgICAgICAgICAgICB2YWx1ZTEgPSBsb2cyZih2YWx1ZTEpOwogICAgICAgICAgICAg ICAgIHZhbHVlMiA9IGxvZzJmKHZhbHVlMik7CiAKLSAgICAgICAgICAgICAgICAvLyBGSVhNRTog b3B0aW1pemUgdG8gbm90IHVzZSBwb3coKSBpbiBpbm5lciBsb29wLCB0aGlzIGlzIGp1c3QgYSBz aW1wbGUgZXhwb25lbnRpYWwgcmFtcC4KKyAgICAgICAgICAgICAgICAvLyBDYWxjdWxhdGUgdGhl IGZpcnN0IHZhbHVlIHdpdGggcG93ZigpCisgICAgICAgICAgICAgICAgZmxvYXQgeCA9IChjdXJy ZW50VGltZSAtIHRpbWUxKSAqIGs7CisgICAgICAgICAgICAgICAgdmFsdWUgPSAoMSAtIHgpICog dmFsdWUxICsgeCAqIHZhbHVlMjsKKyAgICAgICAgICAgICAgICBkb3VibGUgdGVtcCA9IHBvd2Yo Mi4wZiwgdmFsdWUpOworICAgICAgICAgICAgICAgIHZhbHVlc1t3cml0ZUluZGV4XSA9IHRlbXA7 CisgICAgICAgICAgICAgICAgd3JpdGVJbmRleCsrOworCisgICAgICAgICAgICAgICAgLy8gQ2Fs Y3VsYXRlIHRoZSBtdWx0aXBsaWVyIHZhbHVlCisgICAgICAgICAgICAgICAgLy8gQXMgaXQgaXMg anVzdCBhIHNpbXBsZSBleHBvbmVudGlhbCByYW1wLCBqdXN0IG11bHRpcGx5IHRoaXMgdmFsdWUg ZWFjaCB0aW1lCisgICAgICAgICAgICAgICAgZG91YmxlIG11bHRpcGxpZXIgPSBwb3dmKDIuMGYs IGsgKiBzYW1wbGVGcmFtZVRpbWVJbmNyICogKHZhbHVlMiAtIHZhbHVlMSkpOwogICAgICAgICAg ICAgICAgIGZvciAoOyB3cml0ZUluZGV4IDwgZmlsbFRvRnJhbWU7ICsrd3JpdGVJbmRleCkgewot ICAgICAgICAgICAgICAgICAgICBmbG9hdCB4ID0gKGN1cnJlbnRUaW1lIC0gdGltZTEpICogazsK LSAgICAgICAgICAgICAgICAgICAgdmFsdWUgPSAoMSAtIHgpICogdmFsdWUxICsgeCAqIHZhbHVl MjsKLSAgICAgICAgICAgICAgICAgICAgdmFsdWUgPSBwb3dmKDIuMGYsIHZhbHVlKTsKLSAgICAg ICAgICAgICAgICAgICAgdmFsdWVzW3dyaXRlSW5kZXhdID0gdmFsdWU7Ci0gICAgICAgICAgICAg ICAgICAgIGN1cnJlbnRUaW1lICs9IHNhbXBsZUZyYW1lVGltZUluY3I7CisgICAgICAgICAgICAg ICAgICAgIHRlbXAgPSB0ZW1wICogbXVsdGlwbGllcjsKKyAgICAgICAgICAgICAgICAgICAgdmFs dWVzW3dyaXRlSW5kZXhdID0gdGVtcDsKICAgICAgICAgICAgICAgICB9CisKICAgICAgICAgICAg IH0KICAgICAgICAgfSBlbHNlIHsKICAgICAgICAgICAgIC8vIEhhbmRsZSBldmVudCB0eXBlcyBu b3QgcmVxdWlyaW5nIGxvb2tpbmcgYWhlYWQgdG8gdGhlIG5leHQgZXZlbnQuCg== 117522 2011-12-01 16:41:05 -0800 2011-12-12 18:01:17 -0800 optimize to not use pow in the inner loop -- updated opt-pow.patch text/plain 2956 james.wei SW5kZXg6IFNvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2UvV2Vi Q29yZS9DaGFuZ2VMb2cJKHJldmlzaW9uIDEwMTY0MCkKKysrIFNvdXJjZS9XZWJDb3JlL0NoYW5n ZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDE1IEBACisyMDExLTEyLTAxICBKYW1lcyBX ZWkgPGphbWVzLndlaUBpbnRlbC5jb20+IAorCisgICAgICAgIE9wdGltaXplIHRvIG5vdCB1c2Ug cG93KCkgaW4gdGhlIGlubmVyIGxvb3AgaW4gQXVkaW9QYXJhbVRpbWVsaW5lCisgICAgICAgIGh0 dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD03MzUzMAorCisgICAgICAgIFJl dmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIE5vIG5ldyB0ZXN0cy4KKworICAg ICAgICAqIHdlYmF1ZGlvL0F1ZGlvUGFyYW1UaW1lbGluZS5jcHA6CisgICAgICAgIChXZWJDb3Jl OkF1ZGlvUGFyYW1UaW1lbGluZTp2YWx1ZXNGb3JUaW1lUmFuZ2VJbXBsKToKKwogMjAxMS0xMi0w MSAgRmxvcmluIE1hbGl0YSAgPGZtYWxpdGFAZ29vZ2xlLmNvbT4KIAogICAgICAgICBTVkcgR2F1 c3NpYW4gYmx1ciBpbiAxLWRpbWVuc2lvbiBpcyBpbmNvcnJlY3QKSW5kZXg6IFNvdXJjZS9XZWJD b3JlL3dlYmF1ZGlvL0F1ZGlvUGFyYW1UaW1lbGluZS5jcHAKPT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gU291cmNl L1dlYkNvcmUvd2ViYXVkaW8vQXVkaW9QYXJhbVRpbWVsaW5lLmNwcAkocmV2aXNpb24gMTAxNjQw KQorKysgU291cmNlL1dlYkNvcmUvd2ViYXVkaW8vQXVkaW9QYXJhbVRpbWVsaW5lLmNwcAkod29y a2luZyBjb3B5KQpAQCAtMjQ5LDE0ICsyNDksMzMgQEAKICAgICAgICAgICAgICAgICB2YWx1ZTEg PSBsb2cyZih2YWx1ZTEpOwogICAgICAgICAgICAgICAgIHZhbHVlMiA9IGxvZzJmKHZhbHVlMik7 CiAKLSAgICAgICAgICAgICAgICAvLyBGSVhNRTogb3B0aW1pemUgdG8gbm90IHVzZSBwb3coKSBp biBpbm5lciBsb29wLCB0aGlzIGlzIGp1c3QgYSBzaW1wbGUgZXhwb25lbnRpYWwgcmFtcC4KKyAg ICAgICAgICAgICAgICAvLyBUaGUgZm9ybXVsYSBvZiB0aGUgZXhwb25lbnRpYWwgcmFtcCBpcyAK KyAgICAgICAgICAgICAgICAvLyB2KHQpID0gMl4oKDEteCh0KSkqdjEgKyB4KHQpKnYyKSB3aXRo IHgodCkgPSAodC10MSkqaworICAgICAgICAgICAgICAgIC8vCisgICAgICAgICAgICAgICAgLy8g U28gdih0K2R0KSAvIHYodCkgID0gMl5ZIHdpdGgKKyAgICAgICAgICAgICAgICAvLyBZID0gWygx LXgodCtkdCkpKnYxICsgeCh0K2R0KSp2MikgXSAtIFsoMS14KHQpKSp2MSArIHgodCkqdjIpXQor ICAgICAgICAgICAgICAgIC8vICAgPSBbKDEteCh0K2R0KSkqdjEgLSgxLXgodCkpKnYxXSArIFt4 KHQrZHQpKnYyIC0geCh0KSp2Ml0KKyAgICAgICAgICAgICAgICAvLyAgID0gWyh4KHQpIC0geCh0 K2R0KSkqdjFdICsgWygoeCh0K2R0KSAtIHgodCkpICp2Ml0KKyAgICAgICAgICAgICAgICAvLyAg ID0gW3godCtkdCkgLSB4KHQpXSAqICh2MiAtIHYxKQorICAgICAgICAgICAgICAgIC8vIEFzIHgo dCkgPSAodC10MSkqaywgc28geCh0K2R0KSAtIHgodCkgPSBkdCprCisgICAgICAgICAgICAgICAg Ly8gU28gWSA9IGsgKiBkdCAqICh2Mi12MSkgd2l0aCBkdCA9IHNhbXBsZUZyYW1lVGltZUluY3IK KyAgICAgICAgICAgICAgICAvLyBTbyB2KHQrZHQpID0gdih0KSAqIDIgXiAoayAqIHNhbXBsZUZy YW1lVGltZUluY3IgKiAodjIgLSB2MSkpCisKKyAgICAgICAgICAgICAgICAvLyBDYWxjdWxhdGUg dGhlIGZpcnN0IHZhbHVlIHdpdGggcG93ZigpCisgICAgICAgICAgICAgICAgZmxvYXQgeCA9IChj dXJyZW50VGltZSAtIHRpbWUxKSAqIGs7CisgICAgICAgICAgICAgICAgdmFsdWUgPSAoMSAtIHgp ICogdmFsdWUxICsgeCAqIHZhbHVlMjsKKyAgICAgICAgICAgICAgICBkb3VibGUgdGVtcCA9IHBv d2YoMi4wZiwgdmFsdWUpOworICAgICAgICAgICAgICAgIHZhbHVlc1t3cml0ZUluZGV4XSA9IHRl bXA7CisgICAgICAgICAgICAgICAgd3JpdGVJbmRleCsrOworCisgICAgICAgICAgICAgICAgLy8g Q2FsY3VsYXRlIHRoZSBtdWx0aXBsaWVyIHZhbHVlCisgICAgICAgICAgICAgICAgLy8gQXMgaXQg aXMganVzdCBhIHNpbXBsZSBleHBvbmVudGlhbCByYW1wLCBqdXN0IG11bHRpcGx5IHRoaXMgdmFs dWUgZWFjaCB0aW1lCisgICAgICAgICAgICAgICAgZG91YmxlIG11bHRpcGxpZXIgPSBwb3dmKDIu MGYsIGsgKiBzYW1wbGVGcmFtZVRpbWVJbmNyICogKHZhbHVlMiAtIHZhbHVlMSkpOwogICAgICAg ICAgICAgICAgIGZvciAoOyB3cml0ZUluZGV4IDwgZmlsbFRvRnJhbWU7ICsrd3JpdGVJbmRleCkg ewotICAgICAgICAgICAgICAgICAgICBmbG9hdCB4ID0gKGN1cnJlbnRUaW1lIC0gdGltZTEpICog azsKLSAgICAgICAgICAgICAgICAgICAgdmFsdWUgPSAoMSAtIHgpICogdmFsdWUxICsgeCAqIHZh bHVlMjsKLSAgICAgICAgICAgICAgICAgICAgdmFsdWUgPSBwb3dmKDIuMGYsIHZhbHVlKTsKLSAg ICAgICAgICAgICAgICAgICAgdmFsdWVzW3dyaXRlSW5kZXhdID0gdmFsdWU7Ci0gICAgICAgICAg ICAgICAgICAgIGN1cnJlbnRUaW1lICs9IHNhbXBsZUZyYW1lVGltZUluY3I7CisgICAgICAgICAg ICAgICAgICAgIHRlbXAgPSB0ZW1wICogbXVsdGlwbGllcjsKKyAgICAgICAgICAgICAgICAgICAg dmFsdWVzW3dyaXRlSW5kZXhdID0gdGVtcDsKICAgICAgICAgICAgICAgICB9CisKICAgICAgICAg ICAgIH0KICAgICAgICAgfSBlbHNlIHsKICAgICAgICAgICAgIC8vIEhhbmRsZSBldmVudCB0eXBl cyBub3QgcmVxdWlyaW5nIGxvb2tpbmcgYWhlYWQgdG8gdGhlIG5leHQgZXZlbnQuCg== 118942 2011-12-12 19:52:01 -0800 2011-12-12 21:29:24 -0800 Patch bug-73530-20111213115241.patch text/plain 3267 james.wei SW5kZXg6IFNvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2UvV2Vi Q29yZS9DaGFuZ2VMb2cJKHJldmlzaW9uIDEwMjY0NykKKysrIFNvdXJjZS9XZWJDb3JlL0NoYW5n ZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDE1IEBACisyMDExLTEyLTEzICBKYW1lcyBX ZWkgPGphbWVzLndlaUBpbnRlbC5jb20+IAorCisgICAgICAgIE9wdGltaXplIHRvIG5vdCB1c2Ug cG93KCkgaW4gdGhlIGlubmVyIGxvb3AgaW4gQXVkaW9QYXJhbVRpbWVsaW5lCisgICAgICAgIGh0 dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD03MzUzMAorCisgICAgICAgIFJl dmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIE5vIG5ldyB0ZXN0cy4KKworICAg ICAgICAqIHdlYmF1ZGlvL0F1ZGlvUGFyYW1UaW1lbGluZS5jcHA6CisgICAgICAgIChXZWJDb3Jl OkF1ZGlvUGFyYW1UaW1lbGluZTp2YWx1ZXNGb3JUaW1lUmFuZ2VJbXBsKToKKwogMjAxMS0xMi0x MiAgRXJpayBBcnZpZHNzb24gIDxhcnZAY2hyb21pdW0ub3JnPgogCiAgICAgICAgIFtWOF0gQ29k ZUdlbmVyYXRvclY4LnBtIGRvZXMgbm90IGNvcnJlY3RseSB3b3JrIHdpdGggaW5oZXJpdGVkIEhh c0luZGV4R2V0dGVyCkluZGV4OiBTb3VyY2UvV2ViQ29yZS93ZWJhdWRpby9BdWRpb1BhcmFtVGlt ZWxpbmUuY3BwCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT0KLS0tIFNvdXJjZS9XZWJDb3JlL3dlYmF1ZGlvL0F1ZGlvUGFy YW1UaW1lbGluZS5jcHAJKHJldmlzaW9uIDEwMjY0NykKKysrIFNvdXJjZS9XZWJDb3JlL3dlYmF1 ZGlvL0F1ZGlvUGFyYW1UaW1lbGluZS5jcHAJKHdvcmtpbmcgY29weSkKQEAgLTI0NSwxNiArMjQ1 LDI5IEBAIGZsb2F0IEF1ZGlvUGFyYW1UaW1lbGluZTo6dmFsdWVzRm9yVGltZVIKICAgICAgICAg ICAgICAgICBmb3IgKDsgd3JpdGVJbmRleCA8IGZpbGxUb0ZyYW1lOyArK3dyaXRlSW5kZXgpCiAg ICAgICAgICAgICAgICAgICAgIHZhbHVlc1t3cml0ZUluZGV4XSA9IHZhbHVlOwogICAgICAgICAg ICAgfSBlbHNlIHsKLSAgICAgICAgICAgICAgICAvLyBJbnRlcnBvbGF0ZSBpbiBsb2cgc3BhY2Uu Ci0gICAgICAgICAgICAgICAgdmFsdWUxID0gbG9nMmYodmFsdWUxKTsKLSAgICAgICAgICAgICAg ICB2YWx1ZTIgPSBsb2cyZih2YWx1ZTIpOworICAgICAgICAgICAgICAgIGZsb2F0IG51bVNhbXBs ZUZyYW1lcyA9IGRlbHRhVGltZSAqIHNhbXBsZVJhdGU7CisgICAgICAgICAgICAgICAgLy8gVGhl IGZvcm11bGEgb2YgdGhlIGV4cG9uZW50aWFsIHJhbXAgaXMgCisgICAgICAgICAgICAgICAgLy8g dih0KSA9IDJeKCgxLXgodCkpKmxvZyh2MSkgKyB4KHQpKmxvZyh2MikpIHdpdGggeCh0KSA9ICh0 LXQxKSprCisgICAgICAgICAgICAgICAgLy8KKyAgICAgICAgICAgICAgICAvLyBTbyB2KHQrZHQp IC8gdih0KSAgPSAyXlkgd2l0aAorICAgICAgICAgICAgICAgIC8vIFkgPSBbKDEteCh0K2R0KSkq bG9nKHYxKSArIHgodCtkdCkqbG9nKHYyKSkgXSAtIFsoMS14KHQpKSpsb2codjEpICsgeCh0KSps b2codjIpKV0KKyAgICAgICAgICAgICAgICAvLyAgID0gWygxLXgodCtkdCkpKmxvZyh2MSkgLSgx LXgodCkpKmxvZyh2MSldICsgW3godCtkdCkqbG9nKHYyKSAtIHgodCkqbG9nKHYyKV0KKyAgICAg ICAgICAgICAgICAvLyAgID0gWyh4KHQpIC0geCh0K2R0KSkqbG9nKHYxKV0gKyBbKCh4KHQrZHQp IC0geCh0KSkgKmxvZyh2MildCisgICAgICAgICAgICAgICAgLy8gICA9IFt4KHQrZHQpIC0geCh0 KV0gKiAobG9nKHYyKSAtIGxvZyh2MSkpCisgICAgICAgICAgICAgICAgLy8gQXMgeCh0KSA9ICh0 LXQxKSprLCBzbyB4KHQrZHQpIC0geCh0KSA9IGR0KmsKKyAgICAgICAgICAgICAgICAvLyBTbyBZ ID0gayAqIGR0ICogKGxvZyh2MikgLSBsb2codjEpKSB3aXRoIGR0ID0gc2FtcGxlRnJhbWVUaW1l SW5jcgorICAgICAgICAgICAgICAgIC8vIFNvIHYodCtkdCkgPSB2KHQpICogMiBeIChrICogc2Ft cGxlRnJhbWVUaW1lSW5jciAqIChsb2codjIpIC0gbG9nKHYxKSkpCiAKLSAgICAgICAgICAgICAg ICAvLyBGSVhNRTogb3B0aW1pemUgdG8gbm90IHVzZSBwb3coKSBpbiBpbm5lciBsb29wLCB0aGlz IGlzIGp1c3QgYSBzaW1wbGUgZXhwb25lbnRpYWwgcmFtcC4KKyAgICAgICAgICAgICAgICAvLyBB cyBpdCBpcyBqdXN0IGEgc2ltcGxlIGV4cG9uZW50aWFsIHJhbXAsIGp1c3QgbXVsdGlwbHkgdGhp cyB2YWx1ZSBlYWNoIHRpbWUuCisgICAgICAgICAgICAgICAgLy8gU28gbXVsdGlwbGllciA9IDJe KGsgKiBzYW1wbGVGcmFtZVRpbWVJbmNyICogKGxvZyh2MikgLSBsb2codjEpKSkKKyAgICAgICAg ICAgICAgICAvLyAgICAgICAgICAgICAgID0gMl4oKDEgLyAoIGRlbHRhVGltZSAqIHNhbXBsZVJh dGUgKSAqKGxvZyh2MikgLSBsb2codjEpKSkKKyAgICAgICAgICAgICAgICAvLyAgICAgICAgICAg ICAgID0gMl4oKDEgLyBudW1TYW1wbGVGcmFtZXMpICogKGxvZyh2MikgLSBsb2codjEpKSkKKyAg ICAgICAgICAgICAgICAvLyAgICAgICAgICAgICAgID0gKDJeKGxvZyh2MikgLSBsb2codjEpKSkg XiAoMSAvIG51bVNhbXBsZUZyYW1lcykKKyAgICAgICAgICAgICAgICAvLyAgICAgICAgICAgICAg ID0gKHYyIC8gdjEpIF4gKDEgLyBudW1TYW1wbGVGcmFtZXMpCisgICAgICAgICAgICAgICAgZmxv YXQgbXVsdGlwbGllciA9IHBvd2YodmFsdWUyIC8gdmFsdWUxLCAxIC8gbnVtU2FtcGxlRnJhbWVz KTsKICAgICAgICAgICAgICAgICBmb3IgKDsgd3JpdGVJbmRleCA8IGZpbGxUb0ZyYW1lOyArK3dy aXRlSW5kZXgpIHsKLSAgICAgICAgICAgICAgICAgICAgZmxvYXQgeCA9IChjdXJyZW50VGltZSAt IHRpbWUxKSAqIGs7Ci0gICAgICAgICAgICAgICAgICAgIHZhbHVlID0gKDEgLSB4KSAqIHZhbHVl MSArIHggKiB2YWx1ZTI7Ci0gICAgICAgICAgICAgICAgICAgIHZhbHVlID0gcG93ZigyLjBmLCB2 YWx1ZSk7CiAgICAgICAgICAgICAgICAgICAgIHZhbHVlc1t3cml0ZUluZGV4XSA9IHZhbHVlOwor ICAgICAgICAgICAgICAgICAgICB2YWx1ZSAqPSBtdWx0aXBsaWVyOwogICAgICAgICAgICAgICAg ICAgICBjdXJyZW50VGltZSArPSBzYW1wbGVGcmFtZVRpbWVJbmNyOwogICAgICAgICAgICAgICAg IH0KICAgICAgICAgICAgIH0K 118949 2011-12-12 21:29:39 -0800 2011-12-13 20:57:10 -0800 Patch bug-73530-20111213133019.patch text/plain 2146 james.wei SW5kZXg6IFNvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2UvV2Vi Q29yZS9DaGFuZ2VMb2cJKHJldmlzaW9uIDEwMjY0NykKKysrIFNvdXJjZS9XZWJDb3JlL0NoYW5n ZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDE1IEBACisyMDExLTEyLTEzICBKYW1lcyBX ZWkgPGphbWVzLndlaUBpbnRlbC5jb20+IAorCisgICAgICAgIE9wdGltaXplIHRvIG5vdCB1c2Ug cG93KCkgaW4gdGhlIGlubmVyIGxvb3AgaW4gQXVkaW9QYXJhbVRpbWVsaW5lCisgICAgICAgIGh0 dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD03MzUzMAorCisgICAgICAgIFJl dmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIE5vIG5ldyB0ZXN0cy4KKworICAg ICAgICAqIHdlYmF1ZGlvL0F1ZGlvUGFyYW1UaW1lbGluZS5jcHA6CisgICAgICAgIChXZWJDb3Jl OkF1ZGlvUGFyYW1UaW1lbGluZTp2YWx1ZXNGb3JUaW1lUmFuZ2VJbXBsKToKKwogMjAxMS0xMi0x MiAgRXJpayBBcnZpZHNzb24gIDxhcnZAY2hyb21pdW0ub3JnPgogCiAgICAgICAgIFtWOF0gQ29k ZUdlbmVyYXRvclY4LnBtIGRvZXMgbm90IGNvcnJlY3RseSB3b3JrIHdpdGggaW5oZXJpdGVkIEhh c0luZGV4R2V0dGVyCkluZGV4OiBTb3VyY2UvV2ViQ29yZS93ZWJhdWRpby9BdWRpb1BhcmFtVGlt ZWxpbmUuY3BwCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT0KLS0tIFNvdXJjZS9XZWJDb3JlL3dlYmF1ZGlvL0F1ZGlvUGFy YW1UaW1lbGluZS5jcHAJKHJldmlzaW9uIDEwMjY0NykKKysrIFNvdXJjZS9XZWJDb3JlL3dlYmF1 ZGlvL0F1ZGlvUGFyYW1UaW1lbGluZS5jcHAJKHdvcmtpbmcgY29weSkKQEAgLTI0NSwxNiArMjQ1 LDEzIEBAIGZsb2F0IEF1ZGlvUGFyYW1UaW1lbGluZTo6dmFsdWVzRm9yVGltZVIKICAgICAgICAg ICAgICAgICBmb3IgKDsgd3JpdGVJbmRleCA8IGZpbGxUb0ZyYW1lOyArK3dyaXRlSW5kZXgpCiAg ICAgICAgICAgICAgICAgICAgIHZhbHVlc1t3cml0ZUluZGV4XSA9IHZhbHVlOwogICAgICAgICAg ICAgfSBlbHNlIHsKLSAgICAgICAgICAgICAgICAvLyBJbnRlcnBvbGF0ZSBpbiBsb2cgc3BhY2Uu Ci0gICAgICAgICAgICAgICAgdmFsdWUxID0gbG9nMmYodmFsdWUxKTsKLSAgICAgICAgICAgICAg ICB2YWx1ZTIgPSBsb2cyZih2YWx1ZTIpOwotCi0gICAgICAgICAgICAgICAgLy8gRklYTUU6IG9w dGltaXplIHRvIG5vdCB1c2UgcG93KCkgaW4gaW5uZXIgbG9vcCwgdGhpcyBpcyBqdXN0IGEgc2lt cGxlIGV4cG9uZW50aWFsIHJhbXAuCisgICAgICAgICAgICAgICAgZmxvYXQgbnVtU2FtcGxlRnJh bWVzID0gZGVsdGFUaW1lICogc2FtcGxlUmF0ZTsKKyAgICAgICAgICAgICAgICAvLyBUaGUgdmFs dWUgZ29lcyBleHBvbmVudGlhbGx5IGZyb20gdmFsdWUxIHRvIHZhbHVlMiBpbiBhIGR1cmF0aW9u IG9mIGRlbHRhVGltZSBzZWNvbmRzIChjb3JyZXNwb25kaW5nIHRvIG51bVNhbXBsZUZyYW1lcyku CisgICAgICAgICAgICAgICAgLy8gQ29tcHV0ZSB0aGUgcGVyLXNhbXBsZSBtdWx0aXBsaWVyLgor ICAgICAgICAgICAgICAgIGZsb2F0IG11bHRpcGxpZXIgPSBwb3dmKHZhbHVlMiAvIHZhbHVlMSwg MSAvIG51bVNhbXBsZUZyYW1lcyk7CiAgICAgICAgICAgICAgICAgZm9yICg7IHdyaXRlSW5kZXgg PCBmaWxsVG9GcmFtZTsgKyt3cml0ZUluZGV4KSB7Ci0gICAgICAgICAgICAgICAgICAgIGZsb2F0 IHggPSAoY3VycmVudFRpbWUgLSB0aW1lMSkgKiBrOwotICAgICAgICAgICAgICAgICAgICB2YWx1 ZSA9ICgxIC0geCkgKiB2YWx1ZTEgKyB4ICogdmFsdWUyOwotICAgICAgICAgICAgICAgICAgICB2 YWx1ZSA9IHBvd2YoMi4wZiwgdmFsdWUpOwogICAgICAgICAgICAgICAgICAgICB2YWx1ZXNbd3Jp dGVJbmRleF0gPSB2YWx1ZTsKKyAgICAgICAgICAgICAgICAgICAgdmFsdWUgKj0gbXVsdGlwbGll cjsKICAgICAgICAgICAgICAgICAgICAgY3VycmVudFRpbWUgKz0gc2FtcGxlRnJhbWVUaW1lSW5j cjsKICAgICAgICAgICAgICAgICB9CiAgICAgICAgICAgICB9Cg==