tag:blogger.com,1999:blog-1862878851303132605.post7434231044024426374..comments2023-06-07T10:48:04.238-04:00Comments on Notes on the LHC: Grin a little.SamBhttp://www.blogger.com/profile/06560268240719951351noreply@blogger.comBlogger8125tag:blogger.com,1999:blog-1862878851303132605.post-72998052804551254332010-05-11T00:06:06.496-04:002010-05-11T00:06:06.496-04:00rolex 18k
rolex stainless
BREITLING
TAG HEUER
CART...<a href="http://www.watchestoo.com/rolex-c-145.html" rel="nofollow">rolex 18k</a><br /><a href="http://www.watchestoo.com/rolex-c-145.html" rel="nofollow">rolex stainless</a><br /><a href="http://www.watchestoo.com/breitling-c-119.html" rel="nofollow">BREITLING</a><br /><a href="http://www.watchestoo.com/tagheuer-c-156.html" rel="nofollow">TAG HEUER</a><br /><a href="http://www.watchestoo.com/cartier-c-129.html" rel="nofollow">CARTIER</a><br /><a href="http://www.watchestoo.com/omega-c-141.html" rel="nofollow">OMEGA</a><br /><a href="http://www.watchestoo.com/iwc-c-138.html" rel="nofollow">IWC</a><br /><a href="http://www.watchestoo.com/iwc-c-163.html" rel="nofollow">u boat</a><br /><a href="http://www.watchestoo.com/iwcportugueseautomatic7dayslimitededitionlaureusiw500112-p-930.html" rel="nofollow">iw500112</a><br /><a href="http://www.watchestoo.com/iwcaquatimerautomaticchronographglapagosislandsiw376705-p-946.html" rel="nofollow">IW376705</a>ryanhttps://www.blogger.com/profile/03039890117092827523noreply@blogger.comtag:blogger.com,1999:blog-1862878851303132605.post-1405962192127360942009-07-06T22:03:03.676-04:002009-07-06T22:03:03.676-04:00This comment has been removed by a blog administrator.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-1862878851303132605.post-50132935479285446862009-02-14T06:40:00.000-05:002009-02-14T06:40:00.000-05:00jge:We have no plans for improving scalability. In...jge:<BR/>We have no plans for improving scalability. Incremental development can easily be done with other compilers.David Himmelstruphttps://www.blogger.com/profile/12982136700651117492noreply@blogger.comtag:blogger.com,1999:blog-1862878851303132605.post-11390469261854129182009-02-12T02:47:00.000-05:002009-02-12T02:47:00.000-05:00Samule:for example, gcc compiles each c file to ob...Samule:<BR/>for example, gcc compiles each c file to obj first where functions implemented in other c files are "pending", then linker rewrite pointers to "pending" functions with their implementations in other objs. A source file can't too big generally, so each step consumes managable memory.<BR/><BR/>Grin does not have unknown functions and makes all indirections explicit which means a compiled function is also "referential transparent" (you can't pass in a pointer which is absent in its case..of..), you can't have "pending" function pointers. if two modules invokes each together, you have to put both of them in memory, so do 3, 4 or more modules. basically you have to put all modules of a connected graph component in memory. Thinking about compiling GHC in this way.<BR/><BR/>several methods to get scalability came into my mind:<BR/>1. allow runtime unknown functions, an additional indirection, a tradeoff<BR/>2. develop an algorithm which dynamically write disconnected modules(case..of.. is fixed) back to disk to save memory during compilation.<BR/>3. make all case..of.. open, compiler deals with modules one by one and produces .obj and .req for each of them. A module has only one .obj where all case..of.. are open. A module may have a .req for other modules each which stores case..of.. that will be merged to other module. Compiler keeps merging .obj and .req to new .obj until no .req is present. a lot of alrorithm and data structure is needed here, say dependency analysis of modules can speed up convergence.<BR/><BR/>grin is great, so these're my random/rough thoughts to make it scalable, you possibly already have some of them in your code.<BR/><BR/>btw: i'm not native english speaker, some sentences i typed are probably weird to you.Anonymoushttps://www.blogger.com/profile/03377810049038361436noreply@blogger.comtag:blogger.com,1999:blog-1862878851303132605.post-30183236358399664162009-02-07T10:55:00.000-05:002009-02-07T10:55:00.000-05:00jge:I'm not sure what you mean. We don't convert t...jge:<BR/><BR/>I'm not sure what you mean. We don't convert to Grin code until after linking the modules together, anyway.SamBhttps://www.blogger.com/profile/06560268240719951351noreply@blogger.comtag:blogger.com,1999:blog-1862878851303132605.post-59861808117760114012009-02-06T02:22:00.000-05:002009-02-06T02:22:00.000-05:00jge:We haven't yet had to deal with scalability. M...jge:<BR/><BR/>We haven't yet had to deal with scalability. MLton seems to suggests that, with the proper optimizations, control-flow analysis can be done fairly quickly.David Himmelstruphttps://www.blogger.com/profile/12982136700651117492noreply@blogger.comtag:blogger.com,1999:blog-1862878851303132605.post-70765725159648927172009-02-05T21:58:00.000-05:002009-02-05T21:58:00.000-05:00`compilation' and `modules' in my words represent ...`compilation' and `modules' in my words represent low-level ones (not some sort of intermediate code). grin in the paper assumes the whole program is under its grasp, so i'm curious of how to balance scale issue and performance penalty.Anonymoushttps://www.blogger.com/profile/03377810049038361436noreply@blogger.comtag:blogger.com,1999:blog-1862878851303132605.post-41388967593773789782009-02-05T21:43:00.000-05:002009-02-05T21:43:00.000-05:00does LHC's grin scale? i mean the compilation is s...does LHC's grin scale? i mean the compilation is seperated into two phases: modules one by one first(or several together), then linking. otherwise the compiler consumes impractical resources for realworld projects. since you want to inline more apply/eval, the situation seems more crucialAnonymoushttps://www.blogger.com/profile/03377810049038361436noreply@blogger.com