Displaying 20 results from an estimated 4000 matches similar to: "[LLVMdev] newbie question - selecting the write kind of pass"
2005 Apr 03
2
[LLVMdev] newbie question - selecting the write kind of pass
On Sat, Apr 02, 2005 at 11:35:30AM -0600, Chris Lattner wrote:
> On Sat, 2 Apr 2005, Sameer D. Sahasrabuddhe wrote:
> > I want to create a simple map from a Value to the instruction that
> > defines it. Such a map is present inside SchedGraph, but I need it in
> > a much simpler context.
>
> Is this in the context of the code generator?
No ... I am just trying to feel
2005 Apr 02
0
[LLVMdev] newbie question - selecting the write kind of pass
On Sat, 2 Apr 2005, Sameer D. Sahasrabuddhe wrote:
> I want to create a simple map from a Value to the instruction that
> defines it. Such a map is present inside SchedGraph, but I need it in
> a much simpler context.
Is this in the context of the code generator?
> If I got it right, I create a new AnalysisGroup, and write a pass that
> implements it. What kind of pass should I
2005 Nov 12
2
[LLVMdev] building LLVM 1.6 on Debian unstable ...
Hi,
Tried to build the 1.6 release on Debian unstable, but ran into link
problems for some utils ...
Copying the messages when building llc, the same error occurs for lli,
llvm-db and some of the examples. Is this a problem with the compiler
version? LLVMSelectionDAG.o itself builds correctly.
make[2]: Entering directory `/home/sameerds/data/llvm-1.6/tools/llc'
llvm[2]: Linking Debug
2005 Apr 03
0
[LLVMdev] newbie question - selecting the write kind of pass
On Sun, 3 Apr 2005, Sameer D. Sahasrabuddhe wrote:
> On Sat, Apr 02, 2005 at 11:35:30AM -0600, Chris Lattner wrote:
>
>> On Sat, 2 Apr 2005, Sameer D. Sahasrabuddhe wrote:
>>> I want to create a simple map from a Value to the instruction that
>>> defines it. Such a map is present inside SchedGraph, but I need it in
>>> a much simpler context.
>>
2005 Nov 15
4
[LLVMdev] doxygen docs
Hi,
The docs available on illuvium.com are different from the one's present
in the doxygen tarball on the same page ... can the tarball be generated
from the same docs as the browseable version? I considered crawling the
illuvium.com site, but it would be better to simply have a tarball
available.
I tried generating the docs myself ... doxygen simply refuses to create
pages for classes
2005 Nov 12
2
[LLVMdev] building LLVM 1.6 on Debian unstable ...
Marco Matthies wrote:
>> make[2]: Entering directory `/home/sameerds/data/llvm-1.6/tools/llc'
>> llvm[2]: Linking Debug executable llc
>> `.gnu.linkonce.t._ZNK4llvm14TargetLowering12getValueTypeEPKNS_4TypeE'
>> referenced
>> in section `.rodata' of
>> /home/sameerds/data/llvm-1.6/Debug/lib/LLVMSelectionDAG
>> .o: defined in discarded
2005 Apr 29
2
[LLVMdev] about AnalysisUsage
Just noticed that quite a few passes like LoopSimplify are implemented
in a single .cpp file ... this makes it impossible to specify
LoopSimplify using the "addRequired" method. Was there any particular
reason to do it this way? I wouldn't mind doing the splitting myself,
though I am not using the CVS versions right now.
Also, it would be nice to have support for some sort of a
2005 Apr 29
2
[LLVMdev] about AnalysisUsage
On Fri, Apr 29, 2005 at 08:10:17AM -0500, Chris Lattner wrote:
> AU.addRequiredID(LoopSimplifyID);
>
> "LoopSimplifyID" is a marker that is used to identify the pass, which is
> exported from the .cpp file.
I'll have to declare a PassInfo* called LoopSimplifyID inside
namespace llvm, in order for that to compile correctly, right? I was
wondering why not simply
2005 Apr 28
2
[LLVMdev] inserting blocks into a Function
Recently wrote a pass that inserts a preheader for a loop that doesn't
have one. When I tried to run it, I ran into a problem that became
obvious in hindsight - the PHINodes need to be updated in places where
the incoming control-edge has changed. Is there anything else that can
be affected when a block is inserted into the CFG?
Also, planning to write a helper function which will take care
2005 Nov 15
1
[LLVMdev] doxygen docs
Chris Lattner wrote:
>> I tried generating the docs myself ... doxygen simply refuses to
>> create pages for classes defined in anonymous namespaces in cpp files.
>> I enabled options such as EXTRACT_ALL, EXTRACT_PRIVATE and
>> EXTRACT_LOCAL, but no luck. How is the publicly available
>> documentation generated?
>
> They are generated from the Makefile in
2005 Nov 12
0
[LLVMdev] building LLVM 1.6 on Debian unstable ...
Sameer D. Sahasrabuddhe wrote:
> Hi,
>
> Tried to build the 1.6 release on Debian unstable, but ran into link
> problems for some utils ...
>
> Copying the messages when building llc, the same error occurs for lli,
> llvm-db and some of the examples. Is this a problem with the compiler
> version? LLVMSelectionDAG.o itself builds correctly.
>
> make[2]: Entering
2005 Nov 15
0
[LLVMdev] doxygen docs
On Tue, 15 Nov 2005, Sameer D. Sahasrabuddhe wrote:
> The docs available on illuvium.com are different from the one's present in
> the doxygen tarball on the same page ... can the tarball be generated from
> the same docs as the browseable version? I considered crawling the
> illuvium.com site, but it would be better to simply have a tarball available.
Agreed. I have been
2005 Apr 30
1
[LLVMdev] about AnalysisUsage
On Fri, Apr 29, 2005 at 10:32:18PM -0500, Chris Lattner wrote:
> We could definitely expose the pass class itself, but there are no
> implementation details or any other state that is useful.
Yeah ... this, and Transforms/Scalar.h answer my original original
question about why a lot of passes are present as a single CPP. The
whole LLVM code generally abhors including class definitions
2005 Apr 29
1
[LLVMdev] inserting blocks into a Function
On Thu, Apr 28, 2005 at 11:22:26PM -0500, Chris Lattner wrote:
> BasicBlock *SplitBlockPredecessors(BasicBlock *BB, const char *Suffix,
> const std::vector<BasicBlock*> &Preds);
>
> Basically, given a BB with multiple predecessors, it inserts (and returns)
> a new block, moving the predecessors in Preds to the new block and leaving
2005 Jun 29
1
[LLVMdev] null pointer check missing in LoopInfo
In LLVM 1.5, the function isLoopHeader fails to check for a null
pointer after calling getLoopFor. Dunno if this is fixed in CVS, but
attaching a patch anyway ...
Sameer.
--
Research Scholar, KReSIT, IIT Bombay
http://www.it.iitb.ac.in/~sameerds/
-------------- next part --------------
--- LoopInfo.h 2005-06-29 17:42:43.000000000 +0530
+++ llvm/include/llvm/Analysis/LoopInfo.h 2005-06-29
2005 Apr 29
0
[LLVMdev] about AnalysisUsage
On Fri, 29 Apr 2005, Sameer D. Sahasrabuddhe wrote:
> Just noticed that quite a few passes like LoopSimplify are implemented
> in a single .cpp file ... this makes it impossible to specify
> LoopSimplify using the "addRequired" method. Was there any particular
> reason to do it this way? I wouldn't mind doing the splitting myself,
> though I am not using the CVS
2005 Apr 30
0
[LLVMdev] about AnalysisUsage
On Fri, 29 Apr 2005, Sameer D. Sahasrabuddhe wrote:
> On Fri, Apr 29, 2005 at 08:10:17AM -0500, Chris Lattner wrote:
>
>> AU.addRequiredID(LoopSimplifyID);
>>
>> "LoopSimplifyID" is a marker that is used to identify the pass, which is
>> exported from the .cpp file.
>
> I'll have to declare a PassInfo* called LoopSimplifyID inside
>
2005 Apr 29
0
[LLVMdev] inserting blocks into a Function
On Thu, 28 Apr 2005, Sameer D. Sahasrabuddhe wrote:
> Recently wrote a pass that inserts a preheader for a loop that doesn't
> have one. When I tried to run it, I ran into a problem that became
> obvious in hindsight - the PHINodes need to be updated in places where
> the incoming control-edge has changed. Is there anything else that can
> be affected when a block is inserted
2005 Nov 16
1
[LLVMdev] Moving CVS Files
On Tue, 15 Nov 2005, Reid Spencer wrote:
> We should probably review this decision at least once a year. If the basis
> for not moving to svn is that "cvs is the standard", that situation is likely
> to change. Many organizations are now using svn. It is rapidly becoming "the
> new standard". At some point it doesn't make sense for us to continue with
>
2015 Jan 14
3
[LLVMdev] [RFC][PATCH][OPENCL] synchronization scopes redux
On Tue, Jan 13, 2015 at 10:27 PM, Sameer Sahasrabuddhe <
sameer.sahasrabuddhe at amd.com> wrote:
> Ping! We need to close on whether everyone is convinced that symbolic
> memory scopes have a significant advantage over opaque numbers. Either of
> them will be examined by optimizations using a target-implemented API. I
> personally don't think that readability in the LLVM