similar to: MCJIT -- Poor run-time performance for Fibonacci example in LLVM 3.8.1

Dear All, We have a new 32 core machine, so I ran some numbers. These results are for compiling all of LLVM and Clang in a Debug build from scratch. The first number is the -j argument to make, and the rest is the result of the bash builtin time command. 2: 1612.950u 161.293s 15:04.92 196.0% 0+0k 0+0io 0pf+0w 4: 1624.101u 164.121s 8:00.74 371.9% 0+0k 0+0io 0pf+0w 8:

On 11/1/11 8:49 PM, Óscar Fuentes wrote: > Hello John. > > John Criswell<criswell at illinois.edu> writes: > > [snip] > >> I did not use CMake but the standard autoconf + Makefile build. >> >> Not sure if this helps, but here it is, for what it's worth. > Very interesting, thanks! > > CMake introduces more parallelism and it would be great to

Rather than “fixing” it, it might be better to support a separate method for signed extension. My reasoning is as follows: int x = 7; llvm::PointerIntPair<double*, 3, int> pip; pip.setInt(x); There could be code out there that expects pip.getInt() to return 7 and not -1. So if you really want to set a negative and return a negative value, a separate method setSignedInt and getSignedInt

I'd argue that bitfield sign extensions are surprising and are usually a source of bugs. It would be much more explicit and less error prone for the user to write the sign extension if they want it. By extension, it seems good that PointerIntPair doesn't do sign extension when the type happens to be signed. On Wed, Apr 4, 2018 at 9:47 AM David Blaikie via llvm-dev < llvm-dev at

On Thu, 2 Feb 2017 09:36:28 -0800 Chris Stankevitz <chrisstankevitz at gmail.com> wrote: > On Tue, Jan 31, 2017 at 2:45 PM, Rowland Penny via samba > <samba at lists.samba.org> wrote: > >> /etc/resolv.conf: > >> root at nickel:~ # cat /etc/resolv.conf > >> nameserver 192.168.11.5 > >> nameserver 192.168.1.4 > >> domain mydomain.local

define dso_local i32 @f(i32 %0) { %2 = and i32 %0, 7 %3 = icmp eq i32 %2, 7 %4 = zext i1 %3 to i32 ret i32 %4 } I thought instcombine would remove it. It doesn't and nothing else does either. LLVM Version is 10.0.0. /Riyaz -------------- next part -------------- An HTML attachment was scrubbed... URL:

The sign extension is correct. Otherwise setInt(-1) won’t work. If you don’t want sign extension, then use ‘unsigned’ and not ‘int’ in the template arguments. > On Apr 4, 2018, at 14:34, Reid Kleckner via llvm-dev <llvm-dev at lists.llvm.org> wrote: > > I'd argue that bitfield sign extensions are surprising and are usually a source of bugs. It would be much more explicit and

I do agree that sign-extension is the right thing to do. Unlike bit-fields, llvm::PointerIntPair has asserts checking that the int value is within range. So if you assign an out of range value, it should fail with an assertion: llvm::PointerIntPair<SomeType*, 3, int> pip; pip.setInt(7); // can be made to fail as the valid range // of signed 3-bit values is [-4:3] The above

I think it'd be reasonable to model this on the same behavior as int to short to int round-tripping & not to speculate that there might be code relying on the existing behavior until there's evidence of it. I'd suggest changing the behavior & testing to see if anything breaks - and if nothing does, moving to the behavior rather than supporting both. On Wed, Apr 4, 2018 at

It won't move the sign bit, so negative values won't fit, unless you have a 3 bit signed type ;) Note that if you assign negative values to and then read from a signed bit-field, you would do sign extension. So 3-bit signed types do exist in C++. It begs the question why PointerIntPair supports signed int types if it always loses the sign. Is it just to avoid signed/unsigned comparison

int f() { int A[100]; for (int i = 0; i < 100; ++i) A[i] = i; return A[10]; } Neither gcc nor clang eliminates the loop. Clang also blows up the code a bit by unrolling the loop. Why can't this loop and the array be eliminated and the function body turned into ret i32 10 ? Am I missing something? /Riyaz -------------- next part -------------- An HTML attachment was

I have a variadic intrinsic that is defined as something like this: def int_foobar : Intrinsic<[llvm_anyint_ty], [llvm_vararg_ty], [IntrNoMem, IntrSpeculatable]>; When I construct a call to the above intrinsic with IRBuilder::CreateIntrinsic, it mangles the intrinsic name with the return type (i64) and the actual argument

I'd suggest someone try fixing this & see if it breaks anything that can't reasonably be fixed (before we go assuming this is by design/shouldn't be fixed just because it's the way it is today). On Wed, Apr 4, 2018 at 7:16 AM Riyaz Puthiyapurayil via llvm-dev < llvm-dev at lists.llvm.org> wrote: > It won't move the sign bit, so negative values won't fit,

Can someone confirm whether the following is a known clang bug? I am having a lot of trouble compiling some old legacy code due to this error. I know "using namespace" is horrible but that doesn't make it illegal (g++ accepts this). The fix is easy but there are just too many instances of this in the legacy code. The following is not from the code I am trying to compile but is

I have some programs crashing when I upgraded from clang 3.9.1 to clang 4.0.1. Debugging this I found the reason for the crash. This is happening in the following assembly fragment for a piece of code allocating a class object (size: 24 bytes) using operator new and then initializing it: 0x00002aaaafc145f3 <+35>: callq 0x2aaaafdf5f90 <operator new(unsigned long)>

Is there an existing option in X86_64 target code generator to emit a loop for the following code: define i4096 @add(i4096 %a, i4096 %b) alwaysinline { %c = add i4096 %a, %b ret i4096 %c } instead of: movq %rdi, %rax addq 96(%rsp), %rsi adcq 104(%rsp), %rdx movq %rdx, 8(%rdi) movq %rsi, (%rdi) adcq 112(%rsp), %rcx movq %rcx, 16(%rdi) adcq

Ignore the suggested fix in my earlier post. How about this? diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp index 20c81c3..b8ebf42 100644 --- a/lib/Target/X86/X86ISelLowering.cpp +++ b/lib/Target/X86/X86ISelLowering.cpp @@ -1632,10 +1632,11 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM, if (!Subtarget.is64Bit()) { // These

On 4 Apr 2018, at 11:01, Florian Hahn via llvm-dev <llvm-dev at lists.llvm.org> wrote: > > Hi, > > On 04/04/2018 05:34, Riyaz Puthiyapurayil via llvm-dev wrote: >> llvm::PointerIntPair<double*, 3, signed> P; >> P.setInt(-4); >> Ideally, the value range for a 3-bit signed integer should be [-4,3]. But the above call to setInt will fail. Essentially, the

llvm::PointerIntPair<double*, 3, signed> P; P.setInt(-4); Ideally, the value range for a 3-bit signed integer should be [-4,3]. But the above call to setInt will fail. Essentially, the signed int field in PointerIntPair is behaving the same as an 3-bit unsigned field which has the legal value range of [0,7]. Is this by design? Are negative values not allowed in PointerIntPair? /Riyaz

Consider the following C code. When it is compiled using clang with ' -O', the assert fires. Gcc seems to do the opposite. #include <assert.h> int main() { unsigned a; // uninitialized! unsigned b = a; assert(b == a); // shouldn't this be always true? return 0; } Would you consider this a LLVM bug? I think GVN is responsible for this. I don't know if it