[llvm-commits] [llvm] r158014 - in /llvm/trunk: docs/LangRef.html include/llvm/Intrinsics.td include/llvm/Target/TargetLowering.h lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
Hal Finkel
hfinkel at anl.gov
Sat Jun 9 21:51:50 PDT 2012
On Tue, 05 Jun 2012 19:07:46 -0000
Lang Hames <lhames at gmail.com> wrote:
> Author: lhames
> Date: Tue Jun 5 14:07:46 2012
> New Revision: 158014
>
> URL: http://llvm.org/viewvc/llvm-project?rev=158014&view=rev
> Log:
> Add a new intrinsic: llvm.fmuladd. This intrinsic represents a
> multiply-add expression (a * b + c) that can be implemented as a
> fused multiply-add (fma) if the target determines that this will be
> more efficient. This intrinsic will be used to implement FP_CONTRACT
> support and an aggressive FMA formation mode.
>
> If your target has a fast FMA instruction you should override the
> isFMAFasterThanMulAndAdd method in TargetLowering to return true.
Will this apply (or be extended) to other contractions as well (like a *
b - c) [fmsub on PowerPC].
-Hal
>
>
> Modified:
> llvm/trunk/docs/LangRef.html
> llvm/trunk/include/llvm/Intrinsics.td
> llvm/trunk/include/llvm/Target/TargetLowering.h
> llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
>
> Modified: llvm/trunk/docs/LangRef.html
> URL:
> http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/LangRef.html?rev=158014&r1=158013&r2=158014&view=diff
> ==============================================================================
> --- llvm/trunk/docs/LangRef.html (original) +++
> llvm/trunk/docs/LangRef.html Tue Jun 5 14:07:46 2012 @@ -277,6
> +277,11 @@ <li><a
> href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>
> Intrinsics</a></li> </ol> </li>
> + <li><a href="#spec_arithmetic">Specialised Arithmetic
> Intrinsics</a>
> + <ol>
> + <li><a href="#fmuladd">'<tt>llvm.fmuladd</tt>
> Intrinsic</a></li>
> + </ol>
> + </li>
> <li><a href="#int_fp16">Half Precision Floating Point
> Intrinsics</a> <ol>
> <li><a
> href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>'
> Intrinsic</a></li> @@ -7947,6 +7952,52 @@
> <!--
> =======================================================================
> --> <h3>
> + <a name="spec_arithmetic">Specialised Arithmetic Intrinsics</a>
> +</h3>
> +
> +<!--
> _______________________________________________________________________
> --> + +<h4>
> + <a name="fmuladd">'<tt>llvm.fmuladd.*</tt>' Intrinsic</a>
> +</h4>
> +
> +<div>
> +
> +<h5>Syntax:</h5>
> +<pre>
> + declare float @llvm.fmuladd.f32(float %a, float %b, float %c)
> + declare double @llvm.fmuladd.f64(double %a, double %b, double %c)
> +</pre>
> +
> +<h5>Overview:</h5>
> +<p>The '<tt>llvm.fmuladd.*</tt>' intrinsic functions represent
> multiply-add +expressions that can be fused if the code generator
> determines that the fused +expression would be legal and
> efficient.</p> +
> +<h5>Arguments:</h5>
> +<p>The '<tt>llvm.fmuladd.*</tt>' intrinsics each take three
> arguments: two +multiplicands, a and b, and an addend c.</p>
> +
> +<h5>Semantics:</h5>
> +<p>The expression:</p>
> +<pre>
> + %0 = call float @llvm.fmuladd.f32(%a, %b, %c)
> +</pre>
> +<p>is equivalent to the expression a * b + c, except that rounding
> will not be +performed between the multiplication and addition steps
> if the code generator +fuses the operations. Fusion is not
> guaranteed, even if the target platform +supports it. If a fused
> multiply-add is required the corresponding llvm.fma.* +intrinsic
> function should be used instead.</p> +
> +<h5>Examples:</h5>
> +<pre>
> + %r2 = call float @llvm.fmuladd.f32(float %a, float %b, float %c) ;
> yields {float}:r2 = (a * b) + c +</pre>
> +
> +</div>
> +
> +<!--
> =======================================================================
> --> +<h3> <a name="int_fp16">Half Precision Floating Point
> Intrinsics</a> </h3>
>
>
> Modified: llvm/trunk/include/llvm/Intrinsics.td
> URL:
> http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Intrinsics.td?rev=158014&r1=158013&r2=158014&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/Intrinsics.td (original) +++
> llvm/trunk/include/llvm/Intrinsics.td Tue Jun 5 14:07:46 2012 @@
> -266,6 +266,10 @@ def int_fma : Intrinsic<[llvm_anyfloat_ty],
> [LLVMMatchType<0>, LLVMMatchType<0>,
> LLVMMatchType<0>]>;
> +
> + def int_fmuladd : Intrinsic<[llvm_anyfloat_ty],
> + [LLVMMatchType<0>, LLVMMatchType<0>,
> + LLVMMatchType<0>]>;
> }
>
> // NOTE: these are internal interfaces.
>
> Modified: llvm/trunk/include/llvm/Target/TargetLowering.h
> URL:
> http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetLowering.h?rev=158014&r1=158013&r2=158014&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/Target/TargetLowering.h (original) +++
> llvm/trunk/include/llvm/Target/TargetLowering.h Tue Jun 5 14:07:46
> 2012 @@ -1657,6 +1657,14 @@ return false;
> }
>
> + /// isFMAFasterThanMulAndAdd - Return true if an FMA operation is
> faster than
> + /// a pair of mul and add instructions. fmuladd intrinsics will be
> expanded to
> + /// FMAs when this method returns true (and FMAs are legal),
> otherwise fmuladd
> + /// is expanded to mul + add.
> + virtual bool isFMAFasterThanMulAndAdd(EVT) const {
> + return false;
> + }
> +
> /// isNarrowingProfitable - Return true if it's profitable to
> narrow /// operations of type VT1 to VT2. e.g. on x86, it's
> profitable to narrow /// from i32 to i8 but not from i32 to i16.
>
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
> URL:
> http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp?rev=158014&r1=158013&r2=158014&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
> (original) +++
> llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp Tue Jun
> 5 14:07:46 2012 @@ -4932,6 +4932,27 @@ getValue(I.getArgOperand(1)),
> getValue(I.getArgOperand(2)))); return 0;
> + case Intrinsic::fmuladd: {
> + EVT VT = TLI.getValueType(I.getType());
> + if (TLI.isOperationLegal(ISD::FMA, VT) &&
> TLI.isFMAFasterThanMulAndAdd(VT)){
> + setValue(&I, DAG.getNode(ISD::FMA, dl,
> +
> getValue(I.getArgOperand(0)).getValueType(),
> + getValue(I.getArgOperand(0)),
> + getValue(I.getArgOperand(1)),
> + getValue(I.getArgOperand(2))));
> + } else {
> + SDValue Mul = DAG.getNode(ISD::FMUL, dl,
> +
> getValue(I.getArgOperand(0)).getValueType(),
> + getValue(I.getArgOperand(0)),
> + getValue(I.getArgOperand(1)));
> + SDValue Add = DAG.getNode(ISD::FADD, dl,
> +
> getValue(I.getArgOperand(0)).getValueType(),
> + Mul,
> + getValue(I.getArgOperand(2)));
> + setValue(&I, Add);
> + }
> + return 0;
> + }
> case Intrinsic::convert_to_fp16:
> setValue(&I, DAG.getNode(ISD::FP32_TO_FP16, dl,
> MVT::i16,
> getValue(I.getArgOperand(0))));
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
--
Hal Finkel
Postdoctoral Appointee
Leadership Computing Facility
Argonne National Laboratory
More information about the llvm-commits
mailing list