[Mesa-dev] [PATCH v2] nv50/ir: optimize ADD(SHL(a, b), c) to SHLADD(a, b, c)

[Samuel Pitoiset]

On 10/11/2016 11:17 PM, Ilia Mirkin wrote:
> On Tue, Oct 11, 2016 at 5:01 PM, Samuel Pitoiset
> <samuel.pitoiset at gmail.com> wrote:
>> total instructions in shared programs :2286901 -> 2284473 (-0.11%)
>> total gprs used in shared programs    :335256 -> 335273 (0.01%)
>> total local used in shared programs   :31968 -> 31968 (0.00%)
>>
>>                 local        gpr       inst      bytes
>>     helped           0          41         852         852
>>       hurt           0          44          23          23
>>
>> v2: - use visit(Instruction *)
>>     - use getUniqueInsn()
>>     - use getImmediate()
>>     - fix mod for src0
>>
>> Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
>> ---
>>  .../drivers/nouveau/codegen/nv50_ir_peephole.cpp   | 87 ++++++++++++++++++++++
>>  1 file changed, 87 insertions(+)
>>
>> diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp b/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
>> index 6efb29e..6045e8b 100644
>> --- a/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
>> +++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
>> @@ -2132,6 +2132,92 @@ AlgebraicOpt::visit(BasicBlock *bb)
>>
>>  // =============================================================================
>>
>> +// ADD(SHL(a, b), c) -> SHLADD(a, b, c)
>> +class LateAlgebraicOpt : public Pass
>> +{
>> +private:
>> +   virtual bool visit(Instruction *);
>> +
>> +   void handleADD(Instruction *);
>> +   bool tryADDToSHLADD(Instruction *);
>> +};
>> +
>> +void
>> +LateAlgebraicOpt::handleADD(Instruction *add)
>> +{
>> +   Value *src0 = add->getSrc(0);
>> +   Value *src1 = add->getSrc(1);
>> +
>> +   if (src0->reg.file != FILE_GPR || src1->reg.file != FILE_GPR)
>> +      return;
>> +
>> +   if (prog->getTarget()->isOpSupported(OP_SHLADD, add->dType))
>> +      tryADDToSHLADD(add);
>> +}
>> +
>> +// ADD(SHL(a, b), c) -> SHLADD(a, b, c)
>> +bool
>> +LateAlgebraicOpt::tryADDToSHLADD(Instruction *add)
>> +{
>> +   Value *src0 = add->getSrc(0);
>> +   Value *src1 = add->getSrc(1);
>> +   ImmediateValue imm;
>> +   Instruction *shl;
>> +   Modifier mod[2];
>> +   Value *src;
>> +   int s;
>> +
>> +   if (add->saturate || add->usesFlags() || typeSizeof(add->dType) == 8)
>> +      return false;
>
> || isFloatType(add->dType)

This is not really needed because we can't do SHL with floats, so the 
block above handles that, but it won't hurt to add this check earlier.

>
>> +
>> +   if (src0->getUniqueInsn() && src0->getUniqueInsn()->op == OP_SHL)
>> +      s = 0;
>> +   else
>> +   if (src1->getUniqueInsn() && src1->getUniqueInsn()->op == OP_SHL)
>> +      s = 1;
>> +   else
>> +      return false;
>> +
>> +   src = add->getSrc(s);
>> +   shl = src->getUniqueInsn();
>> +
>> +   if (shl->bb != add->bb || shl->usesFlags() || shl->subOp)
>> +      return false;
>> +
>> +   if (!shl->src(1).getImmediate(imm))
>> +      return false;
>> +
>> +   mod[0] = add->src(0).mod;
>> +   mod[1] = add->src(1).mod;
>> +
>> +   add->op = OP_SHLADD;
>> +   add->setSrc(2, add->src(s ? 0 : 1));
>
> src(!s)

The "convention" is really confusing because, for example we use s?0:1 
in tryADDToMadOrSAD().

>
> In other places, we do
>
> int t = !s;
>
> To avoid having to do the !s. The convention is that s = immediate
> source, so t = "other source".
>
>> +   add->src(2).mod = mod[s];
>> +
>> +   add->setSrc(0, shl->getSrc(0));
>> +   add->src(0).mod = mod[!s];
>
> This mod[] array shouldn't be necessary...

Yeah, setSrc() will update the modifier.

>
> With those minor modifications, this is
>
> Reviewed-by: Ilia Mirkin <imirkin at alum.mit.edu>

Thanks.

>
>> +   add->setSrc(1, new_ImmediateValue(shl->bb->getProgram(), imm.reg.data.u32));
>> +   add->src(1).mod = Modifier(0);
>> +
>> +   return true;
>> +}
>> +
>> +bool
>> +LateAlgebraicOpt::visit(Instruction *i)
>> +{
>> +   switch (i->op) {
>> +   case OP_ADD:
>> +      handleADD(i);
>> +      break;
>> +   default:
>> +      break;
>> +   }
>> +
>> +   return true;
>> +}
>> +
>> +// =============================================================================
>> +
>>  static inline void
>>  updateLdStOffset(Instruction *ldst, int32_t offset, Function *fn)
>>  {
>> @@ -3436,6 +3522,7 @@ Program::optimizeSSA(int level)
>>     RUN_PASS(2, AlgebraicOpt, run);
>>     RUN_PASS(2, ModifierFolding, run); // before load propagation -> less checks
>>     RUN_PASS(1, ConstantFolding, foldAll);
>> +   RUN_PASS(2, LateAlgebraicOpt, run);
>>     RUN_PASS(1, LoadPropagation, run);
>>     RUN_PASS(1, IndirectPropagation, run);
>>     RUN_PASS(2, MemoryOpt, run);
>> --
>> 2.10.0
>>
>> _______________________________________________
>> mesa-dev mailing list
>> mesa-dev at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/mesa-dev