isZeroPointZero_ -> isAZeroPointZero_ (#71)

Summary:
Pull Request resolved: https://github.com/pytorch/FBGEMM/pull/71

To distinguish it from B zero point

Reviewed By: dskhudia

Differential Revision: D14021554

fbshipit-source-id: 555fad8342eaaf97a19f22ac0dfbb79df7293ce7

2 files changed, 29 insertions, 33 deletions

diff --git a/src/GroupwiseConv.h b/src/GroupwiseConv.h
index 3605fcd..5681359 100644
--- a/src/GroupwiseConv.h
+++ b/src/GroupwiseConv.h
@@ -39,7 +39,7 @@ using jit_rowoffset_kernel_fp = void (*)(
 template <typename accT = int32_t>
 class GenConvKernel {
  public:
-  GenConvKernel(const conv_param_t<>& conv_param, std::int32_t zero_point)
+  GenConvKernel(const conv_param_t<>& conv_param, std::int32_t a_zero_point)
       : WRegs_avx2_{x86::ymm0,
                     x86::ymm1,
                     x86::ymm2,
@@ -74,11 +74,7 @@ class GenConvKernel {
     // vector width in elements; Each element is int8 or uint8
     VLEN_ = vectorWidth_ / 8;
 
-    if (zero_point == 0) {
-      isZeroPointZero_ = true;
-    } else {
-      isZeroPointZero_ = false;
-    }
+    isAZeroPointZero_ = a_zero_point == 0;
 
     G_ = conv_param.G;
     K_per_G_ = conv_param.OC / conv_param.G;
@@ -105,7 +101,7 @@ class GenConvKernel {
     fileName += "_S-" + std::to_string(S_);
     fileName += "_PADH-" + std::to_string(H_PAD_);
     fileName += "_PADW-" + std::to_string(W_PAD_);
-    fileName += "_isZeroPointZero-" + std::to_string(isZeroPointZero_);
+    fileName += "_isZeroPointZero-" + std::to_string(isAZeroPointZero_);
     if (rowOffsetKernel) {
       fileName += "_rowOffset";
     }
@@ -261,7 +257,7 @@ class GenConvKernel {
   asmjit::X86Gp scratchReg2_;
 
   // Other parameters
-  bool isZeroPointZero_;
+  bool isAZeroPointZero_;
 
   // current conv parameters
   int G_; ///< Number of groups
diff --git a/src/GroupwiseConvAcc32Avx2.cc b/src/GroupwiseConvAcc32Avx2.cc
index 94ca87c..f2ce07c 100644
--- a/src/GroupwiseConvAcc32Avx2.cc
+++ b/src/GroupwiseConvAcc32Avx2.cc
@@ -74,11 +74,11 @@ void calculateRowOffsets(
 
 tuple<bool, int, int, int> getKernelSig(
     const conv_param_t<>& conv_param,
-    bool isZeroPointZero) {
+    bool isAZeroPointZero) {
   int C_per_G = conv_param.IC / conv_param.G;
   int K_per_G = conv_param.OC / conv_param.G;
   auto kernelSig =
-      std::make_tuple(isZeroPointZero, conv_param.G, C_per_G, K_per_G);
+      std::make_tuple(isAZeroPointZero, conv_param.G, C_per_G, K_per_G);
   return kernelSig;
 }
 
@@ -431,7 +431,7 @@ void GenConvKernel<int32_t>::genForTopEdge<inst_set_t::avx2>(
         }
         gen8bitFMA<inst_set_t::avx2>(a, actRegAvx2_, WRegs_avx2_[r * S_ + s]);
       } else {
-        if (!isZeroPointZero_) {
+        if (!isAZeroPointZero_) {
           gen8bitFMA<inst_set_t::avx2>(
               a, zeroPTRegAvx2_, WRegs_avx2_[r * S_ + s]);
         }
@@ -452,7 +452,7 @@ void GenConvKernel<int32_t>::genForTopEdge<inst_set_t::avx2>(
   } else {
     a->vmovups(resultRegAvx2_, x86::dword_ptr(out_acts_R_));
   }
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     for (int r = 0; r < H_PAD_; ++r) {
       for (int s = 0; s < S_; ++s) {
         gen8bitFMA<inst_set_t::avx2>(a, zeroPTRegAvx2_, WRegs_avx2_[s]);
@@ -508,7 +508,7 @@ void GenConvKernel<int32_t>::genForTopEdge<inst_set_t::avx2>(
   } else {
     a->vmovups(resultRegAvx2_, x86::dword_ptr(out_acts_R_));
   }
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     for (int r = 0; r < H_PAD_; ++r) {
       for (int s = 0; s < S_; ++s) {
         gen8bitFMA<inst_set_t::avx2>(
@@ -537,7 +537,7 @@ void GenConvKernel<int32_t>::genForTopEdge<inst_set_t::avx2>(
       }
       gen8bitFMA<inst_set_t::avx2>(a, actRegAvx2_, WRegs_avx2_[r * S_ + s]);
     }
-    if (!isZeroPointZero_) {
+    if (!isAZeroPointZero_) {
       for (int s = S_ - W_PAD_; s < S_; ++s) {
         gen8bitFMA<inst_set_t::avx2>(
             a, zeroPTRegAvx2_, WRegs_avx2_[r * S_ + s]);
@@ -573,7 +573,7 @@ void GenConvKernel<int32_t>::genForLeftEdge<inst_set_t::avx2>(
   a->imul(scratchReg1_, W_R_);
   a->imul(scratchReg1_, static_cast<asmjit::Imm>(C_ * sizeof(uint8_t)));
   for (int r = 0; r < R_; ++r) {
-    if (!isZeroPointZero_) {
+    if (!isAZeroPointZero_) {
       for (int s = 0; s < W_PAD_; ++s) {
         gen8bitFMA<inst_set_t::avx2>(
             a, zeroPTRegAvx2_, WRegs_avx2_[r * S_ + s]);
@@ -663,7 +663,7 @@ void GenConvKernel<int32_t>::genForRightEdge<inst_set_t::avx2>(
       gen8bitFMA<inst_set_t::avx2>(a, actRegAvx2_, WRegs_avx2_[r * S_ + s]);
       a->add(scratchReg1_, static_cast<asmjit::Imm>(C_ * sizeof(uint8_t)));
     }
-    if (!isZeroPointZero_) {
+    if (!isAZeroPointZero_) {
       for (int s = S_ - W_PAD_; s < S_; ++s) {
         gen8bitFMA<inst_set_t::avx2>(
             a, zeroPTRegAvx2_, WRegs_avx2_[r * S_ + s]);
@@ -727,7 +727,7 @@ void GenConvKernel<int32_t>::genForBottomEdge<inst_set_t::avx2>(
   a->imul(scratchReg1_, W_R_);
   a->imul(scratchReg1_, static_cast<asmjit::Imm>(C_ * sizeof(uint8_t)));
   for (int r = 0; r < R_ - H_PAD_; ++r) {
-    if (!isZeroPointZero_) {
+    if (!isAZeroPointZero_) {
       for (int s = 0; s < W_PAD_; ++s) {
         gen8bitFMA<inst_set_t::avx2>(
             a, zeroPTRegAvx2_, WRegs_avx2_[r * S_ + s]);
@@ -756,7 +756,7 @@ void GenConvKernel<int32_t>::genForBottomEdge<inst_set_t::avx2>(
     a->imul(scratchReg2_, W_R_, static_cast<asmjit::Imm>(C_ * sizeof(uint8_t)));
     a->add(scratchReg1_, scratchReg2_);
   }
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     for (int r = R_ - H_PAD_; r < R_; ++r) {
       for (int s = 0; s < S_; ++s) {
         gen8bitFMA<inst_set_t::avx2>(
@@ -806,7 +806,7 @@ void GenConvKernel<int32_t>::genForBottomEdge<inst_set_t::avx2>(
     a->add(scratchReg1_, scratchReg2_);
   }
 
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     for (int r = R_ - W_PAD_; r < R_; ++r) {
       for (int s = 0; s < S_; ++s) {
         gen8bitFMA<inst_set_t::avx2>(
@@ -867,7 +867,7 @@ void GenConvKernel<int32_t>::genForBottomEdge<inst_set_t::avx2>(
     }
     a->imul(scratchReg2_, W_R_, static_cast<asmjit::Imm>(C_ * sizeof(uint8_t)));
     a->add(scratchReg1_, scratchReg2_);
-    if (!isZeroPointZero_) {
+    if (!isAZeroPointZero_) {
       for (int s = S_ - W_PAD_; s < S_; ++s) {
         gen8bitFMA<inst_set_t::avx2>(
             a, zeroPTRegAvx2_, WRegs_avx2_[r * S_ + s]);
@@ -875,7 +875,7 @@ void GenConvKernel<int32_t>::genForBottomEdge<inst_set_t::avx2>(
     }
   }
 
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     for (int r = R_ - H_PAD_; r < R_; ++r) {
       for (int s = 0; s < S_; ++s) {
         gen8bitFMA<inst_set_t::avx2>(
@@ -1064,7 +1064,7 @@ jit_conv_kernel_fp GenConvKernel<int32_t>::getOrCreate<inst_set_t::avx2>(
   loopR1_ = a->gpzRef(14);
   loopR2_ = a->gpzRef(15);
 
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     setToZeroPt<inst_set_t::avx2>(a, zeroPTRegAvx2_);
   }
 
@@ -1095,7 +1095,7 @@ jit_conv_kernel_fp GenConvKernel<int32_t>::getOrCreate<inst_set_t::avx2>(
     std::cout << "Error: in fn add" << std::endl;
     return nullptr;
   }
-  auto kernelSig = getKernelSig(conv_param, isZeroPointZero_);
+  auto kernelSig = getKernelSig(conv_param, isAZeroPointZero_);
   codeCache_[kernelSig] = fn;
 
 #if defined(FBGEMM_LOG_CODE)
@@ -1113,7 +1113,7 @@ void GenConvKernel<int32_t>::genForTopEdgeRowoffset<inst_set_t::avx2>(
   // top-left corner code
   // zero out the results register
   a->vxorps(resultRegAvx2_, resultRegAvx2_, resultRegAvx2_);
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     genZeroPtSum<inst_set_t::avx2>(a, R_ * S_ - (R_ - H_PAD_) * (S_ - W_PAD_));
   }
   for (int r = H_PAD_; r < R_; ++r) {
@@ -1140,7 +1140,7 @@ void GenConvKernel<int32_t>::genForTopEdgeRowoffset<inst_set_t::avx2>(
   a->bind(LoopTopEdge);
   // zero out
   a->vxorps(resultRegAvx2_, resultRegAvx2_, resultRegAvx2_);
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     genZeroPtSum<inst_set_t::avx2>(a, H_PAD_ * S_);
   }
   for (int r = H_PAD_; r < R_; ++r) {
@@ -1174,7 +1174,7 @@ void GenConvKernel<int32_t>::genForTopEdgeRowoffset<inst_set_t::avx2>(
   // top-right corner code
   // zero out
   a->vxorps(resultRegAvx2_, resultRegAvx2_, resultRegAvx2_);
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     genZeroPtSum<inst_set_t::avx2>(a, R_ * S_ - (R_ - H_PAD_) * (S_ - W_PAD_));
   }
   for (int r = H_PAD_; r < R_; ++r) {
@@ -1212,7 +1212,7 @@ void GenConvKernel<int32_t>::genForLeftEdgeRowoffset<inst_set_t::avx2>(
   a->bind(LoopLeftEdge);
   // zero out
   a->vxorps(resultRegAvx2_, resultRegAvx2_, resultRegAvx2_);
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     genZeroPtSum<inst_set_t::avx2>(a, R_ * W_PAD_);
   }
   a->mov(scratchReg1_, loopR1_);
@@ -1264,7 +1264,7 @@ void GenConvKernel<int32_t>::genForRightEdgeRowoffset<inst_set_t::avx2>(
   a->bind(LoopRightEdge);
   // zero out
   a->vxorps(resultRegAvx2_, resultRegAvx2_, resultRegAvx2_);
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     genZeroPtSum<inst_set_t::avx2>(a, R_ * W_PAD_);
   }
   a->mov(scratchReg1_, loopR1_);
@@ -1322,7 +1322,7 @@ void GenConvKernel<int32_t>::genForBottomEdgeRowoffset<inst_set_t::avx2>(
   // bottom-left corner
   // zero out
   a->vxorps(resultRegAvx2_, resultRegAvx2_, resultRegAvx2_);
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     genZeroPtSum<inst_set_t::avx2>(a, R_ * S_ - (R_ - H_PAD_) * (S_ - W_PAD_));
   }
   a->mov(scratchReg1_, H_R_);
@@ -1352,7 +1352,7 @@ void GenConvKernel<int32_t>::genForBottomEdgeRowoffset<inst_set_t::avx2>(
   a->bind(LoopBottomEdge);
   // zero out
   a->vxorps(resultRegAvx2_, resultRegAvx2_, resultRegAvx2_);
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     genZeroPtSum<inst_set_t::avx2>(a, H_PAD_ * S_);
   }
   a->mov(scratchReg1_, H_R_);
@@ -1388,7 +1388,7 @@ void GenConvKernel<int32_t>::genForBottomEdgeRowoffset<inst_set_t::avx2>(
   // bottom-right corner
   // zero out
   a->vxorps(resultRegAvx2_, resultRegAvx2_, resultRegAvx2_);
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     genZeroPtSum<inst_set_t::avx2>(a, R_ * S_ - (R_ - H_PAD_) * (S_ - W_PAD_));
   }
   // input start point
@@ -1538,7 +1538,7 @@ GenConvKernel<int32_t>::getOrCreateRowOffset<inst_set_t::avx2>(
 
   // This uses xmm10 register temporarily. Should come before
   // createVector8BitOne
-  if (!isZeroPointZero_) {
+  if (!isAZeroPointZero_) {
     // we can use xmm11 because ymm11 is used by tmpReg1Avx2_
     asmjit::X86Xmm const_reg_xmm = x86::xmm11;
     a->movq(const_reg_xmm, a_zero_pt_R_);
@@ -1569,7 +1569,7 @@ GenConvKernel<int32_t>::getOrCreateRowOffset<inst_set_t::avx2>(
     std::cout << "Error: in fn add" << std::endl;
     return nullptr;
   }
-  auto kernelSig = getKernelSig(conv_param, isZeroPointZero_);
+  auto kernelSig = getKernelSig(conv_param, isAZeroPointZero_);
   codeCacheRowOffset_[kernelSig] = fn;
 
 #if defined(FBGEMM_LOG_CODE)