Skip to content

Remove identical branches #4494

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 1 commit into from
Closed

Remove identical branches #4494

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions include/fbgemm/Utils.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -447,6 +447,7 @@ void nbit_embedding_sanity_check(
assert(
(input_bit_rate == 2 || input_bit_rate == 4) &&
"input_bit_rate must be 2 or 4");
// NOLINTNEXTLINE(bugprone-branch-clone)
if constexpr (std::is_same_v<OutType, uint8_t>) {
assert(
(no_bag && input_bit_rate == 4 && output_bit_rate == 4) &&
Expand Down
4 changes: 1 addition & 3 deletions src/EmbeddingSpMDMNBit.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -825,9 +825,7 @@ GenEmbeddingSpMDMNBitLookup<
a->vextracti32x4(temp_vreg.xmm(), src_vreg, asmjit::Imm(i));
a->vpmovsxbd(temp_vreg, temp_vreg.xmm());
} else {
if (i == 1) {
a->vpsrldq(src_vreg, src_vreg, asmjit::Imm(8));
} else if (i == 2) {
if (i == 2) {
a->vextractf128(
src_vreg.xmm(), temp2_vreg.ymm(), asmjit::Imm(i >> 1));
} else {
Expand Down
91 changes: 4 additions & 87 deletions src/FbgemmConv.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -149,8 +149,8 @@ int fbgemmConv(
"For depthwise, only requantized output is supported");

if constexpr (
processOutputType::QGRANType == QuantizationGranularity::TENSOR) {
depthwise_3d_same_pad<QuantizationGranularity::TENSOR>(
processOutputType::QGRANType == QuantizationGranularity::TENSOR || processOutputType::QGRANType == QuantizationGranularity::GROUP || processOutputType::QGRANType == QuantizationGranularity::OUT_CHANNEL) {
depthwise_3d_same_pad<processOutputType::QGRANType>(
*reinterpret_cast<const conv_param_t<3>*>(&conv_p),
outProcess.getAZeroPoint(),
activations,
Expand All @@ -165,41 +165,6 @@ int fbgemmConv(
act_times_w_scale,
thread_id,
num_threads);
} else if constexpr (
processOutputType::QGRANType == QuantizationGranularity::GROUP) {
depthwise_3d_same_pad<QuantizationGranularity::GROUP>(
*reinterpret_cast<const conv_param_t<3>*>(&conv_p),
outProcess.getAZeroPoint(),
activations,
B_zero_point,
*(packed_weights.getPackedWForDepthwise()),
C_multiplier,
outProcess.getCZeroPoint(),
out,
outProcess.getColOffsets(),
outProcess.getBias(),
outProcess.RELU_FUSED, // fuse_relu
act_times_w_scale, // act_scale * weight_scale
thread_id,
num_threads);
} else if constexpr (
processOutputType::QGRANType ==
QuantizationGranularity::OUT_CHANNEL) {
depthwise_3d_same_pad<QuantizationGranularity::OUT_CHANNEL>(
*reinterpret_cast<const conv_param_t<3>*>(&conv_p),
outProcess.getAZeroPoint(),
activations,
B_zero_point,
*(packed_weights.getPackedWForDepthwise()),
C_multiplier,
outProcess.getCZeroPoint(),
out,
outProcess.getColOffsets(),
outProcess.getBias(),
outProcess.RELU_FUSED, // fuse_relu
act_times_w_scale, // act_scale * weight_scale
thread_id,
num_threads);
} else {
std::string msg =
"[FBGEMM_CONV_ERROR] This quantization granularity is "
Expand All @@ -208,8 +173,8 @@ int fbgemmConv(
}
} else if constexpr (SPATIAL_DIM == 2) {
if constexpr (
processOutputType::QGRANType == QuantizationGranularity::TENSOR) {
depthwise_2d_same_pad<QuantizationGranularity::TENSOR>(
processOutputType::QGRANType == QuantizationGranularity::TENSOR || processOutputType::QGRANType == QuantizationGranularity::GROUP || processOutputType::QGRANType == QuantizationGranularity::OUT_CHANNEL) {
depthwise_2d_same_pad<processOutputType::QGRANType>(
conv_p.MB, // mini batch
conv_p.IN_DIM[0], // H
conv_p.IN_DIM[1], // W
Expand All @@ -230,54 +195,6 @@ int fbgemmConv(
act_times_w_scale,
thread_id,
num_threads);
} else if constexpr (
processOutputType::QGRANType == QuantizationGranularity::GROUP) {
depthwise_2d_same_pad<QuantizationGranularity::GROUP>(
conv_p.MB, // mini batch
conv_p.IN_DIM[0], // H
conv_p.IN_DIM[1], // W
conv_p.IC, // input channels
conv_p.OC, // output channels
conv_p.stride[0], // stride_h
conv_p.stride[1], // stride_w
outProcess.getAZeroPoint(),
activations,
B_zero_point,
*(packed_weights.getPackedWForDepthwise()),
C_multiplier,
outProcess.getCZeroPoint(),
out,
outProcess.getColOffsets(),
outProcess.getBias(),
outProcess.RELU_FUSED, // fuse_relu
act_times_w_scale, // act_scale * weight_scale
thread_id,
num_threads);
} else if constexpr (
processOutputType::QGRANType ==
QuantizationGranularity::OUT_CHANNEL) {
// The number of input channels == groups for depthwise convolutions
depthwise_2d_same_pad<QuantizationGranularity::OUT_CHANNEL>(
conv_p.MB, // mini batch
conv_p.IN_DIM[0], // H
conv_p.IN_DIM[1], // W
conv_p.IC, // input channels
conv_p.OC, // output channels
conv_p.stride[0], // stride_h
conv_p.stride[1], // stride_w
outProcess.getAZeroPoint(),
activations,
B_zero_point,
*(packed_weights.getPackedWForDepthwise()),
C_multiplier,
outProcess.getCZeroPoint(),
out,
outProcess.getColOffsets(),
outProcess.getBias(),
outProcess.RELU_FUSED, // fuse_relu
act_times_w_scale, // act_scale * weight_scale
thread_id,
num_threads);
} else {
std::string msg =
"[FBGEMM_CONV_ERROR] This quantization granularity is "
Expand Down
4 changes: 0 additions & 4 deletions src/QuantUtilsAvx2.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1265,8 +1265,6 @@ void requantizeOutputProcessingGConvAvx2(
_mm_set1_epi32(r.B_zero_point[quant_param_idx])),
_mm_set1_epi32(r.B_zero_point[quant_param_idx + 1]),
1);
} else if constexpr (C_PER_G == 8) {
B_zero_point_v = _mm256_set1_epi32(r.B_zero_point[quant_param_idx]);
} else {
B_zero_point_v = _mm256_set1_epi32(r.B_zero_point[quant_param_idx]);
}
Expand Down Expand Up @@ -1343,8 +1341,6 @@ void requantizeOutputProcessingGConvAvx2(
_mm_set1_ps(r.C_multiplier[quant_param_idx])),
_mm_set1_ps(r.C_multiplier[quant_param_idx + 1]),
1);
} else if constexpr (C_PER_G == 8) {
multiplier_v = _mm256_set1_ps(r.C_multiplier[quant_param_idx]);
} else {
multiplier_v = _mm256_set1_ps(r.C_multiplier[quant_param_idx]);
}
Expand Down
Loading