Skip to content

GitLab

Commit d8fef836 authored by carlushuang's avatar carlushuang
Browse files

fix bug for ho/wo out-of-bound access

parent 5742d293
Hide whitespace changes
Inline Side-by-side
Showing with 49 additions and 57 deletions
include/ck/tensor_operation/cpu/grid/gridwise_direct_conv_avx2.hpp
View file @ d8fef836
......@@ -425,21 +425,21 @@ struct GridwiseDirectConvNHWCAvx2
const ck::index_t num_works_wo = math::integer_divide_ceil(Wo, m_per_thread);
const ck::index_t num_works_k = math::integer_divide_ceil(K, n_per_thread);
auto distribute_num_threads_nho_wo_k = [&](ck::index_t& num_threads_nho,
ck::index_t& num_threads_wo,
ck::index_t& num_threads_k) {
auto distribute_num_threads_nho_wo_k = [&](ck::index_t& num_threads_nho_,
ck::index_t& num_threads_wo_,
ck::index_t& num_threads_k_) {
// TODO: only consider multiply of 2 to divide threads
ck::index_t num_threads = total_threads;
num_threads_nho = devide_thread(num_threads, num_works_nho, 2);
num_threads = num_threads / num_threads_nho;
num_threads_nho_ = devide_thread(num_threads, num_works_nho, 2);
num_threads = num_threads / num_threads_nho_;
num_threads_wo = devide_thread(num_threads, num_works_wo, 2);
num_threads = num_threads / num_threads_wo;
num_threads_wo_ = devide_thread(num_threads, num_works_wo, 2);
num_threads = num_threads / num_threads_wo_;
num_threads_k = devide_thread(num_threads, num_works_k, 2);
// num_threads = num_threads / num_threads_k;
num_threads_k_ = devide_thread(num_threads, num_works_k, 2);
// num_threads = num_threads / num_threads_k_;
};
ck::index_t num_threads_nho;
......@@ -456,9 +456,9 @@ struct GridwiseDirectConvNHWCAvx2
math::integer_divide_ceil(num_works_k, num_threads_k);
// printf("num_threads_nho:%d, num_threads_wo:%d, num_threads_k:%d |
// num_works_wo_per_thread:%d, num_works_k_per_thread:%d\n",
// num_threads_nho, num_threads_wo, num_threads_k, num_works_wo_per_thread,
// num_works_k_per_thread);
// num_works_nho_per_thread:%d, num_works_wo_per_thread:%d, num_works_k_per_thread:%d\n",
// num_threads_nho, num_threads_wo, num_threads_k, num_works_nho_per_thread,
// num_works_wo_per_thread, num_works_k_per_thread); fflush(stdout);
if(dynamic_tunable.loop_over_spec ==
ck::tensor_operation::cpu::device::ConvolutionForwardBlockLoopOverSpecialization_t::
......@@ -523,7 +523,7 @@ struct GridwiseDirectConvNHWCAvx2
}
};
for(; i_nho < (tid_nho + 1) * num_works_nho_per_thread;
for(; (i_nho < (tid_nho + 1) * num_works_nho_per_thread) && (i_nho < num_works_nho);
i_nho += 1, accumulate_n_ho())
{
// for input
......@@ -573,15 +573,6 @@ struct GridwiseDirectConvNHWCAvx2
// chance to clear zero (like
// padding) we need to manually clear that
/* left corner shift
* when i_wi is negative, need shift i_wo to right to make i_wi
* possitive sx px i_wi steps_wo_turn_possitive 1 0
* 0, 1, 2.... 0 2 0 0, 2, 4... 0 1 1 -1,
* 0, 1.... 1 2 1 -1, 1, 3.... 1 2 2 -2, 0, 2... 1 2
* 3 -3, -1, 1... 2 3 1 -1, 2, 5... 1 3 2 -2,
* 1, 4.... 1 3 3 -3, 0, 3 1 3 4 -4,
* -1, 2... 2
*/
if(i_wi < 0)
{
ck::index_t wi_to_zero_length =
......@@ -603,6 +594,9 @@ struct GridwiseDirectConvNHWCAvx2
Sx; // now i_wi will be a possitive number
}
if(i_wi >= Wi)
continue;
// shrink right wi/wo
if((i_wi + ((current_wo_size - 1) * Sx)) >= Wi)
{
......@@ -623,24 +617,19 @@ struct GridwiseDirectConvNHWCAvx2
if(pad_wo_size != 0)
{
// manually clear zero. this may and only may need once along
// the gemm_k reduction
// ck::index_t i_k = tid_k * num_works_k_per_thread *
// n_per_thread; ck::index_t current_k_block_size =
// ck::math::min(K - i_k, num_works_k_per_thread *
// n_per_thread);
const intptr_t offset_c =
GetCBlockStartOffset(c_grid_desc, i_nho * Wo, i_k);
// printf("pad_wo_size:%d, current_k_block_size:%d, clear
// offset_c:%d\n",
// pad_wo_size, current_k_size * pad_wo_size,
// offset_c);fflush(stdout);
ck::cpu::avx2_util::memset32_avx2(
&c_block_buf.p_data_[offset_c],
0,
current_k_size * pad_wo_size);
for(ck::index_t i_wo_pad = 0; i_wo_pad < pad_wo_size;
i_wo_pad++)
{
const intptr_t offset_c = GetCBlockStartOffset(
c_grid_desc, i_nho * Wo + i_wo_pad, i_k);
// printf("pad_wo_size:%d, current_k_block_size:%d, clear
// offset_c:%d\n",
// pad_wo_size, current_k_size,
// offset_c);fflush(stdout);
ck::cpu::avx2_util::memset32_avx2(
&c_block_buf.p_data_[offset_c], 0, current_k_size);
}
}
const intptr_t offset_a = current_input_offset;
......@@ -652,7 +641,7 @@ struct GridwiseDirectConvNHWCAvx2
// printf("offset_a:%lu, offset_b:%lu, offset_c:%lu, i_n:%d,
// i_hi:%d, i_wi:%d, i_dx:%d, i_dy:%d, i_k:%d, i_ho:%d, i_wo:%d,
// current_wo_size:%d, current_k_size:%d, i_nho:%d, lda:%d, ldb:%d,
// ldc:%d, acc:%d",
// ldc:%d, acc:%d\n",
// offset_a, offset_b, offset_c, i_n, i_hi, i_wi, i_dx, i_dy,
// i_k, i_ho, current_i_wo, current_wo_size, current_k_size,
// i_nho, param.lda / sizeof(FloatA), param.ldb /
......@@ -665,8 +654,6 @@ struct GridwiseDirectConvNHWCAvx2
ThreadwiseGemm_Dispatch::Run(
&param, current_wo_size, current_k_size);
// printf(" ------ \n");fflush(stdout);
}
}
}
......@@ -736,11 +723,10 @@ struct GridwiseDirectConvNHWCAvx2
}
};
for(; i_nho < (tid_nho + 1) * num_works_nho_per_thread;
for(; (i_nho < (tid_nho + 1) * num_works_nho_per_thread) && (i_nho < num_works_nho);
i_nho += 1, accumulate_n_ho())
{
// for input
ck::index_t i_hi_no_y = i_ho * Sy - Py;
for(ck::index_t i_wo = tid_wo * num_works_wo_per_thread * m_per_thread;
......@@ -753,9 +739,11 @@ struct GridwiseDirectConvNHWCAvx2
ck::index_t i_dx = 0;
ck::index_t i_dy = 0;
bool accmulate_c = false;
// printf("-- i_nho:%d, i_wo:%d, num_works_nho:%d, num_threads_nho:%d(Hi:%d,
// Wi:%d)\n", i_nho, i_wo, num_works_nho, num_threads_nho, Hi,
// Wi);fflush(stdout);
// printf("-- [%d] i_nho:%d, i_wo:%d, num_works_nho:%d,
// num_threads_nho:%d(Hi:%d, Wi:%d), current_wo_size_no_dx:%d,
// m_per_thread:%d\n",
// tid, i_nho, i_wo, num_works_nho, num_threads_nho, Hi, Wi,
// current_wo_size_no_dx, m_per_thread);fflush(stdout);
auto accumulate_dy_dx = [&]() {
i_dx += Dx;
......@@ -812,6 +800,9 @@ struct GridwiseDirectConvNHWCAvx2
Sx; // now i_wi will be a possitive number
}
if(i_wi >= Wi)
continue;
// shrink right wi/wo
if((i_wi + ((current_wo_size - 1) * Sx)) >= Wi)
{
......@@ -841,8 +832,10 @@ struct GridwiseDirectConvNHWCAvx2
const intptr_t offset_c =
GetCBlockStartOffset(c_grid_desc, i_nho * Wo, i_k);
// printf("pad_wo_size:%d, current_k_block_size:%d, offset_c:%d\n",
// pad_wo_size, current_k_block_size, offset_c);fflush(stdout);
// printf("[%d] pad_wo_size:%d, current_k_block_size:%d,
// offset_c:%d\n",
// tid, pad_wo_size, current_k_block_size,
// offset_c);fflush(stdout);
ck::cpu::avx2_util::memset32_avx2(&c_block_buf.p_data_[offset_c],
0,
current_k_block_size *
......@@ -861,12 +854,13 @@ struct GridwiseDirectConvNHWCAvx2
const intptr_t offset_c = GetCBlockStartOffset(
c_grid_desc, i_nho * Wo + current_i_wo, i_k);
// printf("offset_a:%lu, offset_b:%lu, offset_c:%lu, i_n:%d,
// printf("[%d] offset_a:%lu, offset_b:%lu, offset_c:%lu, i_n:%d,
// i_hi:%d, i_wi:%d, i_dx:%d, i_dy:%d, i_k:%d, i_ho:%d, i_wo:%d,
// current_wo_size:%d, i_nho:%d, lda:%d, ldb:%d",
// offset_a, offset_b, offset_c, i_n, i_hi, i_wi, i_dx, i_dy,
// i_k, i_ho, current_i_wo, current_wo_size, i_nho, param.lda /
// sizeof(FloatA), param.ldb / sizeof(FloatB)); fflush(stdout);
// current_wo_size:%d, i_nho:%d, lda:%d, ldb:%d\n",
// tid, offset_a, offset_b, offset_c, i_n, i_hi, i_wi, i_dx,
// i_dy, i_k, i_ho, current_i_wo, current_wo_size, i_nho,
// param.lda / sizeof(FloatA), param.ldb / sizeof(FloatB));
// fflush(stdout);
param.p_a = &a_block_buf.p_data_[offset_a];
param.p_b = &b_block_buf.p_data_[offset_b];
......@@ -874,8 +868,6 @@ struct GridwiseDirectConvNHWCAvx2
ThreadwiseGemm_Dispatch::Run(
&param, current_wo_size, current_k_size);
// printf(" ------ \n");fflush(stdout);
}
}
}
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment