diff --git a/script/profile_avg_pool2d_bwd.sh b/script/profile_avg_pool2d_bwd.sh index dc35b41e1f..5272b38314 100755 --- a/script/profile_avg_pool2d_bwd.sh +++ b/script/profile_avg_pool2d_bwd.sh @@ -10,6 +10,6 @@ LOG=$5 TIME=$6 ######## op datatype verify init log time length(NCHW) window size(YX) stride dilation left pad right pad -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 2 32 30 30 --wsize 2 2 --wstride 2 2 --wdilation 1 1 --pad1 1 1 --pad2 1 1 -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 64 256 64 64 --wsize 2 2 --wstride 2 2 --wdilation 1 1 --pad1 1 1 --pad2 1 1 +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 2 32 30 30 --wsize 2 2 --wstride 2 2 --wdilation 1 1 --pad1 1 1 --pad2 1 1 --dmmy 28 29 30 31 32 +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 64 256 64 64 --wsize 2 2 --wstride 2 2 --wdilation 1 1 --pad1 1 1 --pad2 1 1 --dmmy 28 29 30 31 32 diff --git a/script/profile_avg_pool3d_bwd.sh b/script/profile_avg_pool3d_bwd.sh index df987f775a..062d00f184 100755 --- a/script/profile_avg_pool3d_bwd.sh +++ b/script/profile_avg_pool3d_bwd.sh @@ -9,7 +9,7 @@ INIT=$4 LOG=$5 TIME=$6 -######## op datatype verify init log time length(NCDHW) window size(YX) stride dilation left pad right pad -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 2 32 30 30 30 --wsize 2 2 2 --wstride 2 2 2 --wdilation 1 1 --pad1 1 1 1 --pad2 1 1 1 -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 64 64 64 64 64 --wsize 2 2 2 --wstride 2 2 2 --wdilation 1 1 --pad1 1 1 1 --pad2 1 1 1 +######## op datatype verify init log time length(NCDHW) window size(YX) stride dilation left pad right pad +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 2 32 30 30 30 --wsize 2 2 2 --wstride 2 2 2 --wdilation 1 1 1 --pad1 1 1 1 --pad2 1 1 1 +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 64 64 64 64 64 --wsize 2 2 2 --wstride 2 2 2 --wdilation 1 1 1 --pad1 1 1 1 --pad2 1 1 1 diff --git a/script/profile_bnorm.sh b/script/profile_bnorm.sh index 581d14ff17..09a261ccb6 100755 --- a/script/profile_bnorm.sh +++ b/script/profile_bnorm.sh @@ -9,5 +9,5 @@ INIT=$4 TIME=$5 USE=$6 -######## op datatype UseSavedMean verify init log time inOutLengths(nhwc) reduceDims verify - $DRIVER $OP $DATATYPE $USE $VERIFY $INIT $LOG $TIME -D 64,64,280,82 -R 0 $VERIFY \ No newline at end of file +######## op datatype UseSavedMean init time inOutLengths(nhwc) reduceDims verify + $DRIVER $OP $DATATYPE $USE $INIT $TIME -D 64,64,280,82 -R 1,2,3 $VERIFY \ No newline at end of file diff --git a/script/profile_bnorm_fwd.sh b/script/profile_bnorm_fwd.sh new file mode 100755 index 0000000000..008a137bfc --- /dev/null +++ b/script/profile_bnorm_fwd.sh @@ -0,0 +1,14 @@ +#!/bin/bash +## GPU visibility +export HIP_VISIBLE_DEVICES=0 +DRIVER="../build/bin/ckProfiler" +OP=$1 +DATATYPE=$2 +VERIFY="-v $3" +INIT=$4 +TIME=$5 +USE=$6 + +######## op datatype updateMovingAverage saveMeanAndInvVariance init time inOutLengths(nhwc) reduceDims verify +$DRIVER $OP $DATATYPE $USE 0 $INIT $TIME -D 64,64,280,82 -R 1,2,3 $VERIFY +$DRIVER $OP $DATATYPE $USE 1 $INIT $TIME -D 64,64,280,82 -R 1,2,3 $VERIFY \ No newline at end of file diff --git a/script/profile_max_pool2d.sh b/script/profile_max_pool2d.sh deleted file mode 100755 index e0f5d75add..0000000000 --- a/script/profile_max_pool2d.sh +++ /dev/null @@ -1,14 +0,0 @@ -#!/bin/bash -## GPU visibility -export HIP_VISIBLE_DEVICES=0 -DRIVER="../build/bin/ckProfiler" -OP=$1 -DATATYPE=$2 -VERIFY=$3 -INIT=$4 -LOG=$5 -TIME=$6 - -######## op datatype verify init log time length(NCHW) -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 2 32 30 30 -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 64 32 256 256 diff --git a/script/profile_max_pool2d_bwd.sh b/script/profile_max_pool2d_bwd.sh new file mode 100755 index 0000000000..587a8ab350 --- /dev/null +++ b/script/profile_max_pool2d_bwd.sh @@ -0,0 +1,14 @@ +#!/bin/bash +## GPU visibility +export HIP_VISIBLE_DEVICES=0 +DRIVER="../build/bin/ckProfiler" +OP=$1 +DATATYPE=$2 +VERIFY=$3 +INIT=$4 +LOG=$5 +TIME=$6 + +######## op datatype verify init log time length(NCHW) window size(YX) stride dilation left pad right pad +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 2 32 30 30 --wsize 2 2 --wstride 2 2 --wdilation 1 1 --pad1 1 1 --pad2 1 1 --dmmy 28 29 30 31 32 +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 64 32 256 256 --wsize 2 2 --wstride 2 2 --wdilation 1 1 --pad1 1 1 --pad2 1 1 --dmmy 28 29 30 31 32 diff --git a/script/profile_max_pool2d_fwd.sh b/script/profile_max_pool2d_fwd.sh new file mode 100755 index 0000000000..9e7f8293ba --- /dev/null +++ b/script/profile_max_pool2d_fwd.sh @@ -0,0 +1,16 @@ +#!/bin/bash +## GPU visibility +export HIP_VISIBLE_DEVICES=0 +DRIVER="../build/bin/ckProfiler" +OP=$1 +DATATYPE=$2 +VERIFY=$3 +INIT=$4 +LOG=$5 +TIME=$6 + +######## op datatype verify init log time return_idx length(NCHW) window size(YX) stride dilation left pad right pad +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 0 --length 2 32 30 30 --wsize 2 2 --wstride 2 2 --wdilation 1 1 --pad1 1 1 --pad2 1 1 +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 0 --length 64 32 256 256 --wsize 2 2 --wstride 2 2 --wdilation 1 1 --pad1 1 1 --pad2 1 1 +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 1 --length 2 32 30 30 --wsize 2 2 --wstride 2 2 --wdilation 1 1 --pad1 1 1 --pad2 1 1 +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 1 --length 64 32 256 256 --wsize 2 2 --wstride 2 2 --wdilation 1 1 --pad1 1 1 --pad2 1 1 \ No newline at end of file diff --git a/script/profile_max_pool3d.sh b/script/profile_max_pool3d.sh index b4e58b9861..2c8644e2a6 100755 --- a/script/profile_max_pool3d.sh +++ b/script/profile_max_pool3d.sh @@ -9,6 +9,6 @@ INIT=$4 LOG=$5 TIME=$6 -######## op datatype verify init log time length(NCDHW) -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 2 32 30 30 30 -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 8 16 32 256 256 +######## op datatype verify init log time length(NCDHW) window size(YX) stride dilation left pad right pad +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 2 32 30 30 30 --wsize 2 2 2 --wstride 2 2 2 --wdilation 1 1 1 --pad1 1 1 1 --pad2 1 1 1 +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME --length 8 16 32 256 256 --wsize 2 2 2 --wstride 2 2 2 --wdilation 1 1 1 --pad1 1 1 1 --pad2 1 1 1 diff --git a/script/profile_pool3d_fwd.sh b/script/profile_pool3d_fwd.sh index cdf8bd020b..7fa102394b 100755 --- a/script/profile_pool3d_fwd.sh +++ b/script/profile_pool3d_fwd.sh @@ -10,8 +10,8 @@ LOG=$5 TIME=$6 REDUCEOP=$7 -######## op datatype verify init log time return_index reduce_op length(NCDHW) -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 0 $REDUCEOP --length 2 32 30 30 30 -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 0 $REDUCEOP --length 8 16 32 256 256 -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 1 $REDUCEOP --length 2 32 30 30 30 -$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 1 $REDUCEOP --length 8 16 32 256 256 \ No newline at end of file +######## op datatype verify init log time return_index reduce_op length(NCDHW) window size(YX) stride dilation left pad right pad +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 0 $REDUCEOP --length 2 32 30 30 30 --wsize 2 2 2 --wstride 2 2 2 --wdilation 1 1 1 --pad1 1 1 1 --pad2 1 1 1 +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 0 $REDUCEOP --length 8 16 32 256 256 --wsize 2 2 2 --wstride 2 2 2 --wdilation 1 1 1 --pad1 1 1 1 --pad2 1 1 1 +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 1 $REDUCEOP --length 2 32 30 30 30 --wsize 2 2 2 --wstride 2 2 2 --wdilation 1 1 1 --pad1 1 1 1 --pad2 1 1 1 +$DRIVER $OP $DATATYPE $VERIFY $INIT $LOG $TIME 1 $REDUCEOP --length 8 16 32 256 256 --wsize 2 2 2 --wstride 2 2 2 --wdilation 1 1 1 --pad1 1 1 1 --pad2 1 1 1 \ No newline at end of file diff --git a/script/run_ck_tile.sh b/script/run_ck_tile.sh index 88374a4b95..d29a0b064a 100755 --- a/script/run_ck_tile.sh +++ b/script/run_ck_tile.sh @@ -22,24 +22,22 @@ function print_log_header(){ /opt/rocm/bin/amdclang++ --version | grep -e 'InstalledDir' >> $1; } -cd .. - #run mha performance benchmarks export fmha_fwd_log="perf_fmha_fwd_$GPU_arch.log" print_log_header $fmha_fwd_log $env_type $branch $host_name -example/ck_tile/01_fmha/script/benchmark_fwd.sh 2>&1 | tee -a $fmha_fwd_log +./example/ck_tile/01_fmha/script/benchmark_fwd.sh 2>&1 | tee -a $fmha_fwd_log export fmha_bwd_log="perf_fmha_bwd_$GPU_arch.log" print_log_header $fmha_bwd_log $env_type $branch $host_name -example/ck_tile/01_fmha/script/benchmark_bwd.sh 2>&1 | tee -a $fmha_bwd_log +./example/ck_tile/01_fmha/script/benchmark_bwd.sh 2>&1 | tee -a $fmha_bwd_log #run layernorm2d export layernorm2d_log="perf_layernorm2d_$GPU_arch.log" print_log_header $layernorm2d_log $env_type $branch $host_name -example/ck_tile/02_layernorm2d/script/perf_test.sh 2>&1 | tee -a $layernorm2d_log +./example/ck_tile/02_layernorm2d/script/perf_test.sh 2>&1 | tee -a $layernorm2d_log #run gemm -example/ck_tile/03_gemm/script/run_full_test.sh +./example/ck_tile/03_gemm/script/run_full_test.sh #run image2col export img2col_log="perf_img2col_$GPU_arch.log" @@ -54,37 +52,37 @@ print_log_header $reduce_log $env_type $branch $host_name #run permute export permute_log="perf_permute_$GPU_arch.log" print_log_header $permute_log $env_type $branch $host_name -example/ck_tile/06_permute/script/smoke_test.sh 2>&1 | tee -a $permute_log +./example/ck_tile/06_permute/script/smoke_test.sh 2>&1 | tee -a $permute_log #run topk_softmax export topk_softmax_log="perf_topk_softmax_$GPU_arch.log" print_log_header $topk_softmax_log $env_type $branch $host_name -example/ck_tile/09_topk_softmax/script/smoke_test.sh 2>&1 | tee -a $topk_softmax_log +./example/ck_tile/09_topk_softmax/script/smoke_test.sh 2>&1 | tee -a $topk_softmax_log #run rmsnorm2d export rmsnorm2d_log="perf_rmsnorm2d_$GPU_arch.log" print_log_header $rmsnorm2d_log $env_type $branch $host_name -example/ck_tile/10_rmsnorm2d/script/perf_test.sh 2>&1 | tee -a $rmsnorm2d_log +./example/ck_tile/10_rmsnorm2d/script/perf_test.sh 2>&1 | tee -a $rmsnorm2d_log #run add_rmsnorm2d_rdquant export add_rmsnorm2d_rdquant_log="perf_add_rmsnorm2d_rdquant_$GPU_arch.log" print_log_header $add_rmsnorm2d_rdquant_log $env_type $branch $host_name -example/ck_tile/11_add_rmsnorm2d_rdquant/script/perf_test.sh 2>&1 | tee -a $add_rmsnorm2d_rdquant_log +./example/ck_tile/11_add_rmsnorm2d_rdquant/script/perf_test.sh 2>&1 | tee -a $add_rmsnorm2d_rdquant_log #run smoothquant export smoothquant_log="perf_smoothquant_$GPU_arch.log" print_log_header $smoothquant_log $env_type $branch $host_name -example/ck_tile/12_smoothquant/script/perf_test.sh 2>&1 | tee -a $smoothquant_log +./example/ck_tile/12_smoothquant/script/perf_test.sh 2>&1 | tee -a $smoothquant_log #run moe_sorting export moe_sorting_log="perf_moe_sorting_$GPU_arch.log" print_log_header $moe_sorting_log $env_type $branch $host_name -example/ck_tile/13_moe_sorting/script/smoke_test.sh 2>&1 | tee -a $moe_sorting_log +./example/ck_tile/13_moe_sorting/script/smoke_test.sh 2>&1 | tee -a $moe_sorting_log #run moe_smoothquant export moe_smoothquant_log="perf_moe_smoothquant_$GPU_arch.log" print_log_header $moe_smoothquant_log $env_type $branch $host_name -example/ck_tile/14_moe_sorting/script/perf_test.sh 2>&1 | tee -a $moe_smoothquant_log +./example/ck_tile/14_moe_sorting/script/perf_test.sh 2>&1 | tee -a $moe_smoothquant_log #run fused_moe export fused_moe_log="perf_fused_moe_$GPU_arch.log" @@ -100,9 +98,9 @@ print_log_header $fused_moe_log $env_type $branch $host_name #run flatmm export flatmm_log="perf_flatmm_$GPU_arch.log" print_log_header $flatmm_log $env_type $branch $host_name -example/ck_tile/18_flatmm/script/smoke_test_basic.sh 2>&1 | tee -a $flatmm_log +./example/ck_tile/18_flatmm/script/smoke_test_basic.sh 2>&1 | tee -a $flatmm_log #run batched_transpose export batched_transpose_log="perf_batched_transpose_$GPU_arch.log" print_log_header $batched_transpose_log $env_type $branch $host_name -example/ck_tile/35_batched_transpose/script/smoke_test_basic.sh 2>&1 | tee -a $batched_transpose_log +./example/ck_tile/35_batched_transpose/script/smoke_test_basic.sh 2>&1 | tee -a $batched_transpose_log diff --git a/script/run_full_performance_tests.sh b/script/run_full_performance_tests.sh index 4848b81239..930980d44a 100755 --- a/script/run_full_performance_tests.sh +++ b/script/run_full_performance_tests.sh @@ -103,6 +103,9 @@ print_log_header $grouped_conv_bwd_data_log $env_type $branch $host_name ./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 0 1 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log ./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 1 1 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log ./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 2 1 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log +./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 0 0 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log +./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 1 0 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log +./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 2 0 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log #run grouped_bwd_weight tests export grouped_conv_bwd_weight_log="perf_grouped_conv_bwd_weight.log" @@ -193,14 +196,10 @@ print_log_header $batched_gemm_multi_d_log $env_type $branch $host_name #run batched_gemm_reduce tests export batched_gemm_reduce_log="perf_batched_gemm_reduce.log" print_log_header $batched_gemm_reduce_log $env_type $branch $host_name -./profile_batched_gemm.sh batched_gemm_reduce 0 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log -./profile_batched_gemm.sh batched_gemm_reduce 0 1 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log -./profile_batched_gemm.sh batched_gemm_reduce 0 2 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log -./profile_batched_gemm.sh batched_gemm_reduce 0 3 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log -./profile_batched_gemm.sh batched_gemm_reduce 1 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log -./profile_batched_gemm.sh batched_gemm_reduce 1 1 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log -./profile_batched_gemm.sh batched_gemm_reduce 1 2 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log -./profile_batched_gemm.sh batched_gemm_reduce 1 3 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log +./profile_batched_gemm_reduce.sh batched_gemm_reduce 1 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log +./profile_batched_gemm_reduce.sh batched_gemm_reduce 1 1 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log +./profile_batched_gemm_reduce.sh batched_gemm_reduce 1 2 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log +./profile_batched_gemm_reduce.sh batched_gemm_reduce 1 3 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log #run contraction_bilinear tests export contraction_bilinear_log="perf_contraction_bilinear.log" @@ -233,9 +232,6 @@ print_log_header $conv_bwd_data_log $env_type $branch $host_name #run conv_fwd_bias_relu_add tests export conv_fwd_bias_relu_add_log="perf_conv_fwd_bias_relu_add.log" print_log_header $conv_fwd_bias_relu_add_log $env_type $branch $host_name -./profile_conv_fwd_bias_relu_add.sh conv_fwd_bias_relu_add 0 0 $verify 1 0 1 256 2>&1 | tee -a $conv_fwd_bias_relu_add_log -./profile_conv_fwd_bias_relu_add.sh conv_fwd_bias_relu_add 1 0 $verify 1 0 1 256 2>&1 | tee -a $conv_fwd_bias_relu_add_log -./profile_conv_fwd_bias_relu_add.sh conv_fwd_bias_relu_add 0 1 $verify 1 0 1 256 2>&1 | tee -a $conv_fwd_bias_relu_add_log ./profile_conv_fwd_bias_relu_add.sh conv_fwd_bias_relu_add 1 1 $verify 1 0 1 256 2>&1 | tee -a $conv_fwd_bias_relu_add_log #run conv_fwd tests @@ -267,21 +263,6 @@ print_log_header $conv_tensor_rearrange_log $env_type $branch $host_name #run gemm_ab_scale tests export gemm_ab_scale_log="perf_gemm_ab_scale.log" print_log_header $gemm_ab_scale_log $env_type $branch $host_name -./profile_gemm_b_scale.sh gemm_ab_scale 0 0 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 1 0 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 2 0 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 3 0 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 4 0 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 5 0 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 6 0 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 7 0 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 0 1 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 1 1 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 2 1 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 3 1 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 4 1 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 5 1 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log -./profile_gemm_b_scale.sh gemm_ab_scale 6 1 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log ./profile_gemm_b_scale.sh gemm_ab_scale 7 1 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log #run gemm_add_add_fastgelu tests @@ -353,23 +334,14 @@ print_log_header $gemm_add_silu_log $env_type $branch $host_name #run gemm_add tests export gemm_add_log="perf_gemm_add.log" print_log_header $gemm_add_log $env_type $branch $host_name -./profile_gemm_add.sh gemm_add 0 0 $verify 1 0 1 2>&1 | tee -a $gemm_add_log -./profile_gemm_add.sh gemm_add 1 0 $verify 1 0 1 2>&1 | tee -a $gemm_add_log -./profile_gemm_add.sh gemm_add 0 1 $verify 1 0 1 2>&1 | tee -a $gemm_add_log -./profile_gemm_add.sh gemm_add 1 1 $verify 1 0 1 2>&1 | tee -a $gemm_add_log +./profile_gemm_d0_e.sh gemm_add 0 0 $verify 1 0 1 2>&1 | tee -a $gemm_add_log +./profile_gemm_d0_e.sh gemm_add 1 0 $verify 1 0 1 2>&1 | tee -a $gemm_add_log +./profile_gemm_d0_e.sh gemm_add 0 1 $verify 1 0 1 2>&1 | tee -a $gemm_add_log +./profile_gemm_d0_e.sh gemm_add 1 1 $verify 1 0 1 2>&1 | tee -a $gemm_add_log #run gemm_b_scale tests export gemm_b_scale_log="perf_gemm_b_scale.log" print_log_header $gemm_b_scale_log $env_type $branch $host_name -./profile_gemm_b_scale.sh gemm_b_scale 0 0 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log -./profile_gemm_b_scale.sh gemm_b_scale 1 0 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log -./profile_gemm_b_scale.sh gemm_b_scale 2 0 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log -./profile_gemm_b_scale.sh gemm_b_scale 3 0 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log -./profile_gemm_b_scale.sh gemm_b_scale 4 0 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log -./profile_gemm_b_scale.sh gemm_b_scale 5 0 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log -./profile_gemm_b_scale.sh gemm_b_scale 6 0 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log -./profile_gemm_b_scale.sh gemm_b_scale 7 0 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log -./profile_gemm_b_scale.sh gemm_b_scale 8 0 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log ./profile_gemm_b_scale.sh gemm_b_scale 0 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log ./profile_gemm_b_scale.sh gemm_b_scale 1 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log ./profile_gemm_b_scale.sh gemm_b_scale 2 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log @@ -611,8 +583,8 @@ print_log_header $grouped_gemm_multiply_tile_loop_log $env_type $branch $host_na #run grouped_gemm_tile_loop tests export grouped_gemm_tile_loop_log="perf_grouped_gemm_tile_loop.log" print_log_header $grouped_gemm_tile_loop_log $env_type $branch $host_name -./profile_grouped_gemm.sh grouped_gemm_tile_loo 0 0 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_tile_loop_log -./profile_grouped_gemm.sh grouped_gemm_tile_loo 0 1 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_tile_loop_log +./profile_grouped_gemm.sh grouped_gemm_tile_loop 0 0 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_tile_loop_log +./profile_grouped_gemm.sh grouped_gemm_tile_loop 0 1 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_tile_loop_log #run groupnorm tests export groupnorm_log="perf_groupnorm.log" @@ -659,10 +631,10 @@ print_log_header $bnorm_bwd_log $env_type $branch $host_name #run bnorm_fwd tests export bnorm_fwd_log="perf_bnorm_fwd.log" print_log_header $bnorm_fwd_log $env_type $branch $host_name -./profile_bnorm.sh bnorm_fwd 0 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log -./profile_bnorm.sh bnorm_fwd 1 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log -./profile_bnorm.sh bnorm_fwd 5 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log -./profile_bnorm.sh bnorm_fwd 6 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log +./profile_bnorm_fwd.sh bnorm_fwd 0 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log +./profile_bnorm_fwd.sh bnorm_fwd 1 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log +./profile_bnorm_fwd.sh bnorm_fwd 5 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log +./profile_bnorm_fwd.sh bnorm_fwd 6 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log #run bnorm_infer tests export bnorm_infer_log="perf_bnorm_infer.log" @@ -705,19 +677,19 @@ print_log_header $layernorm_fwd_log $env_type $branch $host_name #run max_pool2d_bwd tests export max_pool2d_bwd_log="perf_max_pool2d_bwd.log" print_log_header $max_pool2d_bwd_log $env_type $branch $host_name -./profile_max_pool2d.sh max_pool2d_bwd 0 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log -./profile_max_pool2d.sh max_pool2d_bwd 1 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log -./profile_max_pool2d.sh max_pool2d_bwd 3 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log -./profile_max_pool2d.sh max_pool2d_bwd 5 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log +./profile_max_pool2d_bwd.sh max_pool2d_bwd 0 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log +./profile_max_pool2d_bwd.sh max_pool2d_bwd 1 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log +./profile_max_pool2d_bwd.sh max_pool2d_bwd 3 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log +./profile_max_pool2d_bwd.sh max_pool2d_bwd 5 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log #run max_pool2d_fwd tests export max_pool2d_fwd_log="perf_max_pool2d_fwd.log" print_log_header $max_pool2d_fwd_log $env_type $branch $host_name -./profile_max_pool2d.sh max_pool2d_fwd 0 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log -./profile_max_pool2d.sh max_pool2d_fwd 1 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log -./profile_max_pool2d.sh max_pool2d_fwd 2 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log -./profile_max_pool2d.sh max_pool2d_fwd 3 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log -./profile_max_pool2d.sh max_pool2d_fwd 4 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 0 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 1 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 2 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 3 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 4 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log #run max_pool3d_bwd tests export max_pool3d_bwd_log="perf_max_pool3d_bwd.log" diff --git a/script/run_full_performance_tests_missing.sh b/script/run_full_performance_tests_missing.sh index 475d336190..3794a2e6cc 100755 --- a/script/run_full_performance_tests_missing.sh +++ b/script/run_full_performance_tests_missing.sh @@ -165,8 +165,8 @@ print_log_header $grouped_gemm_multiply_tile_loop_log $env_type $branch $host_na #run grouped_gemm_tile_loop tests export grouped_gemm_tile_loop_log="perf_grouped_gemm_tile_loop.log" print_log_header $grouped_gemm_tile_loop_log $env_type $branch $host_name -./profile_grouped_gemm.sh grouped_gemm_tile_loo 0 0 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_tile_loop_log -./profile_grouped_gemm.sh grouped_gemm_tile_loo 0 1 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_tile_loop_log +./profile_grouped_gemm.sh grouped_gemm_tile_loop 0 0 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_tile_loop_log +./profile_grouped_gemm.sh grouped_gemm_tile_loop 0 1 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_tile_loop_log #run groupnorm tests export groupnorm_log="perf_groupnorm.log" @@ -213,10 +213,10 @@ print_log_header $bnorm_bwd_log $env_type $branch $host_name #run bnorm_fwd tests export bnorm_fwd_log="perf_bnorm_fwd.log" print_log_header $bnorm_fwd_log $env_type $branch $host_name -./profile_bnorm.sh bnorm_fwd 0 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log -./profile_bnorm.sh bnorm_fwd 1 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log -./profile_bnorm.sh bnorm_fwd 5 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log -./profile_bnorm.sh bnorm_fwd 6 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log +./profile_bnorm_fwd.sh bnorm_fwd 0 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log +./profile_bnorm_fwd.sh bnorm_fwd 1 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log +./profile_bnorm_fwd.sh bnorm_fwd 5 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log +./profile_bnorm_fwd.sh bnorm_fwd 6 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log #run bnorm_infer tests export bnorm_infer_log="perf_bnorm_infer.log" @@ -259,19 +259,19 @@ print_log_header $layernorm_fwd_log $env_type $branch $host_name #run max_pool2d_bwd tests export max_pool2d_bwd_log="perf_max_pool2d_bwd.log" print_log_header $max_pool2d_bwd_log $env_type $branch $host_name -./profile_max_pool2d.sh max_pool2d_bwd 0 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log -./profile_max_pool2d.sh max_pool2d_bwd 1 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log -./profile_max_pool2d.sh max_pool2d_bwd 3 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log -./profile_max_pool2d.sh max_pool2d_bwd 5 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log +./profile_max_pool2d_bwd.sh max_pool2d_bwd 0 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log +./profile_max_pool2d_bwd.sh max_pool2d_bwd 1 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log +./profile_max_pool2d_bwd.sh max_pool2d_bwd 3 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log +./profile_max_pool2d_bwd.sh max_pool2d_bwd 5 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log #run max_pool2d_fwd tests export max_pool2d_fwd_log="perf_max_pool2d_fwd.log" print_log_header $max_pool2d_fwd_log $env_type $branch $host_name -./profile_max_pool2d.sh max_pool2d_fwd 0 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log -./profile_max_pool2d.sh max_pool2d_fwd 1 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log -./profile_max_pool2d.sh max_pool2d_fwd 2 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log -./profile_max_pool2d.sh max_pool2d_fwd 3 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log -./profile_max_pool2d.sh max_pool2d_fwd 4 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 0 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 1 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 2 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 3 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 4 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log #run max_pool3d_bwd tests export max_pool3d_bwd_log="perf_max_pool3d_bwd.log" diff --git a/script/run_full_performance_tests_missing2.sh b/script/run_full_performance_tests_missing2.sh new file mode 100755 index 0000000000..27bb8c4362 --- /dev/null +++ b/script/run_full_performance_tests_missing2.sh @@ -0,0 +1,243 @@ +#!/bin/bash +# +# in order to run this script you'd first need to build the ckProfiler executable in ../build/bin/ +# you would also need to set up some environment variables in order to +# post your new test results to the database and compare them to the baseline +# please contact Illia.Silin@amd.com for more details +# +# run the script as "./run_full_performance_tests.sh +# input arguments: +# verification = 0 : do not verify result correctness on CPU +# = 1 : verifuy correctness on CPU (may take a long time) +# environment tag : a string describing the specifics of your test environment +# branch name : name of the branch in git repo (git status | grep -e 'On branch') +# node name : $hostname + +#get the command line arguments: +export verify=$1 +echo 'Verification: ' $verify +export env_type=$2 +echo 'Environment type: ' $env_type +export branch=$3 +echo 'Branch name: ' $branch +export host_name=$4 +echo 'Host name: ' $host_name +function print_log_header(){ + rm -f $1; + echo 'On branch ' $3 &> $1; + echo 'Node name: ' $4 >> $1; + #get GPU_arch and number of compute units from rocminfo + echo -n "GPU_arch: " >> $1; rocminfo | grep "Name:" | grep "gfx" >> $1; + rocminfo | grep "Compute Unit:" >> $1; + hipcc --version | grep -e 'HIP version' >> $1; + echo 'Environment type: ' $2 >> $1; + /opt/rocm/bin/amdclang++ --version | grep -e 'InstalledDir' >> $1; +} + +#run grouped_bwd_data tests +export grouped_conv_bwd_data_log="perf_grouped_conv_bwd_data.log" +print_log_header $grouped_conv_bwd_data_log $env_type $branch $host_name +./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 0 1 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log +./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 1 1 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log +./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 2 1 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log +./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 0 0 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log +./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 1 0 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log +./profile_grouped_conv_bwd_data.sh grouped_conv_bwd_data 2 0 $verify 1 0 1 256 2>&1 | tee -a $grouped_conv_bwd_data_log + +#run batched_gemm_b_scale tests +export batched_gemm_b_scale_log="perf_batched_gemm_b_scale.log" +print_log_header $batched_gemm_b_scale_log $env_type $branch $host_name +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 0 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 1 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 2 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 3 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 4 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 5 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 6 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 7 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 8 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 8 1 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 8 2 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log +./profile_batched_gemm_b_scale.sh batched_gemm_b_scale 8 3 $verify 1 0 1 2>&1 | tee -a $batched_gemm_b_scale_log + +#run batched_gemm_reduce tests +export batched_gemm_reduce_log="perf_batched_gemm_reduce.log" +print_log_header $batched_gemm_reduce_log $env_type $branch $host_name +./profile_batched_gemm_reduce.sh batched_gemm_reduce 1 0 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log +./profile_batched_gemm_reduce.sh batched_gemm_reduce 1 1 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log +./profile_batched_gemm_reduce.sh batched_gemm_reduce 1 2 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log +./profile_batched_gemm_reduce.sh batched_gemm_reduce 1 3 $verify 1 0 1 2>&1 | tee -a $batched_gemm_reduce_log + +#run conv_fwd_bias_relu_add tests +export conv_fwd_bias_relu_add_log="perf_conv_fwd_bias_relu_add.log" +print_log_header $conv_fwd_bias_relu_add_log $env_type $branch $host_name +./profile_conv_fwd_bias_relu_add.sh conv_fwd_bias_relu_add 1 1 $verify 1 0 1 256 2>&1 | tee -a $conv_fwd_bias_relu_add_log + +#run conv_tensor_rearrange tests +export conv_tensor_rearrange_log="perf_conv_tensor_rearrange.log" +print_log_header $conv_tensor_rearrange_log $env_type $branch $host_name +./profile_conv_tensor_rearrange.sh conv_tensor_rearrange 0 0 $verify 1 0 1 0 256 2>&1 | tee -a $conv_tensor_rearrange_log +./profile_conv_tensor_rearrange.sh conv_tensor_rearrange 1 0 $verify 1 0 1 0 256 2>&1 | tee -a $conv_tensor_rearrange_log +./profile_conv_tensor_rearrange.sh conv_tensor_rearrange 2 0 $verify 1 0 1 0 256 2>&1 | tee -a $conv_tensor_rearrange_log +./profile_conv_tensor_rearrange.sh conv_tensor_rearrange 3 0 $verify 1 0 1 0 256 2>&1 | tee -a $conv_tensor_rearrange_log + +./profile_conv_tensor_rearrange.sh conv_tensor_rearrange 1 1 $verify 1 0 1 0 256 2>&1 | tee -a $conv_tensor_rearrange_log +./profile_conv_tensor_rearrange.sh conv_tensor_rearrange 3 1 $verify 1 0 1 0 256 2>&1 | tee -a $conv_tensor_rearrange_log + +./profile_conv_tensor_rearrange.sh conv_tensor_rearrange 1 0 $verify 1 0 1 1 256 2>&1 | tee -a $conv_tensor_rearrange_log +./profile_conv_tensor_rearrange.sh conv_tensor_rearrange 1 1 $verify 1 0 1 1 256 2>&1 | tee -a $conv_tensor_rearrange_log + +#run gemm_ab_scale tests +export gemm_ab_scale_log="perf_gemm_ab_scale.log" +print_log_header $gemm_ab_scale_log $env_type $branch $host_name +./profile_gemm_b_scale.sh gemm_ab_scale 7 1 $verify 1 0 1 2>&1 | tee -a $gemm_ab_scale_log + +#run gemm_add tests +export gemm_add_log="perf_gemm_add.log" +print_log_header $gemm_add_log $env_type $branch $host_name +./profile_gemm_d0_e.sh gemm_add 0 0 $verify 1 0 1 2>&1 | tee -a $gemm_add_log +./profile_gemm_d0_e.sh gemm_add 1 0 $verify 1 0 1 2>&1 | tee -a $gemm_add_log +./profile_gemm_d0_e.sh gemm_add 0 1 $verify 1 0 1 2>&1 | tee -a $gemm_add_log +./profile_gemm_d0_e.sh gemm_add 1 1 $verify 1 0 1 2>&1 | tee -a $gemm_add_log + +#run gemm_b_scale tests +export gemm_b_scale_log="perf_gemm_b_scale.log" +print_log_header $gemm_b_scale_log $env_type $branch $host_name +./profile_gemm_b_scale.sh gemm_b_scale 0 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log +./profile_gemm_b_scale.sh gemm_b_scale 1 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log +./profile_gemm_b_scale.sh gemm_b_scale 2 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log +./profile_gemm_b_scale.sh gemm_b_scale 3 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log +./profile_gemm_b_scale.sh gemm_b_scale 4 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log +./profile_gemm_b_scale.sh gemm_b_scale 5 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log +./profile_gemm_b_scale.sh gemm_b_scale 6 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log +./profile_gemm_b_scale.sh gemm_b_scale 7 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log +./profile_gemm_b_scale.sh gemm_b_scale 8 1 $verify 1 0 1 1 2>&1 | tee -a $gemm_b_scale_log + +#run gemm_multiply_multiply_weight_preshuffle tests +export gemm_multiply_multiply_weight_preshuffle_log="perf_gemm_multiply_multiply_weight_preshuffle.log" +print_log_header $gemm_multiply_multiply_weight_preshuffle_log $env_type $branch $host_name +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply_weight_preshuffle 0 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_weight_preshuffle_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply_weight_preshuffle 1 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_weight_preshuffle_log + +#run gemm_multiply_multiply tests +export gemm_multiply_multiply_log="perf_gemm_multiply_multiply.log" +print_log_header $gemm_multiply_multiply_log $env_type $branch $host_name +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 0 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 1 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 2 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 3 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 4 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 5 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 6 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 7 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 8 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 9 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 10 0 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 0 1 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 1 1 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 2 1 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 3 1 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 4 1 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 5 1 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 6 1 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 7 1 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 8 1 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 9 1 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log +./profile_gemm_d0_d1_e.sh gemm_multiply_multiply 10 1 $verify 1 0 1 2>&1 | tee -a $gemm_multiply_multiply_log + +#run grouped_gemm_fixed_nk tests +export grouped_gemm_fixed_nk_log="perf_grouped_gemm_fixed_nk.log" +print_log_header $grouped_gemm_fixed_nk_log $env_type $branch $host_name +./profile_grouped_gemm_fixed_nk.sh grouped_gemm_fixed_nk 0 0 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_fixed_nk_log +./profile_grouped_gemm_fixed_nk.sh grouped_gemm_fixed_nk 1 0 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_fixed_nk_log +./profile_grouped_gemm_fixed_nk.sh grouped_gemm_fixed_nk 2 0 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_fixed_nk_log +./profile_grouped_gemm_fixed_nk.sh grouped_gemm_fixed_nk 3 0 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_fixed_nk_log +./profile_grouped_gemm_fixed_nk.sh grouped_gemm_fixed_nk 0 1 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_fixed_nk_log +./profile_grouped_gemm_fixed_nk.sh grouped_gemm_fixed_nk 1 1 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_fixed_nk_log +./profile_grouped_gemm_fixed_nk.sh grouped_gemm_fixed_nk 2 1 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_fixed_nk_log +./profile_grouped_gemm_fixed_nk.sh grouped_gemm_fixed_nk 3 1 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_fixed_nk_log + +#run grouped_gemm_tile_loop tests +export grouped_gemm_tile_loop_log="perf_grouped_gemm_tile_loop.log" +print_log_header $grouped_gemm_tile_loop_log $env_type $branch $host_name +./profile_grouped_gemm.sh grouped_gemm_tile_loop 0 0 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_tile_loop_log +./profile_grouped_gemm.sh grouped_gemm_tile_loop 0 1 $verify 1 0 1 2>&1 | tee -a $grouped_gemm_tile_loop_log + +#run avg_pool2d_bwd tests +export avg_pool2d_bwd_log="perf_avg_pool2d_bwd.log" +print_log_header $avg_pool2d_bwd_log $env_type $branch $host_name +./profile_avg_pool2d_bwd.sh avg_pool2d_bwd 0 $verify 1 0 1 2>&1 | tee -a $avg_pool2d_bwd_log +./profile_avg_pool2d_bwd.sh avg_pool2d_bwd 1 $verify 1 0 1 2>&1 | tee -a $avg_pool2d_bwd_log +./profile_avg_pool2d_bwd.sh avg_pool2d_bwd 3 $verify 1 0 1 2>&1 | tee -a $avg_pool2d_bwd_log +./profile_avg_pool2d_bwd.sh avg_pool2d_bwd 5 $verify 1 0 1 2>&1 | tee -a $avg_pool2d_bwd_log +./profile_avg_pool2d_bwd.sh avg_pool2d_bwd 7 $verify 1 0 1 2>&1 | tee -a $avg_pool2d_bwd_log + +#run avg_pool3d_bwd tests +export avg_pool3d_bwd_log="perf_avg_pool3d_bwd.log" +print_log_header $avg_pool3d_bwd_log $env_type $branch $host_name +./profile_avg_pool3d_bwd.sh avg_pool3d_bwd 0 $verify 1 0 1 2>&1 | tee -a $avg_pool3d_bwd_log +./profile_avg_pool3d_bwd.sh avg_pool3d_bwd 1 $verify 1 0 1 2>&1 | tee -a $avg_pool3d_bwd_log +./profile_avg_pool3d_bwd.sh avg_pool3d_bwd 5 $verify 1 0 1 2>&1 | tee -a $avg_pool3d_bwd_log + +#run bnorm_bwd tests +export bnorm_bwd_log="perf_bnorm_bwd.log" +print_log_header $bnorm_bwd_log $env_type $branch $host_name +./profile_bnorm.sh bnorm_bwd 0 $verify 0 1 0 2>&1 | tee -a $bnorm_bwd_log +./profile_bnorm.sh bnorm_bwd 1 $verify 0 1 0 2>&1 | tee -a $bnorm_bwd_log +./profile_bnorm.sh bnorm_bwd 5 $verify 0 1 0 2>&1 | tee -a $bnorm_bwd_log +./profile_bnorm.sh bnorm_bwd 6 $verify 0 1 0 2>&1 | tee -a $bnorm_bwd_log + +#run bnorm_fwd tests +export bnorm_fwd_log="perf_bnorm_fwd.log" +print_log_header $bnorm_fwd_log $env_type $branch $host_name +./profile_bnorm_fwd.sh bnorm_fwd 0 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log +./profile_bnorm_fwd.sh bnorm_fwd 1 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log +./profile_bnorm_fwd.sh bnorm_fwd 5 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log +./profile_bnorm_fwd.sh bnorm_fwd 6 $verify 0 1 0 2>&1 | tee -a $bnorm_fwd_log + +#run bnorm_infer tests +export bnorm_infer_log="perf_bnorm_infer.log" +print_log_header $bnorm_infer_log $env_type $branch $host_name +./profile_bnorm.sh bnorm_infer 0 $verify 0 1 0 2>&1 | tee -a $bnorm_infer_log +./profile_bnorm.sh bnorm_infer 1 $verify 0 1 0 2>&1 | tee -a $bnorm_infer_log +./profile_bnorm.sh bnorm_infer 5 $verify 0 1 0 2>&1 | tee -a $bnorm_infer_log +./profile_bnorm.sh bnorm_infer 6 $verify 0 1 0 2>&1 | tee -a $bnorm_infer_log + +#run max_pool2d_bwd tests +export max_pool2d_bwd_log="perf_max_pool2d_bwd.log" +print_log_header $max_pool2d_bwd_log $env_type $branch $host_name +./profile_max_pool2d_bwd.sh max_pool2d_bwd 0 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log +./profile_max_pool2d_bwd.sh max_pool2d_bwd 1 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log +./profile_max_pool2d_bwd.sh max_pool2d_bwd 3 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log +./profile_max_pool2d_bwd.sh max_pool2d_bwd 5 $verify 1 0 1 2>&1 | tee -a $max_pool2d_bwd_log + +#run max_pool2d_fwd tests +export max_pool2d_fwd_log="perf_max_pool2d_fwd.log" +print_log_header $max_pool2d_fwd_log $env_type $branch $host_name +./profile_max_pool2d_fwd.sh max_pool2d_fwd 0 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 1 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 2 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 3 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log +./profile_max_pool2d_fwd.sh max_pool2d_fwd 4 $verify 1 0 1 2>&1 | tee -a $max_pool2d_fwd_log + +#run max_pool3d_bwd tests +export max_pool3d_bwd_log="perf_max_pool3d_bwd.log" +print_log_header $max_pool3d_bwd_log $env_type $branch $host_name +./profile_max_pool3d.sh max_pool3d_bwd 0 $verify 1 0 1 2>&1 | tee -a $max_pool3d_bwd_log +./profile_max_pool3d.sh max_pool3d_bwd 1 $verify 1 0 1 2>&1 | tee -a $max_pool3d_bwd_log +./profile_max_pool3d.sh max_pool3d_bwd 5 $verify 1 0 1 2>&1 | tee -a $max_pool3d_bwd_log + +#run pool3d_fwd tests +export pool3d_fwd_log="perf_pool3d_fwd.log" +print_log_header $pool3d_fwd_log $env_type $branch $host_name +./profile_pool3d_fwd.sh pool3d_fwd 0 $verify 1 0 1 0 2>&1 | tee -a $pool3d_fwd_log +./profile_pool3d_fwd.sh pool3d_fwd 1 $verify 1 0 1 0 2>&1 | tee -a $pool3d_fwd_log +./profile_pool3d_fwd.sh pool3d_fwd 3 $verify 1 0 1 0 2>&1 | tee -a $pool3d_fwd_log +./profile_pool3d_fwd.sh pool3d_fwd 5 $verify 1 0 1 0 2>&1 | tee -a $pool3d_fwd_log +./profile_pool3d_fwd.sh pool3d_fwd 7 $verify 1 0 1 0 2>&1 | tee -a $pool3d_fwd_log +./profile_pool3d_fwd.sh pool3d_fwd 0 $verify 1 0 1 1 2>&1 | tee -a $pool3d_fwd_log +./profile_pool3d_fwd.sh pool3d_fwd 1 $verify 1 0 1 1 2>&1 | tee -a $pool3d_fwd_log +./profile_pool3d_fwd.sh pool3d_fwd 3 $verify 1 0 1 1 2>&1 | tee -a $pool3d_fwd_log +./profile_pool3d_fwd.sh pool3d_fwd 5 $verify 1 0 1 1 2>&1 | tee -a $pool3d_fwd_log +./profile_pool3d_fwd.sh pool3d_fwd 7 $verify 1 0 1 1 2>&1 | tee -a $pool3d_fwd_log +