Fix restrictions for vecload condition (iGemm v4r1 Bwd, 1x1)

[shaojiewang]

To fix issue: #917

test by :

MIOPEN_FIND_MODE=normal MIOPEN_DEBUG_FIND_ONLY_SOLVER=ConvHipImplicitGemmBwdDataV4R1Xdlops MIOPEN_LOG_LEVEL=5 ./bin/test_conv2d --half --input 8 16 14 14 --weights 32 16 1 1 -v --disable-backward-weights --disable-forward --pads_strides_dilations 1 1 1 1 1 1

and get output:

./bin/test_conv2d --half --cmode conv --pmode default --group-count 1 --disable-forward --disable-backward-weights --input 8, 16, 14, 14 --weights 32, 16, 1, 1 --pads_strides_dilations 1 1 1 1 1 1 --trans_output_pads 0 0 --in_layout NCHW --fil_layout NCHW --out_layout NCHW
error: 0
Max diff: 0
Backward convolution: ConvHipImplicitGemmBwdDataV4R1Xdlops
Input tensor: 8, 16, 14, 14
Weights tensor: 32, 16, 1, 1
Output tensor: 8, 32, 16, 16

[asleepzzz]

./bin/MIOpenDriver conv -n 4 -c 64 --in_d 16 -H 56 -W 56 -k 64 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 1 -u 1 -v 1 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1

./bin/MIOpenDriver conv -n 4 -c 64 --in_d 16 -H 56 -W 56 -k 256 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 1 -u 1 -v 1 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1
./bin/MIOpenDriver conv -n 4 -c 256 --in_d 16 -H 56 -W 56 -k 64 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 1 -u 1 -v 1 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1
./bin/MIOpenDriver conv -n 4 -c 256 --in_d 16 -H 56 -W 56 -k 128 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 1 -u 1 -v 1 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1
./bin/MIOpenDriver conv -n 4 -c 256 --in_d 16 -H 56 -W 56 -k 512 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 2 -u 2 -v 2 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1
./bin/MIOpenDriver conv -n 4 -c 128 --in_d 8 -H 28 -W 28 -k 512 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 1 -u 1 -v 1 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1
./bin/MIOpenDriver conv -n 4 -c 512 --in_d 8 -H 28 -W 28 -k 128 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 1 -u 1 -v 1 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1
./bin/MIOpenDriver conv -n 4 -c 512 --in_d 8 -H 28 -W 28 -k 256 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 1 -u 1 -v 1 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1
./bin/MIOpenDriver conv -n 4 -c 512 --in_d 8 -H 28 -W 28 -k 1024 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 2 -u 2 -v 2 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1
./bin/MIOpenDriver conv -n 4 -c 256 --in_d 4 -H 14 -W 14 -k 1024 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 1 -u 1 -v 1 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1
./bin/MIOpenDriver conv -n 4 -c 1024 --in_d 4 -H 14 -W 14 -k 256 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 1 -u 1 -v 1 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1
./bin/MIOpenDriver conv -n 4 -c 1024 --in_d 4 -H 14 -W 14 -k 512 --fil_d 1 -y 1 -x 1 --pad_d 0 -p 0 -q 0 --conv_stride_d 1 -u 1 -v 1 --dilation_d 1 -l 1 -j 1 --spatial_dim 3 -m conv -g 1 -F 2 -t 1 -s 1

[atamazov]

Converted to draft for now. @shaojiewang please click [Ready for review] when PR is ready.

[atamazov]

Let's make this PR ASAP. It's worth fixing #936 ASAP as well. Thus, we will be ready to deliver a hot fix to the ROCm 4.3 Release branch if urgent need arise.

/cc @asroy @junliume @asleepzzz @shurale-nkn

[shaojiewang]

@atamazov @asleepzzz @shurale-nkn You could review code if available.

[asleepzzz]

LGTM

[shurale-nkn]

LGTM