Skip to content
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

Make quantize shader work for all gpus #23676

Draft
wants to merge 1 commit into
base: main
Choose a base branch
from
+35 −33
Draft

Make quantize shader work for all gpus #23676

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
68 changes: 35 additions & 33 deletions onnxruntime/contrib_ops/webgpu/quantization/matmul_nbits.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -536,40 +536,42 @@ Status DP4AMatMulQuantizeProgram::GenerateShaderCode(ShaderHelper& shader) const
shader.AddOutput("scales", ShaderUsage::UseUniform);

shader.AdditionalImplementation() << R"ADDNL_FN(
var<workgroup> max_values : array<input_a_element_t, 4>;
var<workgroup> a_values : array<input_a_value_t, 64>;
)ADDNL_FN";

shader.MainFunctionBody() << R"MAIN_FN(
var local_a = input_a[global_idx];
var max_val = subgroupMax(abs(local_a));
var max_temp = max(max_val.xy, max_val.zw);
var scale = max(max_temp[0], max_temp[1]);
if (local_idx % sg_size == 0) {
max_values[local_idx / sg_size] = scale;
}
a_values[local_idx] = input_a[global_idx];
workgroupBarrier();

if (sg_size == 8)
{
scale = max(max_values[0], max_values[1]);
scale = max(scale, max_values[2]);
scale = max(scale, max_values[3]);
}
else if (sg_size == 16)
{
scale = max(max_values[0], max_values[1]);
}
else
{
scale = max_values[0];
}

var norm_a = local_a/scale;
output[global_idx] = pack4x8snorm(vec4<f32>(norm_a));
if (local_idx == 0)
{
// 127 is the max value of signed int8 [-127,127] used by pack4x8snorm for 1.0f.
scales[workgroup_idx] = scale/127;
if (local_idx < 32) {
var max_val = input_a_value_t(0);
for (var i = 0; i < 32; i++)
{
max_val = max(max_val, abs(a_values[i]));
}
var max_temp = max(max_val.xy, max_val.zw);
var scale = max(max_temp[0], max_temp[1]);
var norm_a = a_values[local_idx]/scale;
output[global_idx] = pack4x8snorm(vec4<f32>(norm_a));
if (local_idx == 0)
{
// 127 is the max value of signed int8 [-127,127] used by pack4x8snorm for 1.0f.
scales[workgroup_idx].x = scale/127;
}
} else {
var max_val = input_a_value_t(0);
for (var i = 32; i < 64; i++)
{
max_val = max(max_val, abs(a_values[i]));
}
var max_temp = max(max_val.xy, max_val.zw);
var scale = max(max_temp[0], max_temp[1]);
var norm_a = a_values[local_idx]/scale;
output[global_idx] = pack4x8snorm(vec4<f32>(norm_a));
if (local_idx == 32)
{
// 127 is the max value of signed int8 [-127,127] used by pack4x8snorm for 1.0f.
scales[workgroup_idx].y = scale/127;
}
}
)MAIN_FN";
return Status::OK();
Expand Down Expand Up @@ -838,15 +840,15 @@ Status MatMulNBits::ComputeInternal(onnxruntime::webgpu::ComputeContext& context

constexpr uint32_t kBlockSizeA = 128;
DP4AMatMulQuantizeProgram quantize_program;
quantize_program.SetWorkgroupSize(32);
quantize_program.SetDispatchGroupSize(M * K / kBlockSizeA, 1, 1);
quantize_program.SetWorkgroupSize(64);
quantize_program.SetDispatchGroupSize(M * K / 256, 1, 1);
TensorShape a_quant_shape{1, M, K / kU32Components};
Tensor a_quant = context.CreateGPUTensor(DataTypeImpl::GetType<uint32_t>(), a_quant_shape);
TensorShapeVector a_scales_dims({1, 1, M, K / kBlockSizeA});
Tensor a_scale = context.CreateGPUTensor(a->DataType(), a_scales_dims);
quantize_program.AddInputs({{a, ProgramTensorMetadataDependency::TypeAndRank, gsl::narrow<int>(kVec4Components)}})
.AddOutputs({{&a_quant, ProgramTensorMetadataDependency::Rank, a_quant.Shape(), gsl::narrow<int>(1)},
{&a_scale, ProgramTensorMetadataDependency::Rank, a_scale.Shape(), gsl::narrow<int>(1)}});
{&a_scale, ProgramTensorMetadataDependency::Rank, gsl::narrow<int>(2)}});
ORT_RETURN_IF_ERROR(context.RunProgram(quantize_program));

constexpr uint32_t kTileSize = 64;
Expand Down
Loading