[ET-VK] Improve packing format for int4 linear operator + misc improvements

Differential Revision: D72412950

Pull Request resolved: #9883

Author: SS-JIA

backends/vulkan/runtime/api/containers/Tensor.cpp

@@ -497,9 +497,7 @@ vTensor::vTensor(
   VK_CHECK_COND(
       dim_order_is_valid(dim_order_), "computed dim order is invalid");
 
-  if (storage_type != utils::kBuffer) {
-    set_logical_limits(storage_.image_extents_);
-  }
+  set_logical_limits(storage_.image_extents_);
 }
 
 // NOLINTNEXTLINE

backends/vulkan/runtime/graph/ops/glsl/pack_int4_linear_weight_transposed_interleaved.glsl

@@ -0,0 +1,136 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+
+#version 450 core
+
+#define PRECISION ${PRECISION}
+
+${define_required_extensions("uint8")}
+${define_required_extensions("int8")}
+
+layout(std430) buffer;
+
+${layout_declare_tensor(B, "w", "t_qmat2", "uint8", STORAGE, is_scalar_array=False)}
+${layout_declare_tensor(B, "r", "nchw_4x2", "uint8", "buffer")}
+
+layout(push_constant) uniform restrict Block {
+  ivec4 qmat2_sizes;
+};
+
+layout(local_size_x_id = 0, local_size_y_id = 1, local_size_z_id = 2) in;
+
+uint8_t get_first(const uint8_t packed) {
+  return uint8_t((packed & 0xF0) >> 4);
+}
+
+uint8_t get_second(const uint8_t packed) {
+  return uint8_t(packed & 0x0F);
+}
+
+uint8_t combine(const uint8_t first, const uint8_t second) {
+  return uint8_t(first << 4 | second);
+}
+
+/*
+ * This shader packs the weight tensor into a texture.
+ *
+ * The original tensor has a (W, H) shape of (K / 2, N) and each scalar element
+ * is a uint8_t, which contains 2 packed 4 bit uint values.
+ *
+ * The transform performed by this shader is to first transpose the tensor, so
+ * the shape of the packed tensor becomes (N / 2, K). Then, the 4 bit integers
+ * are re-packed in groups of 8. For each 4 uint8_t values, the "left" 4-bits
+ * of each value contain the 0, 1, 2, 3 4-bit values, and the "right" 4-bits of
+ * each value contain the 4, 5, 6, 7 4-bit values.
+ *
+ * As a concrete example, consider the following weight tensor. The | demarks
+ * the packing boundary, so 1| 2 represents a single uint8_t value with 1 in the
+ * leftmost 4 bits and 2 in the rightmost 4 bits.
+ *
+ *  1| 2,  3| 4,  5| 6,  7| 8,
+ *  9|10, 11|12, 13|14, 15|16,
+ * 17|18, 19|20, 21|22, 23|24,
+ * 25|26, 27|28, 29|30, 31|32,
+ * 33|34, 35|36, 37|38, 39|40,
+ * 41|42, 43|44, 45|46, 47|48,
+ * 49|50, 51|52, 53|54, 55|56,
+ * 57|58, 59|60, 61|62, 63|64,
+ *
+ * After packing, the packed tensor would contain
+ *
+ *  1|33,  9|41, 17|49, 25|57,
+ *  2|34, 10|42, 18|50, 26|58,
+ *  3|35, 11|43, 19|51, 27|59,
+ *  4|36, 12|44, 20|52, 28|60,
+ *  5|37, 13|45, 21|53, 29|61,
+ *  6|38, 14|46, 22|54, 30|62,
+ *  7|39, 15|47, 23|55, 31|63,
+ *  8|40, 16|48, 24|56, 32|64,
+ *
+ * The purpose of interleaving is to make it easier to extract the unpacked
+ * values in order using the u8vec4 vectorized type. With the packing in place,
+ * The 4-bit values can be extracted via
+ *
+ * u8vec4 packed;
+ * u8vec4 vals_0123 = (packed & 0xF0) >> 4;
+ * u8vec4 vals_4567 = (packed | 0x0F);
+ */
+void main() {
+  // Each thread writes 2 output texels along the height axis
+  ivec2 packed_pos = ivec2(
+      gl_GlobalInvocationID.x,
+      gl_GlobalInvocationID.y << 1);
+
+  // The packed tensor is width packed
+  if ((packed_pos.x << 2) >= qmat2_sizes.x || packed_pos.y >= qmat2_sizes.y) {
+    return;
+  }
+
+  int out_col = packed_pos.x << 3;
+  int out_row = packed_pos.y;
+
+  int in_col = out_row;
+  int in_int8_col = in_col >> 1;
+  int in_row = out_col;
+
+  int in_numrows = qmat2_sizes.x << 1;
+  int in_numcols = qmat2_sizes.y;
+  int in_num_int8_cols = qmat2_sizes.y >> 1;
+
+  uint8_t in_vals[8][2];
+  for (int r = 0; r < 8; ++r) {
+    if (in_row + r < in_numrows) {
+      uint8_t in_val_packed = nchw_4x2[(in_row + r) * in_num_int8_cols + in_int8_col];
+      in_vals[r][0] = get_first(in_val_packed);
+      in_vals[r][1] = get_second(in_val_packed);
+    } else {
+      in_vals[r][0] = uint8_t(254);
+      in_vals[r][1] = uint8_t(254);
+    }
+  }
+
+  u8vec4 out_tex_1 = u8vec4(
+      combine(in_vals[0][0], in_vals[4][0]),
+      combine(in_vals[1][0], in_vals[5][0]),
+      combine(in_vals[2][0], in_vals[6][0]),
+      combine(in_vals[3][0], in_vals[7][0]));
+
+  u8vec4 out_tex_2 = u8vec4(
+      combine(in_vals[0][1], in_vals[4][1]),
+      combine(in_vals[1][1], in_vals[5][1]),
+      combine(in_vals[2][1], in_vals[6][1]),
+      combine(in_vals[3][1], in_vals[7][1]));
+
+  $if STORAGE == "buffer":
+    int stride = qmat2_sizes.x >> 2;
+    t_qmat2[packed_pos.y * stride + packed_pos.x] = out_tex_1;
+    t_qmat2[(packed_pos.y + 1) * stride + packed_pos.x] = out_tex_2;
+  $else:
+    imageStore(t_qmat2, ivec3(packed_pos.xy, 0), out_tex_1);
+    imageStore(t_qmat2, ivec3(packed_pos.x, packed_pos.y + 1, 0), out_tex_2);
+}

backends/vulkan/runtime/graph/ops/glsl/pack_int4_linear_weight_transposed_interleaved.yaml

@@ -0,0 +1,13 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+pack_int4_linear_weight_transposed_interleaved:
+  parameter_names_with_default_values:
+    STORAGE: texture3d
+  shader_variants:
+    - NAME: pack_int4_linear_weight_transposed_interleaved_texture3d
+    - NAME: pack_int4_linear_weight_transposed_interleaved_buffer
+      STORAGE: buffer

backends/vulkan/runtime/graph/ops/glsl/q_4w_linear.glsl

@@ -8,44 +8,42 @@
 
 #version 450 core
 
-#include "indexing_utils.h"
-
 #define PRECISION ${PRECISION}
 
-#define FOUR 4
-
-#define VEC4_T ${texel_load_type(DTYPE, STORAGE)}
-#define FLOAT_T ${buffer_scalar_type(DTYPE)}
+#define T ${buffer_scalar_type(DTYPE)}
+#define VEC4_T ${buffer_gvec_type(DTYPE, 4)}
 
-${define_active_storage_type(STORAGE)}
-
-${define_required_extensions([DTYPE, "uint8", "uint16"])}
-#extension GL_EXT_control_flow_attributes : require
+${define_required_extensions(DTYPE)}
+${define_required_extensions("int8")}
 
 layout(std430) buffer;
 
-${layout_declare_tensor(B, "w", "ret", DTYPE, STORAGE)}
-${layout_declare_tensor(B, "r", "x", DTYPE, STORAGE)}
-${layout_declare_tensor(B, "r", "weights", "uint8", "buffer")}
-${layout_declare_tensor(B, "r", "qparams", DTYPE, STORAGE)}
-${layout_declare_ubo(B, "ivec3", "ret_limits")}
-${layout_declare_ubo(B, "ivec4", "x_sizes")}
-${layout_declare_ubo(B, "ivec4", "weights_strides")}
-${layout_declare_ubo(B, "ivec4", "qparams_strides")}
+${layout_declare_tensor(B, "w", "t_out", DTYPE, OUT_STORAGE, is_scalar_array=False)}
+${layout_declare_tensor(B, "r", "t_mat1", DTYPE, IN_STORAGE, is_scalar_array=False)}
+${layout_declare_tensor(B, "r", "t_qmat2", "uint8", WEIGHT_STORAGE, is_scalar_array=False)}
+${layout_declare_tensor(B, "r", "t_qparams", DTYPE, "texture3D")}
+
+layout(push_constant) uniform restrict Block {
+  ivec4 out_sizes;
+  ivec4 mat1_sizes;
+  ivec4 qmat2_sizes;
+};
 
 layout(local_size_x_id = 0, local_size_y_id = 1, local_size_z_id = 2) in;
 
-layout(constant_id = 3) const int group_size = 1;
+layout(constant_id = 3) const int group_size = 64;
 
 /*
  * This shader computes a linear operator between a floating point input matrix
  * x and a weights matrix that is quantized to 4 bits.
  *
  * The (W, H, C) shape of each tensor is:
  * - x: (K, M)
- * - weights: (K / 2, N)
+ * - weights: (N / 2, K)
  *   - The weights tensor has a data type of `uint8`. Each element in the tensor
  *     contains 2 4-bit values packed into a uint8.
+ *   - See the pack_int4_linear_weight_transposed_interleave shader to see more
+ *     details on how the weight tensor is stored.
  * - qparams: (2, N, number_of_groups)
  *   - This tensor contains the scales and zeros quantization parameters for the
  *     weights tensor. The weight tensor is quantized group-wise, which means
@@ -57,56 +55,68 @@ layout(constant_id = 3) const int group_size = 1;
  * Note that this shader assumes that all tensors are width packed.
  */
 void main() {
-  // output positions being calculated are (n, m), (n + 1, m), ...
-  // This means multiplying the m-th row of x with the n-th, (n+1)-th, ... rows
-  // of the weights tensor.
-  const u16vec3 ret_pos = u16vec3(gl_GlobalInvocationID);
-  if (any(greaterThanEqual(ret_pos, ret_limits))) {
+  const uint out_row = gl_GlobalInvocationID.y;
+  // Each thread writes out 2 texels along the width axis, equivalent to 8
+  // scalar elements. Therefore multiply the thread_idx.x by 8.
+  const uint out_col = gl_GlobalInvocationID.x << 3;
+  // Similar reasoning to the above, each thread works on 2 texels along the
+  // width axis so multiply thread_idx.x by 2.
+  const int out_col_texel_idx = int(gl_GlobalInvocationID.x) << 1;
+
+  if (out_col >= out_sizes.x || out_row >= out_sizes.y) {
     return;
   }
 
-  // Since ret is width packed, need to multiply by 4
-  const uint16_t n = uint16_t(ret_pos.x * 4);
+  const int num_blocks = mat1_sizes.x / group_size;
 
-  // K is guaranteed to be a multiple of group size
-  const uint16_t num_blocks = uint16_t(x_sizes.x / group_size);
+  VEC4_T sums[2];
 
-  uint16_t k_texel_i = uint16_t(0);
-  vec4 sums = vec4(0.0);
-  for (uint16_t block_idx = uint16_t(0); block_idx < num_blocks; block_idx++) {
-    vec4 scales;
-    vec4 zeros;
+  sums[0] = VEC4_T(0);
+  sums[1] = VEC4_T(0);
 
-    [[unroll]] for (int comp = 0; comp < 4; comp++) {
-      const vec4 scale_and_zero = load_texel(
-          qparams, u16vec3(0, n + comp, block_idx));
-      scales[comp] = scale_and_zero.x;
-      zeros[comp] = scale_and_zero.y;
-    }
+  VEC4_T scales[2];
+  VEC4_T zeros[2];
+
+  $if WEIGHT_STORAGE == "buffer":
+    const int qmat2_stride = qmat2_sizes.x >> 2;
+
+  for (int block_idx = 0; block_idx < num_blocks; ++block_idx) {
+    scales[0] = texelFetch(t_qparams, ivec3(out_col_texel_idx, 0, block_idx), 0);
+    zeros[0] = texelFetch(t_qparams, ivec3(out_col_texel_idx, 1, block_idx), 0);
 
-    for (uint16_t i = uint16_t(0); i < group_size; i += uint16_t(4), k_texel_i++) {
-      const VEC4_T x_texel = load_texel(
-          x, u16vec3(k_texel_i, ret_pos.y, ret_pos.z));
-
-      [[unroll]] for (int comp = 0; comp < 4; comp++) {
-        const int weights_bufi = (n + comp) * weights_strides.y + (k_texel_i * 2);
-        // Need to read 4 unpacked values, which corresponds to 2 packed values
-        const uint8_t weights_val_1 = weights[weights_bufi];
-        const uint8_t weights_val_2 = weights[weights_bufi + 1];
-
-        const u8vec4 weights_texel = u8vec4(
-          (weights_val_1 & 0xF0) >> 4,
-          weights_val_1 & 0x0F,
-          (weights_val_2 & 0xF0) >> 4,
-          weights_val_2 & 0x0F);
-
-        // Note that the unpacked 4-bit values are unsigned, therefore they must
-        // first be "centered" around 0 by subtracting 8 before applying the
-        // scale and zero point.
-        sums[comp] += dot(
-            x_texel, (vec4(weights_texel) - 8.0) * scales[comp] + zeros[comp]);
+    scales[1] = texelFetch(t_qparams, ivec3(out_col_texel_idx + 1, 0, block_idx), 0);
+    zeros[1] = texelFetch(t_qparams, ivec3(out_col_texel_idx + 1, 1, block_idx), 0);
+
+    for (int g_idx = 0; g_idx < group_size; g_idx += 4) {
+      const int k = block_idx * group_size + g_idx;
+
+      $if IN_STORAGE == "buffer":
+        const VEC4_T mat1_tex = t_mat1[(out_row * mat1_sizes.x + k) >> 2];
+      $else:
+        const VEC4_T mat1_tex = texelFetch(t_mat1, ivec3(k >> 2, out_row, 0), 0);
+
+      for (int comp = 0; comp < 4; ++comp) {
+        $if WEIGHT_STORAGE == "buffer":
+          const u8vec4 packed_weight_tex = t_qmat2[(k + comp) * qmat2_stride + gl_GlobalInvocationID.x];
+        $else:
+          const uvec4 packed_weight_tex = texelFetch(
+              t_qmat2,
+              ivec3(gl_GlobalInvocationID.x, k + comp, 0),
+              0);
+
+        const uvec4 weight_tex_1 = (packed_weight_tex & 0xF0) >> 4;
+        const uvec4 weight_tex_2 = packed_weight_tex & 0x0F;
+
+        sums[0] += mat1_tex[comp] * ((vec4(weight_tex_1) - 8.0) * scales[0] + zeros[0]);
+        sums[1] += mat1_tex[comp] * ((vec4(weight_tex_2) - 8.0) * scales[1] + zeros[1]);
       }
     }
   }
-  write_texel(ret, ret_pos, sums);
+
+  $if OUT_STORAGE == "buffer":
+    t_out[(out_row * out_sizes.x + out_col) >> 2] = sums[0];
+    t_out[(out_row * out_sizes.x + out_col + 4) >> 2] = sums[1];
+  $else:
+    imageStore(t_out, ivec3(out_col_texel_idx, out_row, 0), sums[0]);
+    imageStore(t_out, ivec3(out_col_texel_idx + 1, out_row, 0), sums[1]);
 }

backends/vulkan/runtime/graph/ops/glsl/q_4w_linear.yaml

@@ -7,10 +7,17 @@
 q_4w_linear:
   parameter_names_with_default_values:
     DTYPE: float
-    STORAGE: texture3d
-  generate_variant_forall:
-    DTYPE:
-      - VALUE: float
-      - VALUE: half
+    OUT_STORAGE: texture3d
+    IN_STORAGE: texture3d
+    WEIGHT_STORAGE: texture3d
   shader_variants:
-    - NAME: q_4w_linear_texture3d
+    - NAME: q_4w_linear_texture3d_texture3d_texture3d_float
+    - NAME: q_4w_linear_texture3d_buffer_texture3d_float
+      IN_STORAGE: buffer
+    - NAME: q_4w_linear_buffer_buffer_texture3d_float
+      OUT_STORAGE: buffer
+      IN_STORAGE: buffer
+    - NAME: q_4w_linear_buffer_buffer_buffer_float
+      OUT_STORAGE: buffer
+      IN_STORAGE: buffer
+      WEIGHT_STORAGE: buffer