iq2_xxs: quantized CUDA dot product (MMVQ)

We get TG-128 = 153.1 t/s

Author: Iwan Kawrakow

ggml-cuda.cu

@@ -3950,6 +3950,35 @@ static __device__ __forceinline__ float vec_dot_q6_K_q8_1_mul_mat(
     return vec_dot_q6_K_q8_1_impl_mmq(&x_ql[index_x], &y_qs[index_y], sc, x_dmf[i * (WARP_SIZE/QI6_K) + i/QI6_K], &y_df[index_y/QI8_1]);
 }
 
+static __device__ __forceinline__ float vec_dot_iq2_xxs_q8_1(
+    const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int & iqs) {
+#if QK_K == 256
+    const block_iq2_xxs * bq2 = (const block_iq2_xxs *) vbq;
+
+    // iqs is 0...15
+    const int ib32 = iqs/2;
+    const int il = iqs%2;
+    const uint16_t * q2 = bq2->qs + 4*ib32;
+    const uint8_t  * aux8 = (const uint8_t *)q2;
+    const uint8_t  * grid1 = (const uint8_t *)(kgrid_iq2xxs + aux8[2*il+0]);
+    const uint8_t  * grid2 = (const uint8_t *)(kgrid_iq2xxs + aux8[2*il+1]);
+    const uint32_t aux32 = q2[2] | (q2[3] << 16);
+    const float d = (float)bq2->d * (0.5f + (aux32 >> 28)) * (float)bq8_1[ib32].ds.x * 0.25f;
+    const uint8_t signs1 = ksigns_iq2xs[(aux32 >> 14*il) & 127];
+    const uint8_t signs2 = ksigns_iq2xs[(aux32 >> (14*il + 7)) & 127];
+    const int8_t * q8 = bq8_1[ib32].qs + 16*il;
+    int sumi1 = 0, sumi2 = 0;
+    for (int j = 0; j < 8; ++j) {
+        sumi1 += q8[j+0] * grid1[j] * (signs1 & kmask_iq2xs[j] ? -1 : 1);
+        sumi2 += q8[j+8] * grid2[j] * (signs2 & kmask_iq2xs[j] ? -1 : 1);
+    }
+    return d * (sumi1 + sumi2);
+#else
+    assert(false);
+    return 0.f;
+#endif
+}
+
 template <int qk, int qr, int qi, bool need_sum, typename block_q_t, int mmq_x, int mmq_y, int nwarps,
               allocate_tiles_cuda_t allocate_tiles, load_tiles_cuda_t load_tiles, int vdr, vec_dot_q_mul_mat_cuda_t vec_dot>
 static __device__ __forceinline__ void mul_mat_q(
@@ -6044,6 +6073,15 @@ static void mul_mat_vec_q6_K_q8_1_cuda(const void * vx, const void * vy, float *
         <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols, nrows);
 }
 
+static void mul_mat_vec_iq2_xxs_q8_1_cuda(const void * vx, const void * vy, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
+    GGML_ASSERT(ncols % QK_K == 0);
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
+    const dim3 block_nums(block_num_y, 1, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
+    mul_mat_vec_q<QK_K, QI2_XXS, block_iq2_xxs, 1, vec_dot_iq2_xxs_q8_1>
+        <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols, nrows);
+}
+
 static void ggml_mul_mat_q4_0_q8_1_cuda(
     const void * vx, const void * vy, float * dst, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, cudaStream_t stream) {
@@ -7608,6 +7646,9 @@ static void ggml_cuda_op_mul_mat_vec_q(
         case GGML_TYPE_Q6_K:
             mul_mat_vec_q6_K_q8_1_cuda(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, stream);
             break;
+        case GGML_TYPE_IQ2_XXS:
+            mul_mat_vec_iq2_xxs_q8_1_cuda(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, stream);
+            break;
         default:
             GGML_ASSERT(false);
             break;