Better TG performance for GQA models (CPU) (#332)

* Slightly better CPU TG performance for GQA

* Better CPU FA implementation for TG when GQA

* Minor

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>

3 files changed, 134 insertions, 12 deletions

diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c
index 25694fc7..83a48cb6 100644
--- a/ggml/src/ggml.c
+++ b/ggml/src/ggml.c
@@ -21781,19 +21781,27 @@ struct ggml_cplan ggml_graph_plan(const struct ggml_cgraph * cgraph, int n_threa
                     const struct ggml_tensor * q = node->src[0];
                     const struct ggml_tensor * k = node->src[1];
                     if (q->ne[1] == 1 && q->ne[3] == 1 && q->ne[2]/k->ne[2] > 1 && n_tasks > 1 && k->ne[1]/32 > 1) {
-                        int nstep_k = k->ne[1]/32;
-                        int gcd_k   = simple_gcd(nstep_k, n_tasks);
-                        if (gcd_k > 1) {
-                            int nth_k = n_tasks/gcd_k;
-                            int rk2 = q->ne[2]/k->ne[2];
-                            int nq_per_thread = (rk2 + nth_k - 1)/nth_k;
-                            size_t size = (Dv + 16)*nq_per_thread*sizeof(float)*n_tasks;
-                            if (ggml_is_quantized(k->type)) {
-                                enum ggml_type vec_dot_type = type_traits[k->type].vec_dot_type;
-                                size_t row_size = ggml_row_size(vec_dot_type, q->ne[0]);
-                                size += q->ne[2]*row_size;
-                            }
+                        if (k->ne[2] > 1) {
+                            int nk = 32 * (k->ne[2]*k->ne[1]/(32*n_tasks));
+                            int nstep_k = k->ne[2]*k->ne[1]/nk;
+                            size_t result_size = (Dv + 16)*q->ne[2]/k->ne[2]*sizeof(float);
+                            size_t size = nstep_k*result_size;
                             cur = MAX(cur, size);
+                        } else {
+                            int nstep_k = k->ne[1]/32;
+                            int gcd_k   = simple_gcd(nstep_k, n_tasks);
+                            if (gcd_k > 1) {
+                                int nth_k = n_tasks/gcd_k;
+                                int rk2 = q->ne[2]/k->ne[2];
+                                int nq_per_thread = (rk2 + nth_k - 1)/nth_k;
+                                size_t size = (Dv + 16)*nq_per_thread*sizeof(float)*n_tasks;
+                                if (ggml_is_quantized(k->type)) {
+                                    enum ggml_type vec_dot_type = type_traits[k->type].vec_dot_type;
+                                    size_t row_size = ggml_row_size(vec_dot_type, q->ne[0]);
+                                    size += q->ne[2]*row_size;
+                                }
+                                cur = MAX(cur, size);
+                            }
                         }
                     }
 #endif
diff --git a/ggml/src/iqk/iqk_flash_attn.cpp b/ggml/src/iqk/iqk_flash_attn.cpp
index e302c944..3f1d6dc2 100644
--- a/ggml/src/iqk/iqk_flash_attn.cpp
+++ b/ggml/src/iqk/iqk_flash_attn.cpp
@@ -153,6 +153,67 @@ bool iqk_flash_attn_noalibi(int type_q, int type_mask, float max_bias,
         }
     }
 
+    if (neq3 == 1 && rk2 > 1 && rk2 == rv2 && neq1 == 1 && nth >= 1 && nek2*nek1 >= 32*nth) {
+        int nk = 32 * (nek2*nek1/(32*nth));
+        int nkk = (nek1 + nk - 1)/nk;
+        int nstep_k = nek2*nkk;
+        auto result_size = (Dv + 16)*rk2*sizeof(float);
+        //if (ith == 0) printf("rk2 = %d, nek1 = %d, nek2 = %d, nk = %d, nkk = %d, nstep_k = %d\n", (int)rk2, (int)nek1, (int)nek2, nk, nkk, nstep_k);
+        for (int istep_k = ith; istep_k < nstep_k; istep_k += nth) {
+            int ik02 = istep_k/nkk;
+            int ik01 = nk*(istep_k - ik02*nkk);
+            int this_nk = ik01 + nk <= nek1 ? nk : nek1 - ik01;
+            if (this_nk <= 0) break;
+            auto this_result = (float *)((char *)work_buffer + istep_k*result_size);
+            auto this_q = (const float *)((const char *)q + ik02*rk2*nbq2);
+            auto this_k = (const char *)k + ik01*stride_k + ik02*nbk2;
+            auto this_v = (const char *)v + ik01*stride_v + ik02*nbv2;
+            auto this_m = (const char *)mask + ik01*sizeof(uint16_t); // we don't have ggml_half available here
+            if (!iqk_flash_attn_impl(int_type_k, int_type_v,
+                     Dk, Dv, rk2, this_nk, nbq2, stride_k, stride_v, 0, Dv,
+                     this_q, (const void *)this_k, (const void *)this_v, (const void *)this_m,
+                     scale, softcap, this_result, this_result + (Dv+0)*rk2, this_result + (Dv+1)*rk2)) return false;
+        }
+
+        barrier(barrier_data);
+
+        // We have nkk results for each head
+        for (int iq2 = ith; iq2 < neq2; iq2 += nth) {
+            // ik02*rk2 + il = iq2 (il = 0...rk2-1) => ik02 = iq2/rk2, il = iq2%rk2;
+            int ik02 = iq2/rk2;
+            int il = iq2 - ik02*rk2;
+            auto Racc = qkv + iq2*nb1/sizeof(float);
+            std::memset(Racc, 0, Dv*sizeof(float));
+            float M = -INFINITY, S = 0;
+            for (int ikk = 0; ikk < nkk; ++ikk) {
+                int istep_k = ik02*nkk + ikk;
+                auto this_result = (float *)((char *)work_buffer + istep_k*result_size);
+                const float * R  = this_result + il*Dv;
+                const float * Mj = this_result + Dv*rk2;
+                const float * Sj = Mj + rk2;
+                if (Mj[il] == -INFINITY) continue;
+                if (Mj[il] > M) {
+                    if (M == -INFINITY) {
+                        std::memcpy(Racc, R, Dv*sizeof(float));
+                        S = Sj[il];
+                    } else {
+                        float c = exp(M - Mj[il]);
+                        S = c*S + Sj[il];
+                        for (int i = 0; i < Dv; ++i) Racc[i] = c*Racc[i] + R[i];
+                    }
+                    M = Mj[il];
+                } else {
+                    float c = exp(Mj[il] - M);
+                    S += c*Sj[il];
+                    for (int i = 0; i < Dv; ++i) Racc[i] += c*R[i];
+                }
+            }
+            float norm = S > 0 ? 1/S : 1;
+            for (int i = 0; i < Dv; ++i) Racc[i] *= norm;
+        }
+        return true;
+    }
+
     // I keep changing my mind what is the best strategy to split the threads when processing
     // multiple heads. This is my current thinking, the commented out code below was the previous.
     int ntg = nth/simple_gcd(neq2*neq3, nth);
diff --git a/ggml/src/iqk/iqk_mul_mat.cpp b/ggml/src/iqk/iqk_mul_mat.cpp
index 78270f5e..424a65af 100644
--- a/ggml/src/iqk/iqk_mul_mat.cpp
+++ b/ggml/src/iqk/iqk_mul_mat.cpp
@@ -451,6 +451,51 @@ bool iqk_mul_mat_4d(long Nx, long Ny, long ne00,
     auto r3 = ne13 / ne03;
 
     if (ne13 == 1 && Ny == 1 && r2 > 1) {
+        if (Nx >= 256 && Nx%32 == 0) {
+            int nx32 = Nx/32;
+            int nchunk = nx32*ne02;
+            if (r2 <= 8) {
+                MulMat mm;
+                if (!MulMat::prepare(typeA, typeB, ne00, mm, r2)) return false;
+                int nx64 = Nx/64;
+                int nchunk64 = nx64*ne02;
+                for (int ichunk = ith; ichunk < nchunk64; ichunk += nth) {
+                    int i02 = ichunk/nx64;
+                    int ix = 64*(ichunk - i02*nx64);
+                    DataInfo info{C + ix + r2*i02*nb2, (const char *)B + r2*i02*nb12, (size_t)nb2, (size_t)nb12, 0, 1, nullptr, 0};
+                    mm.funcs[r2-1](ne00, (const void *)((const char *)A + ix*strideA + i02*nb02), strideA, info, 64);
+                }
+                int ix0 = 64*nx64;
+                if (ix0 < Nx) {
+                    nx32 -= 2*nx64;
+                    nchunk = nx32*ne02;
+                    for (int ichunk = ith; ichunk < nchunk; ichunk += nth) {
+                        int i02 = ichunk/nx32;
+                        int ix = ix0 + 32*(ichunk - i02*nx32);
+                        DataInfo info{C + ix + r2*i02*nb2, (const char *)B + r2*i02*nb12, (size_t)nb2, (size_t)nb12, 0, 1, nullptr, 0};
+                        mm.funcs[r2-1](ne00, (const void *)((const char *)A + ix*strideA + i02*nb02), strideA, info, 32);
+                    }
+                }
+                //for (int ichunk = ith; ichunk < nchunk; ichunk += nth) {
+                //    int i02 = ichunk/nx32;
+                //    int ix = 32*(ichunk - i02*nx32);
+                //    DataInfo info{C + ix + r2*i02*nb2, (const char *)B + r2*i02*nb12, (size_t)nb2, (size_t)nb12, 0, 1, nullptr, 0};
+                //    mm.funcs[r2-1](ne00, (const void *)((const char *)A + ix*strideA + i02*nb02), strideA, info, 32);
+                //}
+                return true;
+            }
+            for (int ichunk = ith; ichunk < nchunk; ichunk += nth) {
+                int i02 = ichunk/nx32;
+                int ix = ichunk - i02*nx32;
+                if (!iqk_mul_mat(32, r2, ne00,
+                            typeA, (const char *)A + 32*ix*strideA + i02*nb02, strideA,
+                            typeB, (const char *)B + i02*r2*nb12, nb12,
+                            C + 32*ix + r2*i02*nb2, nb2, 0, 1)) return false;
+
+            }
+            return true;
+        }
+        //if (ith == 0) printf("Using this: Nx = %d, r2 = %d, ne02 = %d\n", (int)Nx, (int)r2,(int)ne02);
         int gcd = simple_gcd(ne02, nth);
         int counter = 0;
         for (int64_t i12 = 0; i12 < ne02; i12++) {
@@ -17153,6 +17198,14 @@ inline void iqk_flash_helper(KHelper& kh, VHelper& vh, int nq1, int nk1, int str
         FlashAttn<Dk, Dv, 8, k_step> fa(scale, softcap);
         fa.compute(kh, vh, nq1, nk1, stride_q, stride_m, stride_qkv, q, (const char *)mask, qkv, M, S);
     }
+    else if (nq1 >= 4) {
+        FlashAttn<Dk, Dv, 4, k_step> fa(scale, softcap);
+        fa.compute(kh, vh, nq1, nk1, stride_q, stride_m, stride_qkv, q, (const char *)mask, qkv, M, S);
+    }
+    else if (nq1 >= 2) {
+        FlashAttn<Dk, Dv, 2, k_step> fa(scale, softcap);
+        fa.compute(kh, vh, nq1, nk1, stride_q, stride_m, stride_qkv, q, (const char *)mask, qkv, M, S);
+    }
     else {
         FlashAttn<Dk, Dv, 1, k_step> fa(scale, softcap);
         fa.compute(kh, vh, nq1, nk1, stride_q, stride_m, stride_qkv, q, (const char *)mask, qkv, M, S);