From a15cfa04013bb65b4700e939f3685722da99e46e Mon Sep 17 00:00:00 2001
From: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Date: Mon, 23 Jun 2025 09:59:52 +0200
Subject: [PATCH] iq4_k

46.4 -> 167.2 t/s. iq4_k_r4 is at 115 t/s.
---
 ggml/src/iqk/iqk_gemm_iqk_quants.cpp | 42 +++++++++++++++++++++++++++-
 ggml/src/iqk/iqk_mul_mat.cpp         |  1 +
 2 files changed, 42 insertions(+), 1 deletion(-)

diff --git a/ggml/src/iqk/iqk_gemm_iqk_quants.cpp b/ggml/src/iqk/iqk_gemm_iqk_quants.cpp
index fb2f361e..db0120d5 100644
--- a/ggml/src/iqk/iqk_gemm_iqk_quants.cpp
+++ b/ggml/src/iqk/iqk_gemm_iqk_quants.cpp
@@ -4137,6 +4137,46 @@ void iqk_convert_iq5_ks_q8_k_r8(int n, const void * vx, size_t bx, void * vy, in
     }
 }
 
+void iqk_convert_iq4_k_q8_k_r8(int n, const void * vx, size_t bx, void * vy, int nrc_x) {
+    GGML_ASSERT(n%QK_K == 0);
+    GGML_ASSERT(nrc_x%8 == 0);
+
+    int nb = n/QK_K;
+
+    const block_iq4_k * x8[8];
+
+    block_q8_k_r8 * y = (block_q8_k_r8 *)vy;
+
+    int8x16x2_t values = vld1q_s8_x2(iq4k_values);
+
+    int8x16x2_t xv[8];
+    uint32_t block[8];
+    int8_t   ls[16];
+
+    auto ml = vdupq_n_u8(0xf);
+
+    for (int ix = 0; ix < nrc_x; ix += 8) {
+        for (int k = 0; k < 8; ++k) x8[k] = (const block_iq4_k *)((const char *)vx + (ix+k)*bx);
+        for (int i = 0; i < nb; ++i) {
+            for (int k = 0; k < 8; ++k) {
+                float d = GGML_FP16_TO_FP32(x8[k][i].d);
+                auto extra = x8[k][i].extra;
+                for (int ib32 = 0; ib32 < 8; ++ib32) {
+                    const uint8_t sh = x8[k][i].scales_h[ib32/2] >> 4*(ib32%2);
+                    ls[2*ib32+0] = ((x8[k][i].scales_l[ib32] & 0xf) | ((sh << 4) & 0x30)) - 32;
+                    ls[2*ib32+1] = ((x8[k][i].scales_l[ib32] >>  4) | ((sh << 2) & 0x30)) - 32;
+                    auto bits = vld1q_u8(x8[k][i].qs+16*ib32);
+                    xv[ib32].val[0] = vqtbl1q_s8(values.val[extra & 1], vandq_u8(bits,  ml)); extra >>= 1;
+                    xv[ib32].val[1] = vqtbl1q_s8(values.val[extra & 1], vshrq_n_u8(bits, 4)); extra >>= 1;
+                }
+                float dnew = convert_to_q8_k_r8(1.f/127, xv, ls, block, (uint32_t *)y[i].qs + k);
+                y[i].d[k] = GGML_FP32_TO_FP16(d*dnew);
+            }
+        }
+        y += nb;
+    }
+}
+
 void iqk_convert_iq5_k_q8_k_r8(int n, const void * vx, size_t bx, void * vy, int nrc_x) {
     GGML_ASSERT(n%QK_K == 0);
     GGML_ASSERT(nrc_x%8 == 0);
@@ -4243,7 +4283,7 @@ bool iqk_convert_iqk_quants_q80_r8(int type, int n, const void * vx, size_t bx,
     //    case GGML_TYPE_IQ2_K  : iqk_convert_iq2_k_q8_k_r8 (n, vx, bx, vy, nrc_x); break;
     //    case GGML_TYPE_IQ3_K  : iqk_convert_iq3_k_q8_k_r8 (n, vx, bx, vy, nrc_x); break;
         case GGML_TYPE_IQ4_KS : iqk_convert_iq4_ks_q8_k_r8(n, vx, bx, vy, nrc_x); break;
-    //    case GGML_TYPE_IQ4_K  : iqk_convert_iq4_k_q8_k_r8 (n, vx, bx, vy, nrc_x); break;
+        case GGML_TYPE_IQ4_K  : iqk_convert_iq4_k_q8_k_r8 (n, vx, bx, vy, nrc_x); break;
         case GGML_TYPE_IQ5_KS : iqk_convert_iq5_ks_q8_k_r8(n, vx, bx, vy, nrc_x); break;
         case GGML_TYPE_IQ5_K  : iqk_convert_iq5_k_q8_k_r8 (n, vx, bx, vy, nrc_x); break;
         case GGML_TYPE_IQ6_K  : iqk_convert_iq6_k_q8_k_r8 (n, vx, bx, vy, nrc_x); break;
diff --git a/ggml/src/iqk/iqk_mul_mat.cpp b/ggml/src/iqk/iqk_mul_mat.cpp
index 189170ba..718089a5 100644
--- a/ggml/src/iqk/iqk_mul_mat.cpp
+++ b/ggml/src/iqk/iqk_mul_mat.cpp
@@ -284,6 +284,7 @@ struct MulMat {
             case GGML_TYPE_IQ2_KS : return nrc_y >= 32 ? GGML_TYPE_Q8_K_R8 : type;
             case GGML_TYPE_IQ4_KS : return nrc_y >= 32 ? GGML_TYPE_Q8_K_R8 : type;
             case GGML_TYPE_IQ5_KS : return nrc_y >= 32 ? GGML_TYPE_Q8_K_R8 : type;
+            case GGML_TYPE_IQ4_K  : return nrc_y >= 32 ? GGML_TYPE_Q8_K_R8 : type;
             case GGML_TYPE_IQ5_K  : return nrc_y >= 32 ? GGML_TYPE_Q8_K_R8 : type;
             case GGML_TYPE_IQ6_K  : return nrc_y >= 32 ? GGML_TYPE_Q8_K_R8 : type;
             default: break;