From 5fbaf121db12aa65ca62e5bfe08a5a056ce9759f Mon Sep 17 00:00:00 2001
From: Eve <139727413+netrunnereve@users.noreply.github.com>
Date: Wed, 4 Dec 2024 16:31:14 -0500
Subject: [PATCH] remove a multiply

polynomial iq4_nl test (slower but keep as reference)
---
 ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.comp | 7 ++++++-
 ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp   | 4 +++-
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.comp b/ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.comp
index 6945b0b51..5e917ffd2 100644
--- a/ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.comp
+++ b/ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.comp
@@ -100,6 +100,11 @@ vec4 dequantize4(uint ib, uint iqs, uint a_offset) {
 #endif
 
 #if defined(DATA_A_IQ4_NL)
+float iq_helper(uint i) {
+	const float x = float(i);
+    return round(((0.080958*x-1.875836)*x+25.907107)*x-127.663571);
+}
+
 vec2 dequantize(uint ib, uint iqs, uint a_offset) {
     const float d = float(data_a[a_offset + ib].d);
     const uint vui = uint(data_a[a_offset + ib].qs[iqs]);
@@ -107,6 +112,6 @@ vec2 dequantize(uint ib, uint iqs, uint a_offset) {
 }
 vec4 dequantize4(uint ib, uint iqs, uint a_offset) {
     const uint vui = uint(data_a_packed16[a_offset + ib].qs[iqs/2]);
-    return vec4(kvalues_iq4nl[vui & 0xF], kvalues_iq4nl[(vui >> 4) & 0xF], kvalues_iq4nl[(vui >> 8) & 0xF], kvalues_iq4nl[(vui >> 12) & 0xF]);
+    return vec4(iq_helper(vui & 0xF), iq_helper((vui >> 4) & 0xF), iq_helper((vui >> 8) & 0xF), iq_helper((vui >> 12) & 0xF));
 }
 #endif
diff --git a/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp b/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp
index 433da8593..bca7abf3f 100644
--- a/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp
+++ b/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp
@@ -53,8 +53,10 @@ void iter(inout FLOAT_TYPE temp[NUM_ROWS], const uint first_row, const uint num_
         b1 = FLOAT_TYPE(data_b[b_offset + iybs + iqs + y_offset]);
     }
 #endif
+    uint ibi = first_row*p.ncols;
     [[unroll]] for (uint n = 0; n < num_rows; ++n) {
-        const uint ib = ((first_row + n)*p.ncols + col)/QUANT_K; // block index
+        const uint ib = (ibi + col)/QUANT_K; // block index
+        ibi += p.ncols;
 
 #if K_PER_ITER == 8
         // TODO: can we dequant as f16 instead of as vec?