vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

cleanup

Browse source
This commit is contained in:
Eve 2024-12-07 14:51:05 -05:00
parent f5a15fc69e
commit bd17bc452f
1 changed files with 16 additions and 22 deletions
Download patch file Download diff file Expand all files Collapse all files

38
ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.comp
View file

@ -15,18 +15,12 @@ layout (binding = 0) readonly buffer A_PACKED32 {A_TYPE_PACKED32 data_a_packed32
vec2 dequantize(uint ib, uint iqs, uint a_offset) { vec2 dequantize(uint ib, uint iqs, uint a_offset) {
return vec2(data_a[a_offset + ib], data_a[a_offset + ib + 1]); return vec2(data_a[a_offset + ib], data_a[a_offset + ib + 1]);
} }
vec2 get_dm(uint ib, uint a_offset) {
return vec2(0, 0);
}
#endif #endif
#if defined(DATA_A_F16) #if defined(DATA_A_F16)
vec2 dequantize(uint ib, uint iqs, uint a_offset) { vec2 dequantize(uint ib, uint iqs, uint a_offset) {
return vec2(data_a[a_offset + ib], data_a[a_offset + ib + 1]); return vec2(data_a[a_offset + ib], data_a[a_offset + ib + 1]);
} }
vec2 get_dm(uint ib, uint a_offset) {
return vec2(0, 0);
}
#endif #endif
#if defined(DATA_A_Q4_0) #if defined(DATA_A_Q4_0)
@ -38,9 +32,6 @@ vec4 dequantize4(uint ib, uint iqs, uint a_offset) {
const uint vui = uint(data_a_packed16[a_offset + ib].qs[iqs/2]); const uint vui = uint(data_a_packed16[a_offset + ib].qs[iqs/2]);
return (vec4(vui & 0xF, (vui >> 4) & 0xF, (vui >> 8) & 0xF, vui >> 12) - 8.0f); return (vec4(vui & 0xF, (vui >> 4) & 0xF, (vui >> 8) & 0xF, vui >> 12) - 8.0f);
} }
vec2 get_dm(uint ib, uint a_offset) {
return vec2(float(data_a[a_offset + ib].d), 0);
}
#endif #endif
#if defined(DATA_A_Q4_1) #if defined(DATA_A_Q4_1)
@ -52,9 +43,6 @@ vec4 dequantize4(uint ib, uint iqs, uint a_offset) {
const uint vui = uint(data_a_packed16[a_offset + ib].qs[iqs/2]); const uint vui = uint(data_a_packed16[a_offset + ib].qs[iqs/2]);
return vec4(vui & 0xF, (vui >> 4) & 0xF, (vui >> 8) & 0xF, vui >> 12); return vec4(vui & 0xF, (vui >> 4) & 0xF, (vui >> 8) & 0xF, vui >> 12);
} }
vec2 get_dm(uint ib, uint a_offset) {
return vec2(float(data_a[a_offset + ib].d), float(data_a[a_offset + ib].m));
}
#endif #endif
#if defined(DATA_A_Q5_0) #if defined(DATA_A_Q5_0)
@ -69,10 +57,7 @@ vec4 dequantize4(uint ib, uint iqs, uint a_offset) {
const ivec2 qh0 = ivec2(((uint_qh >> iqs) << 4) & 0x10, (uint_qh >> (iqs + 12)) & 0x10); const ivec2 qh0 = ivec2(((uint_qh >> iqs) << 4) & 0x10, (uint_qh >> (iqs + 12)) & 0x10);
const ivec2 qh1 = ivec2(((uint_qh >> (iqs + 1)) << 4) & 0x10, (uint_qh >> (iqs + 13)) & 0x10); const ivec2 qh1 = ivec2(((uint_qh >> (iqs + 1)) << 4) & 0x10, (uint_qh >> (iqs + 13)) & 0x10);
const uint vui = uint(data_a_packed16[a_offset + ib].qs[iqs/2]); const uint vui = uint(data_a_packed16[a_offset + ib].qs[iqs/2]);
return (vec4(((vui >> 0) & 0xF) | qh0.x, ((vui >> 4) & 0xF) | qh0.y, ((vui >> 8) & 0xF) | qh1.x, (vui >> 12) | qh1.y) - 16.0f); return (vec4((vui & 0xF) | qh0.x, ((vui >> 4) & 0xF) | qh0.y, ((vui >> 8) & 0xF) | qh1.x, (vui >> 12) | qh1.y) - 16.0f);
}
vec2 get_dm(uint ib, uint a_offset) {
return vec2(float(data_a[a_offset + ib].d), 0);
} }
#endif #endif
@ -90,9 +75,6 @@ vec4 dequantize4(uint ib, uint iqs, uint a_offset) {
const uint vui = uint(data_a_packed16[a_offset + ib].qs[iqs/2]); const uint vui = uint(data_a_packed16[a_offset + ib].qs[iqs/2]);
return vec4(((vui >> 0) & 0xF) | qh0.x, ((vui >> 4) & 0xF) | qh0.y, ((vui >> 8) & 0xF) | qh1.x, (vui >> 12) | qh1.y); return vec4(((vui >> 0) & 0xF) | qh0.x, ((vui >> 4) & 0xF) | qh0.y, ((vui >> 8) & 0xF) | qh1.x, (vui >> 12) | qh1.y);
} }
vec2 get_dm(uint ib, uint a_offset) {
return vec2(float(data_a[a_offset + ib].d), float(data_a[a_offset + ib].m));
}
#endif #endif
#if defined(DATA_A_Q8_0) #if defined(DATA_A_Q8_0)
@ -104,9 +86,6 @@ vec4 dequantize4(uint ib, uint iqs, uint a_offset) {
uint32_t v1 = data_a_packed16[a_offset + ib].qs[iqs/2 + 1]; uint32_t v1 = data_a_packed16[a_offset + ib].qs[iqs/2 + 1];
return vec4(int8_t(v0 & 0xFF), int8_t(v0 >> 8), int8_t(v1 & 0xFF), int8_t(v1 >> 8)); return vec4(int8_t(v0 & 0xFF), int8_t(v0 >> 8), int8_t(v1 & 0xFF), int8_t(v1 >> 8));
} }
vec2 get_dm(uint ib, uint a_offset) {
return vec2(float(data_a[a_offset + ib].d), 0);
}
#endif #endif
#if defined(DATA_A_IQ4_NL) #if defined(DATA_A_IQ4_NL)
@ -118,7 +97,22 @@ vec4 dequantize4(uint ib, uint iqs, uint a_offset) {
const uint vui = uint(data_a_packed16[a_offset + ib].qs[iqs/2]); 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]); return vec4(kvalues_iq4nl[vui & 0xF], kvalues_iq4nl[(vui >> 4) & 0xF], kvalues_iq4nl[(vui >> 8) & 0xF], kvalues_iq4nl[vui >> 12]);
} }
#endif
#if defined(DATA_A_F32) || defined(DATA_A_F16)
vec2 get_dm(uint ib, uint a_offset) {
return vec2(0, 0);
}
#endif
#if defined(DATA_A_Q4_0) || defined(DATA_A_Q5_0) || defined(DATA_A_Q8_0) || defined(DATA_A_IQ4_NL)
vec2 get_dm(uint ib, uint a_offset) { vec2 get_dm(uint ib, uint a_offset) {
return vec2(float(data_a[a_offset + ib].d), 0); return vec2(float(data_a[a_offset + ib].d), 0);
} }
#endif #endif
#if defined(DATA_A_Q4_1) || defined(DATA_A_Q5_1)
vec2 get_dm(uint ib, uint a_offset) {
return vec2(float(data_a[a_offset + ib].d), float(data_a[a_offset + ib].m));
}
#endif