Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

vulkan: Support GGML_TYPE_NVFP4 (#21455)

Browse Source 
This adds nvfp4 support for get_rows, dequant, and mul_mat(_id). For
mul_mat, it does not add support for the dp4/q8_1 path, it's all via
fp16/fp32.
This commit is contained in:
Jeff Bolz
2026-04-14 11:34:23 +02:00
committed by GitHub
parent e489a5ca0e
commit 6a6780a232
8 changed files with 171 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs_cm2.glsl
+20
View File
@@ -697,6 +697,24 @@ float16_t dequantFuncMXFP4(const in decodeBufMXFP4 bl, const in uint blockCoords
}
#endif
#if defined(DATA_A_NVFP4)
layout(buffer_reference, std430, buffer_reference_align = 4) buffer decodeBufNVFP4 {
block_nvfp4 block;
};
float16_t dequantFuncNVFP4(const in decodeBufNVFP4 bl, const in uint blockCoords[2], const in uint coordInBlock[2])
{
const uint idx = coordInBlock[1];
const uint sub = (idx & 0x30) >> 4;
const uint iqs = ((idx & 0x30) >> 1) + (idx & 0x7);
const uint shift = (idx & 0x8) >> 1;
const float d = ue4m3_to_fp32(bl.block.d[sub]);
uint qs = uint(bl.block.qs[iqs]);
qs = (qs >> shift) & 0xF;
return float16_t(kvalues_mxfp4[qs] * d * 0.5);
}
#endif
#if defined(DATA_A_Q1_0)
#define dequantFuncA dequantFuncQ1_0
#elif defined(DATA_A_Q4_0)
@@ -743,6 +761,8 @@ float16_t dequantFuncMXFP4(const in decodeBufMXFP4 bl, const in uint blockCoords
#define dequantFuncA dequantFuncIQ4_NL
#elif defined(DATA_A_MXFP4)
#define dequantFuncA dequantFuncMXFP4
#elif defined(DATA_A_NVFP4)
#define dequantFuncA dequantFuncNVFP4
#elif defined(DATA_A_F32)
#define dequantFuncA dequantFuncF32
#endif