Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Files
19620004f59ecbbdcbdad0dfd742500f5971ea61
llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders
T
History
Matt Corallo 19620004f5 vulkan: Block-load Q3_K/Q6_K block data and subtract on 32b ints (#23056) 
Q2_K/Q3_K/Q6_K do much better when using MMVQ on Intel BMG even
though they're only 2-byte aligned, and Q3_K still wins on
NVIDIA as well.

mesa isn't all that great at coalescing back-to-back loads from
alternating arrays, so we force it instead. Further, we can do
subtraction directly on a full int32_t rather than an i8vec4
with bit twiddling because the high bit is always free to start.

On Intel BMG on mesa, the switch to MMVQ provides an immediate
~57% perf increase in tg128 for unsloth/Qwen3.5-9B-GGUF:Q3_K and
~78% perf increase in tg128 for unsloth/Qwen3.5-9B-GGUF:Q6_K.

The futher switch to block loads leads to a ~24% perf increase in
tg128 for unsloth/Qwen3.5-9B-GGUF:Q3_K and a ~48% perf increase in
tg128 for unsloth/Qwen3.5-9B-GGUF:Q6_K.

Finally, Xe2 wins on MMVQ even for small k, so we take the NVIDIA
override for K quants on Xe2 as well.
2026-06-01 11:46:48 +02:00
..
feature-tests
vulkan: use GL_NV_cooperative_matrix_decode_vector for faster matmul (#23541)
2026-05-27 17:18:28 +02:00
abs.comp
vulkan: implement ABS and NEG (#17245)
2025-11-15 12:00:29 +01:00
acc.comp
vulkan: support GGML_OP_SET (#19584)
2026-02-14 06:36:38 +01:00
add1.comp
vulkan: implement ADD1, ARANGE, FILL, SOFTPLUS, STEP, ROUND, CEIL, FLOOR, TRUNC (#17319)
2025-11-19 17:29:45 +01:00
add_id.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
add.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
arange.comp
vulkan: implement ADD1, ARANGE, FILL, SOFTPLUS, STEP, ROUND, CEIL, FLOOR, TRUNC (#17319)
2025-11-19 17:29:45 +01:00
argmax.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
argsort_large.comp
vulkan: support larger argsort (#17313)
2025-11-19 17:25:50 +01:00
argsort.comp
vulkan: support larger argsort (#17313)
2025-11-19 17:25:50 +01:00
ceil.comp
vulkan: implement ADD1, ARANGE, FILL, SOFTPLUS, STEP, ROUND, CEIL, FLOOR, TRUNC (#17319)
2025-11-19 17:29:45 +01:00
clamp.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
CMakeLists.txt
vulkan: use GL_NV_cooperative_matrix_decode_vector for faster matmul (#23541)
2026-05-27 17:18:28 +02:00
concat.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
contig_copy.comp
vulkan: add cpy bf16 -> f32 pipelines (#22677)
2026-05-17 11:31:20 +02:00
conv2d_dw.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
conv2d_mm.comp
vulkan: optimize conv2d and implement coopmat1 support (#22620)
2026-05-26 15:48:05 +02:00
conv_transpose_1d.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
copy_from_quant.comp
vulkan: Support GGML_TYPE_NVFP4 (#21455)
2026-04-14 11:34:23 +02:00
copy_to_quant.comp
vulkan: Programmatically add RoundingModeRTE to all shaders when the device supports it (#21572)
2026-04-14 15:17:45 +02:00
copy_transpose.comp
vulkan: Add copy_transpose shader (#17371)
2025-11-19 16:50:43 +01:00
copy.comp
vulkan: add cpy bf16 -> f32 pipelines (#22677)
2026-05-17 11:31:20 +02:00
cos.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
count_equal.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
count_experts.comp
vulkan: preprocess mul_mat_id experts and discard workgroups more quickly (#18352)
2025-12-26 16:12:58 -06:00
cumsum_multipass1.comp
vulkan: Optimize GGML_OP_CUMSUM (#18417)
2026-01-02 15:32:30 -06:00
cumsum_multipass2.comp
vulkan: Optimize GGML_OP_CUMSUM (#18417)
2026-01-02 15:32:30 -06:00
cumsum.comp
vulkan: Optimize GGML_OP_CUMSUM (#18417)
2026-01-02 15:32:30 -06:00
dequant_f32.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_funcs_cm2.glsl
vulkan: use GL_NV_cooperative_matrix_decode_vector for faster matmul (#23541)
2026-05-27 17:18:28 +02:00
dequant_funcs.glsl
vulkan: Switch MUL_MAT_VEC to 4 K per iteration for F16/32 (#22887)
2026-05-27 17:19:23 +02:00
dequant_head.glsl
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_iq1_m.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_iq1_s.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_iq2_s.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_iq2_xs.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_iq2_xxs.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_iq3_s.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_iq3_xxs.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_iq4_nl.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_iq4_xs.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_mxfp4.comp
Vulkan MMQ Integer Dot Refactor and K-Quant support (#16536)
2025-10-29 14:39:03 +01:00
dequant_nvfp4.comp
vulkan: Support GGML_TYPE_NVFP4 (#21455)
2026-04-14 11:34:23 +02:00
dequant_q1_0.comp
vulkan: Support Q1_0 (#21539)
2026-04-10 08:35:27 +02:00
dequant_q2_k.comp
Vulkan MMQ Integer Dot Refactor and K-Quant support (#16536)
2025-10-29 14:39:03 +01:00
dequant_q3_k.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_q4_0.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_q4_1.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_q4_k.comp
Vulkan MMQ Integer Dot Refactor and K-Quant support (#16536)
2025-10-29 14:39:03 +01:00
dequant_q5_0.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_q5_1.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_q5_k.comp
Vulkan MMQ Integer Dot Refactor and K-Quant support (#16536)
2025-10-29 14:39:03 +01:00
dequant_q6_k.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
dequant_q8_0.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
diag_mask_inf.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
diag.comp
vulkan: Programmatically add RoundingModeRTE to all shaders when the device supports it (#21572)
2026-04-14 15:17:45 +02:00
div.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
elu.comp
ggml-vulkan: Add ELU op support (#20183)
2026-03-08 12:38:17 +01:00
exp.comp
vulkan: Programmatically add RoundingModeRTE to all shaders when the device supports it (#21572)
2026-04-14 15:17:45 +02:00
fill.comp
vulkan: implement ADD1, ARANGE, FILL, SOFTPLUS, STEP, ROUND, CEIL, FLOOR, TRUNC (#17319)
2025-11-19 17:29:45 +01:00
flash_attn_base.glsl
vulkan: add Flash Attention support for BFloat16 KV cache (#23420)
2026-05-30 10:39:31 +02:00
flash_attn_cm1.comp
vulkan: add Flash Attention support for BFloat16 KV cache (#23420)
2026-05-30 10:39:31 +02:00
flash_attn_cm2.comp
vulkan: add Flash Attention support for BFloat16 KV cache (#23420)
2026-05-30 10:39:31 +02:00
flash_attn_dequant.glsl
vulkan: add Flash Attention support for BFloat16 KV cache (#23420)
2026-05-30 10:39:31 +02:00
flash_attn_mask_opt.comp
vulkan: fix OOB check in flash_attn_mask_opt (#20296)
2026-03-12 06:35:49 +01:00
flash_attn_mmq_funcs.glsl
vulkan: Support asymmetric FA in scalar/mmq/coopmat1 paths (#22589)
2026-05-11 12:49:03 +02:00
flash_attn_split_k_reduce.comp
vulkan: support flash attention GQA/split_k with small batches (#18938)
2026-01-21 17:43:43 +01:00
flash_attn.comp
vulkan: Support asymmetric FA in scalar/mmq/coopmat1 paths (#22589)
2026-05-11 12:49:03 +02:00
floor.comp
vulkan: implement ADD1, ARANGE, FILL, SOFTPLUS, STEP, ROUND, CEIL, FLOOR, TRUNC (#17319)
2025-11-19 17:29:45 +01:00
fwht.comp
vulkan: fast path for walsh-hadamard transform (#23687)
2026-05-28 13:18:43 +02:00
gated_delta_net.comp
llama + spec: MTP Support (#22673)
2026-05-16 20:06:23 +08:00
geglu_erf.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
geglu_quick.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
geglu.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
gelu_erf.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
gelu_quick.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
gelu.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
generic_binary_head.glsl
vulkan: Programmatically add RoundingModeRTE to all shaders when the device supports it (#21572)
2026-04-14 15:17:45 +02:00
generic_head.glsl
vulkan: support GGML_UNARY_OP_XIELU (#18062)
2025-12-21 10:17:58 +01:00
generic_unary_head.glsl
Vulkan: MMVQ Integer Dot K-Quant and MUL_MAT_ID support (#16900)
2025-11-29 09:37:22 +01:00
get_rows_quant.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
get_rows.comp
vulkan: support get_rows for i32 (#17941)
2025-12-13 10:12:53 +01:00
glu_head.glsl
vulkan: Programmatically add RoundingModeRTE to all shaders when the device supports it (#21572)
2026-04-14 15:17:45 +02:00
glu_main.glsl
vulkan: add noncontiguous GLU support (#21081)
2026-03-28 08:44:56 +01:00
group_norm.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
hardsigmoid.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
hardswish.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
im2col_3d.comp
vulkan: Programmatically add RoundingModeRTE to all shaders when the device supports it (#21572)
2026-04-14 15:17:45 +02:00
im2col.comp
vulkan: optimize operations in the IM2COL shader (#22685)
2026-05-20 17:15:13 +02:00
l2_norm.comp
vulkan: fix l2_norm epsilon handling (#20350)
2026-03-12 06:39:41 +01:00
leaky_relu.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
log.comp
vulkan: Programmatically add RoundingModeRTE to all shaders when the device supports it (#21572)
2026-04-14 15:17:45 +02:00
mul_mat_split_k_reduce.comp
vulkan: optimize and reenable split_k (#10637)
2024-12-03 20:29:54 +01:00
mul_mat_vec_base.glsl
vulkan: split mul_mat into multiple dispatches to avoid overflow (#19509)
2026-02-18 10:47:10 +01:00
mul_mat_vec_iface.glsl
vulkan: unify type macros to use Vx instead of _VECx (#21605)
2026-04-09 07:31:51 +02:00
mul_mat_vec_iq1_m.comp
vulkan: Use VK_EXT_shader_64bit_indexing to handle large mat_mul(_id) (#18678)
2026-01-12 12:32:13 +01:00
mul_mat_vec_iq1_s.comp
vulkan: Use VK_EXT_shader_64bit_indexing to handle large mat_mul(_id) (#18678)
2026-01-12 12:32:13 +01:00
mul_mat_vec_iq2_s.comp
vulkan: Use VK_EXT_shader_64bit_indexing to handle large mat_mul(_id) (#18678)
2026-01-12 12:32:13 +01:00
mul_mat_vec_iq2_xs.comp
vulkan: Use VK_EXT_shader_64bit_indexing to handle large mat_mul(_id) (#18678)
2026-01-12 12:32:13 +01:00
mul_mat_vec_iq2_xxs.comp
vulkan: Use VK_EXT_shader_64bit_indexing to handle large mat_mul(_id) (#18678)
2026-01-12 12:32:13 +01:00
mul_mat_vec_iq3_s.comp
vulkan: Use VK_EXT_shader_64bit_indexing to handle large mat_mul(_id) (#18678)
2026-01-12 12:32:13 +01:00
mul_mat_vec_iq3_xxs.comp
vulkan: Use VK_EXT_shader_64bit_indexing to handle large mat_mul(_id) (#18678)
2026-01-12 12:32:13 +01:00
mul_mat_vec_nc.comp
vulkan: Fuse mul_mat_id+add_id+mul and mul_mat+add+add. (#17287)
2025-11-15 19:54:23 +01:00
mul_mat_vec_p021.comp
vulkan: Fuse mul_mat_id+add_id+mul and mul_mat+add+add. (#17287)
2025-11-15 19:54:23 +01:00
mul_mat_vec_q2_k.comp
vulkan: unify type macros to use Vx instead of _VECx (#21605)
2026-04-09 07:31:51 +02:00
mul_mat_vec_q3_k.comp
vulkan: Use VK_EXT_shader_64bit_indexing to handle large mat_mul(_id) (#18678)
2026-01-12 12:32:13 +01:00
mul_mat_vec_q4_k.comp
vulkan: unify type macros to use Vx instead of _VECx (#21605)
2026-04-09 07:31:51 +02:00
mul_mat_vec_q5_k.comp
vulkan: unify type macros to use Vx instead of _VECx (#21605)
2026-04-09 07:31:51 +02:00
mul_mat_vec_q6_k.comp
vulkan: Use VK_EXT_shader_64bit_indexing to handle large mat_mul(_id) (#18678)
2026-01-12 12:32:13 +01:00
mul_mat_vec.comp
vulkan: Switch MUL_MAT_VEC to 4 K per iteration for F16/32 (#22887)
2026-05-27 17:19:23 +02:00
mul_mat_vecq_funcs.glsl
vulkan: Block-load Q3_K/Q6_K block data and subtract on 32b ints (#23056)
2026-06-01 11:46:48 +02:00
mul_mat_vecq.comp
vulkan: Use VK_EXT_shader_64bit_indexing to handle large mat_mul(_id) (#18678)
2026-01-12 12:32:13 +01:00
mul_mm_cm2.comp
vulkan: use GL_NV_cooperative_matrix_decode_vector for faster matmul (#23541)
2026-05-27 17:18:28 +02:00
mul_mm_funcs.glsl
vulkan: Coalesce Q4_K/Q5_K scale loads (#21751)
2026-04-28 17:31:04 +02:00
mul_mm_id_funcs.glsl
vulkan: fix data race in mul_mat_id shader (#19790)
2026-02-24 07:43:12 +01:00
mul_mm.comp
vulkan: unify type macros to use Vx instead of _VECx (#21605)
2026-04-09 07:31:51 +02:00
mul_mmq_funcs.glsl
vulkan: unify type macros to use Vx instead of _VECx (#21605)
2026-04-09 07:31:51 +02:00
mul_mmq_shmem_types.glsl
vulkan: Support asymmetric FA in scalar/mmq/coopmat1 paths (#22589)
2026-05-11 12:49:03 +02:00
mul_mmq.comp
vulkan: fix MMQ shader push constants and multi-dispatch (#19732)
2026-02-19 14:59:16 +01:00
mul.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
multi_add.comp
vulkan: Programmatically add RoundingModeRTE to all shaders when the device supports it (#21572)
2026-04-14 15:17:45 +02:00
neg.comp
vulkan: implement ABS and NEG (#17245)
2025-11-15 12:00:29 +01:00
norm.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
opt_step_adamw.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
opt_step_sgd.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
pad.comp
ggml : add circular tiling support to pad, for Vulkan, CUDA, and CPU (used for making seamless textures) (#16985)
2025-12-06 15:07:02 +01:00
pool2d.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
quantize_q8_1.comp
vulkan: handle quantize_q8_1 overflowing the max workgroup count (#18515)
2026-01-05 11:30:14 +01:00
reglu.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
relu.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
repeat_back.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
repeat.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
rms_norm_back.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
rms_norm_partials.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
rms_norm.comp
vulkan: fix non-contig rope (#19299)
2026-02-05 08:38:59 +01:00
roll.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
rope_funcs.glsl
vulkan: Support unaligned tensors for ROPE (#22637)
2026-05-17 11:30:16 +02:00
rope_head.glsl
vulkan: Programmatically add RoundingModeRTE to all shaders when the device supports it (#21572)
2026-04-14 15:17:45 +02:00
rope_multi.comp
vulkan: fix non-contig rope (#19299)
2026-02-05 08:38:59 +01:00
rope_neox.comp
vulkan: fix non-contig rope (#19299)
2026-02-05 08:38:59 +01:00
rope_norm.comp
vulkan: fix non-contig rope (#19299)
2026-02-05 08:38:59 +01:00
rope_params.glsl
vulkan: Support unaligned tensors for ROPE (#22637)
2026-05-17 11:30:16 +02:00
rope_vision.comp
vulkan: fix non-contig rope (#19299)
2026-02-05 08:38:59 +01:00
round.comp
vulkan: implement ADD1, ARANGE, FILL, SOFTPLUS, STEP, ROUND, CEIL, FLOOR, TRUNC (#17319)
2025-11-19 17:29:45 +01:00
scale.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
sgn.comp
ggml-vulkan: add SGN operator, auto-generate Vulkan.csv and ops.md (#20219)
2026-03-09 07:24:16 +01:00
sigmoid.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
silu_back.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
silu.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
sin.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
snake.comp
vulkan: fuse snake activation (mul, sin, sqr, mul, add) (#22855)
2026-05-21 19:39:42 +02:00
soft_max_back.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
soft_max_large1.comp
vulkan: Multi-pass softmax for large number of cols (#17892)
2025-12-13 10:04:29 +01:00
soft_max_large2.comp
vulkan: Multi-pass softmax for large number of cols (#17892)
2025-12-13 10:04:29 +01:00
soft_max_large3.comp
vulkan: Multi-pass softmax for large number of cols (#17892)
2025-12-13 10:04:29 +01:00
soft_max_large_common.glsl
vulkan: Multi-pass softmax for large number of cols (#17892)
2025-12-13 10:04:29 +01:00
soft_max.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
softplus.comp
vulkan: implement ADD1, ARANGE, FILL, SOFTPLUS, STEP, ROUND, CEIL, FLOOR, TRUNC (#17319)
2025-11-19 17:29:45 +01:00
solve_tri.comp
vulkan: support solve_tri with larger N/K values (#17781)
2025-12-06 08:56:45 +01:00
sqrt.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
square.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
ssm_conv.comp
vulkan: fuse SSM_CONV + BIAS + SILU (#22653)
2026-05-17 10:25:50 +02:00
ssm_scan.comp
vulkan: optimize ssm_scan (#18630)
2026-01-08 15:16:54 +01:00
step.comp
vulkan: implement ADD1, ARANGE, FILL, SOFTPLUS, STEP, ROUND, CEIL, FLOOR, TRUNC (#17319)
2025-11-19 17:29:45 +01:00
sub.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
sum_rows.comp
vulkan: Implement GGML_OP_CUMSUM (#17479)
2025-11-26 07:08:10 +01:00
sum_rows.glsl
vulkan: Implement GGML_OP_CUMSUM (#17479)
2025-11-26 07:08:10 +01:00
swiglu_oai.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
swiglu.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
tanh.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
timestep_embedding.comp
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
topk_argsort.comp
vulkan: Reduce temporary memory usage for TOP_K (#17623)
2025-12-02 19:22:04 +01:00
topk_moe.comp
vulkan: fix topk_moe_sigmoid_norm_bias failures in GLM-4.6 (#18582)
2026-01-05 11:51:39 +01:00
topk_nary_search.comp
vulkan: fix top_k bug when there are ties in the input (#17659)
2025-12-05 22:03:19 +01:00
tri.comp
vulkan: Programmatically add RoundingModeRTE to all shaders when the device supports it (#21572)
2026-04-14 15:17:45 +02:00
trunc.comp
vulkan: implement ADD1, ARANGE, FILL, SOFTPLUS, STEP, ROUND, CEIL, FLOOR, TRUNC (#17319)
2025-11-19 17:29:45 +01:00
types.glsl
vulkan: Switch MUL_MAT_VEC to 4 K per iteration for F16/32 (#22887)
2026-05-27 17:19:23 +02:00
upscale.comp
vulkan: Support UPSCALE w/antialias (#18327)
2025-12-26 17:00:57 +01:00
utils.glsl
vulkan : incremental shader builds (#16341)
2025-10-04 11:42:56 +02:00
vulkan-shaders-gen.cpp
vulkan: add Flash Attention support for BFloat16 KV cache (#23420)
2026-05-30 10:39:31 +02:00
wkv6.comp
rwkv6: add wkv6 support for Vulkan backend (#10829)
2024-12-16 22:00:46 +01:00
wkv7.comp
llama: Add support for RWKV v7 architecture (#12412)
2025-03-18 07:27:50 +08:00
xielu.comp
vulkan: support GGML_UNARY_OP_XIELU (#18062)
2025-12-21 10:17:58 +01:00