Skip to content

musa: add GGML_UNUSED_VARS #15446

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Aug 21, 2025
Merged

musa: add GGML_UNUSED_VARS #15446

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions ggml/include/ggml.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -244,6 +244,13 @@
#define GGML_MROPE_SECTIONS 4

#define GGML_UNUSED(x) (void)(x)
#ifdef __CUDACC__
template<typename... Args>
__host__ __device__ constexpr inline void ggml_unused_vars_impl(Args&&...) noexcept {}
#define GGML_UNUSED_VARS(...) ggml_unused_vars_impl(__VA_ARGS__)
#else
#define GGML_UNUSED_VARS(...) do { (void)sizeof((__VA_ARGS__, 0)); } while(0)
#endif // __CUDACC__

#define GGML_PAD(x, n) (((x) + (n) - 1) & ~((n) - 1))

Expand Down
5 changes: 1 addition & 4 deletions ggml/src/ggml-cuda/conv-transpose-1d.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,10 +34,7 @@ static __global__ void conv_transpose_1d_kernel(
}
}
dst[global_index] = accumulator;
GGML_UNUSED(p0); GGML_UNUSED(d0); GGML_UNUSED(src0_ne3);
GGML_UNUSED(src1_ne3); GGML_UNUSED(dst_ne3);
GGML_UNUSED(src1_ne1); GGML_UNUSED(dst_ne1);
GGML_UNUSED(src1_ne2); GGML_UNUSED(dst_ne2);
GGML_UNUSED_VARS(p0, d0, src0_ne3, src1_ne3, dst_ne3, src1_ne1, dst_ne1, src1_ne2, dst_ne2);
}

static void conv_transpose_1d_f32_f32_cuda(
Expand Down
4 changes: 1 addition & 3 deletions ggml/src/ggml-cuda/convert.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,9 +71,7 @@ static __global__ void dequantize_block_q8_0_f16(const void * __restrict__ vx, h
y2[iy/2 + threadIdx.x] = __hmul2(make_half2(qs.x, qs.y), __half2half2(d));
}
#else
GGML_UNUSED(vx);
GGML_UNUSED(y);
GGML_UNUSED(k);
GGML_UNUSED_VARS(vx, y, k);
NO_DEVICE_CODE;
#endif // __CUDA_ARCH__ >= GGML_CUDA_CC_PASCAL
}
Expand Down
3 changes: 1 addition & 2 deletions ggml/src/ggml-cuda/cpy.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -134,8 +134,7 @@ void ggml_cuda_cpy_dest_ptrs_copy(ggml_cuda_graph * cuda_graph, char ** host_des
CUDA_CHECK(cudaMemcpyAsync(cuda_graph->dest_ptrs_d, host_dest_ptrs, host_dest_ptrs_size*sizeof(char *), cudaMemcpyHostToDevice, stream));
cuda_graph->graph_cpynode_index = 0; // reset index
#else
GGML_UNUSED(cuda_graph); GGML_UNUSED(host_dest_ptrs);
GGML_UNUSED(host_dest_ptrs_size); GGML_UNUSED(stream);
GGML_UNUSED_VARS(cuda_graph, host_dest_ptrs, host_dest_ptrs_size, stream);
#endif
}

Expand Down
45 changes: 18 additions & 27 deletions ggml/src/ggml-cuda/fattn-mma-f16.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -767,14 +767,11 @@ static __device__ __forceinline__ void flash_attn_ext_f16_iter(
}
}
#else
GGML_UNUSED(Q_f2); GGML_UNUSED(K_h2); GGML_UNUSED(V_h2);
GGML_UNUSED(mask_h2); GGML_UNUSED(dstk); GGML_UNUSED(dstk_fixup);
GGML_UNUSED(scale); GGML_UNUSED(slope); GGML_UNUSED(logit_softcap);
GGML_UNUSED(ne01); GGML_UNUSED(ne02); GGML_UNUSED(stride_K); GGML_UNUSED(stride_V);
GGML_UNUSED(stride_mask); GGML_UNUSED(tile_K);
GGML_UNUSED(tile_V); GGML_UNUSED(tile_mask); GGML_UNUSED(Q_B);
GGML_UNUSED(VKQ_C); GGML_UNUSED(KQ_max); GGML_UNUSED(KQ_rowsum);
GGML_UNUSED(kb0); GGML_UNUSED(tile_Q);
GGML_UNUSED_VARS(Q_f2, K_h2, V_h2, mask_h2, dstk, dstk_fixup,
scale, slope, logit_softcap, ne01, ne02,
stride_K, stride_V, stride_mask,
tile_Q, tile_K, tile_V, tile_mask,
Q_B, VKQ_C, KQ_max, KQ_rowsum, kb0);
NO_DEVICE_CODE;
#endif // TURING_MMA_AVAILABLE
}
Expand Down Expand Up @@ -1236,14 +1233,10 @@ static __device__ __forceinline__ void flash_attn_ext_f16_process_tile(
}
}
#else
GGML_UNUSED(Q_f2); GGML_UNUSED(K_h2); GGML_UNUSED(V_h2);
GGML_UNUSED(mask_h2); GGML_UNUSED(sinks_f);
GGML_UNUSED(dstk); GGML_UNUSED(dstk_fixup);
GGML_UNUSED(scale); GGML_UNUSED(slope); GGML_UNUSED(logit_softcap);
GGML_UNUSED(ne01); GGML_UNUSED(ne02);
GGML_UNUSED(stride_Q1); GGML_UNUSED(stride_Q2);
GGML_UNUSED(stride_K); GGML_UNUSED(stride_V); GGML_UNUSED(stride_mask);
GGML_UNUSED(jt); GGML_UNUSED(kb0_start); GGML_UNUSED(kb0_stop);
GGML_UNUSED_VARS(Q_f2, K_h2, V_h2, mask_h2, sinks_f, dstk, dstk_fixup,
scale, slope, logit_softcap, ne01, ne02,
stride_Q1, stride_Q2, stride_K, stride_V, stride_mask,
jt, kb0_start, kb0_stop);
NO_DEVICE_CODE;
#endif // TURING_MMA_AVAILABLE
}
Expand Down Expand Up @@ -1397,17 +1390,15 @@ static __global__ void flash_attn_ext_f16(
(Q_f2, K_h2, V_h2, mask_h2, sinks_f, dstk, dst_meta, scale, slope, logit_softcap,
ne01, ne02, stride_Q1, stride_Q2, stride_K, stride_V, stride_mask, jt, kb0_start_kernel, kb0_stop_kernel);
#else
GGML_UNUSED(Q); GGML_UNUSED(K); GGML_UNUSED(V); GGML_UNUSED(mask);
GGML_UNUSED(sinks); GGML_UNUSED(KV_max); GGML_UNUSED(dst); GGML_UNUSED(dst_meta);
GGML_UNUSED(scale); GGML_UNUSED(max_bias); GGML_UNUSED(m0); GGML_UNUSED(m1);
GGML_UNUSED(n_head_log2); GGML_UNUSED(logit_softcap);
GGML_UNUSED(ne00); GGML_UNUSED(ne01); GGML_UNUSED(ne02); GGML_UNUSED(ne03);
GGML_UNUSED(nb01); GGML_UNUSED(nb02); GGML_UNUSED(nb03);
GGML_UNUSED(ne10); GGML_UNUSED(ne11); GGML_UNUSED(ne12); GGML_UNUSED(ne13);
GGML_UNUSED(nb11); GGML_UNUSED(nb12); GGML_UNUSED(nb13);
GGML_UNUSED(nb21); GGML_UNUSED(nb22); GGML_UNUSED(nb23);
GGML_UNUSED(ne31); GGML_UNUSED(ne32); GGML_UNUSED(ne33);
GGML_UNUSED(nb31); GGML_UNUSED(nb32); GGML_UNUSED(nb33);
GGML_UNUSED_VARS(Q, K, V, mask, sinks, KV_max, dst, dst_meta, scale,
max_bias, m0, m1, n_head_log2, logit_softcap,
ne00, ne01, ne02, ne03,
nb01, nb02, nb03,
ne10, ne11, ne12, ne13,
nb11, nb12, nb13,
nb21, nb22, nb23,
ne31, ne32, ne33,
nb31, nb32, nb33);
NO_DEVICE_CODE;
#endif // defined(FLASH_ATTN_AVAILABLE) && defined(TURING_MMA_AVAILABLE)
}
Expand Down
20 changes: 9 additions & 11 deletions ggml/src/ggml-cuda/fattn-tile-f16.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -299,17 +299,15 @@ static __global__ void flash_attn_tile_ext_f16(
}
}
#else
GGML_UNUSED(Q); GGML_UNUSED(K); GGML_UNUSED(V); GGML_UNUSED(mask);
GGML_UNUSED(sinks); GGML_UNUSED(KV_max); GGML_UNUSED(dst); GGML_UNUSED(dst_meta);
GGML_UNUSED(scale); GGML_UNUSED(max_bias); GGML_UNUSED(m0); GGML_UNUSED(m1);
GGML_UNUSED(n_head_log2); GGML_UNUSED(logit_softcap);
GGML_UNUSED(ne00); GGML_UNUSED(ne01); GGML_UNUSED(ne02); GGML_UNUSED(ne03);
GGML_UNUSED(nb01); GGML_UNUSED(nb02); GGML_UNUSED(nb03);
GGML_UNUSED(ne10); GGML_UNUSED(ne11); GGML_UNUSED(ne12); GGML_UNUSED(ne13);
GGML_UNUSED(nb11); GGML_UNUSED(nb12); GGML_UNUSED(nb13);
GGML_UNUSED(nb21); GGML_UNUSED(nb22); GGML_UNUSED(nb23);
GGML_UNUSED(ne31); GGML_UNUSED(ne32); GGML_UNUSED(ne33);
GGML_UNUSED(nb31); GGML_UNUSED(nb32); GGML_UNUSED(nb33);
GGML_UNUSED_VARS(Q, K, V, mask, sinks, KV_max, dst, dst_meta, scale,
max_bias, m0, m1, n_head_log2, logit_softcap,
ne00, ne01, ne02, ne03,
nb01, nb02, nb03,
ne10, ne11, ne12, ne13,
nb11, nb12, nb13,
nb21, nb22, nb23,
ne31, ne32, ne33,
nb31, nb32, nb33);
NO_DEVICE_CODE;
#endif // defined(FLASH_ATTN_AVAILABLE) && defined(FP16_AVAILABLE)
}
Expand Down
29 changes: 18 additions & 11 deletions ggml/src/ggml-cuda/fattn-tile-f32.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,6 +38,15 @@ static __global__ void flash_attn_tile_ext_f32(
return;
#endif // FP16_MMA_AVAILABLE
if (use_logit_softcap && !(D == 128 || D == 256)) {
GGML_UNUSED_VARS(Q, K, V, mask, sinks, KV_max, dst, dst_meta, scale,
max_bias, m0, m1, n_head_log2, logit_softcap,
ne00, ne01, ne02, ne03,
nb01, nb02, nb03,
ne10, ne11, ne12, ne13,
nb11, nb12, nb13,
nb21, nb22, nb23,
ne31, ne32, ne33,
nb31, nb32, nb33);
NO_DEVICE_CODE;
return;
}
Expand Down Expand Up @@ -301,17 +310,15 @@ static __global__ void flash_attn_tile_ext_f32(
}
}
#else
GGML_UNUSED(Q); GGML_UNUSED(K); GGML_UNUSED(V); GGML_UNUSED(mask);
GGML_UNUSED(sinks); GGML_UNUSED(KV_max); GGML_UNUSED(dst); GGML_UNUSED(dst_meta);
GGML_UNUSED(scale); GGML_UNUSED(max_bias); GGML_UNUSED(m0); GGML_UNUSED(m1);
GGML_UNUSED(n_head_log2); GGML_UNUSED(logit_softcap);
GGML_UNUSED(ne00); GGML_UNUSED(ne01); GGML_UNUSED(ne02); GGML_UNUSED(ne03);
GGML_UNUSED(nb01); GGML_UNUSED(nb02); GGML_UNUSED(nb03);
GGML_UNUSED(ne10); GGML_UNUSED(ne11); GGML_UNUSED(ne12); GGML_UNUSED(ne13);
GGML_UNUSED(nb11); GGML_UNUSED(nb12); GGML_UNUSED(nb13);
GGML_UNUSED(nb21); GGML_UNUSED(nb22); GGML_UNUSED(nb23);
GGML_UNUSED(ne31); GGML_UNUSED(ne32); GGML_UNUSED(ne33);
GGML_UNUSED(nb31); GGML_UNUSED(nb32); GGML_UNUSED(nb33);
GGML_UNUSED_VARS(Q, K, V, mask, sinks, KV_max, dst, dst_meta, scale,
max_bias, m0, m1, n_head_log2, logit_softcap,
ne00, ne01, ne02, ne03,
nb01, nb02, nb03,
ne10, ne11, ne12, ne13,
nb11, nb12, nb13,
nb21, nb22, nb23,
ne31, ne32, ne33,
nb31, nb32, nb33);
NO_DEVICE_CODE;
#endif // FLASH_ATTN_AVAILABLE
}
Expand Down
20 changes: 9 additions & 11 deletions ggml/src/ggml-cuda/fattn-vec-f16.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -349,17 +349,15 @@ static __global__ void flash_attn_vec_ext_f16(
dst_meta[((sequence*ne01 + ic0 + tid)*ne02 + head)*gridDim.y + blockIdx.y] = make_float2(kqmax[tid], kqsum[tid]);
}
#else
GGML_UNUSED(Q); GGML_UNUSED(K); GGML_UNUSED(V); GGML_UNUSED(mask);
GGML_UNUSED(sinks); GGML_UNUSED(KV_max); GGML_UNUSED(dst); GGML_UNUSED(dst_meta);
GGML_UNUSED(scale); GGML_UNUSED(max_bias); GGML_UNUSED(m0); GGML_UNUSED(m1);
GGML_UNUSED(n_head_log2); GGML_UNUSED(logit_softcap);
GGML_UNUSED(ne00); GGML_UNUSED(ne01); GGML_UNUSED(ne02); GGML_UNUSED(ne03);
GGML_UNUSED(nb01); GGML_UNUSED(nb02); GGML_UNUSED(nb03);
GGML_UNUSED(ne10); GGML_UNUSED(ne11); GGML_UNUSED(ne12); GGML_UNUSED(ne13);
GGML_UNUSED(nb11); GGML_UNUSED(nb12); GGML_UNUSED(nb13);
GGML_UNUSED(nb21); GGML_UNUSED(nb22); GGML_UNUSED(nb23);
GGML_UNUSED(ne31); GGML_UNUSED(ne32); GGML_UNUSED(ne33);
GGML_UNUSED(nb31); GGML_UNUSED(nb32); GGML_UNUSED(nb33);
GGML_UNUSED_VARS(Q, K, V, mask, sinks, KV_max, dst, dst_meta, scale,
max_bias, m0, m1, n_head_log2, logit_softcap,
ne00, ne01, ne02, ne03,
nb01, nb02, nb03,
ne10, ne11, ne12, ne13,
nb11, nb12, nb13,
nb21, nb22, nb23,
ne31, ne32, ne33,
nb31, nb32, nb33);
NO_DEVICE_CODE;
#endif // defined(FLASH_ATTN_AVAILABLE) && defined(FP16_AVAILABLE)
}
Expand Down
29 changes: 18 additions & 11 deletions ggml/src/ggml-cuda/fattn-vec-f32.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,15 @@ static __global__ void flash_attn_vec_ext_f32(

// Skip unused kernel variants for faster compilation:
if (use_logit_softcap && !(D == 128 || D == 256)) {
GGML_UNUSED_VARS(Q, K, V, mask, sinks, KV_max, dst, dst_meta, scale,
max_bias, m0, m1, n_head_log2, logit_softcap,
ne00, ne01, ne02, ne03,
nb01, nb02, nb03,
ne10, ne11, ne12, ne13,
nb11, nb12, nb13,
nb21, nb22, nb23,
ne31, ne32, ne33,
nb31, nb32, nb33);
NO_DEVICE_CODE;
return;
}
Expand Down Expand Up @@ -334,17 +343,15 @@ static __global__ void flash_attn_vec_ext_f32(
dst_meta[((sequence*ne01 + ic0 + tid)*ne02 + head)*gridDim.y + blockIdx.y] = make_float2(kqmax[tid], kqsum[tid]);
}
#else
GGML_UNUSED(Q); GGML_UNUSED(K); GGML_UNUSED(V); GGML_UNUSED(mask);
GGML_UNUSED(sinks); GGML_UNUSED(KV_max); GGML_UNUSED(dst); GGML_UNUSED(dst_meta);
GGML_UNUSED(scale); GGML_UNUSED(max_bias); GGML_UNUSED(m0); GGML_UNUSED(m1);
GGML_UNUSED(n_head_log2); GGML_UNUSED(logit_softcap);
GGML_UNUSED(ne00); GGML_UNUSED(ne01); GGML_UNUSED(ne02); GGML_UNUSED(ne03);
GGML_UNUSED(ne10); GGML_UNUSED(ne11); GGML_UNUSED(ne12); GGML_UNUSED(ne13);
GGML_UNUSED(ne31); GGML_UNUSED(ne32); GGML_UNUSED(ne33);
GGML_UNUSED(nb31); GGML_UNUSED(nb32); GGML_UNUSED(nb33);
GGML_UNUSED(nb01); GGML_UNUSED(nb02); GGML_UNUSED(nb03);
GGML_UNUSED(nb11); GGML_UNUSED(nb12); GGML_UNUSED(nb13);
GGML_UNUSED(nb21); GGML_UNUSED(nb22); GGML_UNUSED(nb23);
GGML_UNUSED_VARS(Q, K, V, mask, sinks, KV_max, dst, dst_meta, scale,
max_bias, m0, m1, n_head_log2, logit_softcap,
ne00, ne01, ne02, ne03,
nb01, nb02, nb03,
ne10, ne11, ne12, ne13,
nb11, nb12, nb13,
nb21, nb22, nb23,
ne31, ne32, ne33,
nb31, nb32, nb33);
NO_DEVICE_CODE;
#endif // FLASH_ATTN_AVAILABLE
}
Expand Down
19 changes: 9 additions & 10 deletions ggml/src/ggml-cuda/fattn-wmma-f16.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -471,16 +471,15 @@ static __global__ void flash_attn_ext_f16(
dst_meta[j_dst_unrolled] = dst_meta_val;
}
#else
GGML_UNUSED(Q); GGML_UNUSED(K); GGML_UNUSED(V); GGML_UNUSED(mask);
GGML_UNUSED(sinks); GGML_UNUSED(KV_max); GGML_UNUSED(dst); GGML_UNUSED(dst_meta);
GGML_UNUSED(scale); GGML_UNUSED(max_bias); GGML_UNUSED(m0); GGML_UNUSED(m1);
GGML_UNUSED(n_head_log2); GGML_UNUSED(logit_softcap);
GGML_UNUSED(ne00); GGML_UNUSED(ne01); GGML_UNUSED(ne02); GGML_UNUSED(ne03);
GGML_UNUSED(ne10); GGML_UNUSED(ne11); GGML_UNUSED(ne12); GGML_UNUSED(ne13);
GGML_UNUSED(ne31); GGML_UNUSED(ne32); GGML_UNUSED(ne33); GGML_UNUSED(nb31);
GGML_UNUSED(nb32); GGML_UNUSED(nb33); GGML_UNUSED(nb01); GGML_UNUSED(nb02);
GGML_UNUSED(nb03); GGML_UNUSED(nb11); GGML_UNUSED(nb12); GGML_UNUSED(nb13);
GGML_UNUSED(nb21); GGML_UNUSED(nb22); GGML_UNUSED(nb23);
GGML_UNUSED_VARS(Q, K, V, mask, sinks, KV_max, dst, dst_meta, scale,
max_bias, m0, m1, n_head_log2, logit_softcap,
ne00, ne01, ne02, ne03,
nb01, nb02, nb03,
ne10, ne11, ne12, ne13,
nb11, nb12, nb13,
nb21, nb22, nb23,
ne31, ne32, ne33,
nb31, nb32, nb33);
NO_DEVICE_CODE;
#endif // defined(FLASH_ATTN_AVAILABLE) && (__CUDA_ARCH__ == GGML_CUDA_CC_VOLTA || (defined(GGML_HIP_ROCWMMA_FATTN) && defined(FP16_MMA_AVAILABLE)))
}
Expand Down
4 changes: 1 addition & 3 deletions ggml/src/ggml-cuda/ggml-cuda.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1328,9 +1328,7 @@ static void ggml_cuda_op_mul_mat_cublas(
&beta, dst_dd_i, ldc));
}

GGML_UNUSED(dst);
GGML_UNUSED(src1_ddq_i);
GGML_UNUSED(src1_padded_row_size);
GGML_UNUSED_VARS(dst, src1_ddq_i, src1_padded_row_size);
}

static void ggml_cuda_set_peer_access(const int n_tokens, int main_device) {
Expand Down
Loading
Loading