vulkan: add dryrun support to sin and cos ops (ggml/947)

sin and cos failed test-backend-ops because they
tried to dereference a context pointer that is null
on dry runs.

This commit prevents that segfault.

Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
This commit is contained in:
Salvatore Mesoraca 2024-09-06 14:34:25 +02:00 committed by Georgi Gerganov
parent 9cb9260861
commit 406c1a32a1

View File

@ -4616,7 +4616,7 @@ static void ggml_vk_sqr(ggml_backend_vk_context * ctx, vk_context& subctx, const
}, dryrun); }, dryrun);
} }
static void ggml_vk_sin(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst) { static void ggml_vk_sin(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst, bool dryrun = false) {
const uint32_t src0_type_size = ggml_type_size(src0->type); const uint32_t src0_type_size = ggml_type_size(src0->type);
const uint32_t dst_type_size = ggml_type_size(dst->type); const uint32_t dst_type_size = ggml_type_size(dst->type);
@ -4626,10 +4626,10 @@ static void ggml_vk_sin(ggml_backend_vk_context * ctx, vk_context& subctx, const
(uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2], (uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size, (uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2], (uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size,
0, 0,
0.0f, 0.0f, 0.0f, 0.0f,
}); }, dryrun);
} }
static void ggml_vk_cos(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst) { static void ggml_vk_cos(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst, bool dryrun = false) {
const uint32_t src0_type_size = ggml_type_size(src0->type); const uint32_t src0_type_size = ggml_type_size(src0->type);
const uint32_t dst_type_size = ggml_type_size(dst->type); const uint32_t dst_type_size = ggml_type_size(dst->type);
@ -4639,7 +4639,7 @@ static void ggml_vk_cos(ggml_backend_vk_context * ctx, vk_context& subctx, const
(uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2], (uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size, (uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2], (uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size,
0, 0,
0.0f, 0.0f, 0.0f, 0.0f,
}); }, dryrun);
} }
static void ggml_vk_clamp(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst, bool dryrun = false) { static void ggml_vk_clamp(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst, bool dryrun = false) {
@ -5783,11 +5783,11 @@ static void ggml_vk_build_graph(ggml_backend_vk_context * ctx, ggml_tensor * nod
break; break;
case GGML_OP_SIN: case GGML_OP_SIN:
ggml_vk_sin(ctx, compute_ctx, src0, node); ggml_vk_sin(ctx, compute_ctx, src0, node, dryrun);
break; break;
case GGML_OP_COS: case GGML_OP_COS:
ggml_vk_cos(ctx, compute_ctx, src0, node); ggml_vk_cos(ctx, compute_ctx, src0, node, dryrun);
break; break;
case GGML_OP_CLAMP: case GGML_OP_CLAMP: