CUDA: Check PTX version on host side to guard PDL dispatch

[ORippler]

Overview

Checking on __CUDA_ARCH_LIST__ alone is insufficient for JIT, as this variable doesn't differentiate between compiling for say sm_90, sm_90a or sm_90f (so forward-jittable PTX vs. arch/family-specific PTX).

Thus, one can have a bug when compiling with
DCMAKE_CUDA_ARCHITECTURES="89;90a", where current code would wrongly dispatch to PDL on sm_90/sm_120 in forward-JIT mode.

This PR fixes this issue by checking cudaFuncAttributes::ptxVersion of the incoming kernel at runtime. A check on ptxVersion alone is sufficient, as device-codes will always be >= ptxVersion (and any violation of this would be a severe bug in CUDA/nvcc), see:
https://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/#gpu-code-code-code

Additional information

Follow-up to #23471 that implements a more complete fix.

To verify, apply

diff --git a/ggml/src/ggml-cuda/common.cuh b/ggml/src/ggml-cuda/common.cuh
index 20b1846c0..19dd53a72 100644
--- a/ggml/src/ggml-cuda/common.cuh
+++ b/ggml/src/ggml-cuda/common.cuh
@@ -1584,6 +1584,9 @@ static bool ggml_cuda_kernel_can_use_pdl(const void * kernel) {
     // We have to guard on a loaded kernel's PTX version so a kernel forward-JIT'ed
     // from pre-Hopper PTX to a Hopper-or-newer GPU does not opt into PDL.
     const bool can_use_pdl = attr.ptxVersion >= 90;
+    printf("Kernel %p on device %d has PTX version %d -> can_use_pdl = %d, highest compiled_arch = %d\n", kernel,
+           device, attr.ptxVersion, can_use_pdl,
+           ggml_cuda_highest_compiled_arch(ggml_cuda_info().devices[ggml_cuda_get_device()].cc));
     cache.emplace(key, can_use_pdl);
     return can_use_pdl;
 }

& compile with something like

cmake -S . -B build_test_ptx -G Ninja \                                
  -DLLAMA_FATAL_WARNINGS=ON \
  -DCMAKE_BUILD_TYPE=Release \
  -DCMAKE_CUDA_ARCHITECTURES="89;90a" \
  -DGGML_CUDA=ON

Running test-backend-ops -o MUL_MAT on a CC > 9.0 GPU will now correctly disable PDL, and show logs like
Kernel 0x7cf22ed84aa0 on device 0 has PTX version 89 -> can_use_pdl = 0, highest compiled_arch = 900

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: YES - paired with codex on this.

[JohannesGaessler]

The convention of the surrounding code is size_t vs. e.g. std::size_t but either way is fine I think.