vulkan: add exp operation

[ddwkim]

Make sure to read the contributing guidelines before submitting a PR

We have found that exp operation is not supported while implementing Vocos, a neural vocoder. The exp operation implementation in Vulkan results in expected output from our model, and the test results on Apple M4 MAX and NVIDIA A6000 are as follows.

test-backend-ops results on Apple M4 Max

./build/bin/test-backend-ops -o EXP
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Apple M4 Max (MoltenVK) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 32768 | int dot: 0 | matrix cores: none
Testing 4 devices
...
Backend 1/4: Metal
  Device description: Apple M4 Max
  Device memory: 27648 MB (27642 MB free)

  EXP(type=f16,ne_a=[128,2,2,2],v=0): not supported [Metal]
  EXP(type=f16,ne_a=[5,7,11,13],v=0): not supported [Metal]
  EXP(type=f16,ne_a=[128,2,2,2],v=1): not supported [Metal]
  EXP(type=f16,ne_a=[5,7,11,13],v=1): not supported [Metal]
  EXP(type=f32,ne_a=[128,2,2,2],v=0): OK
  EXP(type=f32,ne_a=[5,7,11,13],v=0): OK
  EXP(type=f32,ne_a=[128,2,2,2],v=1): not supported [Metal]
  EXP(type=f32,ne_a=[5,7,11,13],v=1): not supported [Metal]
  10844/10844 tests passed
  Backend Metal: OK
ggml_metal_free: deallocating
ggml_metal_mem_pool_free: freeing memory pool, num heaps = 0 (total = 0)
ggml_metal_mem_pool_free: freeing memory pool, num heaps = 0 (total = 0)
ggml_metal_mem_pool_free: freeing memory pool, num heaps = 0 (total = 0)
ggml_metal_mem_pool_free: freeing memory pool, num heaps = 0 (total = 0)
ggml_metal_mem_pool_free: freeing memory pool, num heaps = 0 (total = 0)
ggml_metal_mem_pool_free: freeing memory pool, num heaps = 0 (total = 0)
ggml_metal_mem_pool_free: freeing memory pool, num heaps = 0 (total = 0)
ggml_metal_mem_pool_free: freeing memory pool, num heaps = 0 (total = 0)
Backend 2/4: Vulkan0
  Device description: Apple M4 Max
  Device memory: 36864 MB (36864 MB free)

  EXP(type=f16,ne_a=[128,2,2,2],v=0): OK
  EXP(type=f16,ne_a=[5,7,11,13],v=0): OK
  EXP(type=f16,ne_a=[128,2,2,2],v=1): not supported [Vulkan0]
  EXP(type=f16,ne_a=[5,7,11,13],v=1): not supported [Vulkan0]
  EXP(type=f32,ne_a=[128,2,2,2],v=0): OK
  EXP(type=f32,ne_a=[5,7,11,13],v=0): OK
  EXP(type=f32,ne_a=[128,2,2,2],v=1): not supported [Vulkan0]
  EXP(type=f32,ne_a=[5,7,11,13],v=1): not supported [Vulkan0]
  10844/10844 tests passed
  Backend Vulkan0: OK
Backend 3/4: BLAS
  Device description: Accelerate
  Device memory: 0 MB (0 MB free)

  EXP(type=f16,ne_a=[128,2,2,2],v=0): not supported [BLAS]
  EXP(type=f16,ne_a=[5,7,11,13],v=0): not supported [BLAS]
  EXP(type=f16,ne_a=[128,2,2,2],v=1): not supported [BLAS]
  EXP(type=f16,ne_a=[5,7,11,13],v=1): not supported [BLAS]
  EXP(type=f32,ne_a=[128,2,2,2],v=0): not supported [BLAS]
  EXP(type=f32,ne_a=[5,7,11,13],v=0): not supported [BLAS]
  EXP(type=f32,ne_a=[128,2,2,2],v=1): not supported [BLAS]
  EXP(type=f32,ne_a=[5,7,11,13],v=1): not supported [BLAS]
  10844/10844 tests passed
  Backend BLAS: OK
Backend 4/4: CPU
  Skipping CPU backend
4/4 backends passed
OK

test-backend-ops results on NVIDIA RTX A6000

./build/bin/test-backend-ops -o EXP
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:    no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
ggml_cuda_init: found 1 CUDA devices:
  Device 0: NVIDIA RTX A6000, compute capability 8.6, VMM: yes
register_backend: registered backend CUDA (1 devices)
register_device: registered device CUDA0 (NVIDIA RTX A6000)
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = NVIDIA RTX A6000 (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: NV_coopmat2
register_backend: registered backend Vulkan (1 devices)
register_device: registered device Vulkan0 (NVIDIA RTX A6000)
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (AMD Ryzen Threadripper PRO 5975WX 32-Cores)
load_backend: failed to find ggml_backend_init in /home/dongwon/prj/llama.cpp/build/bin/libggml-cuda.so
load_backend: failed to find ggml_backend_init in /home/dongwon/prj/llama.cpp/build/bin/libggml-vulkan.so
load_backend: failed to find ggml_backend_init in /home/dongwon/prj/llama.cpp/build/bin/libggml-cpu.so
Testing 3 devices

Backend 1/3: CUDA0
  Device description: NVIDIA RTX A6000
  Device memory: 48539 MB (48223 MB free)

update_cuda_graph_executable: CUDA graph update failed
ggml_backend_cuda_graph_compute: disabling CUDA graphs due to too many consecutive updates
  EXP(type=f16,ne_a=[128,2,2,2],v=0): OK
  EXP(type=f16,ne_a=[5,7,11,13],v=0): OK
  EXP(type=f16,ne_a=[128,2,2,2],v=1): not supported [CUDA0] 
  EXP(type=f16,ne_a=[5,7,11,13],v=1): not supported [CUDA0] 
  EXP(type=f32,ne_a=[128,2,2,2],v=0): OK
  EXP(type=f32,ne_a=[5,7,11,13],v=0): OK
  EXP(type=f32,ne_a=[128,2,2,2],v=1): not supported [CUDA0] 
  EXP(type=f32,ne_a=[5,7,11,13],v=1): not supported [CUDA0] 
  10844/10844 tests passed
  Backend CUDA0: OK
Backend 2/3: Vulkan0
  Device description: NVIDIA RTX A6000
  Device memory: 49140 MB (49140 MB free)

  EXP(type=f16,ne_a=[128,2,2,2],v=0): OK
  EXP(type=f16,ne_a=[5,7,11,13],v=0): OK
  EXP(type=f16,ne_a=[128,2,2,2],v=1): not supported [Vulkan0] 
  EXP(type=f16,ne_a=[5,7,11,13],v=1): not supported [Vulkan0] 
  EXP(type=f32,ne_a=[128,2,2,2],v=0): OK
  EXP(type=f32,ne_a=[5,7,11,13],v=0): OK
  EXP(type=f32,ne_a=[128,2,2,2],v=1): not supported [Vulkan0] 
  EXP(type=f32,ne_a=[5,7,11,13],v=1): not supported [Vulkan0] 
  10844/10844 tests passed
  Backend Vulkan0: OK
Backend 3/3: CPU
  Skipping CPU backend
3/3 backends passed
OK

[jeffbolznv]

LGTM, and I verified the tests pass on my system.

[0cc4m]

Thank you!