feat: add native support for TPUs and multi-vendor hardware accelerators

plugins/nf-k8s/src/main/nextflow/k8s/model/PodSpecBuilder.groovy

@@ -680,12 +680,20 @@ class PodSpecBuilder {
     @PackageScope
     String getAcceleratorType(AcceleratorResource accelerator) {
 
-        def type = accelerator.type ?: 'nvidia.com'
+        // Default to standard NVIDIA GPU if left entirely blank.
+        def type = accelerator.type?.toLowerCase() ?: 'nvidia.com'
 
         if ( type.contains('/') )
             // Assume the user has fully specified the resource type.
             return type
 
+        // Map common vendor shorthands to their standard K8s Extended Resource strings.
+        if (type =~ /\b(nvidia|tesla|ampere|h100|a100)\b/)  return 'nvidia.com/gpu'
+        if (type =~ /\b(amd|radeon|instinct)\b/)            return 'amd.com/gpu'
+        if (type =~ /\b(tpu|google)\b/)                     return 'google.com/tpu'
+        if (type =~ /\b(neuron|inferentia|trainium|aws)\b/) return 'aws.amazon.com/neuron'
+        if (type =~ /\b(intel|gaudi)\b/)                    return 'gpu.intel.com/i915'
+
         // Assume we're using GPU and update as necessary.
         if( !type.contains('.') ) type += '.com'
         type += '/gpu'

plugins/nf-k8s/src/test/nextflow/k8s/model/PodSpecBuilderTest.groovy

@@ -613,47 +613,24 @@ class PodSpecBuilderTest extends Specification {
         given:
         def builder = new PodSpecBuilder()
 
-        when:
-        def res = builder.addAcceleratorResources(new AcceleratorResource(request:2, limit: 5), null)
-        then:
-        res.requests == ['nvidia.com/gpu': 2]
-        res.limits == ['nvidia.com/gpu': 5]
-
-        when:
-        res = builder.addAcceleratorResources(new AcceleratorResource(limit: 5, type:'foo'), null)
-        then:
-        res.requests == ['foo.com/gpu': 5]
-        res.limits == ['foo.com/gpu': 5]
-
-        when:
-        res = builder.addAcceleratorResources(new AcceleratorResource(request: 5, type:'foo.org'), null)
-        then:
-        res.requests == ['foo.org/gpu': 5]
-        res.limits == null
-
-        when:
-        res = builder.addAcceleratorResources(new AcceleratorResource(request: 5, type: 'foo.org'), [requests: [cpu: 2]])
-        then:
-        res.requests == [cpu: 2, 'foo.org/gpu': 5]
-        res.limits == null
-
-        when:
-        res = builder.addAcceleratorResources(new AcceleratorResource(request: 5, limit: 10, type: 'foo.org'), [requests: [cpu: 2]])
-        then:
-        res.requests == [cpu: 2, 'foo.org/gpu': 5]
-        res.limits == ['foo.org/gpu': 10]
-
-        when:
-        res = builder.addAcceleratorResources(new AcceleratorResource(request: 5, type:'example.com/fpga'), null)
-        then:
-        res.requests == ['example.com/fpga': 5]
-        res.limits == null
+        expect:
+        def res = builder.addAcceleratorResources(new AcceleratorResource(request: req, limit: lim, type: inputType), existing)
+        res.requests == expectedReq
+        res.limits == expectedLim
 
-        when:
-        res = builder.addAcceleratorResources(new AcceleratorResource(request: 5, limit: 10, type: 'example.com/fpga'), [requests: [cpu: 2]])
-        then:
-        res.requests == [cpu: 2, 'example.com/fpga': 5]
-        res.limits == ['example.com/fpga': 10]
+        where:
+        inputType              | req | lim | existing              | expectedReq                        | expectedLim
+        null                   | 2   | 5   | null                  | ['nvidia.com/gpu': 2]              | ['nvidia.com/gpu': 5]
+        'foo'                  | null| 5   | null                  | ['foo.com/gpu': 5]                 | ['foo.com/gpu': 5]
+        'foo.org'              | 5   | null| null                  | ['foo.org/gpu': 5]                 | null
+        'foo.org'              | 5   | null| [requests: [cpu: 2]]  | [cpu: 2, 'foo.org/gpu': 5]         | null
+        'foo.org'              | 5   | 10  | [requests: [cpu: 2]]  | [cpu: 2, 'foo.org/gpu': 5]         | ['foo.org/gpu': 10]
+        'example.com/fpga'     | 5   | null| null                  | ['example.com/fpga': 5]            | null
+        'nvidia-tesla-k80'     | 4   | 4   | null                  | ['nvidia.com/gpu': 4]              | ['nvidia.com/gpu': 4]
+        'tpu-v5-lite-podslice' | 8   | 8   | null                  | ['google.com/tpu': 8]              | ['google.com/tpu': 8]
+        'amd-instinct-mi300'   | 2   | 2   | null                  | ['amd.com/gpu': 2]                 | ['amd.com/gpu': 2]
+        'aws-trn1-32xlarge'    | 8   | 8   | null                  | ['aws.amazon.com/neuron': 8]       | ['aws.amazon.com/neuron': 8]
+        'intel-gaudi-3'        | 1   | 1   | null                  | ['gpu.intel.com/i915': 1]          | ['gpu.intel.com/i915': 1]
     }
 
     def 'should add resources limits' () {