Skip to content

Commit 5581c34

Browse files
author
Jakub Chlanda
authored
[SYCL] Correctly spell out SM version macro when AOT compiling (#15615)
There was a typo resulting in the compiler using `*_SM[0-9][9-9]__` format when the SM was not supported: `#define __SYCL_TARGET_NVIDIA_GPU_SM50__ 0` vs: `#define __SYCL_TARGET_NVIDIA_GPU_SM_50__ 1` for the supported SM version. Prefer the underscore (`*_SM_[0-9][9-9]__`), as this is what CUDA uses, for example when specifying `gpu-name` switch for `ptxas`.
1 parent 0d94c74 commit 5581c34

File tree

2 files changed

+71
-71
lines changed

2 files changed

+71
-71
lines changed

‎sycl/doc/design/DeviceIf.md‎

Lines changed: 15 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -169,21 +169,21 @@ one of the following corresponding C++ macro names:
169169
* `__SYCL_TARGET_INTEL_GPU_ACM_G11__`
170170
* `__SYCL_TARGET_INTEL_GPU_ACM_G12__`
171171
* `__SYCL_TARGET_INTEL_GPU_PVC__`
172-
* `__SYCL_TARGET_NVIDIA_GPU_SM50__`
173-
* `__SYCL_TARGET_NVIDIA_GPU_SM52__`
174-
* `__SYCL_TARGET_NVIDIA_GPU_SM53__`
175-
* `__SYCL_TARGET_NVIDIA_GPU_SM60__`
176-
* `__SYCL_TARGET_NVIDIA_GPU_SM61__`
177-
* `__SYCL_TARGET_NVIDIA_GPU_SM62__`
178-
* `__SYCL_TARGET_NVIDIA_GPU_SM70__`
179-
* `__SYCL_TARGET_NVIDIA_GPU_SM72__`
180-
* `__SYCL_TARGET_NVIDIA_GPU_SM75__`
181-
* `__SYCL_TARGET_NVIDIA_GPU_SM80__`
182-
* `__SYCL_TARGET_NVIDIA_GPU_SM86__`
183-
* `__SYCL_TARGET_NVIDIA_GPU_SM87__`
184-
* `__SYCL_TARGET_NVIDIA_GPU_SM89__`
185-
* `__SYCL_TARGET_NVIDIA_GPU_SM90__`
186-
* `__SYCL_TARGET_NVIDIA_GPU_SM90A__`
172+
* `__SYCL_TARGET_NVIDIA_GPU_SM_50__`
173+
* `__SYCL_TARGET_NVIDIA_GPU_SM_52__`
174+
* `__SYCL_TARGET_NVIDIA_GPU_SM_53__`
175+
* `__SYCL_TARGET_NVIDIA_GPU_SM_60__`
176+
* `__SYCL_TARGET_NVIDIA_GPU_SM_61__`
177+
* `__SYCL_TARGET_NVIDIA_GPU_SM_62__`
178+
* `__SYCL_TARGET_NVIDIA_GPU_SM_70__`
179+
* `__SYCL_TARGET_NVIDIA_GPU_SM_72__`
180+
* `__SYCL_TARGET_NVIDIA_GPU_SM_75__`
181+
* `__SYCL_TARGET_NVIDIA_GPU_SM_80__`
182+
* `__SYCL_TARGET_NVIDIA_GPU_SM_86__`
183+
* `__SYCL_TARGET_NVIDIA_GPU_SM_87__`
184+
* `__SYCL_TARGET_NVIDIA_GPU_SM_89__`
185+
* `__SYCL_TARGET_NVIDIA_GPU_SM_90__`
186+
* `__SYCL_TARGET_NVIDIA_GPU_SM_90A__`
187187
* `__SYCL_TARGET_AMD_GPU_GFX700__`
188188
* `__SYCL_TARGET_AMD_GPU_GFX701__`
189189
* `__SYCL_TARGET_AMD_GPU_GFX702__`

‎sycl/include/sycl/ext/oneapi/experimental/device_architecture.hpp‎

Lines changed: 56 additions & 56 deletions
Original file line numberDiff line numberDiff line change
@@ -152,47 +152,47 @@ static constexpr ext::oneapi::experimental::architecture
152152
#ifndef __SYCL_TARGET_INTEL_GPU_LNL_M__
153153
#define __SYCL_TARGET_INTEL_GPU_LNL_M__ 0
154154
#endif
155-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM50__
156-
#define __SYCL_TARGET_NVIDIA_GPU_SM50__ 0
155+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_50__
156+
#define __SYCL_TARGET_NVIDIA_GPU_SM_50__ 0
157157
#endif
158-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM52__
159-
#define __SYCL_TARGET_NVIDIA_GPU_SM52__ 0
158+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_52__
159+
#define __SYCL_TARGET_NVIDIA_GPU_SM_52__ 0
160160
#endif
161-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM53__
162-
#define __SYCL_TARGET_NVIDIA_GPU_SM53__ 0
161+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_53__
162+
#define __SYCL_TARGET_NVIDIA_GPU_SM_53__ 0
163163
#endif
164-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM60__
165-
#define __SYCL_TARGET_NVIDIA_GPU_SM60__ 0
164+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_60__
165+
#define __SYCL_TARGET_NVIDIA_GPU_SM_60__ 0
166166
#endif
167-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM61__
168-
#define __SYCL_TARGET_NVIDIA_GPU_SM61__ 0
167+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_61__
168+
#define __SYCL_TARGET_NVIDIA_GPU_SM_61__ 0
169169
#endif
170-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM62__
171-
#define __SYCL_TARGET_NVIDIA_GPU_SM62__ 0
170+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_62__
171+
#define __SYCL_TARGET_NVIDIA_GPU_SM_62__ 0
172172
#endif
173-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM70__
174-
#define __SYCL_TARGET_NVIDIA_GPU_SM70__ 0
173+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_70__
174+
#define __SYCL_TARGET_NVIDIA_GPU_SM_70__ 0
175175
#endif
176-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM72__
177-
#define __SYCL_TARGET_NVIDIA_GPU_SM72__ 0
176+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_72__
177+
#define __SYCL_TARGET_NVIDIA_GPU_SM_72__ 0
178178
#endif
179-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM75__
180-
#define __SYCL_TARGET_NVIDIA_GPU_SM75__ 0
179+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_75__
180+
#define __SYCL_TARGET_NVIDIA_GPU_SM_75__ 0
181181
#endif
182-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM80__
183-
#define __SYCL_TARGET_NVIDIA_GPU_SM80__ 0
182+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_80__
183+
#define __SYCL_TARGET_NVIDIA_GPU_SM_80__ 0
184184
#endif
185-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM86__
186-
#define __SYCL_TARGET_NVIDIA_GPU_SM86__ 0
185+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_86__
186+
#define __SYCL_TARGET_NVIDIA_GPU_SM_86__ 0
187187
#endif
188-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM87__
189-
#define __SYCL_TARGET_NVIDIA_GPU_SM87__ 0
188+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_87__
189+
#define __SYCL_TARGET_NVIDIA_GPU_SM_87__ 0
190190
#endif
191-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM89__
192-
#define __SYCL_TARGET_NVIDIA_GPU_SM89__ 0
191+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_89__
192+
#define __SYCL_TARGET_NVIDIA_GPU_SM_89__ 0
193193
#endif
194-
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM90__
195-
#define __SYCL_TARGET_NVIDIA_GPU_SM90__ 0
194+
#ifndef __SYCL_TARGET_NVIDIA_GPU_SM_90__
195+
#define __SYCL_TARGET_NVIDIA_GPU_SM_90__ 0
196196
#endif
197197
#ifndef __SYCL_TARGET_AMD_GPU_GFX700__
198198
#define __SYCL_TARGET_AMD_GPU_GFX700__ 0
@@ -343,20 +343,20 @@ static constexpr bool is_allowable_aot_mode =
343343
(__SYCL_TARGET_INTEL_GPU_ARL_H__ == 1) ||
344344
(__SYCL_TARGET_INTEL_GPU_BMG_G21__ == 1) ||
345345
(__SYCL_TARGET_INTEL_GPU_LNL_M__ == 1) ||
346-
(__SYCL_TARGET_NVIDIA_GPU_SM50__ == 1) ||
347-
(__SYCL_TARGET_NVIDIA_GPU_SM52__ == 1) ||
348-
(__SYCL_TARGET_NVIDIA_GPU_SM53__ == 1) ||
349-
(__SYCL_TARGET_NVIDIA_GPU_SM60__ == 1) ||
350-
(__SYCL_TARGET_NVIDIA_GPU_SM61__ == 1) ||
351-
(__SYCL_TARGET_NVIDIA_GPU_SM62__ == 1) ||
352-
(__SYCL_TARGET_NVIDIA_GPU_SM70__ == 1) ||
353-
(__SYCL_TARGET_NVIDIA_GPU_SM72__ == 1) ||
354-
(__SYCL_TARGET_NVIDIA_GPU_SM75__ == 1) ||
355-
(__SYCL_TARGET_NVIDIA_GPU_SM80__ == 1) ||
356-
(__SYCL_TARGET_NVIDIA_GPU_SM86__ == 1) ||
357-
(__SYCL_TARGET_NVIDIA_GPU_SM87__ == 1) ||
358-
(__SYCL_TARGET_NVIDIA_GPU_SM89__ == 1) ||
359-
(__SYCL_TARGET_NVIDIA_GPU_SM90__ == 1) ||
346+
(__SYCL_TARGET_NVIDIA_GPU_SM_50__ == 1) ||
347+
(__SYCL_TARGET_NVIDIA_GPU_SM_52__ == 1) ||
348+
(__SYCL_TARGET_NVIDIA_GPU_SM_53__ == 1) ||
349+
(__SYCL_TARGET_NVIDIA_GPU_SM_60__ == 1) ||
350+
(__SYCL_TARGET_NVIDIA_GPU_SM_61__ == 1) ||
351+
(__SYCL_TARGET_NVIDIA_GPU_SM_62__ == 1) ||
352+
(__SYCL_TARGET_NVIDIA_GPU_SM_70__ == 1) ||
353+
(__SYCL_TARGET_NVIDIA_GPU_SM_72__ == 1) ||
354+
(__SYCL_TARGET_NVIDIA_GPU_SM_75__ == 1) ||
355+
(__SYCL_TARGET_NVIDIA_GPU_SM_80__ == 1) ||
356+
(__SYCL_TARGET_NVIDIA_GPU_SM_86__ == 1) ||
357+
(__SYCL_TARGET_NVIDIA_GPU_SM_87__ == 1) ||
358+
(__SYCL_TARGET_NVIDIA_GPU_SM_89__ == 1) ||
359+
(__SYCL_TARGET_NVIDIA_GPU_SM_90__ == 1) ||
360360
(__SYCL_TARGET_AMD_GPU_GFX700__ == 1) ||
361361
(__SYCL_TARGET_AMD_GPU_GFX701__ == 1) ||
362362
(__SYCL_TARGET_AMD_GPU_GFX702__ == 1) ||
@@ -487,46 +487,46 @@ get_current_architecture_aot() {
487487
#if __SYCL_TARGET_INTEL_GPU_LNL_M__
488488
return ext::oneapi::experimental::architecture::intel_gpu_lnl_m;
489489
#endif
490-
#if __SYCL_TARGET_NVIDIA_GPU_SM50__
490+
#if __SYCL_TARGET_NVIDIA_GPU_SM_50__
491491
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_50;
492492
#endif
493-
#if __SYCL_TARGET_NVIDIA_GPU_SM52__
493+
#if __SYCL_TARGET_NVIDIA_GPU_SM_52__
494494
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_52;
495495
#endif
496-
#if __SYCL_TARGET_NVIDIA_GPU_SM53__
496+
#if __SYCL_TARGET_NVIDIA_GPU_SM_53__
497497
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_53;
498498
#endif
499-
#if __SYCL_TARGET_NVIDIA_GPU_SM60__
499+
#if __SYCL_TARGET_NVIDIA_GPU_SM_60__
500500
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_60;
501501
#endif
502-
#if __SYCL_TARGET_NVIDIA_GPU_SM61__
502+
#if __SYCL_TARGET_NVIDIA_GPU_SM_61__
503503
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_61;
504504
#endif
505-
#if __SYCL_TARGET_NVIDIA_GPU_SM62__
505+
#if __SYCL_TARGET_NVIDIA_GPU_SM_62__
506506
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_62;
507507
#endif
508-
#if __SYCL_TARGET_NVIDIA_GPU_SM70__
508+
#if __SYCL_TARGET_NVIDIA_GPU_SM_70__
509509
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_70;
510510
#endif
511-
#if __SYCL_TARGET_NVIDIA_GPU_SM72__
511+
#if __SYCL_TARGET_NVIDIA_GPU_SM_72__
512512
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_72;
513513
#endif
514-
#if __SYCL_TARGET_NVIDIA_GPU_SM75__
514+
#if __SYCL_TARGET_NVIDIA_GPU_SM_75__
515515
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_75;
516516
#endif
517-
#if __SYCL_TARGET_NVIDIA_GPU_SM80__
517+
#if __SYCL_TARGET_NVIDIA_GPU_SM_80__
518518
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_80;
519519
#endif
520-
#if __SYCL_TARGET_NVIDIA_GPU_SM86__
520+
#if __SYCL_TARGET_NVIDIA_GPU_SM_86__
521521
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_86;
522522
#endif
523-
#if __SYCL_TARGET_NVIDIA_GPU_SM87__
523+
#if __SYCL_TARGET_NVIDIA_GPU_SM_87__
524524
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_87;
525525
#endif
526-
#if __SYCL_TARGET_NVIDIA_GPU_SM89__
526+
#if __SYCL_TARGET_NVIDIA_GPU_SM_89__
527527
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_89;
528528
#endif
529-
#if __SYCL_TARGET_NVIDIA_GPU_SM90__
529+
#if __SYCL_TARGET_NVIDIA_GPU_SM_90__
530530
return ext::oneapi::experimental::architecture::nvidia_gpu_sm_90;
531531
#endif
532532
#if __SYCL_TARGET_AMD_GPU_GFX700__

0 commit comments

Comments
 (0)