[SPIRV] Track id to variable mapping to be able to use original variable name…

[hernanponcedeleon]

This PR adds mappings between SPIRV ids and original variables names. This allows to write the program header (inputs/outputs/specification) in terms of sourcecode variables. One can still used the SPIRV id so there should not be regressions.

I updated the headers in benchmarks/opencl/benchmarks and benchmarks/opencl/patterns so now those can be passed directly as input to dartagnan, e.g.,

dartagnan cat/vulkan.cat --target=vulkan benchmarks/opencl/benchmarks/caslock.cl           
Test: benchmarks/opencl/benchmarks/caslock.cl
Result: PASS
Time: 0.032 secs

Notice I removed many of the // Currently, clspv inserts Coherent decorations only if ... since this has been fixed in google/clspv#1651.

[github-actions]

Performance comparison

Linux x64

Benchmark details

Memory model: vmm

Benchmark	Base branch	PR branch	Improvement (99% CI)	Result
benchmarks/locks/cna.c	11.375 ± 0.552 s	11.478 ± 0.742 s	➖ -0.9% [-11.0%, +9.2%]	UNKNOWN
benchmarks/locks/mutex_musl.c	24.854 ± 3.624 s	23.641 ± 2.805 s	➖ +4.6% [-20.2%, +29.4%]	UNKNOWN
benchmarks/lfds/dglm.c	19.812 ± 0.668 s	21.147 ± 0.764 s	➖ -6.8% [-39.0%, +25.3%]	UNKNOWN
benchmarks/lfds/ms.c	36.524 ± 0.625 s	35.242 ± 1.666 s	➖ +3.5% [-24.5%, +31.5%]	UNKNOWN
benchmarks/lfds/treiber.c	10.490 ± 0.296 s	10.241 ± 0.419 s	➖ +2.4% [-7.7%, +12.5%]	UNKNOWN
benchmarks/lfds/safe_stack.c	6.444 ± 0.260 s	6.568 ± 0.152 s	➖ -2.0% [-13.6%, +9.6%]	UNKNOWN
benchmarks/challenging/cna.c	32.592 ± 2.975 s	29.770 ± 1.086 s	➖ +8.4% [-20.1%, +36.9%]	UNKNOWN

Memory model: aarch64

Benchmark	Base branch	PR branch	Improvement (99% CI)	Result
benchmarks/locks/linuxrwlock.c	6.093 ± 0.045 s	6.062 ± 0.084 s	➖ +0.5% [-3.6%, +4.6%]	UNKNOWN
benchmarks/challenging/cna.c	13.327 ± 1.030 s	12.398 ± 1.428 s	➖ +7.2% [-14.6%, +28.9%]	UNKNOWN

Memory model: power

Benchmark	Base branch	PR branch	Improvement (99% CI)	Result
benchmarks/locks/linuxrwlock.c	15.195 ± 0.544 s	15.467 ± 0.407 s	➖ -1.8% [-10.5%, +6.9%]	UNKNOWN
benchmarks/locks/mutex_musl.c	14.664 ± 1.109 s	13.656 ± 0.178 s	➖ +6.5% [-37.5%, +50.5%]	UNKNOWN
benchmarks/lfds/ms.c	26.253 ± 2.310 s	23.903 ± 1.594 s	➖ +8.8% [-15.6%, +33.2%]	UNKNOWN
benchmarks/lfds/treiber.c	12.202 ± 0.924 s	12.401 ± 0.629 s	➖ -1.8% [-18.1%, +14.6%]	UNKNOWN

Total

Benchmarks	Base branch	PR branch	Improvement (99% CI)
All reported benchmarks	229.824 ± 4.924 s	221.974 ± 6.015 s	➖ +3.4% [-1.3%, +8.1%]

5 benchmark(s) omitted because both averages were below 5 seconds.

macOS ARM64

Benchmark details

Memory model: vmm

Benchmark	Base branch	PR branch	Improvement (99% CI)	Result
benchmarks/locks/cna.c	24.123 ± 1.580 s	25.043 ± 1.406 s	➖ -4.0% [-42.3%, +34.3%]	UNKNOWN
benchmarks/locks/mutex_musl.c	26.502 ± 0.431 s	28.646 ± 0.123 s	➖ -8.1% [-20.1%, +3.9%]	UNKNOWN
benchmarks/lfds/dglm.c	67.333 ± 4.163 s	88.667 ± 10.263 s	➖ -31.4% [-71.6%, +8.8%]	UNKNOWN
benchmarks/lfds/ms.c	78.000 ± 15.000 s	87.000 ± 5.000 s	➖ -13.5% [-103.6%, +76.5%]	UNKNOWN
benchmarks/lfds/treiber.c	16.770 ± 0.503 s	16.091 ± 0.779 s	➖ +4.1% [-9.8%, +17.9%]	UNKNOWN
benchmarks/lfds/safe_stack.c	11.002 ± 0.430 s	11.466 ± 0.888 s	➖ -4.4% [-59.8%, +51.1%]	UNKNOWN
benchmarks/challenging/cna.c	50.038 ± 2.586 s	47.023 ± 5.788 s	➖ +6.3% [-34.1%, +46.6%]	UNKNOWN

Memory model: aarch64

Benchmark	Base branch	PR branch	Improvement (99% CI)	Result
benchmarks/locks/linuxrwlock.c	20.490 ± 1.179 s	19.401 ± 2.194 s	➖ +5.3% [-47.1%, +57.7%]	UNKNOWN
benchmarks/locks/mutex_musl.c	13.495 ± 2.651 s	16.133 ± 0.436 s	➖ -22.4% [-149.9%, +105.1%]	UNKNOWN
benchmarks/lfds/dglm.c	15.349 ± 4.351 s	17.289 ± 2.714 s	➖ -17.7% [-184.8%, +149.4%]	PASS
benchmarks/lfds/ms.c	20.315 ± 1.940 s	18.829 ± 0.718 s	➖ +6.6% [-59.1%, +72.3%]	UNKNOWN
benchmarks/challenging/cna.c	34.707 ± 4.679 s	39.564 ± 2.855 s	➖ -16.2% [-158.7%, +126.2%]	UNKNOWN
benchmarks/challenging/wsq.c	15.602 ± 1.644 s	14.429 ± 0.554 s	➖ +7.0% [-32.9%, +47.0%]	UNKNOWN

Memory model: power

Benchmark	Base branch	PR branch	Improvement (99% CI)	Result
benchmarks/locks/linuxrwlock.c	67.604 ± 14.741 s	64.667 ± 4.509 s	➖ +0.3% [-166.5%, +167.1%]	UNKNOWN
benchmarks/locks/mutex_musl.c	30.863 ± 6.794 s	32.446 ± 4.154 s	➖ -6.5% [-62.1%, +49.1%]	UNKNOWN
benchmarks/lfds/dglm.c	13.989 ± 0.843 s	15.059 ± 1.017 s	➖ -7.7% [-43.3%, +27.8%]	UNKNOWN
benchmarks/lfds/ms.c	64.393 ± 7.015 s	64.020 ± 10.405 s	➖ +0.8% [-56.7%, +58.3%]	UNKNOWN
benchmarks/lfds/treiber.c	26.331 ± 2.373 s	27.869 ± 2.302 s	➖ -6.2% [-63.3%, +50.9%]	UNKNOWN

Total

Benchmarks	Base branch	PR branch	Improvement (99% CI)
All reported benchmarks	596.906 ± 17.535 s	633.640 ± 20.191 s	➖ -6.2% [-35.5%, +23.1%]

[ThomasHaas]

Oh, you found another usage for Metadata. Now I feel like we should create a top-level metadata package with all metadata classes, seeing that with this PR and #1104 we already have 3 different metadata packages that are quite deeply nested.
Alternatively, we could also put Metadata as inner classes of the class they should be associated with, e.g., MemoryObject.VariableName, Program.SourceLocation, and Wmm.CutAnnotation.