bit_cast operator

[CharlieL7]

• Converts between types while keeping the data the same. Basically an extension of reinterpret_cast without undefined behavior
• Needed for conversion pass between fp8e4m3fn to fp8e4m3fnuz.

[codecov]

Codecov Report

Attention: Patch coverage is 95.00000% with 1 line in your changes missing coverage. Please review.

Project coverage is 92.20%. Comparing base (6b886e3) to head (463379f).
Report is 4 commits behind head on develop.

Files with missing lines	Patch %	Lines
src/include/migraphx/op/bit_cast.hpp	95.00%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #3655   +/-   ##
========================================
  Coverage    92.20%   92.20%           
========================================
  Files          513      514    +1     
  Lines        21658    21678   +20     
========================================
+ Hits         19970    19989   +19     
- Misses        1688     1689    +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[pfultz2]

What is the exact error are you getting when doing only one overload for bit_cast like this?

template <typename To,
          typename From,
          MIGRAPHX_REQUIRES(is_trivially_copyable<To>{} and is_trivially_copyable<From>{})>
inline constexpr auto bit_cast(From fr) noexcept
{
    return vec_transform(fr)([](auto x) -> To {
        static_assert(sizeof(To) == sizeof(decltype(x)));
        return __builtin_bit_cast(To, x);
    });
}

[TedThemistokleous]

Looks good. Just add test cases for THROWs

[migraphx-bot]

Test	Batch	Rate new 0b6cae	Rate old 241e24	Diff	Compare
torchvision-resnet50	64	3,259.55	3,256.24	0.10%	✅
torchvision-resnet50_fp16	64	6,982.07	6,984.56	-0.04%	✅
torchvision-densenet121	32	2,433.16	2,434.44	-0.05%	✅
torchvision-densenet121_fp16	32	4,089.56	4,078.22	0.28%	✅
torchvision-inceptionv3	32	1,629.02	1,629.38	-0.02%	✅
torchvision-inceptionv3_fp16	32	2,746.05	2,742.92	0.11%	✅
cadene-inceptionv4	16	764.56	764.47	0.01%	✅
cadene-resnext64x4	16	809.45	812.01	-0.32%	✅
slim-mobilenet	64	7,455.75	7,468.13	-0.17%	✅
slim-nasnetalarge	64	208.44	208.51	-0.03%	✅
slim-resnet50v2	64	3,440.92	3,440.29	0.02%	✅
bert-mrpc-onnx	8	1,148.71	1,148.20	0.04%	✅
bert-mrpc-tf	1	463.56	464.32	-0.16%	✅
pytorch-examples-wlang-gru	1	418.64	426.54	-1.85%	✅
pytorch-examples-wlang-lstm	1	377.99	380.15	-0.57%	✅
torchvision-resnet50_1	1	784.01	775.13	1.14%	✅
cadene-dpn92_1	1	435.23	403.98	7.73%	🔆
cadene-resnext101_1	1	384.63	383.04	0.42%	✅
onnx-taau-downsample	1	345.21	345.85	-0.19%	✅
dlrm-criteoterabyte	1	33.35	33.33	0.06%	✅
dlrm-criteoterabyte_fp16	1	52.74	52.75	-0.03%	✅
agentmodel	1	10,062.90	8,145.38	23.54%	🔆
unet_fp16	2	58.79	58.93	-0.23%	✅
resnet50v1_fp16	1	948.11	935.00	1.40%	✅
resnet50v1_int8	1	1,020.83	1,023.68	-0.28%	✅
bert_base_cased_fp16	64	1,170.84	1,169.83	0.09%	✅
bert_large_uncased_fp16	32	363.48	363.35	0.03%	✅
bert_large_fp16	1	200.17	195.75	2.26%	✅
distilgpt2_fp16	16	2,194.18	2,196.59	-0.11%	✅
yolov5s	1	526.77	534.53	-1.45%	✅
tinyllama	1	43.59	43.39	0.45%	✅
vicuna-fastchat	1	168.51	171.47	-1.73%	✅
whisper-tiny-encoder	1	418.00	417.49	0.12%	✅
whisper-tiny-decoder	1	433.27	427.46	1.36%	✅

Check results before merge 🔆

[migraphx-bot]

     ✅ bert-mrpc-onnx: PASSED: MIGraphX meets tolerance

     ✅ bert-mrpc-tf: PASSED: MIGraphX meets tolerance

     ✅ pytorch-examples-wlang-gru: PASSED: MIGraphX meets tolerance

     ✅ pytorch-examples-wlang-lstm: PASSED: MIGraphX meets tolerance

     ✅ torchvision-resnet50_1: PASSED: MIGraphX meets tolerance

     ✅ cadene-dpn92_1: PASSED: MIGraphX meets tolerance

     ✅ cadene-resnext101_1: PASSED: MIGraphX meets tolerance

     ✅ dlrm-criteoterabyte: PASSED: MIGraphX meets tolerance

     ✅ agentmodel: PASSED: MIGraphX meets tolerance

     ✅ unet: PASSED: MIGraphX meets tolerance

     ✅ resnet50v1: PASSED: MIGraphX meets tolerance

     ✅ bert_base_cased_fp16: PASSED: MIGraphX meets tolerance

🔴bert_large_uncased_fp16: FAILED: MIGraphX is not within tolerance - check verbose output

     ✅ bert_large: PASSED: MIGraphX meets tolerance

     ✅ yolov5s: PASSED: MIGraphX meets tolerance

     ✅ tinyllama: PASSED: MIGraphX meets tolerance

     ✅ vicuna-fastchat: PASSED: MIGraphX meets tolerance

     ✅ whisper-tiny-encoder: PASSED: MIGraphX meets tolerance

     ✅ whisper-tiny-decoder: PASSED: MIGraphX meets tolerance

     ✅ distilgpt2_fp16: PASSED: MIGraphX meets tolerance