diff options
| author | Ray Molenkamp <github@lazydodo.com> | 2020-06-24 19:42:00 +0300 |
|---|---|---|
| committer | Ray Molenkamp <github@lazydodo.com> | 2020-06-24 19:42:00 +0300 |
| commit | 5cfbc722d095664c3d2496f21d9deb21c47f2e12 (patch) | |
| tree | c2faceea65da47da56753690393f33a5802c2a50 /intern | |
| parent | c2ab069dfca3c2c82313d7f5d2fa2c0a97dbe514 (diff) | |
Fix T78047: Fix failing denoiser tests on windows
When we switched to MSVC2019 and C++17 we seemingly
managed to trigger a code-gen bug with MSVC in the
AVX code-path.
This change works around the issue by (hopefully
temporary) disabling the optimizer for the fast_exp2f4
function, given it is only used in a single pass
of the denoiser and nowhere else, this is luckily
not as bad as it could have been.
Once the compiler is fixed or a different fix is
available we'll have to revisit this.
Details and link to the repro posted to MS is
available in T78047
Diffstat (limited to 'intern')
| -rw-r--r-- | intern/cycles/util/util_math_fast.h | 8 |
1 files changed, 8 insertions, 0 deletions
diff --git a/intern/cycles/util/util_math_fast.h b/intern/cycles/util/util_math_fast.h index dbed83ab84d..1e0792859a2 100644 --- a/intern/cycles/util/util_math_fast.h +++ b/intern/cycles/util/util_math_fast.h @@ -446,6 +446,11 @@ ccl_device_inline float fast_expf(float x) } #ifndef __KERNEL_GPU__ +/* MSVC seems to have a codegen bug here in atleast SSE41/AVX + * see T78047 for details. */ +#ifdef _MSC_VER +# pragma optimize( "", off ) +#endif ccl_device float4 fast_exp2f4(float4 x) { const float4 one = make_float4(1.0f); @@ -461,6 +466,9 @@ ccl_device float4 fast_exp2f4(float4 x) r = madd4(x, r, make_float4(1.0f)); return __int4_as_float4(__float4_as_int4(r) + (m << 23)); } +#ifdef _MSC_VER +# pragma optimize( "", on ) +#endif ccl_device_inline float4 fast_expf4(float4 x) { |