Commit 94a61241 authored by Erik de Castro Lopo's avatar Erik de Castro Lopo

libFLAC: Add a workaround for a bug in MSVC2105 update2

MSVC2105 update2 compiles the C code:

    abs_residual_partition_sums[partition] =
                  (FLAC__uint32)_mm_cvtsi128_si32(mm_sum);

into this:

    movq    QWORD PTR [rsi], xmm2

while it should be:

    movd    eax, xmm2
    mov     QWORD PTR [rsi], rax

With this patch, MSVC emits:

    movq    QWORD PTR [rsi], xmm2
    mov     DWORD PTR [rsi+4], r9d

so the price of this workaround is 1 extra write instruction per
partition.

Patch-from: lvqcl <lvqcl.mail@gmail.com>
parent 387992bd
......@@ -84,6 +84,10 @@ void FLAC__precompute_partition_info_sums_intrin_avx2(const FLAC__int32 residual
sum128 = _mm_hadd_epi32(sum128, sum128);
sum128 = _mm_hadd_epi32(sum128, sum128);
abs_residual_partition_sums[partition] = (FLAC__uint32)_mm_cvtsi128_si32(sum128);
/* workaround for a bug in MSVC2015U2 - see https://connect.microsoft.com/VisualStudio/feedback/details/2659191/incorrect-code-generation-for-x86-64 */
#if (defined _MSC_VER) && (_MSC_FULL_VER == 190023918) && (defined FLAC__CPU_X86_64)
abs_residual_partition_sums[partition] &= 0xFFFFFFFF; /**/
#endif
}
}
else { /* have to pessimistically use 64 bits for accumulator */
......
......@@ -98,6 +98,10 @@ void FLAC__precompute_partition_info_sums_intrin_sse2(const FLAC__int32 residual
mm_sum = _mm_add_epi32(mm_sum, _mm_srli_si128(mm_sum, 8));
mm_sum = _mm_add_epi32(mm_sum, _mm_srli_si128(mm_sum, 4));
abs_residual_partition_sums[partition] = (FLAC__uint32)_mm_cvtsi128_si32(mm_sum);
/* workaround for a bug in MSVC2015U2 - see https://connect.microsoft.com/VisualStudio/feedback/details/2659191/incorrect-code-generation-for-x86-64 */
#if (defined _MSC_VER) && (_MSC_FULL_VER == 190023918) && (defined FLAC__CPU_X86_64)
abs_residual_partition_sums[partition] &= 0xFFFFFFFF;
#endif
}
}
else { /* have to pessimistically use 64 bits for accumulator */
......
......@@ -87,6 +87,10 @@ void FLAC__precompute_partition_info_sums_intrin_ssse3(const FLAC__int32 residua
mm_sum = _mm_hadd_epi32(mm_sum, mm_sum);
mm_sum = _mm_hadd_epi32(mm_sum, mm_sum);
abs_residual_partition_sums[partition] = (FLAC__uint32)_mm_cvtsi128_si32(mm_sum);
/* workaround for a bug in MSVC2015U2 - see https://connect.microsoft.com/VisualStudio/feedback/details/2659191/incorrect-code-generation-for-x86-64 */
#if (defined _MSC_VER) && (_MSC_FULL_VER == 190023918) && (defined FLAC__CPU_X86_64)
abs_residual_partition_sums[partition] &= 0xFFFFFFFF;
#endif
}
}
else { /* have to pessimistically use 64 bits for accumulator */
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment