diff options
author | Ian Romanick <[email protected]> | 2008-02-11 15:27:13 -0800 |
---|---|---|
committer | Ian Romanick <[email protected]> | 2008-02-14 10:08:48 -0800 |
commit | c2903679856856e3758ceb744cd7d91af9e2eb45 (patch) | |
tree | 2140a3e63b0ce009b5331003576417ab6f028549 /src | |
parent | 0230c56ed0db16f007e4d1881554c7dbfa3ac3de (diff) |
Vectorize a couple fetch functions
These C-coded vectorized functions are expected to be short lived.
They're basically a proof-of-concept for dynamically generated fetch
routines.
Diffstat (limited to 'src')
-rw-r--r-- | src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 46 |
1 files changed, 42 insertions, 4 deletions
diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index 6e86a919ce4..ec10bb99df0 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -46,6 +46,48 @@ static const vec_float4 defaults = { 0.0, 0.0, 0.0, 1.0 }; +static INLINE qword +fetch_unaligned_qword(const void *ptr) +{ + const int shift = (unsigned)(ptr) & 0x0f; + const qword x = *(qword *)(ptr); + const qword y = *(qword *)(ptr + 16); + + return si_or((qword) spu_slqwbyte(x, shift), + (qword) spu_rlmaskqwbyte(y, shift - 16)); +} + +static qword +fetch_R32G32B32A32_FLOAT(const void *ptr) +{ + return fetch_unaligned_qword(ptr); +} + + +static qword +fetch_R32G32B32A32_USCALED(const void *ptr) +{ + return si_cuflt(fetch_unaligned_qword(ptr), 0); +} + + +static qword +fetch_R32G32B32A32_UNORM(const void *ptr) +{ + qword x = si_cuflt(fetch_unaligned_qword(ptr), 0); + vec_float4 scale = spu_splats(1.0f / 255.0f); + + return si_fm(x, (qword) scale); +} + + +static qword +fetch_R32G32B32A32_SSCALED(const void *ptr) +{ + return si_csflt(fetch_unaligned_qword(ptr), 0); +} + + /** * Fetch a float[4] vertex attribute from memory, doing format/type * conversion as needed. @@ -90,22 +132,18 @@ FETCH_ATTRIB( R64G64B64_FLOAT, 3, CVT_64_FLOAT ) FETCH_ATTRIB( R64G64_FLOAT, 2, CVT_64_FLOAT ) FETCH_ATTRIB( R64_FLOAT, 1, CVT_64_FLOAT ) -FETCH_ATTRIB( R32G32B32A32_FLOAT, 4, CVT_32_FLOAT ) FETCH_ATTRIB( R32G32B32_FLOAT, 3, CVT_32_FLOAT ) FETCH_ATTRIB( R32G32_FLOAT, 2, CVT_32_FLOAT ) FETCH_ATTRIB( R32_FLOAT, 1, CVT_32_FLOAT ) -FETCH_ATTRIB( R32G32B32A32_USCALED, 4, CVT_32_USCALED ) FETCH_ATTRIB( R32G32B32_USCALED, 3, CVT_32_USCALED ) FETCH_ATTRIB( R32G32_USCALED, 2, CVT_32_USCALED ) FETCH_ATTRIB( R32_USCALED, 1, CVT_32_USCALED ) -FETCH_ATTRIB( R32G32B32A32_SSCALED, 4, CVT_32_SSCALED ) FETCH_ATTRIB( R32G32B32_SSCALED, 3, CVT_32_SSCALED ) FETCH_ATTRIB( R32G32_SSCALED, 2, CVT_32_SSCALED ) FETCH_ATTRIB( R32_SSCALED, 1, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32B32A32_UNORM, 4, CVT_32_UNORM ) FETCH_ATTRIB( R32G32B32_UNORM, 3, CVT_32_UNORM ) FETCH_ATTRIB( R32G32_UNORM, 2, CVT_32_UNORM ) FETCH_ATTRIB( R32_UNORM, 1, CVT_32_UNORM ) |