diff options
| author | Alyssa Rosenzweig <[email protected]> | 2020-05-13 13:51:06 -0400 |
|---|---|---|
| committer | Marge Bot <[email protected]> | 2020-05-25 20:03:52 +0000 |
| commit | fcbc022787d4fdfcfdf843d9f720a587e1f0579d (patch) | |
| tree | 7c79ab32faddcb68c8ec17b067c21f1382eaaedf /src/compiler | |
| parent | 46d5b07c5c39d1b8cf10976f6574a63062dea9c4 (diff) | |
nir: Add un/pack_32_4x8 opcodes
Complement the existing un/pack_32_2x16 opcodes. These are useful for
8-bit format packing. On Midgard, they are equivalent to just a 32-bit
move, but other GPUs could lower to other packs if needed.
Signed-off-by: Alyssa Rosenzweig <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5107>
Diffstat (limited to 'src/compiler')
| -rw-r--r-- | src/compiler/nir/nir_opcodes.py | 7 |
1 files changed, 7 insertions, 0 deletions
diff --git a/src/compiler/nir/nir_opcodes.py b/src/compiler/nir/nir_opcodes.py index 03f00d12639..142d7a427d5 100644 --- a/src/compiler/nir/nir_opcodes.py +++ b/src/compiler/nir/nir_opcodes.py @@ -99,6 +99,7 @@ tbool8 = "bool8" tbool16 = "bool16" tbool32 = "bool32" tuint = "uint" +tuint8 = "uint8" tuint16 = "uint16" tfloat16 = "float16" tfloat32 = "float32" @@ -357,6 +358,9 @@ dst.x = (src0.x << 0) | (src0.w << 24); """) +unop_horiz("pack_32_4x8", 1, tuint32, 4, tuint8, + "dst.x = src0.x | ((uint32_t)src0.y << 8) | ((uint32_t)src0.z << 16) | ((uint32_t)src0.w << 24);") + unop_horiz("pack_32_2x16", 1, tuint32, 2, tuint16, "dst.x = src0.x | ((uint32_t)src0.y << 16);") @@ -375,6 +379,9 @@ unop_horiz("unpack_64_4x16", 4, tuint16, 1, tuint64, unop_horiz("unpack_32_2x16", 2, tuint16, 1, tuint32, "dst.x = src0.x; dst.y = src0.x >> 16;") +unop_horiz("unpack_32_4x8", 4, tuint8, 1, tuint32, + "dst.x = src0.x; dst.y = src0.x >> 8; dst.z = src0.x >> 16; dst.w = src0.x >> 24;") + unop_horiz("unpack_half_2x16_flush_to_zero", 2, tfloat32, 1, tuint32, """ dst.x = unpack_half_1x16_flush_to_zero((uint16_t)(src0.x & 0xffff)); dst.y = unpack_half_1x16_flush_to_zero((uint16_t)(src0.x << 16)); |