diff options
| author | Rob Clark <[email protected]> | 2019-09-25 10:10:39 -0700 |
|---|---|---|
| committer | Rob Clark <[email protected]> | 2019-10-18 15:08:54 -0700 |
| commit | 0568761f8e7b9249ec8cffbd7826c24e38bf16d6 (patch) | |
| tree | 174362a432ebe93bf918d9461264a3ffbfec80cd | |
| parent | bc2ccdc45ab6de74d72412244984b4551ece855c (diff) | |
nir: Add a new ALU nir_op_imul24
Some hardware can do 24b multiply in a single instruction, but not 32b.
However in most cases 24b is sufficient for address/offset calculation.
Signed-off-by: Rob Clark <[email protected]>
Acked-by: Jason Ekstrand <[email protected]>
Reviewed-by: Kristian H. Kristensen <[email protected]>
Reviewed-by: Eduardo Lima Mitev <[email protected]>
| -rw-r--r-- | src/compiler/nir/nir_opcodes.py | 3 |
1 files changed, 3 insertions, 0 deletions
diff --git a/src/compiler/nir/nir_opcodes.py b/src/compiler/nir/nir_opcodes.py index b759910cd80..4bb21ea1358 100644 --- a/src/compiler/nir/nir_opcodes.py +++ b/src/compiler/nir/nir_opcodes.py @@ -1061,3 +1061,6 @@ dst.x = ((((src0.x & 0xffff0000) >> 16) * (src1.x & 0x0000ffff)) << 16) + src2.x triop("imad24_ir3", tint32, _2src_commutative, "(((int32_t)src0 << 8) >> 8) * (((int32_t)src1 << 8) >> 8) + src2") +# 24b multiply into 32b result (with sign extension) +binop("imul24", tint32, _2src_commutative + associative, + "(((int32_t)src0 << 8) >> 8) * (((int32_t)src1 << 8) >> 8)") |