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 <robdclark@chromium.org>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>

diff --git a/src/compiler/nir/nir_opcodes.py b/src/compiler/nir/nir_opcodes.py
index b759910cd800bb922fdc53adefd5a6d5016414c4..4bb21ea13587632d2c1c4cdd360cdee27720dca4 100644 (file)
--- 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)")