diff options
| author | Ilia Mirkin <imirkin@alum.mit.edu> | 2014-10-01 01:13:38 -0400 |
|---|---|---|
| committer | Emil Velikov <emil.l.velikov@gmail.com> | 2014-10-29 17:44:21 +0000 |
| commit | 5bc1397bdaf29670a2924b4523cbf418786c2197 (patch) | |
| tree | 9c6587e402cd268a8ce2fa5c42455763283a06e9 /src | |
| parent | 3b1a59259ba470bf78abfddd681269c0c552e646 (diff) | |
freedreno/ir3: add UMAD support
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
(cherry picked from commit c4e2a196c3e699023e2d371d1c698daaa19a6e77)
Diffstat (limited to 'src')
| -rw-r--r-- | src/gallium/drivers/freedreno/ir3/ir3_compiler.c | 19 |
1 files changed, 15 insertions, 4 deletions
diff --git a/src/gallium/drivers/freedreno/ir3/ir3_compiler.c b/src/gallium/drivers/freedreno/ir3/ir3_compiler.c index 83c687f266f..3f1d721e71c 100644 --- a/src/gallium/drivers/freedreno/ir3/ir3_compiler.c +++ b/src/gallium/drivers/freedreno/ir3/ir3_compiler.c @@ -1986,7 +1986,7 @@ trans_cov(const struct instr_translater *t, } /* - * UMUL + * UMUL / UMAD * * There is no 32-bit multiply instruction, so splitting a and b into high and * low components, we get that @@ -1996,6 +1996,8 @@ trans_cov(const struct instr_translater *t, * mull.u tmp0, a, b (mul low, i.e. al * bl) * madsh.m16 tmp1, a, b, tmp0 (mul-add shift high mix, i.e. ah * bl << 16) * madsh.m16 dst, b, a, tmp1 (i.e. al * bh << 16) + * + * For UMAD, replace first mull.u with mad.u16. */ static void trans_umul(const struct instr_translater *t, @@ -2018,9 +2020,17 @@ trans_umul(const struct instr_translater *t, if (is_rel_or_const(b)) b = get_unconst(ctx, b); - /* mull.u tmp0, a, b */ - instr = instr_create(ctx, 2, OPC_MULL_U); - vectorize(ctx, instr, &tmp0_dst, 2, a, 0, b, 0); + if (t->tgsi_opc == TGSI_OPCODE_UMUL) { + /* mull.u tmp0, a, b */ + instr = instr_create(ctx, 2, OPC_MULL_U); + vectorize(ctx, instr, &tmp0_dst, 2, a, 0, b, 0); + } else { + struct tgsi_src_register *c = &inst->Src[2].Register; + + /* mad.u16 tmp0, a, b, c */ + instr = instr_create(ctx, 3, OPC_MAD_U16); + vectorize(ctx, instr, &tmp0_dst, 3, a, 0, b, 0, c, 0); + } /* madsh.m16 tmp1, a, b, tmp0 */ instr = instr_create(ctx, 3, OPC_MADSH_M16); @@ -2404,6 +2414,7 @@ static const struct instr_translater translaters[TGSI_OPCODE_LAST] = { INSTR(NOT, instr_cat2, .opc = OPC_NOT_B), INSTR(XOR, instr_cat2, .opc = OPC_XOR_B), INSTR(UMUL, trans_umul), + INSTR(UMAD, trans_umul), INSTR(UDIV, trans_idiv), INSTR(IDIV, trans_idiv), INSTR(MOD, trans_idiv), |