freedreno/ir3: add UMOD support, based on UDIV

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
This commit is contained in:
Ilia Mirkin
2014-09-28 21:05:05 -04:00
committed by Rob Clark
parent 8f7d01c2cb
commit cab3cb1d71
@@ -1995,9 +1995,10 @@ trans_umul(const struct instr_translater *t,
}
/*
* IDIV / UDIV
* IDIV / UDIV / UMOD
*
* See NV50LegalizeSSA::handleDIV for the origin of this implementation.
* See NV50LegalizeSSA::handleDIV for the origin of this implementation. For
* UMOD, it becomes a - UDIV(a, modulus) * modulus.
*/
static void
trans_idiv(const struct instr_translater *t,
@@ -2005,7 +2006,7 @@ trans_idiv(const struct instr_translater *t,
struct tgsi_full_instruction *inst)
{
struct ir3_instruction *instr;
struct tgsi_dst_register *dst = get_dst(ctx, inst);
struct tgsi_dst_register *dst = get_dst(ctx, inst), *premod_dst = dst;
struct tgsi_src_register *a = &inst->Src[0].Register;
struct tgsi_src_register *b = &inst->Src[1].Register;
@@ -2027,6 +2028,9 @@ trans_idiv(const struct instr_translater *t,
get_immediate(ctx, &negative_2, -2);
get_immediate(ctx, &thirty_one, 31);
if (t->tgsi_opc == TGSI_OPCODE_UMOD)
premod_dst = &q_dst;
/* cov.[us]32f32 af, numerator */
instr = instr_create(ctx, 1, 0);
instr->cat1.src_type = src_type;
@@ -2147,10 +2151,10 @@ trans_idiv(const struct instr_translater *t,
instr->cat2.condition = IR3_COND_GE;
vectorize(ctx, instr, &r_dst, 2, r_src, 0, b_src, 0);
if (t->tgsi_opc == TGSI_OPCODE_UDIV) {
if (t->tgsi_opc != TGSI_OPCODE_IDIV) {
/* add.u dst, q, r */
instr = instr_create(ctx, 2, OPC_ADD_U);
vectorize(ctx, instr, dst, 2, q_src, 0, r_src, 0);
vectorize(ctx, instr, premod_dst, 2, q_src, 0, r_src, 0);
} else {
/* add.u q, q, r */
instr = instr_create(ctx, 2, OPC_ADD_U);
@@ -2174,7 +2178,27 @@ trans_idiv(const struct instr_translater *t,
/* sel.b dst, b, r, q */
instr = instr_create(ctx, 3, OPC_SEL_B32);
vectorize(ctx, instr, dst, 3, b_src, 0, r_src, 0, q_src, 0);
vectorize(ctx, instr, premod_dst, 3, b_src, 0, r_src, 0, q_src, 0);
}
if (t->tgsi_opc == TGSI_OPCODE_UMOD) {
/* The division result will have ended up in q. */
/* mull.u r, q, b */
instr = instr_create(ctx, 2, OPC_MULL_U);
vectorize(ctx, instr, &r_dst, 2, q_src, 0, b, 0);
/* madsh.m16 r, q, b, r */
instr = instr_create(ctx, 3, OPC_MADSH_M16);
vectorize(ctx, instr, &r_dst, 3, q_src, 0, b, 0, r_src, 0);
/* madsh.m16 r, b, q, r */
instr = instr_create(ctx, 3, OPC_MADSH_M16);
vectorize(ctx, instr, &r_dst, 3, b, 0, q_src, 0, r_src, 0);
/* sub.u dst, a, r */
instr = instr_create(ctx, 2, OPC_SUB_U);
vectorize(ctx, instr, dst, 2, a, 0, r_src, 0);
}
put_dst(ctx, inst, dst);
@@ -2341,6 +2365,7 @@ static const struct instr_translater translaters[TGSI_OPCODE_LAST] = {
INSTR(UMUL, trans_umul),
INSTR(UDIV, trans_idiv),
INSTR(IDIV, trans_idiv),
INSTR(UMOD, trans_idiv),
INSTR(SHL, instr_cat2, .opc = OPC_SHL_B),
INSTR(USHR, instr_cat2, .opc = OPC_SHR_B),
INSTR(ISHR, instr_cat2, .opc = OPC_ASHR_B),