tgsi: implement new float comparison instructions returning integer masks

Also while here add a bunch of other forgotten (integer) instructions to
tgsi_util_get_inst_usage_mask() (which isn't used for much except optimizing
away unused input components), though it may still be incomplete.

Reviewed-by: Zack Rusin <zackr@vmware.com>

diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index d991d4b5cdfd00bb9525a81ed90a3072b6dacff2..1ffd9e94e5460294600fba5e8dfb95731aecf71c 100644 (file)
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -3263,6 +3263,50 @@ micro_f2i(union tgsi_exec_channel *dst,
    dst->i[3] = (int)src->f[3];
 }
 
+static void
+micro_fseq(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src0,
+           const union tgsi_exec_channel *src1)
+{
+   dst->u[0] = src0->f[0] == src1->f[0] ? ~0 : 0;
+   dst->u[1] = src0->f[1] == src1->f[1] ? ~0 : 0;
+   dst->u[2] = src0->f[2] == src1->f[2] ? ~0 : 0;
+   dst->u[3] = src0->f[3] == src1->f[3] ? ~0 : 0;
+}
+
+static void
+micro_fsge(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src0,
+           const union tgsi_exec_channel *src1)
+{
+   dst->u[0] = src0->f[0] >= src1->f[0] ? ~0 : 0;
+   dst->u[1] = src0->f[1] >= src1->f[1] ? ~0 : 0;
+   dst->u[2] = src0->f[2] >= src1->f[2] ? ~0 : 0;
+   dst->u[3] = src0->f[3] >= src1->f[3] ? ~0 : 0;
+}
+
+static void
+micro_fslt(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src0,
+           const union tgsi_exec_channel *src1)
+{
+   dst->u[0] = src0->f[0] < src1->f[0] ? ~0 : 0;
+   dst->u[1] = src0->f[1] < src1->f[1] ? ~0 : 0;
+   dst->u[2] = src0->f[2] < src1->f[2] ? ~0 : 0;
+   dst->u[3] = src0->f[3] < src1->f[3] ? ~0 : 0;
+}
+
+static void
+micro_fsne(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src0,
+           const union tgsi_exec_channel *src1)
+{
+   dst->u[0] = src0->f[0] != src1->f[0] ? ~0 : 0;
+   dst->u[1] = src0->f[1] != src1->f[1] ? ~0 : 0;
+   dst->u[2] = src0->f[2] != src1->f[2] ? ~0 : 0;
+   dst->u[3] = src0->f[3] != src1->f[3] ? ~0 : 0;
+}
+
 static void
 micro_idiv(union tgsi_exec_channel *dst,
            const union tgsi_exec_channel *src0,
@@ -4152,6 +4196,22 @@ exec_instruction(
       exec_vector_unary(mach, inst, micro_f2i, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_FLOAT);
       break;
 
+   case TGSI_OPCODE_FSEQ:
+      exec_vector_binary(mach, inst, micro_fseq, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_FLOAT);
+      break;
+
+   case TGSI_OPCODE_FSGE:
+      exec_vector_binary(mach, inst, micro_fsge, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_FLOAT);
+      break;
+
+   case TGSI_OPCODE_FSLT:
+      exec_vector_binary(mach, inst, micro_fslt, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_FLOAT);
+      break;
+
+   case TGSI_OPCODE_FSNE:
+      exec_vector_binary(mach, inst, micro_fsne, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_FLOAT);
+      break;
+
    case TGSI_OPCODE_IDIV:
       exec_vector_binary(mach, inst, micro_idiv, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT);
       break;

diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.c b/src/gallium/auxiliary/tgsi/tgsi_info.c
index 7e930288f556615192268d53c66e292c0080d3d3..7a5d18f59c54e80954e248bde16e21a1c00fc969 100644 (file)
--- a/src/gallium/auxiliary/tgsi/tgsi_info.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_info.c
@@ -145,10 +145,10 @@ static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] =
    { 0, 0, 0, 0, 0, 0, NONE, "", 105 },     /* removed */
    { 0, 0, 0, 0, 0, 0, NONE, "", 106 },     /* removed */
    { 0, 0, 0, 0, 0, 0, NONE, "NOP", TGSI_OPCODE_NOP },
-   { 0, 0, 0, 0, 0, 0, NONE, "", 108 },     /* removed */
-   { 0, 0, 0, 0, 0, 0, NONE, "", 109 },     /* removed */
-   { 0, 0, 0, 0, 0, 0, NONE, "", 110 },     /* removed */
-   { 0, 0, 0, 0, 0, 0, NONE, "", 111 },     /* removed */
+   { 1, 2, 0, 0, 0, 0, COMP, "FSEQ", TGSI_OPCODE_FSEQ },
+   { 1, 2, 0, 0, 0, 0, COMP, "FSGE", TGSI_OPCODE_FSGE },
+   { 1, 2, 0, 0, 0, 0, COMP, "FSLT", TGSI_OPCODE_FSLT },
+   { 1, 2, 0, 0, 0, 0, COMP, "FSNE", TGSI_OPCODE_FSNE },
    { 1, 1, 0, 0, 0, 0, REPL, "NRM4", TGSI_OPCODE_NRM4 },
    { 0, 1, 0, 0, 0, 0, NONE, "CALLNZ", TGSI_OPCODE_CALLNZ },
    { 0, 1, 0, 0, 0, 0, NONE, "", 114 },     /* removed */
@@ -302,6 +302,10 @@ tgsi_opcode_infer_type( uint opcode )
    case TGSI_OPCODE_ARR:
    case TGSI_OPCODE_MOD:
    case TGSI_OPCODE_F2I:
+   case TGSI_OPCODE_FSEQ:
+   case TGSI_OPCODE_FSGE:
+   case TGSI_OPCODE_FSLT:
+   case TGSI_OPCODE_FSNE:
    case TGSI_OPCODE_IDIV:
    case TGSI_OPCODE_IMAX:
    case TGSI_OPCODE_IMIN:
@@ -343,6 +347,10 @@ tgsi_opcode_infer_src_type( uint opcode )
    case TGSI_OPCODE_TXQ_LZ:
    case TGSI_OPCODE_F2I:
    case TGSI_OPCODE_F2U:
+   case TGSI_OPCODE_FSEQ:
+   case TGSI_OPCODE_FSGE:
+   case TGSI_OPCODE_FSLT:
+   case TGSI_OPCODE_FSNE:
    case TGSI_OPCODE_UCMP:
       return TGSI_TYPE_FLOAT;
    default:

diff --git a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h
index e9d0a0554031e6838f4f70e77c9f1adcf22f05f0..b8144a8916b6fce01b54ca628d130f2639c66d94 100644 (file)
--- a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h
@@ -159,6 +159,10 @@ OP01(BREAKC)
 OP01(KILL_IF)
 OP00(END)
 OP11(F2I)
+OP12(FSEQ)
+OP12(FSGE)
+OP12(FSLT)
+OP12(FSNE)
 OP12(IDIV)
 OP12(IMAX)
 OP12(IMIN)

diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.c b/src/gallium/auxiliary/tgsi/tgsi_util.c
index 98c1e6e6e51683d567e0974b7c5217c5b0b4622d..39b605bd071f8edca6509eae5e1a55ecf6e88895 100644 (file)
--- a/src/gallium/auxiliary/tgsi/tgsi_util.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_util.c
@@ -217,6 +217,32 @@ tgsi_util_get_inst_usage_mask(const struct tgsi_full_instruction *inst,
    case TGSI_OPCODE_OR:
    case TGSI_OPCODE_XOR:
    case TGSI_OPCODE_SAD:
+   case TGSI_OPCODE_FSEQ:
+   case TGSI_OPCODE_FSGE:
+   case TGSI_OPCODE_FSLT:
+   case TGSI_OPCODE_FSNE:
+   case TGSI_OPCODE_F2I:
+   case TGSI_OPCODE_IDIV:
+   case TGSI_OPCODE_IMAX:
+   case TGSI_OPCODE_IMIN:
+   case TGSI_OPCODE_INEG:
+   case TGSI_OPCODE_ISGE:
+   case TGSI_OPCODE_ISHR:
+   case TGSI_OPCODE_ISLT:
+   case TGSI_OPCODE_F2U:
+   case TGSI_OPCODE_U2F:
+   case TGSI_OPCODE_UADD:
+   case TGSI_OPCODE_UDIV:
+   case TGSI_OPCODE_UMAD:
+   case TGSI_OPCODE_UMAX:
+   case TGSI_OPCODE_UMIN:
+   case TGSI_OPCODE_UMOD:
+   case TGSI_OPCODE_UMUL:
+   case TGSI_OPCODE_USEQ:
+   case TGSI_OPCODE_USGE:
+   case TGSI_OPCODE_USHR:
+   case TGSI_OPCODE_USLT:
+   case TGSI_OPCODE_USNE:
       /* Channel-wise operations */
       read_mask = write_mask;
       break;