gallium: remove TGSI opcode XPD

use MUL+MAD+MOV instead.

Reviewed-by: Roland Scheidegger <[email protected]>

2 files changed, 0 insertions, 174 deletions

diff --git a/src/gallium/drivers/svga/svga_tgsi_insn.c b/src/gallium/drivers/svga/svga_tgsi_insn.c
index 928330ce147..a325a567c18 100644
--- a/src/gallium/drivers/svga/svga_tgsi_insn.c
+++ b/src/gallium/drivers/svga/svga_tgsi_insn.c
@@ -2196,63 +2196,6 @@ emit_pow(struct svga_shader_emitter *emit,
 
 
 /**
- * Translate/emit TGSI XPD (vector cross product) instruction.
- */
-static boolean
-emit_xpd(struct svga_shader_emitter *emit,
-         const struct tgsi_full_instruction *insn)
-{
-   SVGA3dShaderDestToken dst = translate_dst_register( emit, insn, 0 );
-   const struct src_register src0 = translate_src_register(
-      emit, &insn->Src[0] );
-   const struct src_register src1 = translate_src_register(
-      emit, &insn->Src[1] );
-   boolean need_dst_tmp = FALSE;
-
-   /* XPD can only output to a temporary */
-   if (SVGA3dShaderGetRegType(dst.value) != SVGA3DREG_TEMP)
-      need_dst_tmp = TRUE;
-
-   /* The dst reg must not be the same as src0 or src1*/
-   if (alias_src_dst(src0, dst) ||
-       alias_src_dst(src1, dst))
-      need_dst_tmp = TRUE;
-
-   if (need_dst_tmp) {
-      SVGA3dShaderDestToken tmp = get_temp( emit );
-
-      /* Obey DX9 restrictions on mask:
-       */
-      tmp.mask = dst.mask & TGSI_WRITEMASK_XYZ;
-
-      if (!submit_op2(emit, inst_token( SVGA3DOP_CRS ), tmp, src0, src1))
-         return FALSE;
-
-      if (!submit_op1(emit, inst_token( SVGA3DOP_MOV ), dst, src( tmp )))
-         return FALSE;
-   }
-   else {
-      if (!submit_op2(emit, inst_token( SVGA3DOP_CRS ), dst, src0, src1))
-         return FALSE;
-   }
-
-   /* Need to emit 1.0 to dst.w?
-    */
-   if (dst.mask & TGSI_WRITEMASK_W) {
-      struct src_register one = get_one_immediate( emit );
-
-      if (!submit_op1(emit,
-                      inst_token( SVGA3DOP_MOV ),
-                      writemask(dst, TGSI_WRITEMASK_W),
-                      one))
-         return FALSE;
-   }
-
-   return TRUE;
-}
-
-
-/**
  * Emit a LRP (linear interpolation) instruction.
  */
 static boolean
@@ -2986,9 +2929,6 @@ svga_emit_instruction(struct svga_shader_emitter *emit,
    case TGSI_OPCODE_BRK:
       return emit_brk( emit, insn );
 
-   case TGSI_OPCODE_XPD:
-      return emit_xpd( emit, insn );
-
    case TGSI_OPCODE_KILL:
       return emit_kill( emit, insn );
 
@@ -3604,7 +3544,6 @@ needs_to_create_common_immediate(const struct svga_shader_emitter *emit)
        emit->info.opcode_count[TGSI_OPCODE_SEQ] >= 1 ||
        emit->info.opcode_count[TGSI_OPCODE_EXP] >= 1 ||
        emit->info.opcode_count[TGSI_OPCODE_LOG] >= 1 ||
-       emit->info.opcode_count[TGSI_OPCODE_XPD] >= 1 ||
        emit->info.opcode_count[TGSI_OPCODE_KILL] >= 1)
       return TRUE;
 
diff --git a/src/gallium/drivers/svga/svga_tgsi_vgpu10.c b/src/gallium/drivers/svga/svga_tgsi_vgpu10.c
index e57e78d11a5..9d86f72ea05 100644
--- a/src/gallium/drivers/svga/svga_tgsi_vgpu10.c
+++ b/src/gallium/drivers/svga/svga_tgsi_vgpu10.c
@@ -5210,117 +5210,6 @@ emit_txp(struct svga_shader_emitter_v10 *emit,
 }
 
 
-/*
- * Emit code for TGSI_OPCODE_XPD instruction.
- */
-static boolean
-emit_xpd(struct svga_shader_emitter_v10 *emit,
-         const struct tgsi_full_instruction *inst)
-{
-   /* dst.x = src0.y * src1.z - src1.y * src0.z
-    * dst.y = src0.z * src1.x - src1.z * src0.x
-    * dst.z = src0.x * src1.y - src1.x * src0.y
-    * dst.w = 1
-    */
-   struct tgsi_full_src_register s0_xxxx =
-      scalar_src(&inst->Src[0], TGSI_SWIZZLE_X);
-   struct tgsi_full_src_register s0_yyyy =
-      scalar_src(&inst->Src[0], TGSI_SWIZZLE_Y);
-   struct tgsi_full_src_register s0_zzzz =
-      scalar_src(&inst->Src[0], TGSI_SWIZZLE_Z);
-
-   struct tgsi_full_src_register s1_xxxx =
-      scalar_src(&inst->Src[1], TGSI_SWIZZLE_X);
-   struct tgsi_full_src_register s1_yyyy =
-      scalar_src(&inst->Src[1], TGSI_SWIZZLE_Y);
-   struct tgsi_full_src_register s1_zzzz =
-      scalar_src(&inst->Src[1], TGSI_SWIZZLE_Z);
-
-   unsigned tmp1 = get_temp_index(emit);
-   struct tgsi_full_src_register tmp1_src = make_src_temp_reg(tmp1);
-   struct tgsi_full_dst_register tmp1_dst = make_dst_temp_reg(tmp1);
-
-   unsigned tmp2 = get_temp_index(emit);
-   struct tgsi_full_src_register tmp2_src = make_src_temp_reg(tmp2);
-   struct tgsi_full_dst_register tmp2_dst = make_dst_temp_reg(tmp2);
-   struct tgsi_full_src_register neg_tmp2_src = negate_src(&tmp2_src);
-
-   unsigned tmp3 = get_temp_index(emit);
-   struct tgsi_full_src_register tmp3_src = make_src_temp_reg(tmp3);
-   struct tgsi_full_dst_register tmp3_dst = make_dst_temp_reg(tmp3);
-   struct tgsi_full_dst_register tmp3_dst_x =
-      writemask_dst(&tmp3_dst, TGSI_WRITEMASK_X);
-   struct tgsi_full_dst_register tmp3_dst_y =
-      writemask_dst(&tmp3_dst, TGSI_WRITEMASK_Y);
-   struct tgsi_full_dst_register tmp3_dst_z =
-      writemask_dst(&tmp3_dst, TGSI_WRITEMASK_Z);
-   struct tgsi_full_dst_register tmp3_dst_w =
-      writemask_dst(&tmp3_dst, TGSI_WRITEMASK_W);
-
-   /* Note: we put all the intermediate computations into tmp3 in case
-    * the XPD dest register is that same as one of the src regs (in which
-    * case we could clobber a src reg before we're done with it) .
-    *
-    * Note: we could get by with just one temp register instead of three
-    * since we're doing scalar operations and there's enough room in one
-    * temp for everything.
-    */
-
-   /* MUL tmp1, src0.y, src1.z */
-   /* MUL tmp2, src1.y, src0.z */
-   /* ADD tmp3.x, tmp1, -tmp2 */
-   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) {
-      emit_instruction_op2(emit, VGPU10_OPCODE_MUL, &tmp1_dst,
-                           &s0_yyyy, &s1_zzzz, FALSE);
-      emit_instruction_op2(emit, VGPU10_OPCODE_MUL, &tmp2_dst,
-                           &s1_yyyy, &s0_zzzz, FALSE);
-      emit_instruction_op2(emit, VGPU10_OPCODE_ADD, &tmp3_dst_x,
-                           &tmp1_src, &neg_tmp2_src, FALSE);
-   }
-
-   /* MUL tmp1, src0.z, src1.x */
-   /* MUL tmp2, src1.z, src0.x */
-   /* ADD tmp3.y, tmp1, -tmp2 */
-   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
-      emit_instruction_op2(emit, VGPU10_OPCODE_MUL, &tmp1_dst, &s0_zzzz,
-                           &s1_xxxx, FALSE);
-      emit_instruction_op2(emit, VGPU10_OPCODE_MUL, &tmp2_dst, &s1_zzzz,
-                           &s0_xxxx, FALSE);
-      emit_instruction_op2(emit, VGPU10_OPCODE_ADD, &tmp3_dst_y,
-                           &tmp1_src, &neg_tmp2_src, FALSE);
-   }
-
-   /* MUL tmp1, src0.x, src1.y */
-   /* MUL tmp2, src1.x, src0.y */
-   /* ADD tmp3.z, tmp1, -tmp2 */
-   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Z) {
-      emit_instruction_op2(emit, VGPU10_OPCODE_MUL, &tmp1_dst, &s0_xxxx,
-                           &s1_yyyy, FALSE);
-      emit_instruction_op2(emit, VGPU10_OPCODE_MUL, &tmp2_dst, &s1_xxxx,
-                           &s0_yyyy, FALSE);
-      emit_instruction_op2(emit, VGPU10_OPCODE_ADD, &tmp3_dst_z,
-                           &tmp1_src, &neg_tmp2_src, FALSE);
-   }
-
-   /* MOV tmp3.w, 1.0 */
-   if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_W) {
-      struct tgsi_full_src_register one =
-         make_immediate_reg_float(emit, 1.0f);
-
-      emit_instruction_op1(emit, VGPU10_OPCODE_MOV, &tmp3_dst_w, &one, FALSE);
-   }
-
-   /* MOV dst, tmp3 */
-   emit_instruction_op1(emit, VGPU10_OPCODE_MOV, &inst->Dst[0], &tmp3_src,
-                        inst->Instruction.Saturate);
-
-
-   free_temp_indexes(emit);
-
-   return TRUE;
-}
-
-
 /**
  * Emit code for TGSI_OPCODE_TXD (explicit derivatives)
  */
@@ -5742,8 +5631,6 @@ emit_vgpu10_instruction(struct svga_shader_emitter_v10 *emit,
       return emit_txq(emit, inst);
    case TGSI_OPCODE_UIF:
       return emit_if(emit, inst);
-   case TGSI_OPCODE_XPD:
-      return emit_xpd(emit, inst);
    case TGSI_OPCODE_UMUL_HI:
    case TGSI_OPCODE_IMUL_HI:
    case TGSI_OPCODE_UDIV: