Signed-off-by: Topi Pohjolainen <topi.pohjolai...@intel.com>
---
src/mesa/drivers/dri/i965/brw_blorp_blit.cpp | 340 ++++++++++++------------
src/mesa/drivers/dri/i965/brw_blorp_blit_eu.cpp | 28 +-
src/mesa/drivers/dri/i965/brw_blorp_blit_eu.h | 110 +-------
3 files changed, 189 insertions(+), 289 deletions(-)
diff --git a/src/mesa/drivers/dri/i965/brw_blorp_blit.cpp
b/src/mesa/drivers/dri/i965/brw_blorp_blit.cpp
index 4fc3b7c..0ceb7ed 100644
--- a/src/mesa/drivers/dri/i965/brw_blorp_blit.cpp
+++ b/src/mesa/drivers/dri/i965/brw_blorp_blit.cpp
@@ -1030,8 +1030,8 @@ brw_blorp_blit_program::compute_frag_coords()
* Then, we need to add the repeating sequence (0, 1, 0, 1, ...) to the
* result, since pixels n+1 and n+3 are in the right half of the subspan.
*/
- emit_add(vec16(retype(X, BRW_REGISTER_TYPE_UW)),
- stride(suboffset(R1, 4), 2, 4, 0), brw_imm_v(0x10101010));
+ emit(ADD(vec16(retype(X, BRW_REGISTER_TYPE_UW)),
+ stride(suboffset(R1, 4), 2, 4, 0), brw_imm_v(0x10101010)));
/* Similarly, Y coordinates for subspans come from R1.2[31:16] through
* R1.5[31:16], so to get pixel Y coordinates we need to start at the 5th
@@ -1041,12 +1041,12 @@ brw_blorp_blit_program::compute_frag_coords()
* And we need to add the repeating sequence (0, 0, 1, 1, ...), since
* pixels n+2 and n+3 are in the bottom half of the subspan.
*/
- emit_add(vec16(retype(Y, BRW_REGISTER_TYPE_UW)),
- stride(suboffset(R1, 5), 2, 4, 0), brw_imm_v(0x11001100));
+ emit(ADD(vec16(retype(Y, BRW_REGISTER_TYPE_UW)),
+ stride(suboffset(R1, 5), 2, 4, 0), brw_imm_v(0x11001100)));
/* Move the coordinates to UD registers. */
- emit_mov(vec16(Xp), retype(X, BRW_REGISTER_TYPE_UW));
- emit_mov(vec16(Yp), retype(Y, BRW_REGISTER_TYPE_UW));
+ emit(MOV(vec16(Xp), retype(X, BRW_REGISTER_TYPE_UW)));
+ emit(MOV(vec16(Yp), retype(Y, BRW_REGISTER_TYPE_UW)));
SWAP_XY_AND_XPYP();
if (key->persample_msaa_dispatch) {
@@ -1062,10 +1062,11 @@ brw_blorp_blit_program::compute_frag_coords()
* then copy from it using vstride=1, width=4, hstride=0.
*/
struct brw_reg t1_uw1 = retype(t1, BRW_REGISTER_TYPE_UW);
- emit_mov(vec16(t1_uw1), brw_imm_v(0x3210));
+ emit(MOV(vec16(t1_uw1), brw_imm_v(0x3210)));
/* Move to UD sample_index register. */
- emit_mov_8(S, stride(t1_uw1, 1, 4, 0));
- emit_mov_8(offset(S, 1), suboffset(stride(t1_uw1, 1, 4, 0), 2));
+ emit(MOV(S, stride(t1_uw1, 1, 4, 0)))->force_uncompressed = true;
+ emit(MOV(offset(S, 1), suboffset(stride(t1_uw1, 1, 4, 0), 2)))
+ ->force_uncompressed = true;
break;
}
case 8: {
@@ -1085,14 +1086,15 @@ brw_blorp_blit_program::compute_frag_coords()
struct brw_reg t1_ud1 = vec1(retype(t1, BRW_REGISTER_TYPE_UD));
struct brw_reg t2_uw1 = retype(t2, BRW_REGISTER_TYPE_UW);
struct brw_reg r0_ud1 = vec1(retype(R0, BRW_REGISTER_TYPE_UD));
- emit_and(t1_ud1, r0_ud1, brw_imm_ud(0xc0));
- emit_shr(t1_ud1, t1_ud1, brw_imm_ud(5));
- emit_mov(vec16(t2_uw1), brw_imm_v(0x3210));
- emit_add(vec16(S), retype(t1_ud1, BRW_REGISTER_TYPE_UW),
- stride(t2_uw1, 1, 4, 0));
- emit_add_8(offset(S, 1),
- retype(t1_ud1, BRW_REGISTER_TYPE_UW),
- suboffset(stride(t2_uw1, 1, 4, 0), 2));
+ emit(AND(t1_ud1, r0_ud1, brw_imm_ud(0xc0)));
+ emit(SHR(t1_ud1, t1_ud1, brw_imm_ud(5)));
+ emit(MOV(vec16(t2_uw1), brw_imm_v(0x3210)));
+ emit(ADD(vec16(S), retype(t1_ud1, BRW_REGISTER_TYPE_UW),
+ stride(t2_uw1, 1, 4, 0)));
+ emit(ADD(offset(S, 1),
+ retype(t1_ud1, BRW_REGISTER_TYPE_UW),
+ suboffset(stride(t2_uw1, 1, 4, 0), 2)))
+ ->force_uncompressed = true;
break;
}
default:
@@ -1162,21 +1164,21 @@ brw_blorp_blit_program::translate_tiling(bool
old_tiled_w, bool new_tiled_w)
* X' = (X & ~0b1011) >> 1 | (Y & 0b1) << 2 | X & 0b1 (4)
* Y' = (Y & ~0b1) << 1 | (X & 0b1000) >> 2 | (X & 0b10) >> 1
*/
- emit_and(t1, X, brw_imm_uw(0xfff4)); /* X & ~0b1011 */
- emit_shr(t1, t1, brw_imm_uw(1)); /* (X & ~0b1011) >> 1 */
- emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- emit_shl(t2, t2, brw_imm_uw(2)); /* (Y & 0b1) << 2 */
- emit_or(t1, t1, t2); /* (X & ~0b1011) >> 1 | (Y & 0b1) << 2 */
- emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
- emit_or(Xp, t1, t2);
- emit_and(t1, Y, brw_imm_uw(0xfffe)); /* Y & ~0b1 */
- emit_shl(t1, t1, brw_imm_uw(1)); /* (Y & ~0b1) << 1 */
- emit_and(t2, X, brw_imm_uw(8)); /* X & 0b1000 */
- emit_shr(t2, t2, brw_imm_uw(2)); /* (X & 0b1000) >> 2 */
- emit_or(t1, t1, t2); /* (Y & ~0b1) << 1 | (X & 0b1000) >> 2 */
- emit_and(t2, X, brw_imm_uw(2)); /* X & 0b10 */
- emit_shr(t2, t2, brw_imm_uw(1)); /* (X & 0b10) >> 1 */
- emit_or(Yp, t1, t2);
+ emit(AND(t1, X, brw_imm_uw(0xfff4))); /* X & ~0b1011 */
+ emit(SHR(t1, t1, brw_imm_uw(1))); /* (X & ~0b1011) >> 1 */
+ emit(AND(t2, Y, brw_imm_uw(1))); /* Y & 0b1 */
+ emit(SHL(t2, t2, brw_imm_uw(2))); /* (Y & 0b1) << 2 */
+ emit(OR(t1, t1, t2)); /* (X & ~0b1011) >> 1 | (Y & 0b1) << 2 */
+ emit(AND(t2, X, brw_imm_uw(1))); /* X & 0b1 */
+ emit(OR(Xp, t1, t2));
+ emit(AND(t1, Y, brw_imm_uw(0xfffe))); /* Y & ~0b1 */
+ emit(SHL(t1, t1, brw_imm_uw(1))); /* (Y & ~0b1) << 1 */
+ emit(AND(t2, X, brw_imm_uw(8))); /* X & 0b1000 */
+ emit(SHR(t2, t2, brw_imm_uw(2))); /* (X & 0b1000) >> 2 */
+ emit(OR(t1, t1, t2)); /* (Y & ~0b1) << 1 | (X & 0b1000) >> 2 */
+ emit(AND(t2, X, brw_imm_uw(2))); /* X & 0b10 */
+ emit(SHR(t2, t2, brw_imm_uw(1))); /* (X & 0b10) >> 1 */
+ emit(OR(Yp, t1, t2));
SWAP_XY_AND_XPYP();
} else {
/* Applying the same logic as above, but in reverse, we obtain the
@@ -1185,22 +1187,22 @@ brw_blorp_blit_program::translate_tiling(bool
old_tiled_w, bool new_tiled_w)
* X' = (X & ~0b101) << 1 | (Y & 0b10) << 2 | (Y & 0b1) << 1 | X & 0b1
* Y' = (Y & ~0b11) >> 1 | (X & 0b100) >> 2
*/
- emit_and(t1, X, brw_imm_uw(0xfffa)); /* X & ~0b101 */
- emit_shl(t1, t1, brw_imm_uw(1)); /* (X & ~0b101) << 1 */
- emit_and(t2, Y, brw_imm_uw(2)); /* Y & 0b10 */
- emit_shl(t2, t2, brw_imm_uw(2)); /* (Y & 0b10) << 2 */
- emit_or(t1, t1, t2); /* (X & ~0b101) << 1 | (Y & 0b10) << 2 */
- emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- emit_shl(t2, t2, brw_imm_uw(1)); /* (Y & 0b1) << 1 */
- emit_or(t1, t1, t2); /* (X & ~0b101) << 1 | (Y & 0b10) << 2
- | (Y & 0b1) << 1 */
- emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
- emit_or(Xp, t1, t2);
- emit_and(t1, Y, brw_imm_uw(0xfffc)); /* Y & ~0b11 */
- emit_shr(t1, t1, brw_imm_uw(1)); /* (Y & ~0b11) >> 1 */
- emit_and(t2, X, brw_imm_uw(4)); /* X & 0b100 */
- emit_shr(t2, t2, brw_imm_uw(2)); /* (X & 0b100) >> 2 */
- emit_or(Yp, t1, t2);
+ emit(AND(t1, X, brw_imm_uw(0xfffa))); /* X & ~0b101 */
+ emit(SHL(t1, t1, brw_imm_uw(1))); /* (X & ~0b101) << 1 */
+ emit(AND(t2, Y, brw_imm_uw(2))); /* Y & 0b10 */
+ emit(SHL(t2, t2, brw_imm_uw(2))); /* (Y & 0b10) << 2 */
+ emit(OR(t1, t1, t2)); /* (X & ~0b101) << 1 | (Y & 0b10) << 2 */
+ emit(AND(t2, Y, brw_imm_uw(1))); /* Y & 0b1 */
+ emit(SHL(t2, t2, brw_imm_uw(1))); /* (Y & 0b1) << 1 */
+ emit(OR(t1, t1, t2)); /* (X & ~0b101) << 1 | (Y & 0b10) << 2
+ | (Y & 0b1) << 1 */
+ emit(AND(t2, X, brw_imm_uw(1))); /* X & 0b1 */
+ emit(OR(Xp, t1, t2));
+ emit(AND(t1, Y, brw_imm_uw(0xfffc))); /* Y & ~0b11 */
+ emit(SHR(t1, t1, brw_imm_uw(1))); /* (Y & ~0b11) >> 1 */
+ emit(AND(t2, X, brw_imm_uw(4))); /* X & 0b100 */
+ emit(SHR(t2, t2, brw_imm_uw(2))); /* (X & 0b100) >> 2 */
+ emit(OR(Yp, t1, t2));
SWAP_XY_AND_XPYP();
}
}
@@ -1240,23 +1242,23 @@ brw_blorp_blit_program::encode_msaa(unsigned
num_samples,
* where X' = (X & ~0b1) << 1 | (S & 0b1) << 1 | (X & 0b1)
* Y' = (Y & ~0b1) << 1 | (S & 0b10) | (Y & 0b1)
*/
- emit_and(t1, X, brw_imm_uw(0xfffe)); /* X & ~0b1 */
+ emit(AND(t1, X, brw_imm_uw(0xfffe))); /* X & ~0b1 */
if (!s_is_zero) {
- emit_and(t2, S, brw_imm_uw(1)); /* S & 0b1 */
- emit_or(t1, t1, t2); /* (X & ~0b1) | (S & 0b1) */
+ emit(AND(t2, S, brw_imm_uw(1))); /* S & 0b1 */
+ emit(OR(t1, t1, t2)); /* (X & ~0b1) | (S & 0b1) */
}
- emit_shl(t1, t1, brw_imm_uw(1)); /* (X & ~0b1) << 1
+ emit(SHL(t1, t1, brw_imm_uw(1))); /* (X & ~0b1) << 1
| (S & 0b1) << 1 */
- emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
- emit_or(Xp, t1, t2);
- emit_and(t1, Y, brw_imm_uw(0xfffe)); /* Y & ~0b1 */
- emit_shl(t1, t1, brw_imm_uw(1)); /* (Y & ~0b1) << 1 */
+ emit(AND(t2, X, brw_imm_uw(1))); /* X & 0b1 */
+ emit(OR(Xp, t1, t2));
+ emit(AND(t1, Y, brw_imm_uw(0xfffe))); /* Y & ~0b1 */
+ emit(SHL(t1, t1, brw_imm_uw(1))); /* (Y & ~0b1) << 1 */
if (!s_is_zero) {
- emit_and(t2, S, brw_imm_uw(2)); /* S & 0b10 */
- emit_or(t1, t1, t2); /* (Y & ~0b1) << 1 | (S & 0b10) */
+ emit(AND(t2, S, brw_imm_uw(2))); /* S & 0b10 */
+ emit(OR(t1, t1, t2)); /* (Y & ~0b1) << 1 | (S & 0b10) */
}
- emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- emit_or(Yp, t1, t2);
+ emit(AND(t2, Y, brw_imm_uw(1))); /* Y & 0b1 */
+ emit(OR(Yp, t1, t2));
break;
case 8:
/* encode_msaa(8, IMS, X, Y, S) = (X', Y', 0)
@@ -1264,26 +1266,26 @@ brw_blorp_blit_program::encode_msaa(unsigned
num_samples,
* | (X & 0b1)
* Y' = (Y & ~0b1) << 1 | (S & 0b10) | (Y & 0b1)
*/
- emit_and(t1, X, brw_imm_uw(0xfffe)); /* X & ~0b1 */
- emit_shl(t1, t1, brw_imm_uw(2)); /* (X & ~0b1) << 2 */
+ emit(AND(t1, X, brw_imm_uw(0xfffe))); /* X & ~0b1 */
+ emit(SHL(t1, t1, brw_imm_uw(2))); /* (X & ~0b1) << 2 */
if (!s_is_zero) {
- emit_and(t2, S, brw_imm_uw(4)); /* S & 0b100 */
- emit_or(t1, t1, t2); /* (X & ~0b1) << 2 | (S & 0b100) */
- emit_and(t2, S, brw_imm_uw(1)); /* S & 0b1 */
- emit_shl(t2, t2, brw_imm_uw(1)); /* (S & 0b1) << 1 */
- emit_or(t1, t1, t2); /* (X & ~0b1) << 2 | (S & 0b100)
- | (S & 0b1) << 1 */
+ emit(AND(t2, S, brw_imm_uw(4))); /* S & 0b100 */
+ emit(OR(t1, t1, t2)); /* (X & ~0b1) << 2 | (S & 0b100) */
+ emit(AND(t2, S, brw_imm_uw(1))); /* S & 0b1 */
+ emit(SHL(t2, t2, brw_imm_uw(1))); /* (S & 0b1) << 1 */
+ emit(OR(t1, t1, t2)); /* (X & ~0b1) << 2 | (S & 0b100)
+ | (S & 0b1) << 1 */
}
- emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
- emit_or(Xp, t1, t2);
- emit_and(t1, Y, brw_imm_uw(0xfffe)); /* Y & ~0b1 */
- emit_shl(t1, t1, brw_imm_uw(1)); /* (Y & ~0b1) << 1 */
+ emit(AND(t2, X, brw_imm_uw(1))); /* X & 0b1 */
+ emit(OR(Xp, t1, t2));
+ emit(AND(t1, Y, brw_imm_uw(0xfffe))); /* Y & ~0b1 */
+ emit(SHL(t1, t1, brw_imm_uw(1))); /* (Y & ~0b1) << 1 */
if (!s_is_zero) {
- emit_and(t2, S, brw_imm_uw(2)); /* S & 0b10 */
- emit_or(t1, t1, t2); /* (Y & ~0b1) << 1 | (S & 0b10) */
+ emit(AND(t2, S, brw_imm_uw(2))); /* S & 0b10 */
+ emit(OR(t1, t1, t2)); /* (Y & ~0b1) << 1 | (S & 0b10) */
}
- emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- emit_or(Yp, t1, t2);
+ emit(AND(t2, Y, brw_imm_uw(1))); /* Y & 0b1 */
+ emit(OR(Yp, t1, t2));
break;
}
SWAP_XY_AND_XPYP();
@@ -1329,18 +1331,18 @@ brw_blorp_blit_program::decode_msaa(unsigned
num_samples,
* Y' = (Y & ~0b11) >> 1 | (Y & 0b1)
* S = (Y & 0b10) | (X & 0b10) >> 1
*/
- emit_and(t1, X, brw_imm_uw(0xfffc)); /* X & ~0b11 */
- emit_shr(t1, t1, brw_imm_uw(1)); /* (X & ~0b11) >> 1 */
- emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
- emit_or(Xp, t1, t2);
- emit_and(t1, Y, brw_imm_uw(0xfffc)); /* Y & ~0b11 */
- emit_shr(t1, t1, brw_imm_uw(1)); /* (Y & ~0b11) >> 1 */
- emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- emit_or(Yp, t1, t2);
- emit_and(t1, Y, brw_imm_uw(2)); /* Y & 0b10 */
- emit_and(t2, X, brw_imm_uw(2)); /* X & 0b10 */
- emit_shr(t2, t2, brw_imm_uw(1)); /* (X & 0b10) >> 1 */
- emit_or(S, t1, t2);
+ emit(AND(t1, X, brw_imm_uw(0xfffc))); /* X & ~0b11 */
+ emit(SHR(t1, t1, brw_imm_uw(1))); /* (X & ~0b11) >> 1 */
+ emit(AND(t2, X, brw_imm_uw(1))); /* X & 0b1 */
+ emit(OR(Xp, t1, t2));
+ emit(AND(t1, Y, brw_imm_uw(0xfffc))); /* Y & ~0b11 */
+ emit(SHR(t1, t1, brw_imm_uw(1))); /* (Y & ~0b11) >> 1 */
+ emit(AND(t2, Y, brw_imm_uw(1))); /* Y & 0b1 */
+ emit(OR(Yp, t1, t2));
+ emit(AND(t1, Y, brw_imm_uw(2))); /* Y & 0b10 */
+ emit(AND(t2, X, brw_imm_uw(2))); /* X & 0b10 */
+ emit(SHR(t2, t2, brw_imm_uw(1))); /* (X & 0b10) >> 1 */
+ emit(OR(S, t1, t2));
break;
case 8:
/* decode_msaa(8, IMS, X, Y, 0) = (X', Y', S)
@@ -1348,20 +1350,20 @@ brw_blorp_blit_program::decode_msaa(unsigned
num_samples,
* Y' = (Y & ~0b11) >> 1 | (Y & 0b1)
* S = (X & 0b100) | (Y & 0b10) | (X & 0b10) >> 1
*/
- emit_and(t1, X, brw_imm_uw(0xfff8)); /* X & ~0b111 */
- emit_shr(t1, t1, brw_imm_uw(2)); /* (X & ~0b111) >> 2 */
- emit_and(t2, X, brw_imm_uw(1)); /* X & 0b1 */
- emit_or(Xp, t1, t2);
- emit_and(t1, Y, brw_imm_uw(0xfffc)); /* Y & ~0b11 */
- emit_shr(t1, t1, brw_imm_uw(1)); /* (Y & ~0b11) >> 1 */
- emit_and(t2, Y, brw_imm_uw(1)); /* Y & 0b1 */
- emit_or(Yp, t1, t2);
- emit_and(t1, X, brw_imm_uw(4)); /* X & 0b100 */
- emit_and(t2, Y, brw_imm_uw(2)); /* Y & 0b10 */
- emit_or(t1, t1, t2); /* (X & 0b100) | (Y & 0b10) */
- emit_and(t2, X, brw_imm_uw(2)); /* X & 0b10 */
- emit_shr(t2, t2, brw_imm_uw(1)); /* (X & 0b10) >> 1 */
- emit_or(S, t1, t2);
+ emit(AND(t1, X, brw_imm_uw(0xfff8))); /* X & ~0b111 */
+ emit(SHR(t1, t1, brw_imm_uw(2))); /* (X & ~0b111) >> 2 */
+ emit(AND(t2, X, brw_imm_uw(1))); /* X & 0b1 */
+ emit(OR(Xp, t1, t2));
+ emit(AND(t1, Y, brw_imm_uw(0xfffc))); /* Y & ~0b11 */
+ emit(SHR(t1, t1, brw_imm_uw(1))); /* (Y & ~0b11) >> 1 */
+ emit(AND(t2, Y, brw_imm_uw(1))); /* Y & 0b1 */
+ emit(OR(Yp, t1, t2));
+ emit(AND(t1, X, brw_imm_uw(4))); /* X & 0b100 */
+ emit(AND(t2, Y, brw_imm_uw(2))); /* Y & 0b10 */
+ emit(OR(t1, t1, t2)); /* (X & 0b100) | (Y & 0b10) */
+ emit(AND(t2, X, brw_imm_uw(2))); /* X & 0b10 */
+ emit(SHR(t2, t2, brw_imm_uw(1))); /* (X & 0b10) >> 1 */
+ emit(OR(S, t1, t2));
break;
}
s_is_zero = false;
@@ -1383,24 +1385,24 @@ brw_blorp_blit_program::translate_dst_to_src()
struct brw_reg Yp_f = retype(Yp, BRW_REGISTER_TYPE_F);
/* Move the UD coordinates to float registers. */
- emit_mov(Xp_f, X);
- emit_mov(Yp_f, Y);
+ emit(MOV(Xp_f, X));
+ emit(MOV(Yp_f, Y));
/* Scale and offset */
- emit_mul(X_f, Xp_f, x_transform.multiplier);
- emit_mul(Y_f, Yp_f, y_transform.multiplier);
- emit_add(X_f, X_f, x_transform.offset);
- emit_add(Y_f, Y_f, y_transform.offset);
+ emit(MUL(X_f, Xp_f, x_transform.multiplier));
+ emit(MUL(Y_f, Yp_f, y_transform.multiplier));
+ emit(ADD(X_f, X_f, x_transform.offset));
+ emit(ADD(Y_f, Y_f, y_transform.offset));
if (key->blit_scaled && key->blend) {
/* Translate coordinates to lay out the samples in a rectangular grid
* roughly corresponding to sample locations.
*/
- emit_mul(X_f, X_f, brw_imm_f(key->x_scale));
- emit_mul(Y_f, Y_f, brw_imm_f(key->y_scale));
+ emit(MUL(X_f, X_f, brw_imm_f(key->x_scale)));
+ emit(MUL(Y_f, Y_f, brw_imm_f(key->y_scale)));
/* Adjust coordinates so that integers represent pixel centers rather
* than pixel edges.
*/
- emit_add(X_f, X_f, brw_imm_f(-0.5));
- emit_add(Y_f, Y_f, brw_imm_f(-0.5));
+ emit(ADD(X_f, X_f, brw_imm_f(-0.5)));
+ emit(ADD(Y_f, Y_f, brw_imm_f(-0.5)));
/* Clamp the X, Y texture coordinates to properly handle the sampling of
* texels on texture edges.
@@ -1412,21 +1414,21 @@ brw_blorp_blit_program::translate_dst_to_src()
/* Store the fractional parts to be used as bilinear interpolation
* coefficients.
*/
- emit_frc(x_frac, X_f);
- emit_frc(y_frac, Y_f);
+ emit(FRC(x_frac, X_f));
+ emit(FRC(y_frac, Y_f));
/* Round the float coordinates down to nearest integer */
- emit_rndd(Xp_f, X_f);
- emit_rndd(Yp_f, Y_f);
- emit_mul(X_f, Xp_f, brw_imm_f(1 / key->x_scale));
- emit_mul(Y_f, Yp_f, brw_imm_f(1 / key->y_scale));
+ emit(RNDD(Xp_f, X_f));
+ emit(RNDD(Yp_f, Y_f));
+ emit(MUL(X_f, Xp_f, brw_imm_f(1 / key->x_scale)));
+ emit(MUL(Y_f, Yp_f, brw_imm_f(1 / key->y_scale)));
SWAP_XY_AND_XPYP();
} else if (!key->bilinear_filter) {
/* Round the float coordinates down to nearest integer by moving to
* UD registers.
*/
- emit_mov(Xp, X_f);
- emit_mov(Yp, Y_f);
+ emit(MOV(Xp, X_f));
+ emit(MOV(Yp, Y_f));
SWAP_XY_AND_XPYP();
}
}
@@ -1458,10 +1460,10 @@ brw_blorp_blit_program::single_to_blend()
* that maxe up a pixel). So we need to multiply our X and Y coordinates
* each by 2 and then add 1.
*/
- emit_shl(t1, X, brw_imm_w(1));
- emit_shl(t2, Y, brw_imm_w(1));
- emit_add(Xp, t1, brw_imm_w(1));
- emit_add(Yp, t2, brw_imm_w(1));
+ emit(SHL(t1, X, brw_imm_w(1)));
+ emit(SHL(t2, Y, brw_imm_w(1)));
+ emit(ADD(Xp, t1, brw_imm_w(1)));
+ emit(ADD(Yp, t2, brw_imm_w(1)));
SWAP_XY_AND_XPYP();
}
@@ -1528,7 +1530,7 @@ brw_blorp_blit_program::manual_blend_average(unsigned
num_samples)
s_is_zero = true;
} else {
s_is_zero = false;
- emit_mov(vec16(S), brw_imm_ud(i));
+ emit(MOV(vec16(S), brw_imm_ud(i)));
}
texel_fetch(texture_data[stack_depth++]);
@@ -1547,7 +1549,8 @@ brw_blorp_blit_program::manual_blend_average(unsigned
num_samples)
* Since we have already sampled from sample 0, all we need to do is
* skip the remaining fetches and averaging if MCS is zero.
*/
- emit_cmp_if(BRW_CONDITIONAL_NZ, mcs_data, brw_imm_ud(0));
+ emit_cmp(BRW_CONDITIONAL_NZ, mcs_data, brw_imm_ud(0));
+ emit(BRW_OPCODE_IF);
}
/* Do count_trailing_one_bits(i) times */
@@ -1573,14 +1576,14 @@ brw_blorp_blit_program::manual_blend_average(unsigned
num_samples)
/* Scale the result down by a factor of num_samples */
/* TODO: should use a smaller loop bound for non-RGBA formats */
for (int j = 0; j < 4; ++j) {
- emit_mul(offset(texture_data[0], 2*j),
- offset(vec8(texture_data[0]), 2*j),
- brw_imm_f(1.0/num_samples));
+ emit(MUL(offset(texture_data[0], 2*j),
+ offset(vec8(texture_data[0]), 2*j),
+ brw_imm_f(1.0/num_samples)));
}
}
if (key->tex_layout == INTEL_MSAA_LAYOUT_CMS)
- emit_endif();
+ emit(BRW_OPCODE_ENDIF);
}
void
@@ -1607,12 +1610,12 @@ brw_blorp_blit_program::manual_blend_bilinear(unsigned
num_samples)
s_is_zero = false;
/* Compute pixel coordinates */
- emit_add(vec16(x_sample_coords), Xp_f,
- brw_imm_f((float)(i & 0x1) * (1.0 / key->x_scale)));
- emit_add(vec16(y_sample_coords), Yp_f,
- brw_imm_f((float)((i >> 1) & 0x1) * (1.0 / key->y_scale)));
- emit_mov(vec16(X), x_sample_coords);
- emit_mov(vec16(Y), y_sample_coords);
+ emit(ADD(vec16(x_sample_coords), Xp_f,
+ brw_imm_f((float)(i & 0x1) * (1.0 / key->x_scale))));
+ emit(ADD(vec16(y_sample_coords), Yp_f,
+ brw_imm_f((float)((i >> 1) & 0x1) * (1.0 / key->y_scale))));
+ emit(MOV(vec16(X), x_sample_coords));
+ emit(MOV(vec16(Y), y_sample_coords));
/* The MCS value we fetch has to match up with the pixel that we're
* sampling from. Since we sample from different pixels in each
@@ -1646,31 +1649,32 @@ brw_blorp_blit_program::manual_blend_bilinear(unsigned
num_samples)
* | 6 | 7 | | 7 | 1 |
* --------- ---------
*/
- emit_frc(vec16(t1_f), x_sample_coords);
- emit_frc(vec16(t2_f), y_sample_coords);
- emit_mul(vec16(t1_f), t1_f, brw_imm_f(key->x_scale));
- emit_mul(vec16(t2_f), t2_f, brw_imm_f(key->x_scale * key->y_scale));
- emit_add(vec16(t1_f), t1_f, t2_f);
- emit_mov(vec16(S), t1_f);
+ emit(FRC(vec16(t1_f), x_sample_coords));
+ emit(FRC(vec16(t2_f), y_sample_coords));
+ emit(MUL(vec16(t1_f), t1_f, brw_imm_f(key->x_scale)));
+ emit(MUL(vec16(t2_f), t2_f, brw_imm_f(key->x_scale * key->y_scale)));
+ emit(ADD(vec16(t1_f), t1_f, t2_f));
+ emit(MOV(vec16(S), t1_f));
if (num_samples == 8) {
/* Map the sample index to a sample number */
- emit_cmp_if(BRW_CONDITIONAL_L, S, brw_imm_d(4));
+ emit_cmp(BRW_CONDITIONAL_L, S, brw_imm_d(4));
+ emit(BRW_OPCODE_IF);
{
- emit_mov(vec16(t2), brw_imm_d(5));
+ emit(MOV(vec16(t2), brw_imm_d(5)));
emit_if_eq_mov(S, 1, vec16(t2), 2);
emit_if_eq_mov(S, 2, vec16(t2), 4);
emit_if_eq_mov(S, 3, vec16(t2), 6);
}
- emit_else();
+ emit(BRW_OPCODE_ELSE);
{
- emit_mov(vec16(t2), brw_imm_d(0));
+ emit(MOV(vec16(t2), brw_imm_d(0)));
emit_if_eq_mov(S, 5, vec16(t2), 3);
emit_if_eq_mov(S, 6, vec16(t2), 7);
emit_if_eq_mov(S, 7, vec16(t2), 1);
}
- emit_endif();
- emit_mov(vec16(S), t2);
+ emit(BRW_OPCODE_ENDIF);
+ emit(MOV(vec16(S), t2));
}
texel_fetch(texture_data[i]);
}
@@ -1682,17 +1686,17 @@ brw_blorp_blit_program::manual_blend_bilinear(unsigned
num_samples)
* the color components.
*/
for (int k = 0; k < 8; k += 2)
- emit_lrp(vec8(SAMPLE(index - 1, k)),
+ emit(LRP(vec8(SAMPLE(index - 1, k)),
x_frac,
vec8(SAMPLE(index, k)),
- vec8(SAMPLE(index - 1, k)));
+ vec8(SAMPLE(index - 1, k))));
index -= 2;
}
for (int k = 0; k < 8; k += 2)
- emit_lrp(vec8(SAMPLE(0, k)),
+ emit(LRP(vec8(SAMPLE(0, k)),
y_frac,
vec8(SAMPLE(2, k)),
- vec8(SAMPLE(0, k)));
+ vec8(SAMPLE(0, k))));
#undef SAMPLE
}
@@ -1810,23 +1814,23 @@ brw_blorp_blit_program::texture_lookup(struct brw_reg
dst,
switch (args[arg]) {
case SAMPLER_MESSAGE_ARG_U_FLOAT:
if (key->bilinear_filter)
- emit_mov(retype(mrf, BRW_REGISTER_TYPE_F),
- retype(X, BRW_REGISTER_TYPE_F));
+ emit(MOV(retype(mrf, BRW_REGISTER_TYPE_F),
+ retype(X, BRW_REGISTER_TYPE_F)));
else
- emit_mov(retype(mrf, BRW_REGISTER_TYPE_F), X);
+ emit(MOV(retype(mrf, BRW_REGISTER_TYPE_F), X));
break;
case SAMPLER_MESSAGE_ARG_V_FLOAT:
if (key->bilinear_filter)
- emit_mov(retype(mrf, BRW_REGISTER_TYPE_F),
- retype(Y, BRW_REGISTER_TYPE_F));
+ emit(MOV(retype(mrf, BRW_REGISTER_TYPE_F),
+ retype(Y, BRW_REGISTER_TYPE_F)));
else
- emit_mov(retype(mrf, BRW_REGISTER_TYPE_F), Y);
+ emit(MOV(retype(mrf, BRW_REGISTER_TYPE_F), Y));
break;
case SAMPLER_MESSAGE_ARG_U_INT:
- emit_mov(mrf, X);
+ emit(MOV(mrf, X));
break;
case SAMPLER_MESSAGE_ARG_V_INT:
- emit_mov(mrf, Y);
+ emit(MOV(mrf, Y));
break;
case SAMPLER_MESSAGE_ARG_SI_INT:
/* Note: on Gen7, this code may be reached with s_is_zero==true
@@ -1835,14 +1839,14 @@ brw_blorp_blit_program::texture_lookup(struct brw_reg
dst,
* appropriate message register.
*/
if (s_is_zero)
- emit_mov(mrf, brw_imm_ud(0));
+ emit(MOV(mrf, brw_imm_ud(0)));
else
- emit_mov(mrf, S);
+ emit(MOV(mrf, S));
break;
case SAMPLER_MESSAGE_ARG_MCS_INT:
switch (key->tex_layout) {
case INTEL_MSAA_LAYOUT_CMS:
- emit_mov(mrf, mcs_data);
+ emit(MOV(mrf, mcs_data));
break;
case INTEL_MSAA_LAYOUT_IMS:
/* When sampling from an IMS surface, MCS data is not relevant,
@@ -1858,7 +1862,7 @@ brw_blorp_blit_program::texture_lookup(struct brw_reg dst,
}
break;
case SAMPLER_MESSAGE_ARG_ZERO_INT:
- emit_mov(mrf, brw_imm_ud(0));
+ emit(MOV(mrf, brw_imm_ud(0)));
break;
}
mrf.nr += 2;
@@ -1890,16 +1894,16 @@ brw_blorp_blit_program::render_target_write()
bool use_header = key->use_kill;
if (use_header) {
/* Copy R0/1 to MRF */
- emit_mov(retype(mrf_rt_write, BRW_REGISTER_TYPE_UD),
- retype(R0, BRW_REGISTER_TYPE_UD));
+ emit(MOV(retype(mrf_rt_write, BRW_REGISTER_TYPE_UD),
+ retype(R0, BRW_REGISTER_TYPE_UD)));
mrf_offset += 2;
}
/* Copy texture data to MRFs */
for (int i = 0; i < 4; ++i) {
/* E.g. mov(16) m2.0<1>:f r2.0<8;8,1>:f { Align1, H1 } */
- emit_mov(offset(mrf_rt_write, mrf_offset),
- offset(vec8(texture_data[0]), 2*i));
+ emit(MOV(offset(mrf_rt_write, mrf_offset),
+ offset(vec8(texture_data[0]), 2*i)));
mrf_offset += 2;
}
diff --git a/src/mesa/drivers/dri/i965/brw_blorp_blit_eu.cpp
b/src/mesa/drivers/dri/i965/brw_blorp_blit_eu.cpp
index 38969d8..cd98423 100644
--- a/src/mesa/drivers/dri/i965/brw_blorp_blit_eu.cpp
+++ b/src/mesa/drivers/dri/i965/brw_blorp_blit_eu.cpp
@@ -26,14 +26,14 @@
#include "brw_blorp.h"
brw_blorp_eu_emitter::brw_blorp_eu_emitter(struct brw_context *brw)
- : mem_ctx(ralloc_context(NULL)), c(rzalloc(mem_ctx, struct brw_wm_compile)),
- generator(brw, c, NULL, NULL, false)
+ : fs_emitter(brw, rzalloc(NULL, struct brw_wm_compile), 16),
+ generator(brw, fs_emitter::c, NULL, NULL, false)
{
}
brw_blorp_eu_emitter::~brw_blorp_eu_emitter()
{
- ralloc_free(mem_ctx);
+ ralloc_free(fs_emitter::c);
}
const unsigned *
@@ -43,10 +43,11 @@ brw_blorp_eu_emitter::get_program(unsigned *program_size,
FILE *dump_file)
if (unlikely(INTEL_DEBUG & DEBUG_BLORP)) {
fprintf(stderr, "Native code for BLORP blit:\n");
- res = generator.generate_assembly(NULL, &insts, program_size, dump_file);
+ res = generator.generate_assembly(NULL, &instructions, program_size,
+ dump_file);
fprintf(stderr, "\n");
} else {
- res = generator.generate_assembly(NULL, &insts, program_size);
+ res = generator.generate_assembly(NULL, &instructions, program_size);
}
return res;
@@ -73,9 +74,8 @@ brw_blorp_eu_emitter::emit_kill_if_outside_rect(const struct
brw_reg &x,
emit_cmp(BRW_CONDITIONAL_L, x, dst_x1)->predicate = BRW_PREDICATE_NORMAL;
emit_cmp(BRW_CONDITIONAL_L, y, dst_y1)->predicate = BRW_PREDICATE_NORMAL;
- fs_inst *inst = new (mem_ctx) fs_inst(BRW_OPCODE_AND, g1, f0, g1);
+ fs_inst *inst = emit(BRW_OPCODE_AND, g1, f0, g1);
inst->force_writemask_all = true;
- insts.push_tail(inst);
}
void
@@ -84,14 +84,12 @@ brw_blorp_eu_emitter::emit_texture_lookup(const struct
brw_reg &dst,
unsigned base_mrf,
unsigned msg_length)
{
- fs_inst *inst = new (mem_ctx) fs_inst(op, dst, brw_message_reg(base_mrf));
+ fs_inst *inst = emit(op, dst, brw_message_reg(base_mrf));
inst->base_mrf = base_mrf;
inst->mlen = msg_length;
inst->sampler = 0;
inst->header_present = false;
-
- insts.push_tail(inst);
}
void
@@ -100,15 +98,13 @@ brw_blorp_eu_emitter::emit_render_target_write(const
struct brw_reg &src0,
unsigned msg_length,
bool use_header)
{
- fs_inst *inst = new (mem_ctx) fs_inst(FS_OPCODE_BLORP_FB_WRITE);
+ fs_inst *inst = emit(FS_OPCODE_BLORP_FB_WRITE);
inst->src[0] = src0;
inst->base_mrf = msg_reg_nr;
inst->mlen = msg_length;
inst->header_present = use_header;
inst->target = BRW_BLORP_RENDERBUFFER_BINDING_TABLE_INDEX;
-
- insts.push_tail(inst);
}
void
@@ -119,7 +115,7 @@ brw_blorp_eu_emitter::emit_combine(enum opcode
combine_opcode,
{
assert(combine_opcode == BRW_OPCODE_ADD || combine_opcode ==
BRW_OPCODE_AVG);
- insts.push_tail(new (mem_ctx) fs_inst(combine_opcode, dst, src_1, src_2));
+ emit(combine_opcode, dst, src_1, src_2);
}
fs_inst *
@@ -127,10 +123,8 @@ brw_blorp_eu_emitter::emit_cmp(int op,
const struct brw_reg &x,
const struct brw_reg &y)
{
- fs_inst *cmp = new (mem_ctx) fs_inst(BRW_OPCODE_CMP,
- vec16(brw_null_reg()), x, y);
+ fs_inst *cmp = emit(BRW_OPCODE_CMP, vec16(brw_null_reg()), x, y);
cmp->conditional_mod = op;
- insts.push_tail(cmp);
return cmp;
}
diff --git a/src/mesa/drivers/dri/i965/brw_blorp_blit_eu.h
b/src/mesa/drivers/dri/i965/brw_blorp_blit_eu.h
index c10695e..4026ca4 100644
--- a/src/mesa/drivers/dri/i965/brw_blorp_blit_eu.h
+++ b/src/mesa/drivers/dri/i965/brw_blorp_blit_eu.h
@@ -27,7 +27,7 @@
#include "brw_context.h"
#include "brw_fs.h"
-class brw_blorp_eu_emitter
+class brw_blorp_eu_emitter : public fs_emitter
{
protected:
explicit brw_blorp_eu_emitter(struct brw_context *brw);
@@ -65,9 +65,8 @@ protected:
{
emit_cmp(op, x, y);
- fs_inst *mv = new (mem_ctx) fs_inst(BRW_OPCODE_MOV, dst, src);
+ fs_inst *mv = emit(MOV(dst, src));
mv->predicate = BRW_PREDICATE_NORMAL;
- insts.push_tail(mv);
}
inline void emit_if_eq_mov(const struct brw_reg &x, unsigned y,
@@ -76,114 +75,17 @@ protected:
emit_cond_mov(x, brw_imm_d(y), BRW_CONDITIONAL_EQ, dst, brw_imm_d(src));
}
- inline void emit_lrp(const struct brw_reg &dst,
- const struct brw_reg &src1,
- const struct brw_reg &src2,
- const struct brw_reg &src3)
- {
- insts.push_tail(
- new (mem_ctx) fs_inst(BRW_OPCODE_LRP, dst, src1, src2, src3));
- }
-
- inline void emit_mov(const struct brw_reg& dst, const struct brw_reg& src)
- {
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_MOV, dst, src));
- }
-
- inline void emit_mov_8(const struct brw_reg& dst, const struct brw_reg& src)
- {
- fs_inst *mv = new (mem_ctx) fs_inst(BRW_OPCODE_MOV, dst, src);
- mv->force_uncompressed = true;
- insts.push_tail(mv);
- }
-
- inline void emit_and(const struct brw_reg& dst,
- const struct brw_reg& src1,
- const struct brw_reg& src2)
- {
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_AND, dst, src1, src2));
- }
-
- inline void emit_add(const struct brw_reg& dst,
- const struct brw_reg& src1,
- const struct brw_reg& src2)
- {
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_ADD, dst, src1, src2));
- }
-
- inline void emit_add_8(const struct brw_reg& dst,
- const struct brw_reg& src1,
- const struct brw_reg& src2)
- {
- fs_inst *add = new (mem_ctx) fs_inst(BRW_OPCODE_ADD, dst, src1, src2);
- add->force_uncompressed = true;
- insts.push_tail(add);
- }
-
- inline void emit_mul(const struct brw_reg& dst,
- const struct brw_reg& src1,
- const struct brw_reg& src2)
- {
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_MUL, dst, src1, src2));
- }
-
- inline void emit_shr(const struct brw_reg& dst,
- const struct brw_reg& src1,
- const struct brw_reg& src2)
- {
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_SHR, dst, src1, src2));
- }
-
- inline void emit_shl(const struct brw_reg& dst,
- const struct brw_reg& src1,
- const struct brw_reg& src2)
- {
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_SHL, dst, src1, src2));
- }
-
- inline void emit_or(const struct brw_reg& dst,
- const struct brw_reg& src1,
- const struct brw_reg& src2)
- {
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_OR, dst, src1, src2));
- }
-
- inline void emit_frc(const struct brw_reg& dst,
- const struct brw_reg& src)
- {
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_FRC, dst, src));
- }
-
- inline void emit_rndd(const struct brw_reg& dst,
- const struct brw_reg& src)
- {
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_RNDD, dst, src));
- }
-
- inline void emit_cmp_if(int op,
- const struct brw_reg &x,
- const struct brw_reg &y)
- {
- emit_cmp(op, x, y);
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_IF));
- }
+ fs_inst *emit_cmp(int op, const struct brw_reg &x, const struct brw_reg &y);
- inline void emit_else(void)
+private:
+ virtual void dump_instruction(backend_instruction *)
{
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_ELSE));
}
- inline void emit_endif(void)
+ virtual void invalidate_live_intervals()
{
- insts.push_tail(new (mem_ctx) fs_inst(BRW_OPCODE_ENDIF));
}
-private:
- fs_inst *emit_cmp(int op, const struct brw_reg &x, const struct brw_reg &y);
-
- void *mem_ctx;
- struct brw_wm_compile *c;
- exec_list insts;
fs_generator generator;
};
--
1.8.3.1
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
