agx: Model both sources of sample_mask

We need to control both sources to implement multisampling properly. The
semantic is something like:

   foreach sample in the first mask {
      if correspond bit in second bit set {
         make sample live
      } else {
         make sample dead
      }
   }

But I'm reticent to document more formally until the details are really
understood and properly tested.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23480>

src/asahi/compiler/agx_compile.c

@@ -428,7 +428,7 @@ agx_write_sample_mask_1(agx_builder *b)
        * we are therefore still active and need to write a full sample mask.
        * TODO: interactions with MSAA and gl_SampleMask writes
        */
-      agx_sample_mask(b, agx_immediate(1));
+      agx_sample_mask(b, agx_immediate(0xFF), agx_immediate(1));
       agx_signal_pix(b, 1);
       b->shader->did_sample_mask = true;
 
@@ -688,7 +688,7 @@ agx_emit_discard(agx_builder *b)
    b->shader->did_writeout = true;
 
    b->shader->out->writes_sample_mask = true;
-   agx_sample_mask(b, agx_immediate(0));
+   agx_sample_mask(b, agx_immediate(0xFF), agx_immediate(0));
    return agx_signal_pix(b, 1);
 }
 

src/asahi/compiler/agx_lower_uniform_sources.c

@@ -41,6 +41,7 @@ should_lower(enum agx_opcode op, agx_index uniform, unsigned src_index)
    case AGX_OPCODE_UNIFORM_STORE:
    case AGX_OPCODE_ST_VARY:
    case AGX_OPCODE_LOCAL_ATOMIC:
+   case AGX_OPCODE_SAMPLE_MASK:
       return true;
    default:
       return false;

src/asahi/compiler/agx_opcodes.py

@@ -292,7 +292,7 @@ op("wait", (0x38, 0xFF, 2, _), dests = 0,
 
 op("get_sr", (0x72, 0x7F | L, 4, _), dests = 1, imms = [SR])
 
-op("sample_mask", (0x7fc1, 0xffff, 6, _), dests = 0, srcs = 1, can_eliminate = False)
+op("sample_mask", (0x7fc1, 0xffff, 6, _), dests = 0, srcs = 2, can_eliminate = False)
 
 # Sources: sample mask, combined depth/stencil
 op("zs_emit", (0x41, 0xFF | L, 4, _), dests = 0, srcs = 2,

src/asahi/compiler/agx_pack.c

@@ -502,9 +502,10 @@ agx_pack_instr(struct util_dynarray *emission, struct util_dynarray *fixups,
    }
 
    case AGX_OPCODE_SAMPLE_MASK: {
-      unsigned S = agx_pack_sample_mask_src(I->src[0]);
+      unsigned S = agx_pack_sample_mask_src(I->src[1]);
-      unsigned T = 0xFF;
+      unsigned T = I->src[0].value;
-      bool Tt = true /* immediate */;
+      bool Tt = I->src[0].type == AGX_INDEX_IMMEDIATE;
+      assert(Tt || I->src[0].type == AGX_INDEX_REGISTER);
       uint32_t raw = 0xc1 | (Tt ? BITFIELD_BIT(8) : 0) |
                      ((T & BITFIELD_MASK(6)) << 9) | ((S & 0xff) << 16) |
                      ((T >> 6) << 24) | ((S >> 8) << 26);