panfrost/midgard: NIRify blend load scale/convert

The scale and type-convert can now be expressed in NIR, rather than MIR,
which is significantly more maintainable and demonstrates correctness of
the type conversion patches.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

diff --git a/src/gallium/drivers/panfrost/midgard/midgard_compile.c b/src/gallium/drivers/panfrost/midgard/midgard_compile.c
index b7379426b38e50b06979e59406a0bbd3afa17273..70e9666ce71bb8aa8744d4d9314508007720d3f7 100644 (file)
--- a/src/gallium/drivers/panfrost/midgard/midgard_compile.c
+++ b/src/gallium/drivers/panfrost/midgard/midgard_compile.c
@@ -1255,58 +1255,6 @@ emit_fb_read_blend_scalar(compiler_context *ctx, unsigned reg)
                 ins.load_store.unknown = c;
                 emit_mir_instruction(ctx, ins);
         }
-
-        /* vadd.u2f hr2, zext(hr2), #0 */
-
-        midgard_vector_alu_src alu_src = blank_alu_src;
-        alu_src.mod = midgard_int_zero_extend;
-        alu_src.half = true;
-
-        midgard_instruction u2f = {
-                .type = TAG_ALU_4,
-                .ssa_args = {
-                        .src0 = reg,
-                        .src1 = SSA_UNUSED_0,
-                        .dest = reg,
-                        .inline_constant = true
-                },
-                .alu = {
-                        .op = midgard_alu_op_u2f_rtz,
-                        .reg_mode = midgard_reg_mode_16,
-                        .dest_override = midgard_dest_override_none,
-                        .mask = 0xF,
-                        .src1 = vector_alu_srco_unsigned(alu_src),
-                        .src2 = vector_alu_srco_unsigned(blank_alu_src),
-                }
-        };
-
-        emit_mir_instruction(ctx, u2f);
-
-        /* vmul.fmul.sat r1, hr2, #0.00392151 */
-
-        alu_src.mod = 0;
-
-        midgard_instruction fmul = {
-                .type = TAG_ALU_4,
-                .inline_constant = _mesa_float_to_half(1.0 / 255.0),
-                .ssa_args = {
-                        .src0 = reg,
-                        .dest = reg,
-                        .src1 = SSA_UNUSED_0,
-                        .inline_constant = true
-                },
-                .alu = {
-                        .op = midgard_alu_op_fmul,
-                        .reg_mode = midgard_reg_mode_32,
-                        .dest_override = midgard_dest_override_none,
-                        .outmod = midgard_outmod_sat,
-                        .mask = 0xFF,
-                        .src1 = vector_alu_srco_unsigned(alu_src),
-                        .src2 = vector_alu_srco_unsigned(blank_alu_src),
-                }
-        };
-
-        emit_mir_instruction(ctx, fmul);
 }
 
 static void

diff --git a/src/gallium/drivers/panfrost/midgard/nir_lower_framebuffer.c b/src/gallium/drivers/panfrost/midgard/nir_lower_framebuffer.c
index 08ef290a20b2f556163625ee36efbea351463876..5f3115b6ae8eff5faf2d7d9a668418ea7ef9c35f 100644 (file)
--- a/src/gallium/drivers/panfrost/midgard/nir_lower_framebuffer.c
+++ b/src/gallium/drivers/panfrost/midgard/nir_lower_framebuffer.c
@@ -57,7 +57,13 @@ nir_float_to_native(nir_builder *b, nir_ssa_def *c_float)
 static nir_ssa_def *
 nir_native_to_float(nir_builder *b, nir_ssa_def *c_native)
 {
-   return c_native;
+   /* First, we convert up from u8 to f32 */
+   nir_ssa_def *converted = nir_u2f32(b, nir_u2u32(b, c_native));
+
+   /* Next, we scale down from [0, 255.0] to [0, 1] */
+   nir_ssa_def *scaled = nir_fsat(b, nir_fmul_imm(b, converted, 1.0/255.0));
+
+   return scaled;
 }
 
 void
@@ -122,7 +128,7 @@ nir_lower_framebuffer(nir_shader *shader)
 
                new->num_components = 4;
 
-               unsigned bitsize = 32;
+               unsigned bitsize = 8;
                nir_ssa_dest_init(&new->instr, &new->dest, 4, bitsize, NULL);
                nir_builder_instr_insert(&b, &new->instr);