pan/mdg: Fix auxiliary load/store swizzle packing

It needs to respect the existing swizzle, as well as the type size.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6321>

diff --git a/src/panfrost/midgard/helpers.h b/src/panfrost/midgard/helpers.h
index 6f6f82cc70aa302c981d71abae32c69474fd2964..73f82a5aaaf241c8a5df4739718d783baa256947 100644 (file)
--- a/src/panfrost/midgard/helpers.h
+++ b/src/panfrost/midgard/helpers.h
@@ -310,9 +310,19 @@ mir_is_simple_swizzle(unsigned *swizzle, unsigned mask)
 /* Packs a load/store argument */
 
 static inline uint8_t
 /* Packs a load/store argument */
 
 static inline uint8_t

-midgard_ldst_reg(unsigned reg, unsigned component)
+midgard_ldst_reg(unsigned reg, unsigned component, unsigned size)

 {
         assert((reg == REGISTER_LDST_BASE) || (reg == REGISTER_LDST_BASE + 1));
 {
         assert((reg == REGISTER_LDST_BASE) || (reg == REGISTER_LDST_BASE + 1));

+        assert(size == 16 || size == 32 || size == 64);
+
+        /* Shift so everything is in terms of 32-bit units */
+        if (size == 64) {
+                assert(component < 2);
+                component <<= 1;
+        } else if (size == 16) {
+                assert((component & 1) == 0);
+                component >>= 1;
+        }

 
         midgard_ldst_register_select sel = {
                 .component = component,
 
         midgard_ldst_register_select sel = {
                 .component = component,

diff --git a/src/panfrost/midgard/midgard_emit.c b/src/panfrost/midgard/midgard_emit.c
index 8b022554a94068c0fc32ee540f8a97aa500df1d0..77f2e3b7bf2c7a3da7ea59c2e4704ac45300de69 100644 (file)
--- a/src/panfrost/midgard/midgard_emit.c
+++ b/src/panfrost/midgard/midgard_emit.c
@@ -501,12 +501,14 @@ load_store_from_instr(midgard_instruction *ins)
 
         if (ins->src[1] != ~0) {
                 unsigned src = SSA_REG_FROM_FIXED(ins->src[1]);
 
         if (ins->src[1] != ~0) {
                 unsigned src = SSA_REG_FROM_FIXED(ins->src[1]);

-                ldst.arg_1 |= midgard_ldst_reg(src, ins->swizzle[1][0]);
+                unsigned sz = nir_alu_type_get_type_size(ins->src_types[1]);
+                ldst.arg_1 |= midgard_ldst_reg(src, ins->swizzle[1][0], sz);

         }
 
         if (ins->src[2] != ~0) {
                 unsigned src = SSA_REG_FROM_FIXED(ins->src[2]);
         }
 
         if (ins->src[2] != ~0) {
                 unsigned src = SSA_REG_FROM_FIXED(ins->src[2]);

-                ldst.arg_2 |= midgard_ldst_reg(src, ins->swizzle[2][0]);
+                unsigned sz = nir_alu_type_get_type_size(ins->src_types[2]);
+                ldst.arg_2 |= midgard_ldst_reg(src, ins->swizzle[2][0], sz);

         }
 
         return ldst;
         }
 
         return ldst;

diff --git a/src/panfrost/midgard/midgard_ra.c b/src/panfrost/midgard/midgard_ra.c
index b06eb97ab8ec77f96e3126fc9bb0416a8fb9ca1a..bb95d3854298f9277df3aa815a385eb8f2c650ab 100644 (file)
--- a/src/panfrost/midgard/midgard_ra.c
+++ b/src/panfrost/midgard/midgard_ra.c
@@ -723,19 +723,19 @@ install_registers_instr(
                 unsigned src3 = ins->src[2];
 
                 if (src2 != ~0) {
                 unsigned src3 = ins->src[2];
 
                 if (src2 != ~0) {

-                        struct phys_reg src = index_to_reg(ctx, l, src2, 2);
+                        struct phys_reg src = index_to_reg(ctx, l, src2, src_shift[1]);

                         unsigned component = src.offset >> src.shift;
                         assert(component << src.shift == src.offset);
                         ins->src[1] = SSA_FIXED_REGISTER(src.reg);
                         unsigned component = src.offset >> src.shift;
                         assert(component << src.shift == src.offset);
                         ins->src[1] = SSA_FIXED_REGISTER(src.reg);

-                        ins->swizzle[1][0] = component;
+                        ins->swizzle[1][0] += component;

                 }
 
                 if (src3 != ~0) {
                 }
 
                 if (src3 != ~0) {

-                        struct phys_reg src = index_to_reg(ctx, l, src3, 2);
+                        struct phys_reg src = index_to_reg(ctx, l, src3, src_shift[2]);

                         unsigned component = src.offset >> src.shift;
                         assert(component << src.shift == src.offset);
                         ins->src[2] = SSA_FIXED_REGISTER(src.reg);
                         unsigned component = src.offset >> src.shift;
                         assert(component << src.shift == src.offset);
                         ins->src[2] = SSA_FIXED_REGISTER(src.reg);

-                        ins->swizzle[2][0] = component;
+                        ins->swizzle[2][0] += component;

                 }
  
                 break;
                 }
  
                 break;