vc4: Don't bother masking out the low 24 bits for integer multiplies

[mesa.git] / src / gallium / drivers / vc4 / vc4_program.c

diff --git a/src/gallium/drivers/vc4/vc4_program.c b/src/gallium/drivers/vc4/vc4_program.c
index 56a3a96c1b5c0472c50ed36b6cfa9a698dfea840..9e145e54ccd8ac8bfc117332a20f11f62c849f9f 100644 (file)
--- a/src/gallium/drivers/vc4/vc4_program.c
+++ b/src/gallium/drivers/vc4/vc4_program.c
@@ -366,22 +366,18 @@ tgsi_to_qir_umul(struct vc4_compile *c,
                  struct tgsi_full_instruction *tgsi_inst,
                  enum qop op, struct qreg *src, int i)
 {
-        struct qreg src0_hi = qir_SHR(c, src[0 * 4 + i],
-                                      qir_uniform_ui(c, 16));
-        struct qreg src0_lo = qir_AND(c, src[0 * 4 + i],
-                                      qir_uniform_ui(c, 0xffff));
-        struct qreg src1_hi = qir_SHR(c, src[1 * 4 + i],
-                                      qir_uniform_ui(c, 16));
-        struct qreg src1_lo = qir_AND(c, src[1 * 4 + i],
-                                      qir_uniform_ui(c, 0xffff));
-
-        struct qreg hilo = qir_MUL24(c, src0_hi, src1_lo);
-        struct qreg lohi = qir_MUL24(c, src0_lo, src1_hi);
-        struct qreg lolo = qir_MUL24(c, src0_lo, src1_lo);
+        struct qreg src0 = src[0 * 4 + i];
+        struct qreg src0_hi = qir_SHR(c, src0, qir_uniform_ui(c, 24));
+        struct qreg src1 = src[1 * 4 + i];
+        struct qreg src1_hi = qir_SHR(c, src1, qir_uniform_ui(c, 24));
+
+        struct qreg hilo = qir_MUL24(c, src0_hi, src1);
+        struct qreg lohi = qir_MUL24(c, src0, src1_hi);
+        struct qreg lolo = qir_MUL24(c, src0, src1);
 
         return qir_ADD(c, lolo, qir_SHL(c,
                                         qir_ADD(c, hilo, lohi),
-                                        qir_uniform_ui(c, 16)));
+                                        qir_uniform_ui(c, 24)));
 }
 
 static struct qreg