From: Cesar Strauss <cestrauss@gmail.com>
Date: Sat, 10 Apr 2021 20:27:48 +0000 (-0300)
Subject: Implement 1<<r3 predicate mode
X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=f903e20aff0178bd1bab4c289d2f7e8a9abc6c93;p=soc.git

Implement 1<<r3 predicate mode

The mask bit selected by r3 is set to one.
A possible optimization would be to do step = r3 directly, but this is only
valid in non-zero mode.
The corresponding test cases now pass.
---

diff --git a/src/soc/simple/issuer.py b/src/soc/simple/issuer.py
index dfd0b96a..b2711cbb 100644
--- a/src/soc/simple/issuer.py
+++ b/src/soc/simple/issuer.py
@@ -467,8 +467,14 @@ class TestIssuerInternal(Elaboratable):
                 # store destination mask
                 inv = Repl(dinvert, 64)
                 new_dstmask = Signal(64)
-                # invert mask if requested
-                comb += new_dstmask.eq(self.int_pred.data_o ^ inv)
+                with m.If(dunary):
+                    # set selected mask bit for 1<<r3 mode
+                    dst_shift = Signal(range(64))
+                    comb += dst_shift.eq(self.int_pred.data_o & 0b111111)
+                    comb += new_dstmask.bit_select(dst_shift, 1).eq(1)
+                with m.Else():
+                    # invert mask if requested
+                    comb += new_dstmask.eq(self.int_pred.data_o ^ inv)
                 # shift-out already used mask bits
                 sync += self.dstmask.eq(new_dstmask >> dststep)
                 # skip fetching source mask register, when zero
@@ -485,8 +491,14 @@ class TestIssuerInternal(Elaboratable):
                 # store source mask
                 inv = Repl(sinvert, 64)
                 new_srcmask = Signal(64)
-                # invert mask if requested
-                comb += new_srcmask.eq(self.int_pred.data_o ^ inv)
+                with m.If(sunary):
+                    # set selected mask bit for 1<<r3 mode
+                    src_shift = Signal(range(64))
+                    comb += src_shift.eq(self.int_pred.data_o & 0b111111)
+                    comb += new_srcmask.bit_select(src_shift, 1).eq(1)
+                with m.Else():
+                    # invert mask if requested
+                    comb += new_srcmask.eq(self.int_pred.data_o ^ inv)
                 # shift-out already used mask bits
                 sync += self.srcmask.eq(new_srcmask >> srcstep)
                 m.next = "FETCH_PRED_DONE"