[Qemu-devel] [PULL 30/43] target/ppc: convert vsplt[bhw] to use vector operations

David Gibson posted 43 patches 38 weeks ago
Maintainers: "Hervé Poussineau" <hpoussin@reactos.org>, "Michael S. Tsirkin" <mst@redhat.com>, "Cédric Le Goater" <clg@kaod.org>, Igor Mammedov <imammedo@redhat.com>, Gerd Hoffmann <kraxel@redhat.com>, "Marc-André Lureau" <marcandre.lureau@redhat.com>, Christian Borntraeger <borntraeger@de.ibm.com>, Dmitry Fleytman <dmitry.fleytman@gmail.com>, Paolo Bonzini <pbonzini@redhat.com>, Halil Pasic <pasic@linux.ibm.com>, Fam Zheng <fam@euphon.net>, David Gibson <david@gibson.dropbear.id.au>, Amit Shah <amit@kernel.org>, Collin Walling <walling@linux.ibm.com>, Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>, Marcel Apfelbaum <marcel.apfelbaum@gmail.com>, Richard Henderson <rth@twiddle.net>, David Hildenbrand <david@redhat.com>, Cornelia Huck <cohuck@redhat.com>

[Qemu-devel] [PULL 30/43] target/ppc: convert vsplt[bhw] to use vector operations

Posted by David Gibson 38 weeks ago
From: Richard Henderson <richard.henderson@linaro.org>

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Message-Id: <20190215100058.20015-5-mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
 target/ppc/helper.h                 |  3 --
 target/ppc/int_helper.c             | 19 ------------
 target/ppc/translate/vmx-impl.inc.c | 46 +++++++++++++++++------------
 3 files changed, 27 insertions(+), 41 deletions(-)

diff --git a/target/ppc/helper.h b/target/ppc/helper.h
index 2aa60e5d36..069daa9883 100644
--- a/target/ppc/helper.h
+++ b/target/ppc/helper.h
@@ -246,9 +246,6 @@ DEF_HELPER_3(vrld, void, avr, avr, avr)
 DEF_HELPER_3(vsl, void, avr, avr, avr)
 DEF_HELPER_3(vsr, void, avr, avr, avr)
 DEF_HELPER_4(vsldoi, void, avr, avr, avr, i32)
-DEF_HELPER_3(vspltb, void, avr, avr, i32)
-DEF_HELPER_3(vsplth, void, avr, avr, i32)
-DEF_HELPER_3(vspltw, void, avr, avr, i32)
 DEF_HELPER_3(vextractub, void, avr, avr, i32)
 DEF_HELPER_3(vextractuh, void, avr, avr, i32)
 DEF_HELPER_3(vextractuw, void, avr, avr, i32)
diff --git a/target/ppc/int_helper.c b/target/ppc/int_helper.c
index 2f793a3543..5c9623e952 100644
--- a/target/ppc/int_helper.c
+++ b/target/ppc/int_helper.c
@@ -1869,25 +1869,6 @@ void helper_vslo(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
 #endif
 }
 
-/* Experimental testing shows that hardware masks the immediate.  */
-#define _SPLAT_MASKED(element) (splat & (ARRAY_SIZE(r->element) - 1))
-#define SPLAT_ELEMENT(element) _SPLAT_MASKED(element)
-#define VSPLT(suffix, element, access)                                  \
-    void helper_vsplt##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t splat) \
-    {                                                                   \
-        uint32_t s = b->access(SPLAT_ELEMENT(element));                 \
-        int i;                                                          \
-                                                                        \
-        for (i = 0; i < ARRAY_SIZE(r->element); i++) {                  \
-            r->access(i) = s;                                           \
-        }                                                               \
-    }
-VSPLT(b, u8, VsrB)
-VSPLT(h, u16, VsrH)
-VSPLT(w, u32, VsrW)
-#undef VSPLT
-#undef SPLAT_ELEMENT
-#undef _SPLAT_MASKED
 #if defined(HOST_WORDS_BIGENDIAN)
 #define VINSERT(suffix, element)                                            \
     void helper_vinsert##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \
diff --git a/target/ppc/translate/vmx-impl.inc.c b/target/ppc/translate/vmx-impl.inc.c
index c26c342e16..41ddbd879f 100644
--- a/target/ppc/translate/vmx-impl.inc.c
+++ b/target/ppc/translate/vmx-impl.inc.c
@@ -798,24 +798,32 @@ GEN_VXFORM_NOA(vprtybw, 1, 24);
 GEN_VXFORM_NOA(vprtybd, 1, 24);
 GEN_VXFORM_NOA(vprtybq, 1, 24);
 
-#define GEN_VXFORM_UIMM(name, opc2, opc3)                               \
-static void glue(gen_, name)(DisasContext *ctx)                                 \
-    {                                                                   \
-        TCGv_ptr rb, rd;                                                \
-        TCGv_i32 uimm;                                                  \
-        if (unlikely(!ctx->altivec_enabled)) {                          \
-            gen_exception(ctx, POWERPC_EXCP_VPU);                       \
-            return;                                                     \
-        }                                                               \
-        uimm = tcg_const_i32(UIMM5(ctx->opcode));                       \
-        rb = gen_avr_ptr(rB(ctx->opcode));                              \
-        rd = gen_avr_ptr(rD(ctx->opcode));                              \
-        gen_helper_##name (rd, rb, uimm);                               \
-        tcg_temp_free_i32(uimm);                                        \
-        tcg_temp_free_ptr(rb);                                          \
-        tcg_temp_free_ptr(rd);                                          \
+static void gen_vsplt(DisasContext *ctx, int vece)
+{
+    int uimm, dofs, bofs;
+
+    if (unlikely(!ctx->altivec_enabled)) {
+        gen_exception(ctx, POWERPC_EXCP_VPU);
+        return;
     }
 
+    uimm = UIMM5(ctx->opcode);
+    bofs = avr64_offset(rB(ctx->opcode), true);
+    dofs = avr64_offset(rD(ctx->opcode), true);
+
+    /* Experimental testing shows that hardware masks the immediate.  */
+    bofs += (uimm << vece) & 15;
+#ifndef HOST_WORDS_BIGENDIAN
+    bofs ^= 15;
+    bofs &= ~((1 << vece) - 1);
+#endif
+
+    tcg_gen_gvec_dup_mem(vece, dofs, bofs, 16, 16);
+}
+
+#define GEN_VXFORM_VSPLT(name, vece, opc2, opc3) \
+static void glue(gen_, name)(DisasContext *ctx) { gen_vsplt(ctx, vece); }
+
 #define GEN_VXFORM_UIMM_ENV(name, opc2, opc3)                           \
 static void glue(gen_, name)(DisasContext *ctx)                         \
     {                                                                   \
@@ -858,9 +866,9 @@ static void glue(gen_, name)(DisasContext *ctx)                         \
         tcg_temp_free_ptr(rd);                                          \
     }
 
-GEN_VXFORM_UIMM(vspltb, 6, 8);
-GEN_VXFORM_UIMM(vsplth, 6, 9);
-GEN_VXFORM_UIMM(vspltw, 6, 10);
+GEN_VXFORM_VSPLT(vspltb, MO_8, 6, 8);
+GEN_VXFORM_VSPLT(vsplth, MO_16, 6, 9);
+GEN_VXFORM_VSPLT(vspltw, MO_32, 6, 10);
 GEN_VXFORM_UIMM_SPLAT(vextractub, 6, 8, 15);
 GEN_VXFORM_UIMM_SPLAT(vextractuh, 6, 9, 14);
 GEN_VXFORM_UIMM_SPLAT(vextractuw, 6, 10, 12);
-- 
2.20.1